UNIVERSITY OF
ILLINOIS LIBRARY
AT URBANA-CHAMPAIGN
MODERN LANGUAGES
UNIVERSITY LIBRARY
UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN
The person charging this material is responsible for its
renewal or return to the library on or before the due
date. The minimum fee for a lost item is $125.00,
$300.00 for bound journals.
Theft, mutilation, and underlining of books are reasons
for disciplinary action and may result in dismissal from
the University. Please note: self-stick notes may result
in torn pages and lift some inks.
Renew via the Telephone Center at 217-333-8400,
846-262-1510 (toll-free) orcirclib@uiuc.edu.
Renew online by choosing the My Account option at:
http://www.library.uiuc.edu/catalog/
oderf] Lane
•ii.
2690
OCT i 2 2006
" f ^
Studies in
The Linguistic Sciences
PAPERS /N GENERAL L/NGU/ST/CS
EYAMBA G. BOKAMBA Language and National Develop in
Sub-Saharan Africa (A Progress Report) i
GEORGIA M. GREEN Pragnibtics and Syniactic Description 27
GEORGIA M. GREEN Competence for Implicit Text Analysis:
Literary Style Discrimination in Five- Year-Olds 39
CHIN-W . KIM Epenthesis and Elision in Metrical Phonologv 57
CHARLES W. KISSEBERTH Displaced Tones in Digo
(Part 2) ,; 73
YEN LING LEE A Study on Code-Switching in Taiwan 121
JERRY L. MORGAN Some Observations on Discourse and
Sentence Grammar 137
DAVID ODDEN Evidence for the Elsewhere Condition m Shona 145
M ICH lO TSUTSU I Topic Marker Ellipsis in Japanes 1 63
CHIN-CHUAN CHENG AND CHARLES W. KISSEBERTH
Ikorovere Makua Tonology (Part 3} 181
Department of Linguistics
University of Illinois
STUDIES IN THE LINGUISTIC SCIENCES
PUBLICATION OF THE DEPARTMENT OF LINGUISTICS
UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN
EDITORS: Charles W. Kisseberth, Braj B. Kachru, Jerry L. Morgan
REVIEW EDITORS: Chin-W. Kim and Ladislav Zgusta
EDITORIAL BOARD: Eyamba G. Bokamba, Chin-chuan Cheng, Peter
Cole, Alice Davison, Georgia M. Green, Hans Henrich Hock, Yamuna
Kachru, Henry Kahane, Michael Kenstowicz and Howard Maclay.
AIM: SLS is intended as a forum for the presentation of the latest original
research by the faculty and especially students of the Department of
Linguistics, University of Illinois, Urbana-Champaign. Especially invited
papers by scholars not associated with the University of Illinois will also be in-
cluded.
SPECIAL ISSUES: Since its inception SLS has devoted one issue each
year to restricted, specialized topics. A complete list of such special issues is
given on the back cover. The following special issues are under preparation:
South Asian Linguistics: Syntax and Semantics, edited by Yamuna Kachru;
Studies in Language Variation: Nonwestern Case Studies, edited by Braj B.
Kachru; Papers on Diachronic Syntax, edited by Hans Henrich Hock.
BOOKS FOR REVIEW: Review copies of books (in duplicate) may be sent to
the Review Editors, Studies in the Linguistic Sciences, Department of
Linguistics, University of Illinois, Urbana, lUinois, 61801.
SUBSCRIPTION: There will be two issues during the academic year. Requests
for subscriptions should be addressed to SLS Subscriptions, Department of
Linguistics, 4088 Foreign Languages Building, University of Illinois, Urbana,
Illinois, 61801.
Price: $5.00 (per issue)
STUDIES IN THE LINGUISTIC SCIENCES
EDITORS
Charles W. Kisseberth
Braj B. Kachru, Jerry L. Morgan
REVIEW EDITORS
Chin-W, Kim and Ladislav Zgusta
EDITORIAL BOARD
Eyamba G. Bokamba, Chin-chuan Cheng, Peter Cole, Alice Davison,
Georgia M. Green, Hans Henrich Hock, Yamuna Kachru, Henry Kahane,
Michael J. Kenstowicz and Howard Maclay.
VOLUME 11, NUMBER 1
SPRING, 1981
DEPARTMENT OF LINGUISTICS, UNIVERSITY OF ILLINOIS
URBANA, ILLINOIS 61801
v.M
TABLE OF CONTENTS
Eyamba G. Bokamba: Language and National Development in Sub-Saharan
Africa (A Progress Report) 1
Georgia M. Green: Pragmatics and Syntactic Description 27
Georgia M. Green: Competence for Implicit Text Analysis: Literary
Style Discrimination in Five-Year-Olds 39
Chin-W. Kim: Epenthesis and Elision in Metrical Phonology 57
Charles W. Kisseberth: Displaced Tones in Digo (Part 2) 73
Yen Ling Lee: A Study on Code-Switching in Taiwan 121
Jerry L. Morgan: Some Observations on Discourse and Sentence Grammar. 137
David Odden: Evidence for the Elsewhere Condition in Shona 145
Michio Tsutsui: Topic Marker Ellipsis in Japanese 163
Chin-Chuan Cheng and Charles W. Kisseberth: Ikorovere Makua Tonology
(Part 3) 181
Studies In the Linguistic Sciences
Volume 11, Number 1, Spring 1981
LANGUAGE AND NATIONAL DEVELOPMENT IN SUB-SAHARAN AFRICA:
A Progress Report*
Eyamba G. Bokamba
If the 1960s can be considered as the decade of the struggle
for political Independence, and the 1970s as the decade of the
total liberation of Africa and the consolidation of political
power, the 1980s will and should certainly be the decade of de-
velopment in the continent. Certain questions arise, however,
regarding the direction and nature of this development. For
example, what kind of development does Sub-Saharan Africa need
and for whom? In what areas should these developments occur?
Who should initiate them and in what manner?
This paper takes up these questions, among others, and
discusses them with particular reference to the role of language
in personal and national development in Sub-Saharan Africa. The
study is subdivided into two major parts. The first part dis-
cusses development in general terms, and presents a critical
evaluation of African developmental goals against the background
of the 1961 UNESCO/UNECA Conference of Ministers of African
Education, held in Addis Ababa, Ethiopia, using comparative sta-
tistics on literacy and educational attainment in an attempt to
assess the current level of general development. The 2nd part
proposes a developmental model which takes the university and
other institutions of higher learning as the central agents of
development. This model calls for a fundamental restructuring
of the entire educational system, an increased emphasis on
functional literacy for adults, and the reformulation of the
current language policies as to concord with the new objectives.
The proposed model is argued to be more compatible with and
realistic to Sub-Saharan African needs than the current European-
based approach.
0. INTRODUCTION
The continent of Africa is suffering today from various aspects of under-
development, and theories by Western scholars abound as to why this is the
case. Some of these scholars attribute African underdevelopment to the capi-
talist, imperialist, and colonialist exploitation of the continent by Europe
and its allies (Rodney 1972) , and others attribute it to ethnicity and the
lack of trained personnel (Dumont 1963; Balandler 1966). In reading such
studies one gets the impression that if colonialism, ethnicity, and related
facets were removed African development would become an easy goal to achieve.
The fact of the matter is that the situation is much more complex than this:
there are many dimensions and unknowns to the issue.
In this paper I shall consider one of these dimensions, viz. the role
of language in development. I will attempt to demonstrate here how language
is critically involved in personal and societal development. More specifi-
cally, I would like to maintain that language, just like education, is
central to personal and national development. To demonstrate the validity
of this hypothesis, I will describe the manner in which language is involved
in educational, political, socio-cultural, and personal developments; and
will review in passing the language policies of African states vis-a-vis
education in an attempt to ascertain the extent to which they have affected
over-all development in Sub-Saharan Africa.
The paper is essentially sub-divided into two major parts. The first
part discusses development and the role of language in it, and evaluates
critically the current level and direction of development in Sub-Saharan
Africa. This evaluation is presented against the background of the goals
set out by the conference of ministers of African education held in Addis
Ababa in 1961 under the auspices of the UNESCO and the United Nations
Economic Commission for Africa (UNECA) . Statistical data on educational
developments and literacy are considered in making this evaluation. The
second part of the paper proposes and defends a developmental model that
responds to the objective realities of the region, and would permit real
development/progress therein.
Since most of the studies on developmental problems in Africa have
been largely presented from an external perspective, i.e., one which
attempts to locate the causes of these problems outside of Africa, the
present analysis focuses attention on internal causes. We have pursued
this perspective not because of methodological reasons', but because we
are persuaded that African underdevelopment is NOW an African problem.
Before we embark on the analysis proper, however, a word about the
role of language in society is in order. Language, most social scientists
will agree, is the most fundamental feature of the human society: it is
the quintessential of the homo sapiens. Ronald Langacker, one of the
leading American linguists, characterized language as follows back in 1967:
(1) Language is everywhere. It permeates our thoughts, mediates
our relations with others, and even creeps into our dreams.
Most human knowledge and culture is stored and transmitted
in language, which is so ubiquitous that we take it for
granted. Without it, however, society as we know it would
be impossible. (Langacker 1967: 3)
Langacker went on to observe:
(2) Despite its undeniable importance in human affairs, language
is poorly understood. Misconceptions about it are legion,
even among well-educated people andnot even professional
linguists can claim to understand it fully.
Langacker 's observation made fourteen years ago in the height of the
Chomskian linguistic revolution still applies today, despite the fact that
the linguistic sciences have achieved phenomenal progress since that time.
Educators, linguists and other social scientists have not yet fully appre-
ciated the nature of language and its role in society. One of these mis-
conceptions involves the role of language in general development. That is,
what role does language play in personal and national development? To
answer this question, let us first define what development is.
1.0 LANGUAGE AND DEVELOPMENT
1.1 Although there is no commonly agreed upon definition of development,
most social scientists recognize that development is a multifaceted phenomenon
that involves changes in various areas of human life: economic, educational,
cultural, and socio-political. Because of the dominant role played by econo-
mists in social sciences, the definition of development has often been pre-
sented in terms of economic development as measured by indicators such as
per capita income, GNP, and industrial outputs. Such a definition, however,
is only a partial reflection of general development.
As used in this study, the term development will be understood in the
Frankian sense. That is:
(3) Real development involves a structural transformation of the
^ economy, society, and culture of the satellite (nation) that
permits the self-generating and self-perpetuating use and
development of the people's potential. Development comes about
as a consequence of a people's frontal attack on oppression,
exploitation, and poverty that they suffer at the hands of the
dominant classes and their system (Gunder Frank, et al. 1972:
xvi).
|C At the individual/personal level, development implies increased skill and
capacity in manipulating one's environment, greater freedom and opportunities,
creativity, self-discipline, responsibility, and material well-being (Rodney
1972). In other words, it involves the mental faculty as well as physical
factors.
Frank et al.'s (1972: xvi) definition given in (3) not only contrasts
significantly from most single-issue oriented definitions of development, but
also considers human beings, rather than the aggregate of given factors of
production (e.g., land, population, capital, technology, specialization and
mass production), as the key factor in development. Notice, e.g., that by
referring to the exploitation and utilization of "the people's potential,"
Frank et al.'s definition presupposes that the people are and ought to be
capable of perceiving the societal ills that affect them, and are/ought to
be in a position to propose solutions to them. The question that arises at
this point is whether a significant segment of the Sub-Saharan Inhabitants,
i.e., about 60% or more, has such perception and knowledge to improve its
mode of life. My contention is that the vast majority of the people in this
region do not possess, at this point in time, the knowledge needed to
ameliorate their life. Let us examine why this is so.
In order to make the changes called for in Frank et al.'s definition of
development, the target population must have what has been termed "develop-
mental knowledge" (Boulding 1966, Uchendu 1980). According to Boulding
(1966), developmentally relevant knowledge consists of three stages: (1) folk
knowledge, which involves mainly the ability to run one's family and carry
out basic interactions and relations with others; (2) literary knowledge,
which consists essentially of the diffusion of knowledge from one generation
and/or group to another through the written word; and (3) scientific age
knowledge, which involves technological knowledge and its application.
Sub-Saharan African states, according to this subcategorization of de-
velopmental knowledge, fall largely into the first category: folk knowledge,
-i^^is classification is dictated by several factors, most of which are related
to language -and education. First, Sub-Saharan African states are at the folk
kn3wle^gS~~sTage becai/Be of rampant illiteracy. According to the latest avail-
able UNESCO statistics, illiteracy in Africa ranges on the average from 70-90
percent, as can be seen in Table I. The over-all illiteracy rate for the
continent is 73.7 percent, the highest percentage for any continent or sub-
continent, as indicated in Table II where this figure compares with 46.8
percent for Asia and 23.6 for Latin America. What is particularly revealing
and disturbing about the statistics for Africa is that illiteracy is con-
sistently higher among women throughout the continent ;■'• yet it is the women
who are the main custodians of the early educational development of the
children in most African nations. Because of this rampant illiteracy situa-
tion, the ability of the people to bring about significant changes in their
modes of life is considerably limited. The limitation is both from the point
of view of initiating original changes and applying transferred knowledge
from other societies to transform their own.
4^
j^;^cond, the -ajjnost exclusive use of European languages (i.e., English,
French, and PortugueseJ^as the media of instruction militates against the
stablishment of mass-education and permanent literacy. In spite of the
efforts made by many African nations since the advent of political inde-
pendence in the early 1960s, the provision of education to secondary and
university age students remains an impossible dream as access to these
levels of education continues to be severely circumscribed by the language
factor (cf. Champion 1974, Bokamba 1976, Palm and Beloncle 1976, Maloba
1977). The statistics given in Table III partly reflect this problem. The
low enrollments in the secondary and university education also reflect, his-
torically, colonial policies which were designed to restrict post-primary
education to a handful of Africans in order to minimize their participation
in the administration of the colonies.
Third, although the number of university graduates in the region has
significantly increased in absolute terms since the advent of political
independence, scholarly research emanating from these intellectuals has
remained insignificant insofar as its contribution to development is
concerned. The reason for this is that a few African governments devote
any funds to research; there is, therefore, no encouragement for scholars
to undertake either practical or theoretical research. Related to this is
the fact that, unlike in developed countries, most African universities
and other institutions of higher education are not considered and/or used
as agents of development. As a result, most African scholars are not
making effective use of the literary and technological knowledge that is
available to them.
(Mr>.O0«OO9>OOO«OiAOOO<-4i0e4«^O
•«0eo0<0000000r«r«000
• •••*•••■•«.«• — — ^^ ^^ ' ^
0»oo»0«Ooooor«.«00>r^t*in«o<»>r>»oovOirno«»x,Ooo«Ar«.eoo^^SS"»5S>S???»
*00«nrHsO^>»e>iOO«cir(^Oirt,-««n«r>a>m<o«o>ov0^mooooo»^ff>^r«.r>.
_|t-t-(i-l.-lrHl-l.-l^rHrHr^>0,-lr-IMr-lr-l.-t.-4r^r-l,-l^rHr-(f-tr^t-I.H.H,-l,Hr^.-trHr^
,^,H,H^.H,HiHr-li-(iHiHMr-tiHMt-lr-tt-liH.HiHiHf-l,HrHiHiHrHrHFHiHi-trHf-t.-lr-(f-l
a B
a a a ^ -H
O O O w 4J .^
i-i r-i •-) a a u
a & Oi a o. ij
ri 3 « M U □
• u «i o ■ 5
W a -H r-< -^
Table II: Estimated Adult Population £■ Literacy by Continents, Major Areas
& Groups of Countries Around 1960 & 1970
ContincnU. Contincnti.
mj)or areas and grandcs nc'onx et
groups of couatries^ groupes d« pays*
Around I960 / Ven I9<0
Adull
popubtion
IS ytars IMitcrala llliieracy
•nd ov«r adults parcentafe
Arauad 1970/ Vcn 1*70
Adull
poputalfon
13 rears Illil
Adult*
(000)
Euainbl* du monde -
Afrique
Amtrique
Europe and U.S.S.R. Europe e( U.R.S.S.
Developed counlriei Pays diveloppis
Uevelopingcounlrid Pays en voie de
develop pemen I
Africa (excluding Afrique (non compris
Arab Stales) las £tals arabes)
Northern America Am<rique scptentrlonale
Latin America
Asia (excluding
Arab Stales)
Amirique laline
Asie (non compris
les £tats arabes)
£tats orabes
MF
1 869 000
735 000
39.3
1 134 000
2 287 000
783 000
34.2
1 304 000
M
916 000
307 000
33.5
609 000
1 127 000 :
315000
28.0
812000
F
953 000
428 000
44.9
525 000
1 160000 '
468 000
40.3
692 000
MF
153 000
124 000
81.0
29 000
194 000
153 000
73.7
31 too
M
76 000
55 800
73.4
20 200
96 000
60 900
C3.4
33100
F
77 000
68 200
88.5
8 800
98 000
82 000
83.7
16000
MF
259 800
43 300
16.7
216 500
324 000
41 100
12.7
283 000
M
128 300
18 700
14.6
109 600
159 000
16 950
10.7
142 000
F
151500
24 600
18.7
106 900
165 000
24 100
14.6
141000
MF
982 000
542 000
55.2
440 000
1 237 000
579 000
46.8
658 000
M
494 000
224 000
45.3
270 000
624 000
231 000
37.0
393 000
F
488 000
318 000
65.1
170 000
614 000
348 000
56.7
266 000
MF
463 500
24 500
5.3
439 000
S2I000
18 700
3.6
302000
M
212 700
7 700
4.6
205 000
243 000
S800
2.4
237 000
F
250 800
16 800
6.7
234 000
278 000
12900
4.7
263000
MF
10 600
1 200
11.5
9 400
13 200
1400
10.3
11 800
M
5 300
530
9.9
4 800
6 600
380
8.8
6000
F
5 200
680
13.0
4 500
6600
780
11.9
3800
MF
685 400
33 900
4.9
651 500
785 600
27 300
3.J
738 300
M
321 000
11 600
3.6
309 400
371600
9 300
2.5
362 300
F
364 400
22 300
6.2
342 100
414 000
18000
4.3
396 000
MF
1 183 600
701 100
59.2
482 500
1 502 400
756 000
30.2
746 000
M
595 000
295 400
49.6
299 600
756 400
306 000
40.4
431000
F
588 600
405 700
68.9
182 900
746 000
450 000
60J
296 000
MF
116 300
94 200
81.0
22 100
146 200
108 000
73.9
43 300
M
57 600
42 800
74.3
14 800
72 100
46 600
64.6
31 100
F
58 700
51 400
87.6
7 300
74 100
61300
82.7
14 400
MF
136 600
3 300
2.4
133 300
161 000
2300
1.3
138 000
M
66 900
1 300
1.9
65 600
78 000
830
1.1
77 200
F
69 700
2 000
2.8
67 700
82 800
1600
1.9
81200
MF
123 200
40 000
32.5
83 200
163 000
18 600
23.6
125 000
M
61400
17 400
28.4
44 000
81000
16 100
19.9
63 000
F
61 800
22 600
36.6
39 200
82 200
22 500
27.3
60 000
MF
966 000
529 000
54 8
437 000
1 217 500
564 100
46.3
653 400
M
48S90U
218 0(10
44.9
267 900
613 500
224 500
36.6
389 000
F
480 100
311000
64.8
169 100
604 000
339 600
36.2
264 400
MF
52 700
42 800
81. 1
9 900
68 300
49 900
73.0
18 400
M
26 500
19 000
71.6
7 500
34 400
20 800
60.3
13600
F
26 200
23 800
90.7
2 400
33 900
29 100
«5.7
4 800
Source: UNESCO Statistical Yearbook 1972. Paris: Unesco, 1973.
TABLE III: Primary and Secondary School Enrollnents In Africa on the Eve
of Political Independence
Country fi Year of Survey Primary Sch. X Secondary Sch. %
Basutoland 1958
CLesotho)
Bechuanaland 1958
(Botswana)
Central African Rep. 1958
Chad 1959
Congo-Brazzaville 1958
Congo-Kinshasa 1959
(Zaire)
Dahomey 1959
Ethiopia 1958
Gabon 1957
Gambia 1958
Ghana 1959
Guinea 1959
Ivory Coast 1957
Kenya 1958
Liberia 1959
Mali 1957
Niger 1957
Nigeria 1958
Ruanda-Urundi 1958
Senegal 1957
Sierra Leone 1959
Somalia 1958
Sudan 1959
Swaziland 1958
Tanganyika 1958
Uganda 1959
Zambia 1958
Source: Conference of African States on the Development of Education.
in Africa, Addis Ababa. 15-25 May. 1961: Final Report. Paris:
UNESCO, 1962.
119,478
90.5
3,042
4.5
31,193
46.4
485
1.4
45,774
27.2
1,480
0.9
53,973
13.8
1,473
0.4
78,962
70.3
3,259
3.0
1.460,753
71.5
51,671
3.0
81,107
31.3
3,618
1.4
158,005
3.8
8,144
0.5
39,763
65.7
1,156
2.0
4,595
10.7
794
2.2
483,425
66.7
178,581
29.4
79,373
19.7
4,563
1.1
125,727
32.7
5,104
1.4
651,758
52.1
20,291
3.9
55,026
22.4
3,397
3.3
42,053
7.7
2,749
0.5
11,811
3.3
395
0.1
2,545,336
42.9
117,414
2.9
246,149
35.5
5,480
0.9
80,473
23.8
6,102
1.9
74,481
21.0
8,277
2.8
16,485
10.2
1,828
0.8
288.395
12.8
60,941
6.5
29,934
55.9
1,066
4.5
422,832
24.1
15,315
2.1
501,699
52.2
41,653
4.4
243,926
53.9
4,948
2.6
1.2 Having thus discussed in general terms the prerequisites of de-
velopment and the current status of Sub-Saharan African nations, let us
consider in more specific terms the developmental problems that these
nations are confronting. I should like to focus my attention in this
section on the developmental goals set out in the 1961 UNESCO/UNECA
Conference of African Ministers of Education held at Addis Ababa, Ethiopia,
and on the effect that the language policies has had on the non-achievement
of these goals.
When the UNESCO and UNECA convened a conference of Ministers of
Education in Addis Ababa in 1961, a year after the advent of political
independence of most African states, the primary and secondary school
enrollments in the Sub-Saharan countries were as in Table III; and most
of the countries had only a handful of university graduates. As a result
of this and the prevailing economic situation at that time, the conference
outlined an educational plan which was keyed to provide both economic growth
and social progress in the continent during the next twenty years, i.e.
1961-80. The participants made the following recommendations, among others:
(A) a. that the development of natural resources and
b. the content of education should be related to economic needs,
with greater weight being given to science and its applica-
tions,
c. that African nations should aim at providing universal ■
primary education while at the same time giving special
attention to adult and on-the-job training (UNESCO 1962: 16).
The long term objectives for the continent included the following:
■ (5) a. universal free and compulsory primary education by 1980;
b. the enrollment of 30 percent of the children who complete
elementary education, i.e., 23 percent of the age cohorts,
into secondary schools;
c. the provision of higher education mostly in Africa itself
to about 20 percent of those who complete secondary schools,
i.e., two percent of the age group.
d. constant improvement of the quality of African schools and
universities (UNESCO 1962).
The Conference strongly recommended reforms in the content of education as to
better respond to African needs and realities. African leaders were further
urged to combat illiteracy with all available means.
While enrollments at all levels (i.e., primary to university) have sig-
nificantly increased, in many cases doubled and tripled (see Table IV),
since the UNESCO/UNECA conference, none of the four long range goals has
been met. Further, illiteracy continues to be in the 80 percentile as seen
Table IV: Enrollment
at 1st &
2nd leve
ils: growth rates 6. p
ercentage distribt
ition
% Distribution .
of enrollment
Average annua
1 growth
I rate
between
1st &
2nd 1
evel
First
level
Second
1960-65
level
1965-70
First level
1960 1965 1970
Second le
1960 1965
vel
1960-65
1965-70
1970
Algeria
11.5
6.8
2.0
12.4
86.8
91.1
88.8
13.2
8.9
11.2
Angola
15.7
14.7
18.0
13.4
88.7
87.6
88.3
11.3
12.4
11,7
Botswana
12.7
4.7
22.0
23.0
98.0
97.3
94.1
2.0
2.7
5.9
Burundi
9.6
4.3
12.8
9.2
96.5
96.1
95.2
3.5
3.9
4.8
Cameroon
8.8
5.3
21.0
13.8
96.8
94.6
92.2
3.2
3.4
7.8
Central African Republic
14.2
6.5
17.0
19.7
97.0
96.1
94.0
3.0
3.9
6.0
Chad
17.8
2.2
29.0
10.0
97.4
96.0
94.3
2.6
4.0
5.7
Congo
10.0
5.2
22.0
16.7
95.1
92,2
87.5
4.9
7.8
12.5
Dahomey
8.0
5.9
19.4
12.7
94.8
91.6
88.9
5.2
8.4
11.1
Egypt
4.1
1.6
13.3
7.2
84.0
77.4
72.3
16.0
22.6
27.7
Equatorial Guinea
2.7
2.3
24.0
21.0
96.9
92.4
84.0
3.1
7.6
16
Ethiopia
8.6
11.6
42.0
19.5
96.3
87.2
82.9
3.7
12.8
17.1
Gabon
6.8
4.9
17.1
9.3
95.0
82.3
90.7
5.0
7.7
9.3
Gambia
12.4
6.3
17.2
6.5
81.0
77.5
77.4
19.0
22.5
22.6
Ghana
16.3
0.1
27.0
5.1
96.5
94.8
93.5
3.5
5.2
6.5
Guinea
11.2
3.4
19.8
22.0
91.1
87.5
75.1
8.9
12.5
24.9
Guinea Bissau
-0.8
10.3
-8.4
32.0
91.6
94.2
86.9
8.4
5.8
13.1
Ivory Coast
8.2
7.4
21.0
16.6
95.2
91.7
88.1
4.8
8.3
11.9
Kenya
5.9
6.5
14.9
20.0
96.6
95.0
91.9
3.4
5.0
8.8
Lesotho
4.3
1.7
5.1
13.3
97.8
97.7
96.2
2.2
2.3
3.8
Liberia
6.2
-2.0
26.0
12.0
94.7
88.4
82.7
5.3
11.6
17.3
Libyan Arab Republic
9.4
12.2
12.8
12.2
88.3
86.7
86.7
11,7
13.3
13.3
Madagascar
8.3
6.9
17.9
11.0
93.9
90.9
89.2
6,1
9.1
10.8
Malawi
3.0
1.4
27.0
2.2
98.9
96.9
96.8
1,1
3.1
3.2
Mali
20,0
-7.3
-7.3
22.0
93.3
98.1
96.5
6,7
1.9
3.5
Mauritania
12.3
9.8
30.0
14.0
95.3
90.8
89.1
4.7
9.2
10.9
Mauritius
3.7
2.3
8.2
5.3
82.6
79.4
76.9
17.4
20.6
23.1
Morocco
7.0
1.0
19.7
7.2
90.2
84.1
79.7
9.8
15.9
20.3
Mozambique
0.3
3.3
8.6
*12.5
96.5
94.8
*92.3
3.5
5.2
*7.7
Niger
18.4
-7.1
19.2
15.5
94.9
94.8
92.7
5.1
5.2
7.3
Nigeria
-0.1
3.8
8.5
7.3
94.6
92.1
90.8
5.4
7.9
9.2
Rwanda
4.6
4.7
5.2
1.4
97.6
97.6
97.9
2.4
2.4
2.1
Senegal
. . .
3.7
23.0
9.4
91.3
86.3
82.9
8.7
13.7
17.1
Sierra Leone
7.9
5.7
12.6
17.3
90.8
88.9
82.6
9.2
11.1
17.4
Somalia
6.1
3.4
25.0
18.2
86.0
73.4
58.5
14.0
26.6
41.5
Sudan
6.1
14.1
6.6
6.5
81.9
81.5
86.2
18.1
18.5
13.8
Swaziland
7.5
6.9
13.3
22.0
95.3
93.9
89.0
4.7
6.1
11.0
Tanzania, United Republic
9.3
3.8
4.7
10.1
95.3
96.2
95.0
4.7
3.8
5.0
Togo
8.6
8.0
19.1
10.8
94,9
92.2
91.2
5.1
7.8
8.8
Tunisia
10.3
5.0
11.0
12.2
88,2
87.9
83.9
11.8
12,1
16.1
Uganda
1.3
4.8
-3.0
7.7
93,2
94.4
93.7
6.8
5,5
6.3
Upper Volta
9.8
3.2
21.0
7.0
95.0
92,2
90.8
5.0
7,8
9.2
Zaire
5.9
8.4
17.6
17.3
96.7
94.6
92.2
3.3
5,4
7.8
Zambia
7.4
11.1
33.0
21.0
98.2
94.9
92.5
1.8
5.1
7.5
Total
6.9
5.0
13.0
9.8
91.8
89.3
87.0
8.2
10.7
13.0
Source: Conference of 1
Ministers
; of Education o
f Afri
can Member States,
, Lagos,
January 27-Feh
iruary 4,
1976.
Final Report.
Pari!
3: Unei
SCO, 1976.
10
previously. Several questions naturally arise at this point. First, why
have these nations failed to achieve the goals set; out in 1961? Second,
what specific factors have contributed to the continuation of* high illit-
eracy rates in the reg^ioa? And third, how can African states realistically
approach th§, question of general development? That is, what steps should
be taken to overcome current "underdevelopment," to use a popular but loaded
term.
These are complex questions that require more detailed answers than time
and space will allow me to discuss here, but I shall present what I consider
to be the most fundamental pitfalls of the African educational and develop-
mental plans. liCt us examine, first of all, the first two questions, viz.
why African nations have failed to achieve the 1961 conference objectives
and why illiteracy has continued unabated.
There are two but related answers to these questions. First, it is an
accepted fact that the social, educational, political, and economic diffi-
culties that African states are currently experiencing have some of their
roots in the colonial era. For example, the balkanization of Africa at the
Berlin conference of 1885 created new countries without regard to cultural
and linguistic boundaries. The integration of these nation-states into
nations during the colonial period was less than successful as the colo-
nizing powers failed to develop the needed socio-economic infrastructure,
including transportation systems, to link various parts of the country to-
gether. Further, their policies of restricting educational opportunities
to a small minority of the population prevented the development of an
effective elite that would have governed these countries after the advent
of political independence.
r_^ >-|feSecond, over and above these historical considerations, the current
difficulties are largely due to the failure of African leaders to articulate
and pursue coherent educational plans and language policies keyed to over-all
development. With the exception of Guinea (in West Africa) and Tanzania,
: for instance, most Sub-Saharan nations have kept intact the colonial edu-
cational systems. The few changes that have been made have been cosmetic,
rather than substantial. These European-inspired educational systems do
not respond to the developmental needs of Africa, yet these countries
continue to devote between 25 to 35 percent of their national budgets on
education. Secondary education, for instance, continues to be restricted
ly^on the average to less than 10 percent of the primary school graduates,
and university education to less than one percent of the secondary school
graduates In most African nations.
Worse yet, is the fact that educational and political leaders have yet
to articulate realistic goals for each cycle of education. In a field
research conducted in 1976 in Botswana for his doctoral dissertation, Josiah
Tlou (in personal communication) found that most of the parents and teachers
questioned about the goals of primary, secondary, and university education
for their children/students responded as follows:
(6) a. Goal of elementary education: prepare for secondary education
b. Goal of secondary education: preparation for university
education
11
c. Goal of university education: preparation for a degree and
a job.
That is, all pre-university education is preparatory to a university educa-
tion; and none of these, except the university education, is terminal in
itself. These answers, which reflect the prevalent thinking in many African
countries, would be impeccable if African needs were analogous to those of
Western Europe, and if a large number of primary and secondary school
graduates had access to secondary and university training. As it is now,
none of these is the case (see Table IV) .
Secondary and university educational opportunities are restricted to
a small percentage of the primary and secondary school graduates not because
of the availability of jobs, but rather by virtue of the testing system and
the language policies practiced by Sub-Saharan African states. In particular,
African educational systems require a student to pass all subjects taught at
his/her level with a certain percentage. If the student fails one of these
subjects, he/she must take another examination on that subject regardless of
his/her high performance on other areas. Failure in the retake examination
forces the student to repeat the class regardless of his/her performance on
other subjects. As a result, about 50 percent of the students in many
African educational institutions are class-repeaters.
If a student is successful on all subjects at each level and completes
his/her education, e.g., primary education, he or she must take an admission
examination into the secondary school cycle. After the successful completion
of this cycle, he/she must submit to a university admission examination. In
most countries each examination consists of two subjects: mathematics and
the language of instruction (e.g., English or French or Portuguese). The
student must pass both subjects in order to be admitted into the educational
cycle for which he/she is being tested.
A number of studies have shown that this kind of educational system is
exorbitantly expensive in that it produces high attrition rates and a high
percentage of class-repeaters (Rideout , et al. 1969; Champion 1974;
Belloncle and Palm 1976) . The loss of potential human resources is con-
siderably high as drop-outs are not channelled to other educational facili-
ties. While the system of education and the quality of teaching contribute
also to low school outputs, there is considerable consensus among African
scholars that the language policies of Sub-Saharan states are largely
responsible for the poor performance of the education sector (cf. e.g.,
Bot Ba Njock 1974, Belloncle and Palm 1976; Champion 1974; Bokamba 1976;
Bokaraba and Tlou 1977; Kahombo 1980). To see why this is the case we need
to review briefly the language policies of African states, the conditions
under which the official languages are learned and used, and discuss the
problems inherent in second language acquisition in a non-native context.
V- 1.3 Language policies in Africa can be viewed from two different
perrods: the colonial and the post colonial eras. With the benefit of
hindsight, it can be said with little hesitation that language policies
during the colonial era in much of Sub-Saharan Africa were highly restric-
tive in the sense that they were formulated to best serve the interests of
the colonial masters, rather than those of the colonized people.
12
In particular, two different approaches , which may be characterized
as assimilatory and accommodating, were adopted. The French and the
Portuguese, whose colonial philosophies called for the assimilation of the
African people to a higher "civilization," espoused the first approach:
assimilation; while the Belgians and British opted for the second, accommoda-
tion. French and Portuguese colonialists felt that they had a civilizing
mission to carry out in Africa, and that the use of their languages in
education and administration was an obligatory and integral part of this
mission (Spencer 1971; Bokamba and Tlou 1977).
In this connection, the French, e.g., regulated the use of languages
in education through a series of ordinances. The first of these was the
metropolitan ordinance of Villers-Cotteret of 1539 which forbade the use of
languages other than French in all official functions within the territories
of France, including its colonies (Spencer 1971; Bokamba and Tlou 1977).
The second ordinance, promulgated on February 14, 1922, regulated private
education and religious teaching in overseas territories. It stipulated
that,
(7) General education must be carried in French... The Coranic
schools and Cathechist schools are authorized to provide ex-
clusively a religious education in the vernaculars. Such
schools are not considered as institutions of public education
(Spencer 1971: 543).
Similar ordinances governed the use of Portuguese in the former Portuguese
colonies of Angola and Mozambique. As a result of these linguistic policies,
the French and Portuguese colonies never developed indigenous language
policies whereby African languages were used as either media or subjects
of instruction in formal education.
In the former Belgian and British colonies-, in contrast, indigenous
language policies advocating the use of African languages as media and
subjects of instruction in primary and secondary education developed thanks
largely to the dominant influence of missionaries in education. The
Belgian and British educators encouraged such a policy, because they sub-
scribed to the principle advocated during the first decade of this century,
and reechoed by the UNESCO in 1953, that the most effective medium of in-
struction in the preliminary stage of a child's education is his mother
tongue (UNESCO 1953; Lewis 1962; Gorman 1974; Bokamba and Tlou 1977).
Accordingly, missionary linguists in countries like Ghana, Nigeria,
Cameroon, Zaire, Rwanda and Burundi, Tanzania, and Kenya described and
produced books in several major languages which became subsequently the
linguae francae of each of these nations (see Bokamba and Tlou 1977 for
details). In almost all the Belgian and British colonies the language
policy called for the use of such linguae francae as the media of in-
struction in the first three years of elementary education, and for their
use as subjects of instruction up to the secondary school and teachers'
colleges.
13
While theoretically the Belgian and British language policies re-
stricted the use of African languages in education to the first three
years of elementary school and Introduced the use of the official language,
French or English, in the fourth year, the inadequacy of qualified teachers
forced the schools to extend the use of African languages to the entire
elementary cycle (Bokamba 1976; Bokamba and Tlou 1977). These policies,
however, permitted and accelerated the development of several major
languages in each country in the Belgian and British colonies; whereas
the French and Portuguese policies prevented such developments.
The dichotomy which existed between the French-Portuguese policies and
the Belgian-British policies during the colonial era disappeared, however,
at the advent of political independence in the early 1960s as African nations
adopted policies which required the use of the official languages as the sole
media of instruction at all levels of formal education. in particular, the
former British colonies of Ghana, Nigeria, Sierra Leone, Kenya, Uganda,
Tanzania, Zambia, Malawi went for what was then known as "straight for
English" from grade one onwards. The former French colonies and/or pro-
tectorates adopted the same type of policy with regard to the use of French
(Bokamba and Tlou 1977).
This total Europeanization of the language policies was not surprising,
as it had been occurring gradually in the late 1950s. The teaching of the
major African languages in the former Belgian and British colonies continued,
however, until around the mid 1960s when the programs were phased out in
some of the countries (e.g., Ghana, Nigeria, and Zaire).
Three main arguments have been advanced in favor of the retention of
the European languages as the sole media of instruction in the former French
and Portuguese colonies and for the total Europeanization of the policies
in the former Belgian and British colonies. These are: (1) efficiency and
expediency; (2) national integration; and (3) national progress (cf. Bokamba
and Tlou 1977). In the words of Dr. Dowuona (1969: 3):
(8) The reasons behind all of these changes were partly political
and partly practical. On the one hand, politicians striving
for national unity, for the supression of tribalism, for rapid
industrialization and accelerated economic jievelopment, saw in
'^h^nian ^'''ig.'ig£gg j. hg^rXJ-gX- to prngrpss. On the other hand,
the vast majority of the people themselves wanted to enter
quickly into the new material civilization to which a knowledge
of English provided one of the keys. Rapid development, it was
felt, could be achieved through a knowledge of English, and new
experiments in English as a medium of instruction right from
the first year of school were begun in the so-called Experimental
Schools.
Duwuona, who was only speaking about the language policy of Ghana, reflected
accurately the prevalent thinking of that period in the continent.
I have argued in detail elsewhere (Bokamba 1976; Bokamba and Tlou 1977)
that while some of the reasons cited in favor of the use of European languages
in education are true, the conclusion that African languages cannot adequately
14
serve the educational needs of Africans is unwarranted. It suffices to
point out here that the fallacy here stems from two basic assumptions in
the argument of the proponents of the Europeanization of the language
policies in Africa: First, it was and is still assumed that the adoption
of an African language as the medium of- instruction automatically precludes
the teaching of international languages such as French, English, and
Portuguese. This, of course, is not true. Second, it is assumed erroneously
that because major African languages have not yet been fully developed in
all ranges or registers of discourse, they cannot eventually be so developed.
Ethiopia, Somalia, and Tanzania where indigenous national languages (i.e.,
Amharic, Somali, and (Ki-)Swahili, respectively) have been adopted as the
media of administration and education are showing that this is possible.
The main reasons for the preference of Europeanized language policies
over Africanized ones, it seems to me, are assimulation and expediency:
it was both practically and politically expedient for African leaders to
continue with and/or adopt their former colonial master's languages as the
sole media of instruction. This short term solution, in my considered
opinion, also had its political benefits: it permitted the new small elite
to retain power and other privileges they had without competition. (see
Scotton 1978 for an interesting discussion of this point.) Whether or not
this is accurate is not at issue here; what concerns us is the over all
evaluation of the present policies vis-I-vis development. How successful
have the current policies been and on what basis do we evaluate them?
Although I do not have empirical evidence regarding the actual effect
of the current language policies vis-^-vis education in Sub-Saharan Africa,
there is strong circumstantial evidence to suggest that the failure of
African states to achieve the developmental goals set out at the 1961 Addis
Ababa Conference is largely attributable to the current languag^ policies.
There'^re severalT reasons for thisl '
^ (9) a. The European languages which are currently serving as
the official languages and sole media of instruction in
Africa are learned as third or fourth languages in multi-
lingual societies where illiteracy is rampant and where
over 90 percent of the population does not speak them;
b. The vast majority of the elementary teachers who are
i~ required to teach every subject in these languages have
a poor command of them at best. Their secondary school
colleagues do not fare any better.
V
Since the official language is rarely, if at all spoken
at home, the student receives no reinforcement at home
nor at the community level. The combination of these
factors, therefore, places the learner in the most un-
favorable condition imaginable.
Needless to point out, this situation has impacted negatively on the
students' academic performance. Several African and Africanist educators
and researchers have in fact observed that the foreign language skills
of primary and secondary school graduates are highly inadequate as to
permit them to function well in the next cycle of studies (see, e.g..
15
Verbeke 1966, Ansre 1968, Champion 1974, Bot Ba Njock 1974). In an extensive
experiment conducted in the elementary schools of Kinshasa, Zaire in 1965,
for example, Verbeke (1966: 456-57) found that elementary school graduates
had a maximum vocabulary baggage of 1,000 words after ten hours of French
studies per week f6r six years. According to Verbeke, the minimum vocabulary
acquisition that an elementary school graduate should have in order to
function adequately in secondary school 16 between 2,000 and 3,000 words.
Findings similar to Verbeke 's have been reported from other African nations
(Bot Ba Njock 1974, Champion 1974).
We know from studies on first language acquisition that the passive
vocabulary of a 24 year child ranges between 2,000 and 3,000 words, and that
by the time he/she reaches the age of 45 months his/her vocabulary has
increased significantly and his/her grammatical competence approximates
that of a native speaker. By the time that this child begins school at the
age of five years he/she has almost complete mastery of his/her language.
If Verbeke 's findings are taken as an accurate reflection of the level of
linguistic development of the average African using an official language, it
becomes evident why, given the learning conditions described above, attrition
rates are extremely high in African schools (cf. Rideout 1969, Bot Ba Njock
1974). Students simply have increasing difficulties in assimilating materials
in a language that they control so poorly.
This fact is clearly attested at the end of the primary and secondary
education cycles where the greatest attrition rates occur. These areas
constitute the greatest bottlenecks precisely because the admission exami-
nations are given at these junctures. As stated previously, the examina-
tions involve superficially two subject areas: mathematics and the language
of instruction (i.e., English or French or Portuguese); but in fact what
the students are tested on are two aspects of the same subject: the language
of instruction. In one case they are tested on the language skills per se,
and on the other they are examined on the application of that language to
mathematics.
There is a general consensus that this system of selection perpetuates
unnecessary elitism, while at the same time barring educational opportuni-
ties to otherwise competent individuals. This is particularly true for
students who are weak in language skills but strong in other subjects, e.g.,
sciences and mathematics (cf. Maloba 1977). One of the consequences of
this selection method is that there are more humanity and social science
majors in African universities than (pure) scientists. That is, potential
science majors get weeded out before they reach the university level. What
is deplorable about this situation is that government leaders do not under-
stand why there are so few students specializing in sciences at a time that
expertise in these fields is so much in demand.
If the lack of opportunities for higher education is deplorable at
the national and group level, it is even more so at the personal level.
This is the case because education is the key to a better life. Education
provides an individual with economic rewards, careers, social prestige,
other social values, and a new way of thinking and dealing with life. The
level of education that an individual attains often determines his upward
mobility in society and his ability to participate actively in its
16
transformation. Denial of educational opportunities to an individual deprives
him not only of these privileges, but also robs the society of the contribu-
tion that such a person could make. In short, the present educational systems
in Africa are inadequate, because they are ill-adapted to the continent's
needs and contemporary world realities. That this should be the case is not
at all surprising for, as Bot Ba Njock (1974) correctly observed, "the school
of yesterday is an imported product, conceived under different circumstances,
for different societies, and is a different answer to different problems."
That is, this kind of education was conceived to respond to a colonial situa-
tion, but not to an independent Africa.
In addition to these areas of development, the current language policies
have also affected political development in Sub-Saharan Africa in a number
of respects. Consider, for instance, the question of participatory democracy.
The advent of political independence in the 1960s was expected to inaugurate
a new era in which the principles of equality before the law and government
by and for the people would be the guiding lights. But this expectation has
not materialized for several reasons which we need not go into in this paper,
except one: the role of language in participatory democracy.
Kelman (1971) and others (e.g., Fishman 1972, Edelman 1977) have argued
persuasively that language is a uniquely powerful force in unifying a diverse
population and in involving individuals/subgroups in the national political
system by providing sentimental and instrumental attachments, and often by
inducing acquiescence (Edelman, 1977). According to Kelman (1971: 25),
(10) An individual is sentimentally attached to the national system
to the extent that he sees it as representing him - as being,
in some way, a reflection and an extension of himself.
Similarily,
(11) An individual is instrumentally attached to the national system
to the extent that he sees it as an effective vehicle for
achieving his own ends and the ends of members of other systems.
_, Assuming that under the best conditions the legitimacy of a national political
system resides on the consent of the governed, Kelman goes on to observe that
such a system is seen as deserving of an individual's loyalty to the extent
that it "provides for a smoothly running society in which individuals can
participate to their mutual benefit and have some assurance that their needs
and interests will be met."
V The concepts of sentimental and instrumental attachments suggested by
Kelman (1971) and the role they play in facilitating national integration
are obviously crucial for political development. As is well known, most
African nations have a single political party system. In spite of the
apparent difficulties that such a system presents in the exercise of politi-
cal rights, it can be said with little hesitation that citizens of liberated
Sub-Saharan nations have the right to exercise the principle of one-man-one
vote. Their other political rights, however, are considerably curtailed,
if not completely denied, by the current European-oriented language policies.
17
Specifically, given that English, French, and Portuguese, rather than
African languages, are the official media of administration, commerce, and
diplomacy, and that the average citizen does not speak any of these lan-
guages, he/she cannot meaningfully participate in the political affairs of
his/her country. In particular, the average citizen does not know what his/
her constitutional rights are partly because the constitution is written in
a European language. He/she does not know what his/her legal rights are,
because the laws are also written in a European language that he/she cannot
speak or understand; he/she cannot understand the party ideology, political
goals and directives, nor can he/she follow parliamentary debates on im-
portant issues of his/her nation because they are conducted in a foreign
language. Further, the average citizen cannot become a parliamentarian,
because of the official language barrier. In short, Mr. or Mrs. Citizen
has very little sentimental and instrumental attachments to the national
political system largely due to the language barrier.
The implications that this situation have for participatory democracy
and integrative political developments are rather obvious. Given the fact
that illiteracy is rampant throughout the region, and that an estimated
70-80 percent of the population of most African states is rural, the
pursuit of the current language policies not only makes participatory
democracy an impossible dream, but also denies the society as a whole the
possible contribution that the vast majority of the population can make
towards nation-building and development. One clear consequence of this
situation has been the questioning of the legitimacy of the national
political system as evidenced in the various civil wars and coups d' etat
which have occurred in the past two decades. These conflicts have in
turn deepened the dependency of the state to the former colonial powers
and their allies. Another implication of the lack of sentimental and
instrumental attachments is that political development and integration
are made difficult to achieve, and an internal colonial situation is
created as a small ruling political elite maintains power without the
consent of the governed through coercion. As will be discussed shortly,
this situation has often led to the lack of meaningful or people-oriented
economic development.
The inability of the average citizen to participate meaningfully in
the formulation of political policies also translates into an inability to
contribute to the formulation of the economic ones. This naturally follows
from the fact that economic policies and/or decisions are derived from
political options. Consider in this regard, for instance, the case of the
economic Infrastructure found in most African nations. This system is
based on what is referred to in economic literature as the ."trickle down
theory" or "big is beautiful" whereby industrialization is seen as pro-
viding wealth and economic development from the urban centers to the rural
areas, and from the rich urban dweller to the average citizen. This is
the economic structure which has brought big tractors to African farms, big
extractive industries in urban centers, and export-import businesses to
such centers.
As we know today, however, this type of economic infrastructure,
largely inherited from the colonial era, has produced mainly economic
■A
18
growth but not economic development. What appears to be often ignored by
African proponents of the large-scale industrialization approach is that
such a system is highly dependent on western technology and technicians,
which African nations do not have, and also based on what John Weeks (1975:
92-93) terms "elite consumption development dynamics." For example, the
manufacturing industries drain important resources from the rural regions
and produce goods that cater largely to the elite class and other urban
dwellers without providing corresponding returns to the rural inhabitants.
The second factor often ignored or taken for granted by African leaders is
that industrialization which is based on the capital-intensive approach
seeks, first of all, to maximize profit; and as such, will largely restrict
employment to the literate and best qualified job seekers. One's oppor-
tunities for employment and upward mobility in this kind of economic system
depends on one's knowledge of the official (European) language; thus
language becomes, once more, a barrier to personal and national development.
The trickle down economic approach contrasts with what E. G. Schumacher
(1973) has aptly termed "small is beautiful" approach where economic de-
velopment results from the education, organization, and discipline of the
people themselves wherever they may be. As we have seen, however, formal
education continues to be restricted to a small percentage of the African
population. This situation is worsened by the relapses into illiteracy
which occur when elementary school graduates are unable to continue their
schooling and the use of the official language after being out of school
for several years. Further, since mass or adult literacy remains only a
dream for most African states, the kind of economic development proposed by
Schumacher cannot occur in the near future.
It should be evident from the discussion presented thus far that real
development, as defined in (3) above, cannot be achieved in Sub-Saharan
Africa under these conditions. •Fuirdamentai changes must occur in the areas
of education and language policies before any genuine development can be
expected. This brings us to the third question raised at the beginning of
this part of the paper, viz. what specific steps should Sub-Saharan African
nations take to overcome underdevelopment.
2.0 TOWARDS A REALISTIC DEVELOPMENT MODEL
Sub-Saharan Africa, like the rest of the continent and other developing
regions of the world, is fraught with all sorts of difficulties, and one
cannot hope to solve them over night. Basic to these difficulties, however,
are developmental problems or questions of realistic reorientation of ob-
jectives for the Africa of the 1980s.
I am not an expert in development, and have no illusion regarding the
magnitude of the problems involved in changing a socio-economic system.
It is my humble and considered opinion, however, that in light of the facts
detailed in this paper the answers to the problems discussed here lie in a
fundamental restructuring_of .two related areas of -development : the school
system^^aSJ the language policies. Sub-Saharan African nations must, either
separately or together, articulate a prioritized and realistic set of de-
velopmental goals. These goals must be formulated on the basis of an
inventory of needs and resources in each country, and should be articulated
19
in a coherent and comprehensive manner. Once these objectives have been set,
the schools and other institutions of higher learning, e.g., research
Institutes, must be restructured as to help meet these objectives. And
naturally, the language policies of the region must be reformulated as to be
consonant with such goals.
2.1 The restructuring can be undertaken essentially in two directions
once the goals for each country or the region have been determined. The
first change must be a transformation of the entire educational system.
Instead of the current European-based system which emphasizes theoretical
training and is keyed to preparing all students for university education,
the educational system of the Africa of the 1980s must emphasize practical
training. It must set up goals for each educational cycle: primary,
secondary, and university. Following Nyerere (1967), I would like to suggest
that the goals of elementary education ought to be to provide the pupils the
type of training that will prepare them for community service at the local
level soon after their graduation, just in case they do not continue their
studies at the next cycle. Given the fact that between 70-80 percent of the
population of most Sub-Saharan African nations is rural, the new primary
school system would have to Include basic training in agriculture, carpentry,
masonry, arts and crafts, and family hygiene. Such a program would enable
youngsters to become gainfully employed either upon graduation or after a
one or two year post primary training in an area of their choice, just in
case they cannot be admitted into the general secondary school.
With regard to secondary school education, its goals ought to be to
provide an education that will prepare the graduates for national service
(Nyerere 1967). That is, secondary school training should lead to employ-
ment, in different sectors, at the national level. The graduates can, e.g.,
serve as civil servants, elementary school teachers, electricians, plumbers,
auto mechanics, masons and carpenters, and clerks. What this means is that
the secondary school system will have to include, apart from the usual uni-
versity preparatory curriculum, programs that will lead to technical or
professional specializations. The advantage of this type of program is
not simply that it will provide secondary education to more citizens, but
it will also provide badly needed low-level technicians. Programs such as
these were available in many African nations during the colonial era as an
option for students who were not qualified for admission into the general
secondary school, but they lost their attraction after independence. The
only training of this type of >ij> that is available at the secondary education
level are the so-called Teachers' Training Colleges and Professional Schools
(for mechanics and carpenters) .
University education, in contrast, must have as a goal the training of
people for national and international services (Nyerere 1967). Specifically,
university training, with its multiple disciplines, should provide the kind
of education that will lead to employment at both the national and inter-
national community levels. Over and above this goal, the university must
become the central agent of national development by engaging in both practical
and theoretical research that will enable the country or region concerned to
become increasingly self-reliant.
20
2.2 Once the objectives of each educational cycle have been set, it
becomes relatively easy to formulate a language policy that is consistent
with these objectives. Assuming that the nation or region concerned is
sincerely determined to pursue the educational plan just outlined, the
language policy that would be consonant to it would be a multilingual one.
That is, this language policy will include (1) the use of the appropriate
regional language(s)-' as medium or media of instruction at the elementary
and secondary school levels; if a country/region has already adopted a
national language, it will serve this purpose; (2) the teaching of an
appropriate international language (e.g., French for francophone African
nations, English for anglophone countries, and Portuguese for lusophone
countries) as an obligatory subject in all three educational cycles; and
(3) the use of an appropriate international language as the main medium
of instruction at the university and post-university levels. If a country
or region has already adopted a national language, as in the case of
Ethiopia, Somalia, and Tanzania, for example, this language would also be
used as medium of instruction for selected subjects.
2.3 If these educational and language policy plans are successfully
adopted and implemented, the transformation of the political and economic
sectors will follow more or less automatically in that an increasingly
literate populace, using an indigenous national language and participating
actively in the job market, will become more and more involved in the
political and economic affairs of their country. That is, the population
will develop sentimental and instrumental attachments to the national
political system. A language policy which calls for the use of national
linguae francae or a national language, for instance, will likely result
in the writing of important political and judicial documents in these
languages. Hence, the people will be able to participate to a certain
extent meaningfully in the affairs of their country.
Similarly, the availability of more medium and high level national
cadres at different areas of the national economy will give them more say
and influence in the development of the economy than previously. Three of
the difficulties facing Sub-Saharan African economies are the lack of
diversification, processing industries, and an adequate transportation
system. These problems persist largely because African nations do not
have at the moment adequate national technicians. If the educational
plan proposed here were to be adopted, these problems would be solved in
a relatively shorter time than if the current educational systems were to
be maintained.
2.4 In order to generalize and accelerate development in Sub-Saharan
Africa there must be increased emphasis on adult literacy; formal education
alone will not considerably reduce or eradicate illiteracy for a long time
to come. Illiteracy, as Dr. A. M. M'Bow, the General Secretary of the
UNESCO, pointed out in his annual message in 1979, is a world problem of
the same magnitude as malnutrition and poverty. As the comparative
statistics in Tables I and II show, this problem is particularly acute in
Africa. To combat it, African governments and cooperating organizations
must devote considerable sums of moneys and conjugate their efforts well.
In particular, African governments, churches, private organizations and
21
international agencies such as UNESCO and the World Bank must cooperate in
the development of programs which will be keyed to personal and community
development, especially for the rural areas. Adult literacy programs such
as those carried out by the Summer Institute of Linguistics (SIL) and those
of the Afrolit Society, a UNESCO affiliate based in Nairobi Kenya, ought to
be particularly encouraged by African governments. Such organizations, given
their long experience in the field, could provide a model for large-scale
programs for the continent.
I should like to ^point out here that Sub-Saharan Africa, or any part of
the continent for that matter, does not want any type of adult literacy
program: it needs what I would like to call a functional and developmentally
relevant literacy program. A functional and developmentally literacy program
is one which responds to the daily needs of the society/sub-society in
question by incorporating into the teaching of reading and writing develop-
mental themes, information, and techniques on how to improve the people's
way of life. Specifically, reading lessons in any adult literacy manual
should be based on topics deemed to be of interest to the community, as de-
termined by its representatives. For instance, if the community or society
in question is involved in agriculture, fishing, and trade, the reading
materials should provide the kind of information that will encourage the
people to introduce relevant modern techniques to improve these areas of
their life. Notions of elementary/basic mathematics, personal and community
health care should also be incorporated in such manuals. It is this type of
program, rather than one that is aimed at what might be called "pure literacy,"
that will attract and retain the learners. The Afrolit Society, under the
able direction of its Secretary-General, Dr. Charles T. Hein, has been using
this approach very successfully. In short, it is the type of changes which
have been suggested all along this paper that constitute what I regard as a
comprehensive and yet realistic developmental model for (Sub-Saharan) Africa.
3.0 CONCLUSION
I began this paper by claiming that language, just like education, is
critical to personal and national development. -In the course of the pre-
sentation I have attempted to show, with facts and figures, how intricate
the relationship is between these three domains: language, education, and
development. In the absence of empirical findings, I have used circum-
stantial evidence to conclude that the failure of African states to achieve
the developmental goals set out at the 1961 Addis Ababa conference was
largely due to the current language policies which call for the exclusive
use of European languages as media of instruction. The discussion of this
issue led to the presentation of what I consider a realistic and compre-
hensive model of development involving the transformation of the school
systems and language policies, and the incorporation of the schools and
literacy programs as agents of development.
Bearing in mind the magnitude of the task to be undertaken and the
difficulties involved in transforming a society, I have assumed throughout
this presentation that African scholars, government and non-government
leaders are aware of their state of underdevelopment and are desirous to
overcome it as quickly as possible. My assumption is not a fortuitous
one: it is based on expressed and published works of African leaders and
22
scholars (cf. Kalanda 1965, Nkrumah 1965, Nyerere 1967, Mobutu 1973, Bot-Ba-
Njock 1974, Kempf and Mudimbe 1977, Kashoki 1979, Kahombo 1980, Ansre 1980),
among others. If the desire to modernize and improve the people's mode of
life is sincere, the apparent difficulties that might exist in the adoption
of the developmental plan proposed here would not constitute an insurmountable
obstacle: the will of the people will finally prevail.
At the beginning of this paper I also stated that the current socio-
economic difficulties confronting African states are largely attributable
to the Africans themselves, especially the politically powerful elite. The
more I look at the African socio-economic systems, the more I become con-
vinced that J. Weeks (1975: 99) was right when he observed that:
(12) ...the pattern of development or economic growth chosen (in
any country) arises from the economic interests of the
politically powerful. The fact that a pattern of development
generates intractable social problems is not sufficient to
stimulate redress of the situation; this will occur only if
the politically powerful see their interests being served as
well in some alternative arrangement as in the existing one,
or if the problems deepen into a crisis which fundamentally
alters the concentration of power.
To avoid either the continuation of underdevelopment or the disruptions that
accompany revolutions, the full potentialities of the African people must
be developed. Education, literacy, and the use of selected African languages
in these tasks will be critical ingredients to a peaceful development in
the Africa of the 1980s.
NOTES
*A shorter version of this paper was presented at the 9th International
Conference on the Unity of the Sciences under the title: "Language Policies
and National Development in Sub-Saharan Africa: Issues for the 1980s," at
Miami Beach, Florida, November 1980. The present version, with the title:
"Language and National Development in Sub-Saharan Africa," was written for
and read at the 7th Annual Third World Conference, organized by Governors'
State University and held at Chicago, Illinois, March 1981. The paper is
considered a progress report, because of the lack of up-to-date statistics
on educational and literacy developments, and of published work on the use
and impact of Amharic, Somali, and Swahili which have been adapted as the
national languages of Ethiopia, Somalia, and Tanzania, respectively. I am,
therefore, continuing the research. In the meantime, I am grateful to
Nzongola Ntalaja (Howard University) and Rudolph Troike (University of
Illinois) for their comments on this paper. I alone am responsible for any
errors of facts or interpretation.
we are aware of the fact that these figures are outdated, but we
believe that the situation has not changed significantly even if recent
statistics were made available.
23
^These restrictions are not dictated by the availability of jobs; most
African nations can create at this moment as many jobs as they wish without
stretching their resources. One of the main economic problems facing
Africa today is under-exploltation of natural resources, including human
resources, rather than unemployment.
^Rideout, et al. (1969: 61) show, for instance, that the average
wastage rate in the secondary schools in Zaire in 1966-1967 was 29.5 percent;
and in 1967-1968 it rose to 35.1 percent for the classes of 1961-1962 and
1962-1963, respectively. The wastage rates between years were extremely high.
For example, between the fifth and sixth year of secondary school the wastage
rate was 42.4 percent during the academic year 1966-1967, and 65.9 percent in
1967-1968 for the same class period. These figures are paralleled elsewhere
in Africa.
^It is also recognized that language, especially in a multilingual
society, can be a disruptive force (Kelman 1971, Fishman 1972). As Kelman
(1971: 21) aptly observes:
However, some of the very features of language that give it this
(unifying) power under some circumstances may, under other circum-
stances, become major sources of disintegration and internal
conflict within a national system.
^All African nations have recognized regional languages which are often
referred to as "national languages" or linguae francae. Nigeria, e.g., has
three such languages: Hausa, Igbo, and Yoruba; Zaire has four: Kikongo,
Lingala, Swahill, and Tshiluba. The adoption of such languages as media of
instruction would, therefore, create no special problems (cf. Bokamba 1976;
Bokamba and Tlou 1977 for detailed discussions of language policies and
planning) .
I have read some of the materials prepared by the Afrolit Society, and
participated as a consultant in one of their workshops for a month, September
1980, in Zaire to develop materials for the Lingala speaking regions of Zaire.
That workshop provided me first hand information on the society's approach
to literacy work, and convinced me of the effectiveness of their materials.
I am grateful to Dr. Hein, the society's secretary-general, for affording
me this opportunity.
REFERENCES
ANSRE, G. 1969. The need for a specific and comprehensive policy on the
teaching of Ghanaian languages, in J. R. Birnie and G. Ansre, eds..
Proceedings of the Conference on the Study of Ghanaian Languages,
pp. 5-11. Legon: Ghana Publishing Corp.
ANSRE, Gilbert, 1980. Language and Development. Paper prepared for the
Ninth International Conference on the Unity of the Sciences, Miami
Beach, Florida, November 27-December 1, 1980.
BELLONCLE and Palm, 1976(?). Le systeme educatif Mallen: analyse et
recommendations. Rapport Provlsolre de la mission UNESCO. Mimeographed.
24
BIRNIE, J. R. and G. Ansre, eds., 1969. Proceedings of the Conference on
the Study of Ghanaian Languages. Legon: Ghana Publishing Corp.
BOKAMBA, E. G. 1976. Authenticity and the choice of a national language:
the case of Zaire. Studies in the Linguistic Sciences 6, 2: 23-64.
Also in Presence Africalne 99/100: 104-42.
BOKAMBA, E. G. , and J. S. Tlou, 1977. The consequences of the language
policies of African states vis-a-vis education. In Paul A. Kotey and
Haig Der-Houssikian, eds., Language and Linguistic Problems in Africa:
Proceedings of the Seventh Conference on African Linguistics, pp. 35-53.
Columbia, S.C.: Hornbeam Press. Also in Kahombo Mateene and J. Kalema,
eds., 1980, Reconsideration of African Linguistic Policies, pp. 45-66.
Kampala: OAU Bureau of Languages.
BOT-BA-NJOCK, Henri M. 1974. Les langues africaines, facteur de developpe-
ment. In Les Langues Africaines, Facteur de Developpement. Actes du
Seminaire pour I'Enseignement des Langues Africaines. Douala, Cameroun:
Le College Liberman. pp. 15-25.
BOULDING, Kenneth E. , 1966. The Impact of the Social Sciences. New
Brunswick: Rutgers University Press.
CHAMPION, Jacques. 1974. Les Langues Africaines et la Francophonie: Essai
d'une Pedagogie du Francais en Afrique Noire par one Analyse Typologique
de Fautes. Paris: Mouton.
DOWUONA, M. 1969. Opening address. In Birnie and Ansre, eds., pp. 1-4.
EDELMAN, Murray, 1977. Political Language: Words that Succeed and Policies
that Fail. New York: Academic Press.
FISHMAN, Joshua A. 1972. Language and Nationalism: Two Integrated Essays.
Rowley, Mass.: Newbury House.
FRANK, A. Gunder, J. D. Crockcroft and D. L. Johnson, eds., 1972. Dependence
and Underdevelopment : Latin America's Political Economy. Garden City:
Anchor Books.
GORMAN, T. P. 1974. The development of language policy in Kenya with par-
ticular reference to the educational system. In W. H. Whiteley, eds.,
Language in Kenya, pp. 397-453. Nairobi: Oxford University Press.
KAHOMBO, Mateene, 1980. Failure in the obligatory use of European languages
in Africa and the advantages of a policy of linguistic independence.
In Kahombo and Kalema, eds., pp. 9-44.
KAHOMBO, M. and J. Kalema, eds., 1980. Reconsideration of African Linguistic
Policies. Kampala: OAU Bureau of Languages.
KALANDA, M. 1965. La Remise en Question: Base de la Decolonisation Mentale.
Paris: Remarques Africaines.
KASHOKI, Mubanga E. 1979. The African language as a tool of development.
Mimeographed .
KELMAN, Herbert C. 1971. Language as an aid and barrier to involvement in
the national system. In Joan Rubin and Bjorn H. Jernudd, eds.. Can
Language be Planned? Sociolinguistic Theory and Practice for Developing
Nations. Honolulu: The University Press of Hawaii.
KEMPF, B. and V. Y. Mudimbe. 1977. Langue et developpement. In Les
Relations Entre les Langues Negro-Af ricaines et la Langue Franpaise,
pp. 502-9. Paris: Conseil International de la Langue Franjaise.
LANGACKER, Ronald W. 1967. Language and its Structure: Some Fundamental
Linguistic Concepts. New York: Harcourt Brace Jovanovich.
LEWIS, L. J. 1962. Phelps-Stokes Reports on Education in Africa. London:
Oxford University Press.
25
MALOBA, Ngoba E. 1977. Predicting Schools' Performance In the State Examina-
tions: A Model for Shaba Province in Zaire. Unpublished doctoral
dissertation, Indiana State University, Terre Haute, Indiana.
M'BOW, M. A. 1979. L' analphabetisme entrave la croissance de 1' enfant.
(Message a I'occasion de la journee Internationale de I'alphabetisation.)
Nouvelles d'Afrollt, Juillet-Decembre, 1979.
MOBUTU, S. S. 1973. Discours k I'Assemblde Generale des Nations Unies, New
York, le 4 octobre, 1973. Kinshasa: Departement de 1 'Orientation
Natlonale.
NKRUMAH, K. 1965. Neo-Colonialism: The Last Stage of Imperialism. New York:
International Publishers.
NYERERE, J. 1967. Education for Self-Reliance. Dar-es-Salaam; Government
Printer.
RIDEOUT, W. M. , D. N. Wilson, M. C. Young, et al., 1969. Survey of Education
in the Democratic Republic of the Congo. Washington, D.C.: American
Council on Education.
SCOTTON, Carol Myers, 1978. "Elite Closure" and Language Policy in Africa.
Paper presented at the Symposium on Language Policies in African Educa-
tion, Linguistic Institute of the Linguistic Society of America, Univer-
sity of Illinois at Urbana-Champaign, July 6-7, 1978. To appear in
E. G. Bokamba, ed., Language Policies in African Education. Washington,
D.C.: University Press of America.
SPENCER, J. 1971. Colonial language policies and their legacies. In
T. Sebeok, ed.. Current Trends in Linguistics, vol. 7: Linguistics in
Sub-Saharan Africa, pp. 537-A7. The Hague: Mouton.
TLOU, J. S. 1976. The Primary Teachers Education Program's Role in National
Development in Botswana. Unpublished Doctoral Dissertation, University
of Illinois.
UCHENDU, V. C. 1980. The great debates in African development: rural
poverty and the challenge to social theory. Africana Studies Occasional
Paper No. 80-2. West Lafayette: Africana Studies and Research Center.
UNESCO. 1953. The Use of Vernaculars in Education. UNESCO Monograph on
Fundamental Education 8. Paris: UNESCO.
. 1962. Conference of African States on the Development of Educa-
tion in Africa, Addis Ababa, 15-25 May, 1961. Final Report. Paris:
UNESCO.
. 1973. UNESCO Statistical Yearbook 1972. Paris: UNESCO.
1976. Conference of Ministers of Education of African Member
States, Lagos, January 27-February 4, 1976. Final Report. Paris: UNESCO.
. 1977. Statistics of Educational Attainment and Illiteracy, 1945-74.
Paris: UNESCO
VERBERKE, R. 1966. Langues Vehicularies de I'enseignement en Afrique Noire.
Revue Internationale de Pedagogie 12, 4: 450-66.
WEEKS, John. 1975. Imbalance Between the Centre and Periphery and the
"Employment Crisis" in Kenya. In Ivar Oxaal, Tony Barnett, and David
Booth, eds. , Beyond the Sociology of Development: Economy and Society
in Latin America and Africa, pp. 86-104. Boston, Mass: Routledge and
Kegan Paul.
Studies in the Linpiiistir. Sciences
Volume 11, Number I, Spring 1981
PRAGMATICS AND SYNTACTIC DESCRIPTION*
Georgia M. Green
It has long been recognized that the acceptability of
example sentences under test conditions depends on what the
informant believes about the "real world," and particularly
about the beliefs and intentions of the presumed speaker
of the example sentences. After providing several detailed
examples of the phenomenon, a history is given of how it
has been dealt with by generative grammarians. It is argued
that the problem is not strictly within the province of
grammar; rather, the grammar should be designed to generate
freely such forms as she knows who did I appoint and a
general theory of communication, as sketched here, should
provide principles regulating its potential use.
The working title for this paper was "On the Nature of 'Pragmatic
Rule Government.'" It was abandoned because it might be understood as
giving the (false) impression that 1 believe there is any such thing.
What I can offer on this subject is 1) a description of the state of
affairs which this unfortunate term has been used to refer to, and after
giving a brief history of how generative grammarians have attempted to
deal with this phenomenon, 2) a sketch of what seems to me a more
reasonable way of describing such states of affairs.
I . The phenomenon
In its most general form, the problem is that sometimes the
acceptability of example sentences under test conditions depends on
the informant's assumptions about the beliefs and intentions of the
presumed speaker. Stated this broadly, this is nothing new. Generative
grammarians knew even in the 1960s that the acceptability of a sentence
like (1) depends on whether or not one assumes that the speaker meant
for he and John to refer to the same person.
(1) He thinks John is stupid.
Likewise, they knew (Lakoff 1969) that the relative acceptability of
questions and conditionals with some and any as in (2) depends on
assumptions about the attitude of the presumed speaker.
(2a) If you eat some spinach, I'll cook hamburgers all week.
(2b) If you eat any spinach, I'll cook hamburgers all week.
Sentence (2a) is considered acceptable if it is taken as offering a
bribe, and it is understood that the speaker believes that the addressee
considers having hamburgers all week desirable; sentence (2b) is con-
sidered acceptable if it is taken as a warning not to eat spinach,
and cooking hamburgers all week is understood to be something that the
speaker assumes the addressee would consider undesirable.
28
Other cases involve "obviously" syntactic phenomena such as question
formation and relativization. Thus, in many dialects of English, Subject-
Auxiliary Inversion (SAI) is found not only in direct questions such as
those in (3), but also in embedded questions, as in (4).
(3a) Who did you appoint?
(3b) Did you appoint Tony?
(4a) She wants to know who did I appoint.
(4b) He asked did I appoint Tony.
However, the optional inversion in embedded questions is possible only
if the referent of the subject (or experiencer) of the verb embedding
the question is presumed not to know the answer. Thus, the sentences
in (5) are unlikely to be considered acceptable.
(5a) She already knows who did I appoint, so you needn't
dissemble any longer.
(5b) He's figured out did I appoint Tony, but he promised not
to tell.
Or, to take another example, sluicing (Ross 1969), as in (6), is
also sensitive to whether or not the individual who is implied to want
to know the answer to the corresponding question does in fact know
the answer. This individual is often the speaker (6a) or the addressee
(6b), occasionally the referent of the subject (6c) (or dative experiencer
(6d)) of the predicate implying (and embedding) the question, but it
isn't always, as (6e) testifies. Thus, my saying (6a) implies that I
don't know what John broke; (6b) is acceptable only if it is assumed
that the presumed speaker believes the addressee doesn't know what
John broke, and (6c) implies that John doesn't know what he broke.
Sentences (6d) and (6e) imply that the speaker, perhaps among others,
doesn't know what John broke, even though the subject of the verb
embedding the sluiced question is not a NP referring to the speaker.
(6a) John broke something, and I'm afraid to ask what.
(6b) John broke something- -guess what.
(6c) John broke something and he needs to know what.
C6d) John broke something, but it's not clear what.
(6e) John broke something, but he won't say what.
If we construct corresponding sentences where these assumptions
seem to be denied, we see that they are less acceptable, or induce
implicatures.
C7a) John broke something, and it's clear what.
(7b) John broke something and you know what.
(7c) John broke something and he told me what.
Thus (7a) and (7c) are either unacceptable or implicate that the speaker
assumes that the addressee doesn't know what John broke, inviting or
daring him to ask. And (7b) will either be judged unacceptable or as
implicating that the speaker doesn't know and would like the addressee
to tell her. Their unreduced counterparts do not have these properties.
The discourses in (7'), where there is no sluicing, are not as bizarre
as the sluiced ones in (7'').
29
(7a') John broke something, and it's clear what he broke.
You needn't investigate further.
(7b') John broke something, and you know what he broke.
I don't want to know, but you'd better get it replaced.
(7c') John broke something, and he told me what he broke.
You needn't harass him any more.
(7a'') John broke something, and it's clear what.
You can stop harassing him about it.
(7b'')John broke something, and you know what.
Don't tell me; just replace it.
(7c'')John broke something, and he told me what.
You can stop harassing him about it.
One can construct examples such as (8) where the semantics of
the utterance seems to preclude the assumption that anyone relevant
is ignorant of the answer:
(8a) John went out with someone, and we all know who.
(8b) John went out with someone, and I don't have to say who.
But even here, these will be judged acceptable if the informant
supposes the presumed speaker to be ironically inviting the addressee
to make a guess (which is almost assured of being correct) .
So--if we assume that these sluicing constructions are derived
by a rule of grammar, and that the above examples judged unacceptable
are as ungrammatical as ones like (9), then we seem to be put in the
position of having to say that the applicability of a rule of grammar
is conditioned by the speaker's beliefs and intentions, or by assump-
tions about the speaker's beliefs and intentions.
(9) *John bought something and I washed what.
One final example. The extraposition of relative clauses as
in (10) is sensitive to the speaker's judgment that the proposition
implied by the relative clause is "new" to the addressee, and to a
concomitant desire to assert it (Ziv 1976).
(10) A man came in who had three ears.
Thus, a relative clause cannot be extraposed when its content is intended
to function restrictively by virtue of its representing a proposition
which the speaker assumes to be presupposed by the addressee. Consequently,
(lib) cannot be used as an answer to (11a), although (lie) can.
(11a) I heard some of Susan's relatives were coming this week,
(lib) Yes, the uncle arrived yesterday who's so rich he could
buy Chicago,
(lie) Yes, the uncle who's so rich he could buy Chicago arrived
yesterday.
Of course, these are not the only cases linguists have been led
to describe in terms of "pragmatic control". Apparently almost all
rules ever proposed by a generative grammarian either have pragmatic
control (e.g. pronominalization, SAl in questions) or pragmatic conditions
on them (like the coexistence presupposition requirement on the items
30
affected by Dative Movement (Green 1974) and Subject-to-Object Raising
(Postal 1974)) or pragmatic motivation in that they seem to serve a
clear rhetorical or processing function, like Extraposition and Relative
Extraposition. Many rules, including Passive, seem to involve a combi-
nation of kinds of pragmatic conditions. This means that the problem
of the proper role of pragmatics with respect to syntactic description
cannot be dismissed as merely an interesting puzzle involving a few
insignificant, unrelated, and restricted phenomena. The involvement of
pragmatics is pervasive.
11 . A bit of history
But dismissal is exactly what the first reaction of generative
grammarians appears to have been--a sort of linguistic Benign Neglect.
The first description of rules for pronominalization acknowledge the
undeniably pragmatic condition of intended coreference of individuals
named by the mentioned NPs. People said things like
(12) Pronominalization applies to structures like John hurt
John unless the Johns are different people.
promiscuously mixing categories. But the grammars they wrote took
no account of this, and they were silent on the question of how
this constraint was related to linguistic conpetence.
Beginning in about 1965, attempts were made to incorporate various
kinds of pragmatic conditions directly into the system then available
for syntactic description. Thus, Postal, McCawley, and Chomsky all
made proposals for incorporating a system of indexing syntactic nodes
for coreference so that the kind of identity required for pronominali-
zation and relative clause formation could be represented as syntactic
information, and represented directly in the syntactic structures.
George Lakoff, in his dissertation (Lakoff 1965) also supposed that
restrictions on usage were all syntactic in nature. Thus, he claimed
that the unacceptability of (13) reflected the fact that beware bore
a syntactic feature which said that Question-Formation could not apply
to it.
(13) Did you beware of John?
But Lakoff s system would break down when confronted with cases like
(4) and (5): one cannot say that know is marked [-SA1 in Complement]
to account for the unacceptability of (5a) because SAI in the complement
of know is fine in (4a) . Since the real restriction on beware is that
it be used to convey or refer to the conveying of a warning, it won't
do to say it is [-R: Q] to account of the unacceptability of (13), and
[-R: WH-Rel] or [-SD: WH-Rel] to account for the unacceptability of (14),
because in (IS) and (16) those rules presumably apply to give acceptable
results.
(14) John met a man who can beware of dogs,
(15) What do I need to beware of?
(16) You must appoint as dean a man who will beware of outside
agitators.
31
In the late 1960s, under the rubric Linguistic Anarchy Notes,
Postal circulated some observations which seemed to entail that rules
of grammar were sensitive to facts of arithmetic (Postal 1977) . Postal
was concerned with how grammatical theory could provide descriptions
which would distinguish the (a) sentences in (17-19) from the (b)
sentences.
(17a) Six Arabs murdered a seventh.
(17b) Six Arabs murdered an eleventh.
(18a) The four Stuart kings were James I, Charles I, Charles II,
and James II.
(18b) The five Stuart kings were James I, Charles I, Charles II,
and James II.
(19a) There were five or ten deans standing around.
(19b) There were five or eleven deans standing around.
Such distributions were taken, in some quarters at least, to represent
a difficult and interesting puzzle for the then relatively monolithic
theory of transformational grammar, but no attempt was ever published
that attempted to give a motivated account of such facts.
About the same time, a second year graduate student made some
observations about the occurrence of the emphatic particles too
and either in conjoined sentences which compare or contrast properties
of the subjects, like those in (20).
(20a) Jan's a lawyer and Ron's a lawyer too.
(20b) Jan's a lawyer, and her husband's a professional too.
(20c) Ron isn't interested in sports, and Bill couldn't tell
a zone press from a fast break either.
Specifically, she observed that for this usage of too/either to be
considered appropriate, the predicate of the second clause has to be
taken as being implied by the predicate of the first clause, the
simplest case of this being the identity of predicates found in (20a) .
Examples like (20c) are more complex cases where the implication is
not a strictly logical or semantic one, but one which depends on contin-
gent assumptions about real-world relations among properties (e.g.
whether being interested in sports entails or implies the ability to
tell a zone press from a fast break). Now her approach was to try to
describe this relationship as part of the particle insertion rule
(Green 1968) or the deep structure (Green 1973), and no one told her
that this was wrong or dumb, though the former proposal clearly required
making the rules of the grammar (too/either insertion) sensitive to
properties of speakers (whether they believed that some predication
implies some other predication), and this involves a gross category error.
Indeed, this was just the first of many cases touted as demonstrating
the role of presuppositions in syntax, still without explicit attention
to how description of langue (a formal description of a static system)
could sensibly require reference to properties of individual language
32
Interestingly, between the time George Lakoff first delivered a paper
on this topic in 1969, and the time it appeared in print (Lakoff 1971)
in 1971, the title changed from "Presuppositions and Relative Grammati-
cality" to "Presuppositions and Relative Well-formedness," indicating an
appreciation of the logical problem. Well-formedness was usable to refer
to appropriateness with respect to a context; grammaticality was an
absolute, and a function of the properties of a linguistic description.
In any case, the assumption here was that since the presupposition-
dependent and context-dependent phenomena were rule-governed just as
much as presupposition-free and context-independent syntactic phenomena,
then the system provided by linguistic theory for the description of
that aspect of linguistic competence that is displayed in grammaticality
judgment would have to countenance rules relating linguistic forms to
language users' beliefs and intentions. Thus, it was assumed that grammar
would include such statements as (21), or equivalents thereof.
(21) The Sluiced construction can be generated (The rule of Sluicing
applies) if it is understood that the referent of the subject
or experiencer of the predicate embedding the sluiced
construction is ignorant of the answer to the question
corresponding to the sluiced construction.
Meanwhile, Grice's paper "Logic and Conversation" (Grice 1975),
which had been circulating in unpublished form for several years began
to have an effect on how linguists thought about the relations between
grammar and usage. Grice's paper, addressed to logicians, suggested
that the conflicts between ordinary language and formal logic could be
resolved by showing how the annoying connotations and implications of
ordinary language usage could be attributed to inferences from the
assumption that speakers were conforming to a principle of cooperation
in participating in conversation, tailoring their contributions to their
conception of the task at hand, according to several corollary principles.
George Lakoff, in a much-quoted paper written with David Gordon (Gordon
and Lakoff 1971), interpreted Grice's proposal as sanctioning the codifi-
cation of likely implicatures into conversational "postulates", and the
incorporation of speech act participants' beliefs and intentions into
derivations in the guise of transderivational constraints. Thus, Gordon
and Lakoff claim that a rule of You-Tense Deletion, responsible for
deriving sentences like (22a) from structures similar to that of (22b)
applies if and only if the logical structure L of the sentence, taken in
conjunction with Con., a class of contexts, add the set of Conversational
Postulates entails 'Dnless you have some good reason for doing X, you
should not do X. '
(22a) Why paint your house purple?
(22b) Why do you paint your house purple?
But their Con. is just an obscure way of referring to the speech act
participants '"""beliefs and intentions about the context of the speech
act as a set of propositions, each of which is true of the context.
This strategy of describing matters of speakers' intentions and beliefs
as if they were semantic matters of truth, carried further in Lakoff
1975, only obscures the relation between form and its interpretation.
33
III. An alternative
The approaches sketched so far to the problem of giving an account
of the role of pragmatics in linguistic description have been either 1)
to interpret distributions of linguistic expressions that depend on users'
beliefs and intentions as reflecting arbitrary formal characteristics of
the expressions, or 2) to redefine the task of syntactic description as
one of description of communicative behavior generally, so that it could
sensibly accommodate rules relating syntactic derivations to language
users' beliefs and intentions.
But there is a third plausible approach, which seems like it ought
to be called interpretive pragmatics, though I'm sure that would be an
unwise name to give it. It says basically that the fault is not in our
stars, dear Brutus, but in ourselves. This approach, which is consistent
with every name-brand syntactic theory that I know of, whether the major
portion of their expressive power is in transformation rules, surface
filters, or phrase structure rules, is to keep the pragmatics strictly
out of the syntax, and describe pragmatic control, constraints, and
motivations within a general theory of communication. Thus syntactic
forms would be generated freely, according to the syntactic rules of
the language, without regard to their possible uses. Description of
pragmatic restrictions on the use of such forms would be provided by
theory of communication which would integrate such more or less universal
principles as Grice's Cooperative Principle and its corollaries, strategies
for referring, asserting, focussing, etc., as well as culture-specific
rules of politeness and conventions of usage (Morgan 1978), all of which
freely make direct referenee to language users' intentions and beliefs.
For example, the convention governing the use of the Sluicing construction
might be something like (23) .
(23) Use of the Sluicing construction implies that the user
believes that the individual implied to want to know
the answer to the question corresponding to the sluiced
construction is in fact ignorant of the answer.
This treatment, in contrast to the others, does not claim that
sentences like (5) or (7) are ungrammatical . Rather, it predicts
1) that the use of such sentences will cause hearers to make certain
inferences about the speaker, 2) that some of these may result in the
sentence being considered perhaps inappropriate, given what else the
hearer knows about the speaker and the subject matter, or contradictory,
or ineffective for the purpose the hearer presumes to be intended by
the speaker, or to use a technical term I introduced a couple of
years ago (Green 1976), dumb, and 3) that the speaker is aware (at
some level) of (1) and (2).
Thus, while the approach of the '60s entailed claiming that a
sentence like (5b) was ungrammatical because the SAI rule for
questions had applied in the complement of a verb bearing the syntactic
feature [-SAI in Complement], and the approach of the '70s entailed
claiming such a sentence was ungrammatical because the application of
SAI in the complement, implying that the speaker believes that the
referent of the subject of the embedding verb didn't know whether
the questioned proposition was true, contradicts the assertion of
the whole sentence, that the referent of the subject has figured it
out, and thus does know. What I am proposing entails claiming that
34
sentences like (5b) are perfectly grammatical --they conform in every
respect to the rules of syntactic combination comprising the grammar,
but, nonetheless, such sentences are inappropriate, communicatively
ineffective, or dumb, for reasons which are essentially the same as
those cited by the second approach in describing what was wrong with
them.
The conventions of usage I am referring to are not something I had
to invent so as to have a place to put the pragmatic restrictions on
the use of various constructions. In fact, they were originally proposed
(Morgan 1978) to give an adequate account of the grammaticized implica-
tures of whimperative requests like (24a) and the use-conditions for
formulae like (24b) .
(24a) Can you pass the salt.
(24b) God bless you.
They seem to be appropriate to describe register differences among lexical
items and constructions as well --it is only an arbitrary convention that
vulgar words are vulgar, while the corresponding technical terms are not,
that Quotation Preposing as in (24c) is restricted to literary language,
while inverted exclamations are strictly colloquial usage, as in (24d) .
(24c) "Blah, blah, blah," said the speaker.
(24d) Is that ever silly!
Idioms would seem also to find a natural description as conventions of
usage: kick the bucket is a slightly disrespectful way of saying what
you mean when you use die literally. And in fact, when you get right
down to it, describing even so-called literal meanings of individual
words as conventions of usage has a certain attraction: cow is the name
conventionally used for members of a certain species; run, cry, tease,
and wash are conventional names for certain states, processes, and
goal-directed activities--but that's another story.
An of course, whimperative requests and formulae like God bless you
and goodbye are not the only sentence-level constructs which seem to
deserve an account in terms of conventions of usage. In 1975, Arnold
Zwicky compiled a list of over 700 "Funny Speech Act Examples", sentences
like those in (25), which were ordinarily understood as conveying something
quite different from what would be predicted from their "literal meaning"
and ostensible illocutionary force (if any).
(25a) Eat your heart out, Paul Newman.
(25b) Talk about leaping to conclusions.
(25c) The hell it is.
(25d) Nyaah, nyaah, nyaah.
Most of these are not fixed expressions, but like (25a-c) , open formulae,
as indicated in (26) .
(26a) Eat your heart out, NP.
(26b) Talk about VP-ing.
(26c) The hell NP V.
35
IVhat they conununicate is conventionally fixed, as are, occasionally, the
set of possible exponents for the open slots. Thus, (26a) is used to
convey that some evident fact or situation is enough to make the referent
of the NP jealous, though this is not a function of the literal meaning
of the component words2and phrases and any theorems derived from the
Cooperative Principle. And (26c) conveys that the speaker considers
the proposition which is derived by completing "NP V" with its recon-
structed complement to be false, preposterous, unlikely, offensive, or
otherwise unacceptable, but the NP must always be an anaphoric pronoun,
and the V apparently must be an auxiliary verb or say, as shown in (27).
(27a) The hell she/*Shawn/*IBM/*SHE does.
(27b) The hell she says/*things/*promises/does/will .
Here I anticipate the objection that if such diverse phenomena as
idiomatic meanings of expressions, register restrictions, appropriate
uses of social formulae, pragmatic restrictions on the use of syntactic
constructions, and maybe even so-called literal meanings of lexical
items themselves, are to be accounted for by a single descriptive
mechanism, as conventions of usage, then why not syntax too? No
principle prevents one from claiming that the rules of syntax and
morphology (e.g. phrase- structure rules, verb- agreement rules) are
just conventions of usage too. Under such a proposal, there would be
no grammatical/ungrammatical distinction; what is wrong with a sentence
like (28) is that cultural conventions about the use of am ensure that
it could never be uttered sincerely by a rational speaker fluent in
English.
(28) John am sick.
The objection strikes me as rather flimsy, because I think what
one would want to say about am is^ that ijt is^ a_ linguistic convention,
an arbitrary fact about the English language, that verbs agree in
person and number with their subjects, and that /aem/ is the first
person singular form of he_, not that there is some arbitrary rule of
the culture of English speakers that says that am is used to express
the existential predicate in the present when the subject is a pronominal
expression referring to the speaker.
In addition, treating syntax as just more conventions of usage
denies the difference that Morgan claims exists between conventions of
language and conventions about language, and seems to me to gain general-
ity only at the cost of losing the distinction between unmotivated and
absolute bizarreness ("ungrammaticality") , and explainable and conditional,
situation-dependent bizarreness. Moreover, empirical study of language
disorders supports the position that the distinction between grammatical
well-formedness and (pragmatic) conventionality of usage is a valid and
psychologically real one. Van Lancker and Canter (1979) summarize
evidence from the clinical and experimental aphasiological literature
which suggests that conventions about particular linguistic expressions
"are represented by a different cerebral organization and are processed
differently than are novel prepositional expressions."
36
IV. Implications
I have suggested that cases of so-called pragmatic control are
really best treated as (pragmatic) conventions governing the using
of syntactic constructions, letting the syntactic rules generate
the constructions freely, without regard to pragmatic constraints.
If the account I have suggested of pragmatic restrictions on syntactic
combinations is indeed more adequate than previous accounts which either
disguised pragmatic restrictions as formal syntactic restrictions, or
enlarged the theory of grammar to include most of psychology and a
good deal of logic, then it means that syntactic theory can go about
its business as just a theory of syntax, as strict and simple and
formal and mathematically manipulable as you like, and syntactic
descriptions will not be cluttered up with psychological, anthropo-
logical, sociological, or statistical constraints. Such syntactic
descriptions won't be very informative about language use, however- -
about the habits of usage and patterns of distribution among the
people who speak a language; description of such matters will be
the task of linguistics and grammarians who choose to describe the
conventions of language use. This will be just as difficult a
task as it ever was, but it needn't be stigmatized because it's
not syntax. No one need feel embarrassed because the constraints
on usage that they're describing aren't syntactic; they're rule-
governed constraints on linguistic behavior, and every bit as
linguistic as regressive assimilation or verb agreement.
And there is a good chance that if description of pragmatic
conditions, and conventions of usage generally, comes out of the
closet, and descriptions of usage are available, they could be a
boon not only to teachers and writers of pedagogical and reference
grammars for speakers of other languages, but in the teaching and
practice of rhetoric, journalism, literary analysis, psychiatry,
and any other fields or enterprises that depend on an accurate and
insightful analysis of discourse.
NOTES
*
This work was supported by the National Institute of Education
under Contract No. US-NIE-C-400-76-0116.
Admittedly, presupposition was still assumed in some circles
to be a semantic relation (among linguistic or logical entities)
rather than a pragmatic one, between speakers and utterances, so
the problem did not appear in these circles to be so severe.
2
The implied "meaning" of (26a) does, however, seem to follow
as a conversational implicature of appearing to address such an order
to an absent celebrity and taking the idiomatic meaning of eat one's
heart out to be 'be jealous.'
REFERENCES
GORDON, David and George Lakoff. 1971. Conversational postulates.
Papers from the 7th Regional Meeting of the Chicago Linguistic
Society, 63-84. Chicago: University of Chicago. Reprinted in
Syntax and semantics, volume 3: speech acts, ed. by Peter Cole
and Jerry L. Morgan, 83-106. New York: Academic Press.
37
GREEN, Georgia M. 1968. On too and either. Papers from the 4th
Regional Meeting of the Chicago Linguistic Society, 22-39.
Chicago: University of Chicago.
. 1973. The lexical expression of emphatic conjunction.
Foundations of Language 10.197-248.
1974. Semantics and syntactic regularity. Bloomington,
IN: Indiana University Press.
1976. Main clause phenomena in subordinate clauses,
Language 52.382-397.
GRICE, H.P. 1975. Logic and conversation. In Peter Cole and Jerry
L. Morgan, eds.: Syntax and semantics, volume 3: speech acts,
pp. 41-58. New York: Academic Press.
LAKOFF, George. 1965. On the nature of syntactic irregularity.
Indiana University Ph.D. dissertation. Published as Irregularity
in syntax. New York: Holt. 1971.
. 1971, Presuppositions and relative well-formedness.
In Danny Steinberg and Leon Jakobovits, eds.: Semantics: an
interdisciplinary reader, pp. 329-340. Cambridge University Press.
1975. Pragmatics in natural logic. In E.L. Keenan, ed.:
Formal semantics of natural languages, pp. 253-286. Cambridge
University Press. Reprinted in Proceedings of the Texas Conference
on Performatives, Presuppositions, and Implicatures, ed. by A.
Rogers et al., 107-134. Arlington, VA: Center for Applied
Linguistics.
LAKOFF, Robin. 1969. Some reasons why there can't be any some-any
rule. Language 45.608-615.
MORGAN, J.L. 1978. Two types of convention in indirect speech acts.
In Peter Cole, ed. : Syntax and semantics, volume 9: pragmatics,
pp. 261-280. New York: Academic Press.
POSTAL, Paul M. 1974. On raising, Cambridge, MA: MIT Press.
. 1977. Linguistic anarchy notes. In James D. McCawley,
ed. : Syntax and semantics, volume 7: notes from the linguistic
underground, pp. 203-226. New York: Academic Press.
ROSS, John R. 1969. Guess who? In R.I. Binnick, A.L. Davison, G.M.
Green, and J.L. Morgan, eds.: Papers from the 5th Regional
Meeting of the Chicago Linguistic Society, pp. 252-286, Chicago:
Chicago Linguistic Society,
VAN LACKER, Diane and Gerald Canter. 1979, Holistic and analytic
phrase structure in speech performance; perception of idiomatic
vs. literal productions of ambiguous sentences. Unpublished ms,
ZIV, Yael, 1976, Functions of relative clauses in English and Hebrew
University of Illinois Ph,D, dissertation.
Studies in the Linguistic Sciences
Volume 11, Number 1, Spring 1981
COMPETENCE FOR IMPLICIT TEXT ANALYSIS:
LITERARY STYLE DISCRIMINATION IN FIVE-YEAR-OLDS
Georgia M. Green
A small-scale experiment is described in which
kindergarteners demonstrated an ability to recognize the
authorship of unfamiliar texts by authors whose other
works they had been exposed to, apparently by attending
to the linguistic and rhetorical characteristics that
constitute literary style. A number of these character-
istics are described and exemplified with citations from
the texts used in the experiment. This implicit ability
to analyze text with respect to a number of different
linguistic and rhetorical properties has implications
for the teaching of reading, and these are outlined in
the final section.
Why on earth would anyone imagine that five-year-olds could
tell one literary style from another? I certainly wouldn't have if
someone hadn't asked me if they could.
A couple of years ago, I remarked to someone that when my daughter
Robin was 2 1/2, she claimed to recognize illustrations she had never
seen before, claiming that we already had books we had just gotten. In
fact, what we already had was books illustrated by the same artist, but
the illustrations were of course not identical: what she recognized was
the illustrator's artistic style. My interlocutor, who was a well-known
expert in language acquisition, asked if Robin also thought she recognized
stories she'd never heard before when they were by the same author as
ones she had heard, if she recognized verbal style as easily as artistic
style. My response was that not only did I not know if she had been able
to do that at 2 1/2, I had no idea if she could do it at the time, when
she was 4 1/2.
By the time I had figured out how to find out, I realized that it
would be about as easy to investigate the abilities of a more represen-
tative population of children as it would be to explore the abilities
of one child, and there would be a lot to be gained, for the question
is of more than passing interest, and my interest in it was more than
idle curiosity: if children at the age when reading instruction
typically begins are sensitive to stylistic properties of texts, then
this has far-ranging implications for their text-processing abilities,
and these in turn have implications for diverse aspects of the practice
of reading instruction. I'll return to these eventually.
So, the question was: given that one child, and I presumed many
others, interpreted the similarities in the illustrations of artists
like Lionel Kalish and Tom O'Sullivan, as shown in (1-2), as identities
of a sort at age 2 1/2, could children recognize similarities in verbal
(or literary) style as indicating identity of authorship.
40
(l.ij
(lb)
THE SHEEP
OF THE
LAL BAGH
by David Mark
m
Pictures by Lionel Kalish
/ Lionel Kaltsh ii-
C2a) ,/i.
'4
C2b) i\}^y
41
I arranged to carry out a small-scale experiment, with the aid of a
research assistant, Margaret Laff, in the kindergarten class of a day-care
center in a midwestern university community of 95,000. The participants
were five girls and eight boys, ranging in age from 5.0 to 6.1 years.
These children had not begun formal reading instruction, although two
of them could read unfamiliar texts with some facility.
At our request, the regular classroom teacher read ten books to
the class at times normal for such an activity and in the way she
normally would read to the children, showing the illustrations and
answering questions. It took fourteen days for the books, two by
each of five authors, to be read once. The ten books, read in the order
in which they are listed, are indicated in (3).
(3) Exposure Books
1. Dr. Seuss: The Lorax. New York: Random House, 1971.
2. Margaret Wise Brown: Wait Till the Moon is Full.
New York: Harper 5 Row, 1948
3. Bill Peet: The Ant and the Elephant. Boston: Houghton
Mifflin, 1972.
4. Virginia Kahl : The Habits of Rabbits. New York: Charles
Scribner's Sons, 1957
5. Beatrix Potter: The Tale of Mr. Jeremy Fisher. New
York: Warne, 1906.
6. Dr. Seuss: Happy Birthday to You. New York: Random
House, 1959.
7. Margaret Wise Brown: The Runaway Bunny. New York:
Harper 5 Row, 1942.
8. Bill Peet: Big Bad Bruce. Boston: Houghton Mifflin,
1977.
9. Virginia Kahl: The Baron's Booty. New York: Charles
Scribner's Sons, 1963.
10. Beatrix Potter: The Tale of Peter Rabbit. New York:
Shortly after the last book was read to the group, we prepared the
group for the task of indicating their identification of new stories with
an activity where they indicated whether they recognized which book an
illustration was from. Five-page booklets were distributed to the children.
On each page of the booklets five pictures had been photocopied in black
and white. Each picture represented a major character from a book by a
different one of the five authors mentioned above. In every case the
character came from one of the books read to the children in class, and
with only one exception, the character's name occurred in the title of
the book. The same five pictures appeared on each page but they were
arranged in different orders. For each page, the children were asked to
put a crayon mark on "the picture that looks like it was drawn by the
person who drew the pictures in (title) and (title)": the two titles by
each author were cited in turn. This was an unusual task for the
children and a few seemed puzzled by it. Though most seemed to know the
correct answers, some may have been distracted by wondering why we would
ask something so obvious. We also observed in at least one case that a
child would point to the correct answer, but for some reason could not be
persuaded to mark it. The children got from 2-5 correct as indicated in
(4); nine got 3 or more correct.
42
(4) Illustrations Task Results
Number correct 2 3 4 5 Total
Number of children 4 5 13 13
Then, five tape recordings of other stories by the same authors were
played individually to each child. The stories on the tapes are indicated
in (5).
(5) Test Books
1. Dr. Seuss: I Had Trouble in Getting to Solla Sollew.
New York: Random House, 1965
2. Beatrix Potter: The Tale of Two Bad Mice. New York:
Wame, 1904
3. Bill Peet: Eli. Boston: Houghton Mifflin, 1978.
4. Margaret Wise Brown: Fox Eyes. New York: Pantheon
Books. 1951.
5. Margaret Wise Brown: The Little Fur Family. New York:
Harper § Row, 1946.
Each child heard the tapes in a different order. Some children
heard one story by each author; some heard two stories by one author
and one by each of three others. Thus, not all the children heard all
the authors. This was intended to serve as a check on guessing strategies.
Unfortunately, one third of the children in the second condition did not
complete the task, so we did not draw any conclusions about guessing
strategies.
Before each story, the children were told that at the end of the
story they would be asked to think about which of the books read by
teacher the new story most reminded them of. The children were also
told that when the story was over, they would be asked to make a mark
on a picture in a booklet similar. or identical to one used in the
illustration identification task. Not all booklets were identical:
the children who heard two stories by the same author had five 4-item
pages, while those who heard one story by each author had five 5-item pages.
When each story was over, the interviewer read these instructions to
the child:
If you think this story was written by Beatrix Potter, who wrote
the stories about Peter Rabbit and Jeremy Fisher, put a mark on
the picture of Peter Rabbit.
If you think the story you just heard was written by Virginia
Kahl, who wrote the stories about Gunhilde and the rabbits, put
a mark on the picture of Gunhilde.
If you think that the story was written by Margaret Wise Brown,
who wrote the stories about the runaway bunny and the raccoon
who wanted to go out at night, put a mark on the little raccoon's
picture.
43
If you think the story was written by Dr. Seuss, who wrote the
stories about the Lorax and the Birthday Bird, put a mark on the
picture of the Lorax.
(5-item group only) If you think the story was written by Bill
Peet, who wrote the stories about Big Bad Bruce and the ant and
the elephant, put a mark on the picture of the bear.
After the child had marked a choice, the interviewer asked the child
three questions:
1) Have you ever heard this story before?
2) How did you know it was that one?
3) Tell me something about the story that made you know who
wrote it.
We did not expect to get much in the way of revealing or even true
answers to such questions (5-year-olds have been observed to have no
qualms about making up answers to such questions out of whole cloth) ,
but we were prepared to consider anything indicating awareness of any
stylistic property to be significant.
Responses fell into one of three categories. Many were either "off
the wall" or simply uninformative. For example, in response to question
2, How did you know it was that one?, we got such responses as:
Well my dad told me.
I just knew. I was just thinking in my head. I remembered in
my mind who it was always written by.
Some of these children had correctly matched the author. Some had matched
incorrectly. A good number of responses, however, seemed to indicate at
least a vague awareness of style. For instance, in response to the same
question. How did you know it was that one?, children who had correctly
identified the authorship of the story said things like:
Because . . . uh . . . because they were talking the same.
Urn, because of how they were talking.
Well, it sounds like she's the one (pause) that was talking. It
really sounds like the Lorax girl. See, in little parts of it
it sounded like she was talking. And she was talking in the
Lorax, I think, because she sounds the same as the Lorax girl.
And a few comments showed that at least one child was conscious of certain
determinants of style. For example, responding to the same question, this
child said:
Because I heard the story of Big Bad Bruce and they said something
about the s . . . nort, and they said it too.
Most of the children, predictably, did not have the concentration to
perform the entire task at a single sitting (about 55 minutes), and did one
or two stories at a time. Three or four children did have the concentration
to do this, however, (two of these were readers) and several were so
intrigued with the task of guessing the authorship that they interrupted
the tape to tell us the author (usually correctly) and preferred, contrary
to our expectations, to go on to the next tape, rather than hear the end
of the story.
44
This part of the experiment was not conducted under the best of cir-
cumstances: The tapes were unfortunately excessively "noisy," and the
listening accommodations were not particularly comfortable--usually the
floor of a small room that was not in use.
Children were allowed to discontinue the experiment at any time if
they did not wish to go on. Three children did not complete the task.
One listened to 4 out of 5 stories, one to 3 out of 5, one to 2 out of 5.
Thus the experiment was performed under a number of conditions that
could be expected to bias the results against the hypothesis that children
can identify stylistic traits of texts well enough to match the authorship
of novel texts to texts they have already heard:
1) Children were exposed to only two exemplars by each author
prior to the testing.
2) Children were exposed to each exemplar only once.
3) Exposure stretched over 14 days. The two books by
each author were read for the most part 7 days apart.
4) The testing task was lengthy.
5) The testing was conducted under uncomfortable and dis-
tracting conditions.
Nonetheless, when they listened to tapes of a third work by each
of the five authors, six of the thirteen children who participated in
the interviews were able to correctly identify the authorship of three
or more of the five stories, as indicated in (6).
(6) Style Test Results
Number
of
Children
1 2 3
Number Correct
The probability of randomly choosing the correct item out of five is
0.2. The probability of doing this three or more times in five trials is
around 0.06. This means that six children performed at a level of accuracy
45
highly unlikely to be attributable to chance. The other four who
completed the interviews performed with far below chance accuracy.
In other words, a large percentage of the children performed in such
a fashion as to imply that their comprehension of stories was not
limited to vague outlines of plot and characterization, but extended
to appreciation of the subtler rhetorical and linguistic aspects of
style. Apparently the other part of the group either (a) misunderstood
the task, (b) did not attend to the discriminants of style, or (c) fixed
upon arbitrary guessing strategies.
Correlations
There was no apparent correlation of the percentage correct with
the participants' age or sex as shown in (7).
(7) Comparison of Number of Correct Responses with Age and with Sex
Average Age of:
Total Group 64.3 months
3-5 Correct Group 64.8 months
0-1 Correct Group 63.8 months
Percentage of Girls in:
Total Group .38 (5/13)
3-5 Correct Group .33 (2/6)
0-1 Correct .43 (3/7)
Furthermore, there was no direct correlation between the children's
ability to do well on the illustration pretest and their ability to perform
the style recognition task. This indicates that performance on the style
recognition task is not a simple function of intelligence or ability to
follow directions. Specifically, of the 10 children who completed the
style recognition task, the three children who did best on the illustration
recognition task (matched all 5 pictures correctly) got 0 or 1 correct on
the style recognition task. The children who did poorest (2 correct) on
the illustration task, with one exception, got 0 or 1 correct on the style
recognition task. But the children who did moderately well on the illus-
tration task (3-4 correct) got 3-5 correct on the style recognition task.
A possible explanation for this is that the group that got 100%
correct on the illustration task were accustomed to attending much more
to the illustrations in listening to stories than to rhetorical and
linguistic properties of the text, and that most of the children in
the group that did poorest on the illustration task simply were not
accustomed to attending to either style or illustrations in listening to
stories. But the reason that the children who did best on the style
recognition task did only moderately well in recognizing illustrations
may be that their concentration on the aspects of literary style that
allowed them to recognize authorship precluded their paying more attention
to the illustrations.
In the absence, however, of confirmatory observations of the individual
children, it seems just as justifiable to attribute the gap between the 0-1
46
correct group and the 3-5 correct group to individual differences (e.g.,
sensitivity to language) or linguistic maturity. Another possibility is
that the children in the 0-1 correct group simply had less prior experience
with the authors whose style we chose to investigate. Logically, this
would seem to be a significant variable only if in being read to before the
experiment these children were made aware of the names of the authors of
the relevant books. I have no idea whether this was true in the case of
the children tested. I would guess that the practice of reading the title
page is not widespread, but I know of no definitive investigations.
Personally, I never used to read aloud even the titles of the books I read
to my children, and as a consequence, they developed their own designations
for books. (Thus, my daughter's name for The Cat Who Stamped His Feet
((2a), by Betty Ren Wright, Golden Press, 1974) was "the cat-in-the-attic
book" and her name for The Sheep of the Lai Bagh ((lb), by David Mark,
Parents Magazine Press, 1967) was "the Ramesh book.")
On the other hand, Cazden suggests (personal communication) that prior
exposure to other books by the same authors, even when the author's name is
not mentioned, might provide a child with a frame in which to assimilate
and categorize stylistic properties of texts.
Before I go on to describe with more specificity the linguistic and
rhetorical aspects of text that these children must have been attending to
in order to make the correct judgments that they made, I will describe how
we selected the materials for this task, because we took great pains to
avoid using materials that would allow a participant to make correct answers
based on text properties that we considered not particularly linguistic, such
as similarities of subject matter, or familiarly named protagonists.
Selection of materials was not a matter to be taken lightly. We knew
that children might use subject matter or characters' names to decide author-
ship. For example, in a similar forced-choice task, one child correctly
chose "the author of the Babar books" as the author of an unfamiliar para-
graph referring to an individual named Arthur, and "the author of Hi, Cat
and Whistle for Willie" as the author of an unfamiliar paragraph referring
to a dog named Willie. When questioned, she replied that she had made her
judgements on the basis of the name Arthur and the name Willie, respectively.
Thus, our materials had to meet all of the following criteria:
1. The author had to have a distinct style. If we were not able,
intuitively, to identify an author's works as stylistically unique, we did
not consider her or his works as candidates for inclusion in the study.
This eliminated a number of celebrated children's authors, including Ezra
Jack Keats and Robert McCloskey.
2. The author had to have written at least three books which were not
all about the same unique subject matter. This ruled out, e.g.. Jay Williams,
among whose books we could find only one that was not about princesses or kings.
3. The author had to have written at least two books with nonover-
lapping sets of characters. This, regrettably, ruled out many authors with
strongly individual styles, for example, the de Brunhoffs, authors of the
Babar books. We considered including such authors, and changing the
characters' names so as not to "give away" the authorship. We rejected
this strategy however, on the grounds that (a) the kinds of names an author
47
chooses are an aspect of style, and we did not want to compromise the
integrity of the experiment by meddling with even one aspect of an author's
style, and (b) if a child did know such an author's works well, it might
be unfairly confusing to ask for judgment on a work that both is and isn't
that author's.
4. We had to have access to at least three books by the author that
shared a distinct style. This eliminated such stylistically interesting
authors as Maurice Sendak and Rosemary Wells, since we couldn't find three
books (on the shelf at the local library) that met our other criteria and
shared the same style.
5. At least one of the books, and preferably all three, had to have
a text which could present the story independently of the illustrations,
so that (a) the familiarization stories could be equally well assimilated
by children sitting farther from the teacher and by children clustered
closely around her, and (b) the taped story would not be incomprehensible.
The testing had to be done with tapes of the books rather than
exemplars, even exemplars that obliterated the author's name, in order to
eliminate the possibility that the children might identify the authorship
by identifying the illustrations, which in most cases here were done by the
author. Also, we wanted to eliminate the graphics (type face, layout) as
a possible source of identification, since we had observed that at least
some two-year-olds can recognize these things and "read" the Crest, K-Mart,
Sears, and Special K logos. (One two-year-old insisted for months that a
certain supermarket was an ice cream store, despite regular correction.
Eventually his mother noticed that the lettering on the store's sign was
very similar to that used by the Baskin-Robbins chain, and made some headway
in clearing up the confusion.) We figured that five-year-olds might also
use such cues, and we wanted to eliminate them.
What we eventually ended up with was the following: two authors who
wrote in rhymed couplets and used lots of long words: Dr. Seuss and
Virginia Kahl; and three authors who wrote about anthropomorphized
animals: Beatrix Potter, Margaret Wise Brown, and Bill Peet.
Finding five authors who met all of our criteria was very difficult.
In the initial planning of the study, we feared that including Dr. Seuss
might bias the experiment in favor of the hypothesis. However, the discovery
of Virginia Kahl allowed us to include both authors in the study, as both
write verse fantasy in similar meter. Samples are reproduced in (8) and (9).
C8a) "Oh, help!" cried the Duchess. "Our children are gone!
They're not in the castle--they're not on the lawn--
They're not in the gardens. Are they down in the moat?'
"If they are," said the Duke, "let us hope they can float."
"They have vanished, they've all disappeared from our sight.
Our dear little daughters give one such a fright.
(Virginia Kahl: The Baron's Booty)
(8b) The message told what the men had seen:
An enormous beast of yellowy-green
With a sinuous neck and a small fierce head
That had no hair but had horns instead.
(How Do You Hide a Monster?)
48
(8c) And everyone cried, "There's been an error.
That beast is never a cause for terror.
He'd never harm us; he's kind and true.
We must protect him; what shall we do?"
At last they announced, after due reflection,
"We'll send the men off in the wrong direction"
(How Do You Hide a Monster?)
(8d) They all ate their pancakes--the very last crumb;
But when they had finished, they all remained dumb.
Then said the Good Wife, "Now, why don't you praise me?
Your manners are dreadful--you really amaze me.
You know that my pancakes are fluffy and flavory.
Tender and toothsome, incredibly savory- -
Served with a syrup so pure and delightful.
That you've often swooned when you've bitten a biteful."
(The Perfect Pancake)
(9a) Down slupps the Whisper-ma-Phone to your ear
and the old Once-ler's whispers are not very clear,
since they have to come down through a snergelly hose,
and it sounds as if he had smallish bees up his nose.
(Dr. Seuss, The Lorax)
(9b) But I'm also in charge of the brown Bar-ba-Loots
Who played in the shade in their Bar-ba-loot suits
and happily lived, eating Truffula fruits.
NOW . . . Thanks to your hacking my trees to the ground,
there's not enough Truffula fruit to go 'round.
And my poor Bar-ba-Loots are all getting the crummies
because they have gas, and no food in their tummies.
(The Lorax)
(9c) I was real happy and carefree and young
and I lived in a place called the Valley of Vung
And nothing, not anything, ever went wrong
Until . . . well, one day I was walking along
And I guess I got careless, I guess I got gawking
At daisies and not looking where I was walking.
(I Had Trouble in Getting to Solla SoUew)
(9d) I dreamed I was sleeping in Solla Sol lew
On the banks of the beautiful River Wah-Hoo
Where they never have troubles. At least very few.
Then I woke up. And it just wasn't true.
I was crashing downhill in a flubbulous flood
With suds in my eyes and my mouth full of mud.
(Solla Sollew)
(9e) Our camel, he said, had a bad case of gleeks
and should be flat in bed for at least twenty weeks.
(Solla Sollew)
49
(9f) I listened all night to the growls and the yowls
And the chattering teeth of those mice and those owls.
While the Midwinter Jicker howled horrible howls.
I tossed and I flipped and I flopped and I flepped.
It was quarter past five when 1 finally slept.
(Solla Sollew)
(9g) We're marching to battle. We need you, my boy.
We're about to attack. We're about to destroy
The Perilous Poozer of Pamplemousse Pass!
So, get into line! You're a Private, First Class,
CSolla Sollew)
(9h) They smell like licorice! And cheese!
Send forty Who-Bubs up the trees
To snip with snippers! Nip with nippers!
Clip and clop with clapping clippers!
Nip and snip with clipping cloppers !
Snip and snop with snipping snoppers!
(Happy Birthday to You)
Similarly, by choosing three animal story authors, we hoped to eliminate
topic as a cue to authorship, and force the judgments to depend on subtler
cues. Indicative samples of the three authors' texts are reproduced in
(10-12).
(10a) "If you are a gardener and find me," said the little bunny,
"I will be a bird and fly away from you."
"If you become a bird and fly away from me," said his mother,
"I will be a tree that you come home to."
(Margaret Wise Brown: The Runaway Bunny)
(10b) Once upon a time in the dark of the moon there was a little
raccoon.
(Wait Till the Moon is Full)
(10c) "Does everyone sleep at night?" asked the little raccoon.
"No," said his mother, "not everyone."
"Who doesn't?' asked the little raccoon.
"All things that love the night," said his mother. "Wait
till the moon is full."
"Is the moon a rabbit?" asked the little raccoon.
"No," said his mother. "The moon is a moon. A big round golden moon.
"Will I see it soon?"
"Wait," said his mother. "Wait till the moon is full."
(Wait Till the Moon is Full)
(lOd) There was a little fur family
warm as toast
smaller than most
in little fur coats
and they lived in a warm wooden tree.
(Little Fur Family)
50
(lOe) Then the little fox climbed an apple tree. Along the bark of
the tree the eye of a tree toad closed suddenly. The fox
coughed, "Whiskerchew!" And the tree toad knew that someone
had seen him hiding there in plain sight against the bark of
the tree. Some children who were supposed to be taking a nap
in the afternoon weren't sleeping at all. "Whiskerchew!"
the fox coughed. And the children knew that the fox knew
that they were not sleeping. All this the fox noted, and he
went on his way,
(Fox Eyes)
(11a) Peter gave himself up for lost, and shed big tears; but his
sobs were overheard by some friendly sparrows, who flew to
him in great excitement, and implored him to exert himself.
(Beatrix Potter: The Tale of Peter Rabbit)
(lib) 1 am sorry to say that Peter was not very well during the
evening.
(Peter Rabbit)
(lie) So that is the story of the two Bad Mice, --but they were not
so very very naughty after all, because Tom Thumb paid for
everything he broke.
(The Tale of Two Bad Mice)
(lid) "What a mercy that was not a pike!" said Mr. Jeremy Fisher.
"I have lost my rod and basket; but it does not much matter
for I am sure I should never have dared to go fishing again!."
(The Tale of Mr. Jeremy Fisher)
(lie) And instead of a nice dish of minnows--they had a roasted grass-
hopper with lady-bird sauce; which frogs consider a beautiful
treat; but I think it must have been nasty!
(Jeremy Fisher)
(12a) "Where in blazes did you come from?!!" she shrieked, giving
the boulder a vicious kick.
(Bill Peet: Big Bad Bruce)
(12b) Once upon a time there was a lion named Eli who lived in the
faraway land of Kumbumbazango. He was a decrepit old cat with
a scruffy mop of mane, and most of the thunder had gone out of
his roar. Now, after many long years as a proud king of beasts
the old lion had finally become as meek as a mouse.
(Eli)
(12c) In one frantic leap, and with a wild swing of a paw, Eli caught
the jackal with a clout to the snout that sent the little rascal
yelping away with his tail between his legs.
(Eli)
51
(12d) Raising his voice to a rumble to make sure all the birds could
hear, the lion let them have it. "You good-for-nothing grubby
old bone-pickers! You flea-bitten beggars! You ugly old coots!
You give me the creeps! Skedaddle! Take off! Get a tree of
your own! Leave me be!"
(Eli)
(12e) "Wade out into that soup and scrunch down in the gunkazunk grass.
The Zoobangas will never look for you there."
(Eli)
What might children have been picking up on to make the correct
identifications that they made? Let us begin with the verse selections.
At first, the similarities between Kahl and Seuss may seem more striking
than the differences. Both write obvious fantasy with a strong 4-foot
meter, mostly anapests. And both do not hesitate to use words likely to
be unfamiliar to young children. But here the similarity ends. Seuss'
unfamiliar words tend to be unfamiliar because they're invented (slupps,
snergelly, gleeks, flubbulous, snop) , whereas Kahl's are likely to be
unfamiliar because they are drawn from the formal, academic register of
language, to which few young children have been exposed, and hardly any
have attended. Sometimes she uses basically academic or literary words
in her verses [e.g., error; toothsome, swooned) , but much of the unfamiliar
word usage is just academic senses of words in common usage in children's
books (e.g, reflection in the sense 'thought,' due in the sense 'sufficient,
true in the sense 'loyal,' dumb in the sense 'mute'). Although the plots
are comparatively simple and predictable, the whole tone of Kahl's stories
is old-fashioned and/or mock-academic, and this is reflected in the syntax
as well, in such phrases as cause for terror, and after due reflection,
and in the non-anaphoric use of the pronoun one to mean 'a person,' and
the Germanic verb-second syntax of Then said the Good Wife.
In contrast, the tone of the Dr. Seuss stories is very intimate and
conversational. This is reflected in the vocabulary, where we find such
colloquial items as smallish, tummies, the crummies, real used to itensify
an adjective, and the contraction go ' round . The conversational tone
shows up just as strikingly in the syntax, in such locutions as the
introductory well, the hedge I guess, and the get + present participle
construction (got walking) .
Then there are the Seuss trademarks--the made-up species (Bar-ba-
Loots, Truffula, Who-Bubs), and the coined place names (Valley of Vung,
River Wah-Hoo) , and the novel compound nouns (Bar-ba-loot suits, Super-
Axe-Hacker, Key-Slapping Slippard) . Finally alliteration, assonance, and
consonance, as in selections (9g-9i), are much more characteristic of
Seuss' verse than of Kahl's.
There are differences in length (the Seuss stories are longer) and
in plot construction: the Seuss stories involve more episodes, are less
predictable, and generally involve a human protagonist in interaction with
non-human species (or only non-human characters), whereas Kahl's stories
involve almost exclusively human protagonists (the sole exception is a
Loch Ness-type monster). However, I suspect that these global properties
of the texts were less salient to the children than the more linguistic
differences, and this feeling is supported by the fact that several children
made judgments (usually correct) before they had heard one-tenth of a
52
story. Without hearing a longer selection, they couldn't easily have
formed correct judgments about such global properties as length and plot
construction.
Furthermore, when the children mentioned reasons for their choices,
they were usually framed in terms like "it sounded like . . . ," although
one child, justifying an incorrect choice mentioned particular actions:
I think it's (by the author of) Peter Rabbit because they
were planting things and stuff. They were planting carrots.
As it turned out, the Dr. Seuss story was identified correctly 7 out
of 12 times; one Kahl story was identified correctly 4 out of 11 times,
the other once in two trials. Among the six children who identified the
authorship of three or more stories correctly, the Dr. Seuss story was
misidentified only once (as being written by Kahl), and the Kahl story
was misidentified twice.
What cues allowed the children to recognize stories as being written
by Brown, Potter, and Peet? First of all, although all three begin their
stories traditionally enough with Once upon a time or Once there was or
There once was, there are striking differences in the register used to tell
the stories. Peet's stories have a colloquial (scrunch, clout, snout),
even earthy tone. He minces n words; his characters are scruffy, decrepit,
crafty. They don't just say or cry or even shout things, they shriek and
let them have it. And his characters, who tend to be rather bad-tempered,
don't mince words either. Roxy comes as close to cursing in (12a) as you
can in a picture book, and Eli sounds almost like a Marine drill instructor
when he calls the vultures all those colorfully rude names in (12d).
On the other hand. Potter's stories, written in Edwardian England,
sound like it. When Mr. Jeremy Fisher curses, it's "What a mercy that was
not a pike!" Some of the vocabulary is very formal and literary (implored,
exert) . Many of the phrases strike the modem ear as old-fashioned or
maiden -auntish, for example shed big tears, so very very naughty, it does
not much matter, I should never have dared to.
The register of Brown's stories is that of bedtime storytelling. As
in the Bank Street College's "Here and Now" stories, for which Brown was
a principal writer, these stories are almost exclusively dialog, with a
little bit of narration and description, and the description is exclusively
literal. In this way, her comparatively plain prose contrasts with Peet's,
which makes copious use of figurative language: Eli has a mop of mane,
and the thunder had gone out of his roar. When the vultures urge Eli to
wade into the swamp (12e), they call it soup. It also contrasts with
Potter's, in that Potter almost always interrupts her narrative at the end
and makes her presence felt with comments like those in (lib, lid, and lie).
I do not mean, by saying that Brown's prose is plain, to imply that it
is either colorless and boring, or lacking in style. It has a lyrical
rhythmicity, clear in the refrain "Wait," said his mother. "Wait till the
moon is full," as well as in the selections in (10b, lOd and lOe) . And there
is so much internal rhyme and half-rhyme (as in 10b, lOd, and lOe) , that
some of the passages almost seem to be in verse. Furthermore, Brown's prose
has a cyclical structure that also mark it as unique, at least among this
group of authors. This structural cyclicity shows up plainly in the conditional-
counterconditional repartee (exemplified in lOa) that constitutes almost the
53
whole of The Runaway Bunny, and it is no less clear in the repeated requests
in Wait Till the Moon is Full that are answered, every page or so, with the
refrain "Wait," said his mother. "Wait tillthe Moon is full." as in (10c).
All of the Brown stories used in the experiment are quiet, calm storieSj
with no violence and a comparatively low level^of suspense--what is going
to happen is never a matter of life and death. In contrast to Brown's
simple, almost plotless stories, Peet's and Potter's stories involve un-
predictable chains of episodes, and in Peet's these involve embedded and
conflicting plans. All of the stories by these three authors that the
children heard have animals as the main protagonists, but Brown's are
almost always presented as juvenile and "pedomorphized" while many of
Potter's and all of Peet's are full-grown, though not grown-up--they act
and react like children. At least one of the stories by each author also
involved human beings, though always as minor characters.
The Potter book in the text (The Tale of Two Bad Mice) was correctly
identified 4 times out of 11; the Brown books (Fox Eyes and Little Fur
Family) 7 times out of 13, and one time out of 4, respectively; and the
Peet book (Eli) , two times out of 6. Among the six children who correctly
identified the authorship of three or more books. Fox Eyes was correctly
identified 5 out of 6 times, and Little Fur Family 1 time out of 2;
The Tale of Two Bad Mice was correctly identified 4 times out of 6, and
Eli, 2 times out of 4.
Let us turn for a moment to address the question of accounting for
the errors that were made. What might have caused some of the confusions?
We can identify a number of cross-author similarities that might account
for some of the errors. Both Peet and Seuss use very colloquial vocabulary
and syntax. And Peet, like Seuss, refers to obviously invented species
and places (gunkazunk grass, Zoobangas, Kumbumbazango) , though Seuss'
are more often compounded of familiar morphemes than Peet's. Both Potter
and Kahl use a fairly formal and literary vocabulary and syntax. Seuss
as well as Potter intrudes into the narrative and makes the author's
presence felt. Happy Birthday to You and The Lorax are specifically
addressed to the reader, the former as an extended wish, the latter as
a sort of reverie. I Had Trouble in Getting to Solla Sollew unsurprisingly
is a first-person narrative. Both Kahl and Brown tell simple stories,
with relatively predictable plots, though Kahl's are more complex, and
some of Brown's have hardly any plot at all. Finally, while Potter's
stories are not as lullaby-like as Brown's, the suspense is muted, the
action damped, by the calm, matter-of-fact tone of the telling.
How well do these similarities account for the errors that were
actually made? If the errors had been random and evenly distributed,
half of them would have been in cells predicted by these similarities.
In fact, 56% of the errors were in these cells (56.5% of the errors by
the 0-2 correct group, and 55% of the errors by the 3-5 correct group) .
And four children made symmetrical errors, for example identifying the
Potter story as by Kahl and vice versa, suggesting that the errors were
not random, but were based on some perception of similarity.
Implications
This study appears to show that at least some five-year-olds have
the ability to appreciate and discriminate among the literary styles
54
available in books intended for young children. Indeed, several children
found the challenge of testing this ability exhilarating.
I cannot show that what the children were attending to when they
correctly identified the authorship of stories they had not heard before
was in fact the linguistic and rhetorical aspects of literary style that
i have indicated. (I couldn't prove that, even if the experiment had
been conducted with well-read and highly articulate adults), but it
seems a good bet. In any case, it means that the children understood
a whole lot more than the bare outlines (or even dressed-out outlines)
of plot. Making the correct judgments almost certainly entailed not
only noticing and abstracting from very fine details of wordcraft, but
also attending to and abstracting from global structural matters of form
and content.
If it is true that five-year-olds generally, and by extension, six-
and seven-year-olds, have the ability to make such fine discriminations,
then it seems likely that they would be able to tell the difference between
the prose in ordinary children's books of the sort I have been discussing,
and the prose in their readers, a sample of which is given in (13).
(13) Rabbit said, "I can run. I can run fast. You can't run fast."
Turtle said, "Look Rabbit, See the Park. You and I will
run. We'll run to the park."
Rabbit said, "I want to stop. I'll stop here. I can run,
but Turtle can't. I can get to the park fast."
Turtle said, "I can't min fast. But I will not stop. Rabbit
can't see me. I'll get to the park."
Such prose is edited to conform to readability formulae which impose
strict limits on sentence length and vocabulary. Owing to the strict
constraints imposed by the publishers of basal readers on sentence
length, vocabulary, and story length, these works end up being designed
in such a way that they are devoid of most characteristics of individual
style. If it is generally true that at the age when reading instruction
begins, children attend to and appreciate stylistic differences, then it
would seem to follow that expecting them to read such basal readers is,
to say the least, inconsiderate. At best it is pointless; at worst it
is counterproductive. It wastes valuable time that could be spent in
more profitable ways and risks boring the children and conveying to
them that there is nothing interesting to be learned in books, or even
in school. Is it possible that Johnny doesn't learn to read because
there is no thrill in being able to read texts like (13), which is from
what is supposed to be a version of Aesop's fable about the hare and the
tortoise?
The objection is likely to be raised that the fact that 5-year-olds
can appreciate the differences between works by Beatrix Potter and
Margaret Wise Brown does not mean that 7 -year-olds could read the works
of either author independently, that 7-year-olds have enough trouble
reading the admittedly anemic prose in the basals. It is certainly
true that there is no direct entailment from what 5-year-olds can compre-
hend orally to what 7-year-olds can independently read, but I think
this study suggests that 7-year-olds might be able to read Margaret Wise
55
Brown and Beatrix Potter; the fact that some have trouble with second-
grade basals might be due to stylistic properties of the basals that are
introduced in the process of writing a graded reader. Work at the Center
for the Study of Reading (Davison, Kantor, et al . , 1980) has shown that
many of the devices used in adapting a text to meet sentence-length,
vocabulary, and passage-length requirements contribute to a marked decrease
in the coherence and interest of the text. In addition, it is a basic
principle of attention theory that perceptual activities which demand more
mental processing tend to be favored over less demanding activities
(Hardiman and Zernich, 1978). Successfully meeting a challenge is itself
a source of pieasure and satisfaction. If some seven-year-olds have
trouble with grade-level basal readers, it may be a problem of motivation;
it may be that they would do better on more complex, more difficult,
more challenging, more rewarding material.
If the ability to discriminate literary styles is general among
primary-grade children, then it may be that by editing their readers to
meet someone's preconceived notions of what is easy, we are depriving
children of the satisfaction of meeting a challenge, and contributing to
making learning to read an unpleasant experience.
NOTES
*This papers was presented at the 32nd Georgetown University Roundtable
on Languages and Linguistics, March 20, 1981.
For testing kindergarteners' ability to recognize literary style,
we considered a number of tasks. A simple recognition task, wherein a
child would be asked if a passage had been heard before, was rejected
as not directly tapping the abilities we wanted to test. A 2 x 2
forced-choice task (matching unfamiliar (or familiar) passages with
familiar authors' names two at a time) was rejected as not very informa-
tive, since making one incorrect answer practically entailed making
another, and vice versa, one correct answer practically entailed making
another correct answer. A 2-out-of-3 (or more) matching task, where a
child would be asked to say which two passages out of a group were by
the same author, was rejected as logistically unfeasible for nonreaders:
The passages would have to be presented orally, and we judged that it
would be asking too much to ask children to remember three or more passages
and their order of presentation, in order to say which two were most alike.
We wanted to make the task as difficult as we coold and still get
better-than-chance performance so that it would test the limits of the
children's ability and so that the results would be as informative as
we could manage. For this reason, we settled on a l-out-of-5 multiple
choice style-matching task, with the test materials containing as few non-
style-related clues as possible.
2
Brovm did not limit herself to "lullaby" stories. A bizarre and
aggressive picture-book called The Steamroller: A Fantasy (Walker
Publishing Company, 1974) shows an entirely different side of her.
In response to perceived demands from text book selections bodies.
REFERENCES
DAVISON, A., R.N. Kantor, J. Hannah, G. Hermon, R. Lutz, and R. Salzillo.
March 1980. Limitations of readability formulas in guiding adaptation
of texts. Technical Report No. 162. Urbana: University of Illinois,
Center for the Study of Reading. (ERIC Document Reproduction Service
No. ED 184 090)
56
HARDIMAN, G.,and T. Zemich. 1978. Basic research: Aesthetic behavior
in the visual arts. Viewpoints 6:1,29-38.
Studies In the Linguistic Sciences
Volume 11, Number 1, Spring 1981
EPENTHESIS AND ELISION IN METRICAL PHONOLOGY
Chln-W, Kim
In metrical phonology (Halle and Vergnaud 1978; Clements and
Keyser 1980) , the syllable structure has been proposed as having
the role of monitoring the well-formedness of the phonological
string. Thus, when the syllable template uncovers a "standard"
consonant, a minimal syllable structure is assigned to It, and
a vowel is Inserted under the empty node. For example, if the
permissible syllable structures in a language are CV and CVC, a
string CCVCC would leave the first and final Cs stranded, i.e.,
A A A
C CVC c
and a vowel would be inserted under the empty nodes to construct
minimal syllables. Three problems relating to this will be
discussed.
1. On what principle is the syllable template assigned?
Given permissible syllable structures of CV, VC, and CVC, is a
A A /^ A /X
string of CC assigned the template C C , C C, or C V C?
2. Can a phonological rule violating the syllable structure
be allowed to apply, only to apply another rule whose sole
function is to repair the damage done by the previous rule?
(e.g. in Klamath, a vowel is elided to yield an ill-formed
syllable structure. An adjacent glide is then "vocalized" to
rectify it.)
3. An impermissible C cluster can be resolved either by a
vowel epenthesis or by a cluster simplification (i.e. a C elision).
Is there a principle that governs this choice?
Recent work in metrical phonology has stressed the importance of syllabic
structure in the application of certain segmental rules as well as in the
assignment of stress. In particular it has been argued that the syllable
structure plays a role in monitoring the well-formedness of phonological
strings. Thus, when during the construction of a metrical tree, a string
is uncovered which does not conform to a language-specific canonical sylla-
ble structure, certain rules apply to restore a canonical form.
The earliest exemplification of this role of metrical phonology was
given by Halle and Vergnaud 1978. Citing data from Kenstowicz 1977 (see also
Kenstowicz and Kisseberth 1979: 22A) , they argued that the phenomenon of vowel
epenthesis in Harari, a Semitic language in Ethiopia, can be handled in a
predictable manner by appealing to a mapping procedure between the nodes in
the metrical tree and the segments dominated by the nodes.
58
For example, the underlying string /t+sSbr/ 'to break' (2 masc. Imp.) is
phonetically realized as [tisSbri] . The epenthetic vowel [i] is inserted
just where an empty slot occurs when the segments in the string are assigned
the permissible syllable structures in the language, i.e., CV and CVC , as
shown below
(1)
A
A / /\ A
t + s a b r
There is something intuitively appealing and elegantly simple about this
treatment of vowel epenthetis. Yet there are some indeterminate procedural
problems in the handling of epenthesis in this way, and this paper is
addressed to these problems.
The question is this. Given a string of segments, on what principle(s)
are the syllable trees built or syllable templates assigned? Are they built
from left to right or in the opposite direction? When a given string is
amenable to more than one way of syllabification (template assignment), what
governs the choice of one structure over another? For example, given per-
missible (canonical) syllable structures of CV , VC, and CVC in a language,
how is the impermissible consonant cluster CC to be resolved? By way of a
vowel epenthesis or a consonant elision? ^f by epenthesis, is a vowel
a;
epenthesized between the two C's as in - ^
T
A A ^
in . C . C, or to the right of each C as in C
T T
V V
or to the left of each C as
+ ^ +
V V
consonant elision, which one deletes and which one survives?
principle governs the choice?
And if by a
Again what
A pair of principles guiding the syllable-tree building emerge from the
literature. They are:
(2) a. Assign a MAXIMAL syllable structure , if possible, to the string
from the beginning of the utterance (i.e., from left to right); and
b. If any segment is still left stranded, assign a MINIMAL structure
to it.
Thus, returning to the Harari example, the string CCVCC (t+sabr) is
assigned the template shown in (1) where the middle maximal stYucture is
flanked by a minimal structure on both sides. Note that if one has the option
of building a minimal structure on the existing segments and a maximal structure
on stranded segments, one can have the following result, which is unacceptable
in Harari.
(3)
A A
C V
*ltisMbir]
59
There are situations, however, in which these principles (which I will
call "Maximum" and "Minimum") do not give an unambiguous resolution. To
see this, suppose that a language allows VC as well as CV and CVC. Then,
given CCVCC, the stranded initial and final consonants can be resolved in
several ways as shown in (4) .
A
(4) a. C
A A
CVC c
A
b. c
A A
CVC c
c. C^
A A
CVC c
A
d. C
A A
CVC c
On what basis do we choose one over the other? Here, one may take either
a universal approach or a language-specific approach, Clements and Keyser
1980 appear to take the latter approach when they say that a theory of the
syllable in a given language contains, in addition to a schema defining the
set of possible syllables, what they call "rules of disambiguation" which
select one syllable structure for a given phonetic string whenever the
schema permits more than one syllabification of such a string. On the other
hand, one can plausibly entertain taking the first path. One can for instance
invoke Kiparsky's (1980) notion of Universal Syllabic Template and say that,
according to this universal template, resolving of a stranded consonant to
CV rather than to VC, or CC to CVC rather than to VCC or CCV, is to be
preferred. This universal approach at least fits the observed Harari data.
Other available data indicate, however, that options for different
syllabifications for different languages and/or dialects should be permitted.
Broselow 1980 for example describes a difference between Egyptian Arabic and
Iraqi Arabic in terms of the way a word-medial tri-consonant cluster is
resolved. In Egyptian Arabic, the vowel i^ is epenthesized to the right of
the middle stranded consonant, while in Iraqi Arabic, ± is epenthesized to
the left of the stranded consonant, as shown below:
A
A A
(5) Egyptian: CVCC CV ?ul+t + l + u->
\ said I to him
?ultnu 'I said to him'
A A A
Iraqi: CVC CCV gil + t + 1+a-*
-h giUtla
'I said to him'
Halle and Vergnaud 1978 also cite two dialects of Berber differing in a
similar way. For example, in order to account for two different pronunciations
of the underlying form xdmx 'I worked' in two dialects (styles?) of Berber,
they posit two different syllabifications of the string as shown in (6) :
60
(6) a. X d m x -> [xsdmex]
+ +
a
A
b. X d mx ->■ [axdamx] ->• [xdemx]
+ +
The nature of the epenthetic vowel (i^ vs. o) is undoubtedly language-specific,
but it is not clear that we can say the same with regard to the position of
epenthesis. The lack of canonical syllabic structure VC in Harari apparently
dictates the right-handed epenthesis of a vowel, but Berber has both CV and
VC, and considering the fact that CV is a universally less marked syllable
paradigm than VC, it is not easy to explain the left-handed epenthesis of a
vowel in Berber.
Instead of talking about the right- or left-handed epenthesis, one might
look on this as interpreting the "stranded" consonant as a part of the onset
or the rhyme of a "degenerate" syllable with a null nucleus. Selkirk 1980,
also noting this problem, calls the right-handed epenthesis the onset analysis,
since the consonant becomes the onset of the resulting syllable, and the
left-handed epenthesis the rime analysis, since the stranded consonant becomes
a part of the rime in this case. She then goes on to say:
"The answer is less easy to come by when we compare the forms
with the C (=onset) and ^C (=rime) analysis. In all cases,
either one would in principle have been possible. I do not in
fact have a satisfactory answer to the question here. Following
Aoun 1979, I will assume that, in the unmarked case, it is the
rime analysis of a lone C that is chosen — this because the rime
rather than the onset is the essential element of the syllable.
We could assume that, by universal principle, a lone C is given
the rime analysis." (Selkirk 1980:14)
The marked case of the onset analysis (i.e., the right-handed epenthesis) is
derived by a rule of what Selkirk calls "Onset switch".
I am not sure that I agree with the Selkirk's statement that the rime
analysis is unmarked because the rime rather than the onset is the essential
element of the syllable. I should think that CV is a less marked syllable
shape than VC. In the Universal Syllabic Template, the first bifurcation is
into onset and rime, a further bifurcation of rime into nucleus and coda
being secondary, and Kiparsky 1979 has given a metrical account of why VCV
should be syllabified as V.CV not as VCV, and why it is more natural for
iu to become ^H (CV) rather than iw (VC) , etc.
The Berber example raises another problem. Recall that we earlier
adopted the "Maximum" and "Minimum" principles and the left-to-right direc-
tion for assigning the syllable structures on a phonological string. If
61
we follow these principles, the syllable tree for the Berber example should
not be as given in (6) but as given in (7) below.
y\
(7) X dm X
+ +
Given the fact that in Berber the maximal syllable structure is CVCC, that
the minimal syllable involving a single consonant is CV or VC , that a vowel
is epenthesized to the left of the stranded consonant, thus favoring VC
over CV, and that the syllable structures are assigned to the string from
left to right, neither tree in (6) is derivable. (6a) happens to be pho-
netically identical with (7), but (6b) is derivable only if the syllable
structure assignment proceeds from right to left. In view of the fact that
in most languages stress assignment is sensitive to syllable structures
starting from the end of a word, the right-to-left direction of syllable
template assignment is altogether reasonable. It also implies that in a
language whose stress assignment counts syllables from the beginning of a
word, the syllable trees should also be built from left to right.
Epenthesis in Palestinian Arabic is particularly instructive in that
not only both left and right epenthesis may be found in the same language
but also even the "Minimum" principle does not appear to work. Examine
the following data from Abu-Salim 1980 where the underlined i^ indicates
the epenthetic vowel.
(8) a. kalam i^ rsaas 'a pencil'
b. sami^n baladi 'home-made butter'
According to Abu-Salim, pe'rmissible syllable structures in Palestinian
Arabic are CV, CW, CVC, and CWC. Now, if one assigns these syllable
structures on the underlying form of (8a) , the initial consonant of rsaas
is left stranded, i.e.,
/
A A
(9) kalam rsaas
If a minimal syllable structure is assigned to this stranded consonant,
one should get [kalam rl^ saas] , for the minimum syllable structure in
Palestinian Arabic is CV, and therefore a syllable tree r 0 must be
built on the stranded consonant _r, epenthesizing a vowel to its right.
But this is not an acceptable form. The acceptable form given in (8a)
suggests that the vowel must be epenthesized to the left of the stranded
consonant, i.e., is interpreted as part of rhyme. This means that the
A A
minimum structure to be built on the stranded consonant is p C, not C 0.
But this is not possible, for Palestinian Arabic does not have a syllabic
structure VC. One might argue that in this case a vowel was epenthesized
to the left of the stranded consonant so that the epenthetic vowel comes
between the two compound elements, i.e., at the compound boundary, thus
62
preserving the unity of compound elements. This argument however does not
work, for the forms like (8b) show that the vowel is epenthesized to the
left of the stranded consonant regardless of the position of the compound
boundary. The following example, again from Abu-Salira 1980, shows that
the same word behaves in a parallel way.
(10) a. ?akl i ktiir 'a lot of food'
b. ?aki^l 9arabi 'Arabic food'
The form given in (10a) presents yet another problem. When the syllabic
templates are assigned to the underlying form of (10a), two consonants, the
final consonant of ?akl and the initial consonant of ktiir, are left stranded,
as shown in (11) :
/A /\ A
(11) ? a k 1 ^ k
Assigning a minimal syllable structure to these stranded consonants and
epenthesizing vowel i^ to the empty nodes to the left of the stranded conso-
nants (despite the problem just mentioned), one should get [?aki^li^ ktiir].
But this is incorrect on account of the presence of the first epenthetic
vowel. Although it is possible to argue that both vowels are epenthesized
and that the superfluous vowel is deleted by a syncope rule that deletes
unstressed high vowels in non-final open syllables (cf. Kenstowicz 1980),
a simpler alternative is to say that the two stranded consonants form a
permissible syllable CVC by inserting a vowel between them. That is, what
is assigned is not a minimal syllable structure on each and every successive
stranded consonant, but rather a minimal syllable structure incorporating
as many of the stranded consonants as possible, so long as this syllable
satisfies the canons of the language. A single consonant still left
stranded would be assigned a minimal syllable structure on that consonant.
There is some evidence that this approach, rather than a solution via
a syncope rule, is correct. This evidence emerges when one examines a
case that contains at least three stranded consonants in succession, for
example, /jjisr 1-kbiir/ 'the big bridge'. Assignment of maximal and
minimal syllable structures on this string yields
AX A A /\ /^
(12) jjis r 1 k biir
Epentheis of vowel ± to the left of each stranded consonant gives
[jjisir 1^ likbiir] . Application of the syncope rule that deletes unstressed
high vowels in non-final open syllables, i.e., when followed by CV, will
delete all but the final epenthetic vowel, yielding the ill-formed
*[jjisr li^kbiir] , the correct form being [jjisir li^kbiir] . The correct
form can be derived if one assigns the minimal syllable structure on the
stranded consonants incorporating as many of them as possible, as shown
in (13).
63
/>- /K
(13) jjis r 1 k biir -»■ [jjisi^r likbiir]
This incidentally shows that the syllabic template assignment should
proceed from left to right, not from right to left, for the latter pro-
cession yields an incorrect form
X>.. A A />A. .^
(14) jjis r 1 k biir -»- [ j j isri^ljjcbiir J
One can still derive the correct form via a syncope rule by assuming that
epenthesis and sycope rules are cyclic rules, applying at the word level
first, then at the phrasal level:
(15) [[jjisr] [Ikbiir]]
[jjisi^r] [j^li^kbiir] epenthesis: 1st cycle
[likbiir] syncope: 1st cycle
epenthesis: 2nd cycle
syncope: 2nd cycle
[jjisir likbiir] final form
But since the status of cyclic rules is so unstable, and since Abu-Salim
1980 rejects the cyclic approach on other grounds, I will not adopt it here
either."
The above discussion then calls for a revision in the "Minimum"
principle:
(16) Assign a minimal syllable structure on stranded consonants
incorporating as many successive segments as possible.
A moment of reflection will show that this is a reasonable assumption.
The purpose of vowel eppenthesis is to make an unpronounceable string in
a language pronounceable. In that role, an epenthetic vowel assumes an
obscure and non-prominent profile as much as possible. For example,
epenthetic vowels are normally the least resonant high vowels, they do
not receive stress, etc. Minimizing the number of epenthetic vowels is
another way of achieving this effect, for it distorts the underlying form
as little as possible. Thus, given a succession of three stranded consonants,
there is no reason to proliferate epenthetic vowels by inserting one to the
side of each consonant. It is only natural that the sequence is resolved
into a pronounceable form, i.e., one of canonical syllable structures, using
the minimum number of epenthetic vowels. Thus, a succession of two stranded
consonants should be resolved into CVC, if the language allows such a
syllable structure, not necessarily into CVCV or VCVC.
64
The revised principle of syllable template assignment given in (16)
does not however explain why the forms given in (8b) and (10b) should
appear as they do, not as *[samni^ baladi] , *[?akli^ 9arabi] , etc., despite
the fact that the minimal syllable in Palestinian Arabic is CV, not VC.
Thus, there is nothing that can reject the ill-formed strings in which a
vowel is epenthesized to the right of the stranded consonant, for the re-
sulting syllable structure CVC.CV (followed by CV.CV.CV) perfectly conforms
to the canonical syllable shapes. What might explain this phenomenon is that
the language prefers, in disyllabic forms, the CV.CVC to CVC.CV sequence of
CV's.
I now turn to a brief discussion of a related phenomenon. Recall that
an impermissible consonant cluster can also be resolved by way of a conso-
nant elision, instead of by way of a vowel epenthesis. Korean provides a
good example of such a language. Like many languages, Korean does not
permit consonant clusters word-initially and finally, and three-consonant
clusters word-medially. Thus, a word-final cluster is reduced to a single
consonant, and a word-medial tri-consonant cluster is reduced to a cluster
of two consonants. For example,
(17) /kaps/ 'price', kaps-i 'the price is', but kap//, kap-to 'price also'
/naks/ 'soul', naks-i 'the soul is', but nak#, nak-to 'soul also'
/calm/ 'young', calm-ira 'youth', but cam-ta 'be young'
/anc/ 'sit', anc-ara 'sit downI',but an-ko 'sit and'
/ilk/ 'read', ilk-ara 'read!', but ik-ca 'let's read'
/yatalp/ 'eight', yatalp-iro 'with eight', but yatal#
/halth/ 'lick', halth-ara 'lick!, but hat-ko 'lick and'
/ilph/ 'recite', ilph-ara 'recite!', but 4p-ca 'let's recite'
One might ask if there are any intrinsic or extrinsic reasons as to why
Korean resorts to consonant ellipsis rather than to vowel epenthesis in
order to resolve impermissible consonant clusters. But there doesn't seem
to be any principled reason for it. That is to say, the choice of vowel
epenthesis or consonant ellipsis appears to be an entirely language-specific
matter. However, even this much generalization seems to be too strong, for
vowel epenthesis is also found in Korean, notably in nativizing loan-words,
i.e., in resolving impermissible consonant clusters in borrowed words, for
example.
(18)
golf
-
k'olpl^u
dance
->
t ' ansi
strike
-v
sit'^iraik^i (in baseball)
trio
->■
t^irio
print
■^
p'^urlnthi^
Christmas
->- k^irisimasi
65
Obviously, clusters are broken up with epenthetic vowels, rather than simpli-
fied, so that as much original pronunciation may be kept in the nativized
pronunciation. This is especially clear in the last example. Since Korean
allows word-medial consonant clusters and has the CVC syllabic structure, all
that is needed in Christmas is an epenthetic vowel to break up the initial
cluster to yield k^jrrismas as a loan-word pronunciation. But Korean does
not allow the obstruent-nasal cluster nor a word-final s^ on the surface.
The syllable-final or word-final £ becomes an unreleased stop _t, and ob-
struents preceding nasals are nasalized (cf. Kim 1976). Application of
these rules to k^Jrrismas would then render k" trinmat , making it all but un-
recognizable as standing for Christmas .
One might say that the Sonority Hierarchy given in (19) (Kiparsky 1979)
below determines in a principled way at least some types of elision. Hankamer
and Aissen 197A have in fact reported that the rule of vowel contraction in
Greek operated in terms of a sonority hierarchy in such a way that when two
unlike vowels come to abut together, one of them deletes, and the survivor
is the one which is higher on the ladder of the sonority hierarchy, which
they define as a, £, _u, e^, i^ for Greek vowels.
(19) Sonority Hierarchy (Kiparsky 1979: 432)
High Sonority Low Sonority
a, e, o, i, u, y, w, r, 1, nasals, fricatives, stops
With a minor variation, the universal vowel hierarchy proposed by Kiparsky
matches the Greek hierarchy. It is easy to be persuaded that it is natural
for the less sonorous vowels to be deleted, for they are phonetically weaker
in their acoustic output, and therefore less audible, everything being equal,
and therefore should disappear first.
But the Korean data given in (17) shows exactly the opposite phenomenon.
It is the less sonorous of the two consonants (the first two in the case of
a word-medial tri-consonantal cluster) that survives, and the more sonorous
deletes. (In Kim 1972, I called this the principle of close articulation
in Korean, whereby the segment with narrower aperture survives. Degrees of
aperture correspond closely to the sonority hierarchy, the maximum aperture
entailing the maximum sonority, and the minimum aperture, the minimum
sonority.)
What this implies is not only that the choice of vowel epenthesis and
consonant elision is a language-dependent matter but also that the choice
of segments to be deleted is a language-specific matter. It also implies
that weak and strong labels of metrical nodes has little role to play here,
because in Korean it is strong nodes that are deleted (if one labels more
sonorous segment as strong and less sonorous as weak)."
It is not always easy to define the teleology of a phonological rule,
although some rules, e.g., vowel epenthesis in Harari or loan-words in
Korean, are easy enough to be identified as having solely the purpose and
function of making a deviant phonological string conform to a canonical
form. By definition, every phonological rule deleting or adding a segment
(or vocalizing and devocalizing a segment) is a syllable modification rule.
66
But it is not clear that every one of such rules is only motivated to preserve
or restore canonical syllable structures, as syllable-dependent phonology
makes it appear to. In fact, many rules have the effect of violating and
destroying the existing perfectly well-formed canonical structures, sometimes
necessitating a subsequent rule to repair the damage. An example or two
should illustrate this point. The following example is from Klamath (Clements
and Keyser 1980) :
(20) C C V C C
A A A A /X
:v cvc ^-ccv cv cvc
nys swi so: nis 'a singer'
Here, elision of vowel 8^ from wa leaves w stranded, which is made to be re-
affiliated with the preceding s^ after undergoing vocalization to oj^. Another
example is again from Palestinian Arabic (Kenstowicz 1980)!
(21) /ti-hrik-i/ 'burn' (2P, f., sg.)
ti-hr_k-i syncope
ti-hi^rk-i epenthesis
Here, the root vowel 1_ is elided by the syncope rule mentioned earlier, and
then the resulting unacceptable cluster is resolved by epenthesizing vowel
i^ between the first two root consonants.
The beauty of a metrical treatment of the phenomenon of epenthesis,
as that of Harari by Halle and Vergnaud 1978, stems from the theory that
a linear string of segments has a non-linear hierarchical structure, and
that epenthesis or elision follows naturally and directly from the fact
that the segments are under the control of this constituent hierarchy to
such an extent that empty nodes must be filled and the excess nodes must
be pruned, so that anomalous syllable structures are brought into line
with optimal canonical syllable structures. If this is the case, then how
is it possible that in the examples given in (20) and (21) , a vowel elides
from a perfectly well-formed syllable structure, even creating an ill-formed
syllable structure in the process? The answer to this might be that vowel
elision here is not a rule dependent on or associated with syllable structure
but a rule independently motivated in the language. But this is an over-
statement, for the syncope is clearly dependent on syllable structure: it
applies only in open syllables. Rather, what this illustrates is, I think,
the holistic, interconnecting, sometimes conflicting, and cross-purposeful
nature of phonological processes emphasized by Stampe 1973, among others.
So it is natural for a "weak" vowel to be dropped from a "weak" position,
for example, in unstressed open syllables. But this can give rise to
violations of syllable structure canons. In Palestinian Arabic, such a
vowel is deleted and the resulting CCC is rectified by vowel epenthesis.
On the other hand, in Egyptian Arabic, the rule of syncope is blocked if
its application gives rise to CCC.
67
While one may speculate that in deriving such forms as given in (22)
via vowel elision,
(22) lagamor ->■ lagmor 'to finish' (Biblical Hebrew. Borer 1979)
?ip'eq'a ->■ ?ip'q'a 'put some things on face' (Klamath.
Clements and Keyser 1980)
elakilakapan ->■ elkilkapan 'he was big' (Passamaquoddy.
Stowell 1979)
the motivation is to reduce the number of syllables by resyllabifying an
existing string consisting of minimal syllables into one of maximal syllable
structures, i.e., CV.CV.CV(C) into CVC.CV(C), one may be able to find a more
principled reason for it. In this attempt, I turn to a portion of a recent
article by Halle and Vergnaud 1980.
In analyzing verb morphology of Arabic, they rearranged all the verb
stem shapes into the following table consisting of what they called
syllable skeleton which is the sequential arrangement of consonants and
vowels in a word. (Dot notations in the table are mine.)
(23) I II III
1. CV.CVC C.CV.CVC
2. CVC.CVC C.CVC.CVC CV.CVC. CVC
3. CW.CVC COW. CVC cv.cw.cvc
The following patterns are readily visible in this table:
(a) Rows are different from each other in the shape of the first
syllable, the last syllable always being CVC
(b) Column II is C+€olumn I, and Column III is CV+Column I, except that
(c) the expected skeleton for Row 1, Column III, i.e., CV. CV.CVC, does
not occur. In Arabic, this expected skeleton becomes identical with the
skeleton in Column I, Row 2, i.e., CVC.CVC, with elision of the middle vowel.
It is this last point (c) that I wish to dwell upon briefly. In order
to explain this curious phenomenon, Halle and Vergnaud 1980 postulate the
following rule (p. 94, their rule 8b)).
(24) Delete a penultimate nonbranching rime if preceded by a non-
branching rime.
A nonbranching rime means a single vowel. This rule says then that a single
vowel in the penultimate syllable should be deleted when it follows a syllable
whose nucleus is a single vowel.
68
This is a description, hardly an explanation, but it points to an
important fact, that is, elision of a vowel will apply if the resulting
skeleton conforms to an existing skeleton.
I believe that this is intimately related to an earlier discussion of
Arabic examples (8b and lOB) . We wondered then why in resolving an im-
permissible tri-consonantal cluster in the forms like samn baladi and ?akl
9arabi, a vowel was epenthesized to the left of the stranded middle conso-
nant, despite the fact that the minimal syllable structure CV in the language
required the minimal tree to be built on the stranded consonant to be
C rather than C, and suggested that it is probably because the language
preferred the sequence CVCVC to CVCCV in disyllabic words, that is, the
preferred syllable skeleton is CVCVC.
When one combines this observation with the observation made earlier
regarding a succession of two or more stranded consonants, it points to a
conclusion that the domain of the mapping procedures for epenthesis and
elision is not a single segment or a syllable but a sequence of CV's called
a syllable skeleton.
I have examined in this paper the mapping procedures employed in the
treatment of epenthesis and elision in the framework of metrical, syllable-
based phonology. The following conclusions can be drawn from this examina-
tion.
(1) Assignment of syllable templates proceeds from right to left (at
least in languages where, for the purpose of stress assignment, syllables
are counted from the end of a word) .
(2) Minimal styllable structures are built on stranded consonants
incorporating as many successive segments as possible.
(3) Choice of vowel epenthesis or consonant elision seems to be a
language-specific matter.
(4) The sonority hierarchy appears to play a minor role in deter-
mining the choice of elliptical segments.
(5) In some languages at least, the domain of vowel elision and
epenthesis is a sequence of CV's called a syllable skeleton.
NOTES
This is an expanded and revised version of the paper of the same
title presented at the 55th Annual Meeting of the Linguistic Society of
America, December 28-30, 1980, San Antonio, TX. I have benefited from
the comments made by George N. Clements and John J. MacCarthy at the
meeting, and from Michael Kenstowicz who read a draft version of this
paper and gave me valuable comments. I thank them all, but any faulty
arguments in the paper are mine.
69
2
This principle is my inference. References to this procedure in the
literature are rather vague, e.g.,
"When we attempt to syllabify the underlying representations..."
(Halle and Vergnaud 1978:3)
"Syllabify according to (5)" (Borer 1979:152), where (5) refers to
a simple list of syllable structures allowed in the language.
The only explicit statement is found in a parallel situation. Hayes
(1980), for example, proposes a "maximal foot construction principle" by
which maximal trees are to be built first when building foot trees on a
string of syllables. "Foot construction rules construct the largest foot
compatible with their conditions." (181). Cf. also: "A foot must univer-
sally be mapped onto as much of a word as possible. . . all feet must be
maximal." (McCarthy 1980:80)
■^Cf. "To obtain the correct output strings we postulate that minimal
syllable structure is assigned to the segments circled" [=stranded] . (Halle
and Vergnaud 1978:3).
"Assign minimal syllabic structure to every "stranded" consonant."
(Borer 1979:152)
In Berber, the following syllable structures are possible: V, VC,
VCC, CV, CVC, CVCC. The final form of (5b) is derived by an additional
rule of initial schwa deletion. The two different forms are attributed to
Saib 1976 (5a) and Guerssel 1977 (5b).
The two types of epenthesis may exist in the same language. Imper-
missible initial consonant clusters in loan-words in Persian are resolved
in both ways, e.g., esmit 'Smith', kelas 'class'. Borer 1978:170 also
notes that in Tiberian (Biblical) Hebrew, the initial consonant cluster is
broken up by inserting a vowel to the right of the stranded consonant,
while the final cluster is broken up by inserting a vowel to the left of
the stranded consonant. Amharic is said to exhibit a similar phenomenon
in that a schwa is inserted to the right of the stranded consonant in the
/\ /\ /\ /\ /\
environment of V C C C V, but to the left in the environment of V C C.
+ +
Borer 1979 persuasively argued that cyclic application of syncope
and epenthesis rules are necessary in Biblical Hebrew. Equally persuasively,
J. McCarthy argued at the San Antonio meeting that the case of Biblical
Hebrew could be handled without resorting to a cyclic mechanism.
Note the following statements in this regard:
"We make the natural assumption that two adjacent stranded consonants
form a CVC syllable." (Kenstowicz and Abdul-Karim 1980:72)
"As for the forms with C ^C, the answer is straightforward: no other
analysis is available, which also minimizes the occurrence of dummy elements
in the syllabification." (Selkirk 1980:14).
70
A similar point has been made by Abu-Salim (1980) who states:
"This means that epenthesis has to refer to the whole structure not
only to the stranded consonant, to determine the position of the epen-
thetic vowel." (6)
8
One case in which the node labeling was used in the operation of well-
formedness conventions on syllable structure is compensatory lengthening.
Ingria 1980, in particular, proposed what is called "Empty Node Convention"
(471) by which empty weak nodes which are part of a syllaic coda are to be
associated with the immediately preceding syllabic nucleus, while other
empty nodes are to be pruned. This convention, however, is for the empty
nodes created by deletion of segments by phonological rules, not for those
created during the process of building syllable trees, and is not relevant
to what we are discussing here. Furthermore, Clements 1980 demonstrated
the inadequacy of the convention even in handling compensatory lengthening.
REFERENCES
ABU-SALIM, I. M. 1980. Epenthesis and geminate consonants in Palestinian
Arabic. Studies in the Linguistic Sciences 10:2.1-12.
AOUN, Y. 1979. Is the syllable or the supersyllable a constituent? MIT
Working Papers in Linguistics, vol. 1, pp. 140-148.
BORER, H. 1979. Schwa's in Tiberian Hebrew. MIT Working Papers in
Linguistics, vol. 1, pp. 149-171.
BROSELOW, E. 1980. Syllabic structure in two Arabic dialects. Studies
in the Linguistic Sciences 10:2.13-24.
CLEMENTS, G. N. 1980. Compensatory lengthening; An independent mechanism
of phonological change. 55th Annual Meeting of LSA.
and S. J. Keyser. 1980. The hierarchical nature of the Klamath
syllable, ms.
GUERSSEL, M. 1977. Constraints on phonological rules. Linguistic Analysis
3.287-305.
HALLE, M. and J. R. Vergnaud. 1978. Metrical structures in phonology, ms.
and . 1980. Three dimensional phonology. Journal
of Linguistic Research 1.83-106.
HANKAMER, W. J. and J. Aissen. 1974. The sonority hierarchy. Papers from
the CLS Parasession on Natural Phonology, pp. 131-145.
HAYES, B. 1980. Aklan stress: Disjunctive ordering or metrical feet?
Cahiers Linguistiques D' Ottawa 9 (=NELS X), pp. 179-89.
INGRIA, R. 1979. Compensatory lengthening as a metrical phenomenon.
Linguistic Inquiry 11.465-495.
KENSTOWICZ, M. 1977. Functional explanations in phonology, ms.
. 1980. Vowel harmony in Palestinian Arabic: A suprasegmental
analysis, ms. (to appear in Linguistics)
and K. Abdul-Karim. 1980. Cyclic stress in Levantine Arabic.
Studies in the Linguistic Sciences 10:2.55-76.
and C. W. Kisseberth. 1979. Generative phonology. Academic
Press.
KIM, C-W. 1972. Two phonological notes: A-sharp and B-flat. In M. Brame
ed. Contributions to Generative Phonology, pp. 155-170. University
of Texas Press.
. 1976. Rule ordering in Korean phonology. Linguistic Journal
of Korea 1.60-83.
71
KIPARSKY, P. 1979. Metrical structure assignment is cyclic. Linguistic
Inquiry 10.421-441.
McCarthy, J. J. 1980. a note on the accentuation of Damscene Arabic.
Studies in the Linguistic Sciences 10:2.77-98.
SAIB, J. 1977. A phonological study of Tamazight Berber. Unpublished
doctoral dissertation, UCLA.
SELKIRK, E. 1980. Epenthesis and degenerate syllables in Cainene Arabic, ms.
STAMPE, D. 1973. A dissertation on natural phonology. Doctoral dissertation,
University of Chicago.
STOWELL, T. 1979. Stress systems of the world, unite! MIT Working Papers
in Linguistics, vol. 1, pp. 51-76.
Studies In the Linguistic Sciences
Volume 11, Number 1, Spring 1981
DISPLACED TONES IN DIGO (PART 2)*
Charles W. Kisseberth
In this paper the analysis of Digo tonology developed
in Kisseberth and Wood (1980) is extended to include some
additional verb tenses, the most common types of nouns, and
phrases consisting of a verb plus a following noun. The
analysis presented in Kisseberth and Wood (1980) is modified
in several places, and evidence is developed in support of
the hypothesis that some of the tone rules involved apply
cyclically. The data from verb-noun phrases establish very
clearly that, despite the considerable opacity of the Digo
tone system, the rules are fully productive and the proposed
set of underlying contrasts is highly motivated.
In Kisseberth and Wood (1980), henceforth K§W, a preliminary analysis
was given of the tonal pattern of three verb tenses in Digo (E.73 in
Guthrie's 1971 classification of Bantu languages). This analysis was
based entirely on the tonal shapes of these verbal forms as they are
pronounced in isolation (i.e. when no word precedes or follows them).
The present paper considerably expands the data base for the analysis
of Digo tonology; I include some additional verbal tenses, I analyze
the most common types of nouns, and I consider phrases consisting of a
verb plus a single noun.
Although my inclination at the present time is to provide an essentially
accentual rather than a purely tonal analysis for Digo (cf. Goldsmith 1981),
the present paper will follow the tonal treatment of Digo given in K6W,
modifying this analysis in some places and expanding it in others. I have
not sought to recast the analysis in an accentual framework since the
essential points I wish to develop here can be brought out adequately
within the general framework of K§W. The interested reader is referred to
my forthcoming paper, "An Accentual Account of Digo Tonology", for an
exploration of an accentual treatment of Digo tone patterns.
The additional data examined here reveal the essential correctness of
many of the principles established in K5W, at the same time demonstrating
the need for modifications at a number of points. Perhaps the most signifi-
cant result deriving from the exploration of these new data is the conclusion
that many of the principles uncovered in K$W are phrasal in scope rather
than being applicable only at the word level. Furthermore, I will show
that there is an elegant argument that (at least some of) the principles
are cyclic in nature, though the details of their cyclic application must
await the collection of much additional data.
74
The paper is divided into five parts. Section 1 summarizes KfjW and
also proposes some revisions of that analysis; section 2 examines the tonal
properties of the most common noun types in Digo; section 3 considers some
verb tenses not treated in K5W; section 4 focuses on the tonology of phrases
consisting of a verb plus a following noun; and section 5 presents my
conclusions.
1. A review of K^W, with some suggested revisions.
K^W assume an essentially autosegmental approach to Digo tonology
whereby tonal specifications occur on a "tier" separate from the segmental
tier, the two tiers being connected by means of "association lines" (cf.
Goldsmith 1976). One major departure from most autosegmental analyses is
made in KSW — namely, only high tones are specified in underlying representa-
tions, low tones being represented as the absence of a high tone. 1 think
that the fact that this particular assumption is a useful one to make will
eventually count as support for the accentual treatment of Digo tonology;
in any case, I shall continue to follow this assumption about low tones
(at the underlying level) in the present paper.
Verb stems and grammatical morphemes in Digo are either low-toned or
bear a single high tone. K5W assume that high tones are not associated
(underlyingly) with any particular vowel in the morpheme. The underlying
representation of a high-toned verb like ku-sindik-^ would thus be
H
/ku-sindik-a/
with no association line connecting the tonal tier and the segmental tier.
It is usually claimed for other Bantu languages that in verb stems it is
the first vowel of the stem that is either high-toned or low-toned, the
other vowels of the stem being assigned their surface tone by rule. I
will provide some motivation in this paper for assuming that in the under-
lying structure of Digo high tones should be associated with a particular
vowel (in contrast to the K5W approach) , and that in the case of verb stems
it is the first vowel of the stem that is specified as high-toned (in
accordance with the usual treatment of Bantu languages) . Under this re-
analysis, the underlying representation of ku-sindik-a would be
H
/ku-sindik-a/
with an association line connecting the H of the tonal tier with the first
stem vowel .
It is a pervasive feature of Digo, however, that high tones do not
remain associated with the vowel that they are associated with in under-
lying structure. The most essential rule in the language that effects the
reassociation of high tones is referred to in K§W as High Tone Displacement
(HTD) . In K§W this rule is formulated so as to associate the rightmost high
tone in a word with the penult vowel of the word. K5W refer to this vowel
as the "starred" vowel, since it was assumed that a general rule of the
language assigns a "star" (cf. Goldsmith 1976 for some discussion of this
device) to the penultimate vowel of a word. I prefer to dispense with the
use of stars and instead to refer directly to the position of the vowel in
the word. However, I would like to suggest that perhaps a better analysis
of High Tone Displacement is to say that it reassociates the rightmost high
tone of the word with the ultimate vowel rather than the penultimate.
75
There are various reasons why it is not iimnediately obvious which
vowel the rightmost high tone should be said to have displaced to. One
factor that can be mentioned here is the fact that there are two distinct
phonetic realizations of the high tone in question. If there is a voiced
obstruent, or a cluster containing a voiced obstruent, before the ultimate
vowel, then a level high tone appears on the penult vowel — e.g. ku-korog-a
'to stir', ku-tanyirlz-a 'to drive off predators'. If any other consonant
or consonant cluster, or no consonant at all, precedes the ultimate vowel,
then the penult is pronounced with a rising pitch and the ultimate vowel
with a falling pitch — e.g. ku-sindik-a 'to close the door', ku-garag^r-a
'to toss around in pain'. It should be emphasized that I am speaking here
of the pronunciation of these items when they occur at the end of an
utterance (and when not preceded by a word that would affect their tonal
structure — cf. section 4 for a demonstration that words earlier in a phrase
can affect the tonal structure of words later in the phrase) .
If we assume that HTD reassociates the rightmost high tone in a word
with the ultimate vowel, then the pronunciation of items such as ku-sindik-a
can be accounted for in terms of a rule (call it Rise-Fall) that converts
f I?
V c ^y $
(where V means that the vowel is not associated with any specification on
the tonal tier and $=utterance-f inal position) into the structure:
i^^^
Such a rule has analogues in other Bantu languages (cf. Becker and Massamba
(1980)) and thus appears to be well-motivated on cross-linguistic grounds.
But what about items such as ku-korog-a? If we start with the assumption
that the high tone on the first stem vowel reassociates with the ultimate
vowel, then these cases seem to require a rule retracting the high tone back
to the penult vowel. But I shall attempt to demonstrate in this paper that
what is actually involved is a process of High Doubling (HD) . This rule
says that a high tone associated with the ultimate vowel of a word causes
a high tone to be assigned to the penult vowel as well, provided that a
voiced obstruent occurs between the two vowels; in other words, the sequence
V Cx V # (where C =a voiced obstruent or a
cluster containing a voiced obstruent)
is converted to
where there are two high tones in the representation rather than just one.
(I will discuss at a later point why I have opted for introducing a second
H on the tonal tier rather than simply adding an association line between
the H on the tonal tier and the penult vowel.) Given such a rule of High
Doubling, then to derive the surface form ku-korog-a, we require a rule of
Utterance-Final Lowering (UFL) . This rule will change a H that is associated
with an utterance-final vowel into a low tone (which I represent now with the
76
symbol L; see below for discussion of this point). UFL will apply, however,
just in case the H in question is not followed by any other tone. If we
then order Rise-Fall before UFL, we can account for why UFL does not have
any effect in the case of items like ku-sindik-a. Since Rise-Fall introduces
a L on the tonal tier after the H that is associated with the ultimate vowel,
UFL will be prevented from applying to any form that has undergone Rise-Fall.
It should be pointed out here that if high tones are associated with a
particular vowel in underlying representations, then the application of
HTD must not only associate the rightmost H with the ultimate vowel but
also disassociate that H from the vowel with which it was underlyingly
associated. The same will be true for the other tonal displacement rules
described below.
V A
The proposed derivations for ku-sindik-a and ku-korog-a are given in (1) .
(1) 1^
H
$ku-sindik-a$ $ku-k6rog-a$
ku-sindik-A ku-korog-d HTD
H H
inapplicable ku-korog-a HD
ku-sindilc-a inapplicable Rise-Fall
H
inapplicable ku-korog-a UFL
A few remarks about the relative ordering of these rules are called for.
Obviously, HTD applies first, since it associates a high tone with the
ultimate vowel and thus establishes the situation that triggers the other
rules. I assume that HD applies before the rule of Rise-Fall. This ordering
is based on the assumption that Rise-Fall applies to the sequence
V Co V $
and consequently must be blocked from applying in the case of a representation
like
$ku-korog-a$
which satisfies the structural description of Rise-Fall but fails to undergo
it. By applying HD first, we obtain the representation
^ H
$ku-korog-i$
which does not meet the structural description of Rise-Fall.
The motivation for applying Rise-Fall before UFL was given earlier.
Since HD must precede Rise-Fall and Rise-Fall must precede UFL, it follows
(given the requirement that ordering relationships must be transitive) that
HD must also precede UFL. This ordering is independently motivated on the
grounds that if UFL were to precede HD, it would change the H in the
representation H
$ku-korog-a$
to L and consequently (incorrectly) prevent the application of HD since there
would no longer be a H on the ultimate vowel that would trigger doubling onto
the penult vowel.
77
Recall that in underlying representations low tones have been represented
as the absence of a high tone. Both Rise-Fall and Utterance-Final Lowering
have the function of introducing a low tone (L) into the tonal tier. Ultimately,
all of the vowels that do not have a tone associated with them will be assigned
the tonal value L.
There is another rule in Digo besides HTD that associates a H with a vowel
other than the vowel that the H is underlyingly associated with. K5W refer to
this rule as Displacement-to-Stem (D-S) . In K§W this rule is phrased to say
that a H which is followed by another H in the word will be displaced as far
to the right as it can be without passing across a voiced obstruent or a stem
vowel. In effect, this rule moves a H to the first stem vowel, unless the H
is prevented from getting that far by the presence of a voiced obstruent.
K5W show, for example, that a high tone originating on the third singular
subject prefix a- in the present tense has the property of displacing to the
stem when there is a high tone later in the word. For instance, underlyingly
a-na-ezek-a 'he is thatching' has a H associated with the prefix a^- and a H
associated with the first stem vowel. HTD displaces the stem H to the
ultimate vowel (thus triggering the Rise-Fall rule) and D-S associates the
prefix H with the first vowel of the stem -ezek-. Underlyingly a-na-dunduriz-a
'he is putting s.t. aside' has a H associated with the prefix and a H
associated with the first stem vowel. HTD displaces the stem H to the
ultimate vowel (eventually triggering HD and UFL) and D-S displaces the
prefix K to the vowel that immediately precedes the stem, since the presence
of a voiced obstruent in stem-initial position prevents the prefix H from
displacing onto the first stem vowel.
Non-final high tones in Digo trigger another process — namely, they
associate not just with the vovjel to which they are displaced, but also
with all subsequent vowels in the word up to (but not including) a vowel
that is preceded by a voiced obstruent. I will call this process High
Tone Spreading (HTS). If we assume that HTS applies iteratively, it can be
formulated very simply to convert the structure
?(C )V
into the structure
hcy^
(where Cy=a consonant that is not a voiced obstruent or a cluster that does
not contain a voiced obstruent). Notice that, given this formulation, HTS
can only be triggered by a high tone associated with a non-final vowel
since high tones associated with an ultimate vowel have no subsequent vowel
to spread onto. Since a high tone can appear on a non-final vowel just in
case there is a subsequent high tone in the word, it follows that HTS
will always be a phenomenon manifested by high tones that are followed by
another high tone in the word. Examples of the application of HTS include:
a-na-sindik-S 'he is shutting the door', a-na-tany£rfz-a 'he is driving off
predators', a-na-tamanfr-^ 'he is putting a spell on s.o.', a-na-pepesuk-a
'he is staggering'. (It should be pointed out here that in K5W the falling
78
quality of the final vowel of items like a-na-sindik-a as opposed to
a-na-tanyiriz-a , where the final vowel is clearly low-pitched, was
erroneously regarded as an insignificant phonetic detail and consequently
omitted from the transcriptions. Within the analysis developed in K^W,
this falling quality of the last vowel did not follow in any way from the
analysis; in the present analysis, however, the phonetic details follow
straightforwardly . )
I can now illustrate the derivation of examples such as a-na-sindik-a
and a-na-tanyiriz-a .
(2) H H H H
$A-na-slndik-a$ $a-na-tAnyiriz-a$
H H H H
A-na-sindik-a A-na-tanyiriz-a HTD
H H H H
a-na-slndik-a a-na-tanyiriz-a D-S
H IJ H
inapplicable a-na-tanyiriz-a HD
a-na-sinSk^ a-na-tinyir4z-4 HTS
inapplicable inapplicable Rise-Fall
a-na-sindik^ a-na-tanylrlz-a UFL
A few remarks about these derivations are in order. First, notice
that at an intermediate stage in the derivation (namely, after the appli-
cation of HTS) , two different high tones on the tonal tier are associated
with the same vowel on the segmental tier. In the case of a-na-sindlk-a ,
the second of these high tones must later in the derivation be changed to
low as a consequence of UFL. It has sometimes been suggested that there
should be a universal constraint on representations that would automatically
change the configuration
T. T.
V
(where T=any tonal autosegment) to one where just a single tone exists on
the tonal tier. The evidence from (2) suggests that such an automatic
reduction of identical tones must not take place if one of the tones is
also associated with some other vowel on the segmental tier.
Second, some remarks about the relative ordering of the rules are
required. D-S must apply before HTS. If the reverse order obtained, a
structure such as the following (where HTD has already been applied)
H H
$a-na-sindik-a$
would be converted to
$ a - na -slnSTk^ $
by HTS. If D-S is then applied to this structure, the result would presumably
be that the association lines between the first H and the segmental tier would
be erased and a new association line drawn between the H and the first stem
vowel. But this would yield the wrong result, since there would be then
no association lines between the H and the following vowels — and such
association lines are necessary to account for the surface form of
a-na-slndik-a. If, however, D-S precedes HTS, the first H will undergo
79
displacement to the stem vowel and then will spread onto the following
vowels. The ordering of HTS and HD is immaterial. HTS must precede Rise-
Fall since its application is responsible for the failure of Rise-Fall to
apply in the case of items like a-na-sindik-a. Recall that Rise-Fall
applies just in case the penult vowel is not associated with the tonal
tier; HTS has the effect of associating the penult vowel with a H on the tonal
tier in cases like a-na-sfndik-'a and thus blocks the application of Rise-Fall.
One of the most interesting phenomena in Digo is what K5W refer to
as the "neutralization" of certain high tones. A considerable amount of
new data bearing on this phenomenon is incorporated in the present paper,
thus it is important to review this process in some detail here. In
Digo, high-toned verb stems of the shape -CqVCq- regularly fail to show
any reflex of this high tone in surface structure when the high tone is
the only high tone in the word (and there is no earlier word in the phrase
that might affect the tonal realization of the verb in question — see
section 4 for evidence that earlier words in a phrase can affect the tonal
shape of later words). For instance, the verb ku-som-a 'to read' is
pronounced exactly like ku-gur-a 'to buy'. That the former word is
fundamentally different from the latter word is shown by a large number
of facts, all of which point to ku-som-a having a high-toned stem under-
lyingly whereas ku-gur-a has a low-toned stem. For example, suf fixation
of a tonally-neutral suffix like -ir-/-er- (which conveys meanings such
as 'to' or 'for' or 'with') results in ku-som^r-a versus ku-gurir-a.
The fact that ku-som-a loses all trace of its high tone but ku-somer-a
does not suggests that the position of the high tone in the word is a
relevant factor. Indeed, I will attempt to accumulate some evidence that
only a penult high tone can neutralize. Other evidence for claiming
that a verb such as ku-som-a is high-toned under lyingly, even though it
does not appear to be, comes from (a) cases where another high appears
earlier in the word and (b) cases where another word follows. Below
I will review the evidence from (a). In section 4 the (b) cases will
be dealt with in detail.
When the root -som- appears in a word where there is a high tone
associated with a prefix, a unique tonal shape arises, one that is not
found in any other circumstance in Digo. For example, in the 3 sg.
present tense form (where there is a high tone associated with the prefix
a-) we get the surface form a-na-sofa-a, with a high-falling tone on the
penult vowel. It is extremely crucial to note, however, that this high-
falling tone on the penult arises just in case the consonant before the
penult is not a voiced obstruent. If it is a voiced obstruent, the penult
vowel is pronounced just as though it were neutralized (i.e. like a low
tone) and the vowel before the voiced obstruent is pronounced as a level
high tone; e.g. a-na-vug-a 'he is cooking' (cf. ku-vug-a 'to cook', but
ku-vug3fr-^ 'to cook for'). The high tone on the vowel preceding the
verb stem in a-na-vug-a appears to be a high tone that originates on
the a- prefix; the H that I have suggested is associated with the stem
-vug- is not directly realized on the surface. However, by inference
we can see that -vug- must be high-toned. If it were low-toned, then
the H associated with the prefix should have been displaced to the ultimate
vowel by virtue of HTD; cf. the case of the low-toned verb -gur-: a-na-gur-a.
Similarly, there is every reason to believe that in the case of a-na-s^-a,
the high part of the high-fall on the penult must originate on the prefix
a- and that the low part of the high-fall must be somehow a reflex of the
neutralization of the stem high tone. Again, -som- must be high-toned
80
underlyingly, otherwise we would expect the high tone of the prefix to
displace to the ultimate vowel.
I have now reviewed some of the evidence that supports the claim that
certain verbs (namely, those of the shape -CqVCq-) which are underlyingly
high-toned undergo a process that (in some instances) neutralizes them
with low-toned verbs. I now proceed to a consideration of the form that
this neutralization process must be assumed to possess.
Suppose that we assume that what the rule of Neutralization (Neut) does
is to convert a H associated with some vowel (to be defined below) to L;
it is, of course, important that this rule change the H to L, rather than
simply deleting the H, since we would like to be able to see the low part
of the high-fall in a-na-som-a as being a reflex of the application of
Neutralization. It is not clear how we could obtain that result if
Neutralization simply deleted the H that is associated with the verb stem
underlyingly. Thus Neutralization will be the third rule that we have
posited where the output includes introducing a L into the tonal tier.
Turning to the question of which vowel a H must be associated with
in order for it to undergo Neutralization, we can first observe that
only H's that are underlyingly associated with the verb stem are subject
to this rule. Thus we never find a H associated with a subject prefix or
a tense-marker or an object prefix that undergoes Neutralization.
Furthermore, the H's that neutralize are always associated (underlyingly)
with the penult vowel. This remark is based, of course, on the assumption
that H's are associated in underlying structure with a particular vowel
and that in the case of the verb stem that vowel is the first one in the
stem.^
Given the assumption that Neutralization affects a penult high tone
that derives from the stem, I will suggest an analysis of Neutralization
here that differs significantly from that proposed in K§W. The motivation
for this reanalysis derives largely from new data presented later in the
paper, though the viability of the proposals also depends on other
revisions made in the present paper (specifically, the revision of HTD
so that it displaces a high tone to the ultimate rather than the penult
vowel ) •
I propose that HTD is restricted so that it does not apply to a high
tone (originating in the stem) that is associated with a penult vowel. As
a result of this restriction, a structure like
H
$ku-som-a$
will fail to undergo HTD. At some later point in the derivation. Neutraliza-
tion will apply to change the H to L. (There are a number of high-toned verbs
of the structure -CgVCQ- that fail to neutralize — cf. ku-n^n-S 'to speak',
ku-r^h-a 'to bring', ku-ror-S 'to marry'. In the present analysis, these
verbs will be regarded as being exceptional in that they are exempted from
the restriction on HTD. Since they are permitted to undergo HTD, their
high tone will displace to the ultimate vowel and thus escape falling under
the scope of Neutralization.) In the case of examples like a-na-som-a and
a-na-vug-a, the underlying forms will be
81
H H H H
$a-na-som-a$ $a-na-vug-a$
and again HTD will be blocked due to the restriction that bars it from
affecting a H on the penult (that originates in the stem). Ultimately,
Neutralization will also apply, changing the second H to L. But in addition
it is necessary that the first H displace to the penult vowel in the former
case and to the antepenult in the latter case. K§W tacitly assume that the
displacement in these instances is to be accomplished by the rule of Displace-
ment-to-Stera. This assumption is a natural one, given the data so far examined,
since in a-na-som-a and a-na-vug-a the neutralized vowel is in fact the
first stem vowel and the prefix H does attempt to displace to that vowel
(succeeding in the former case, but being blocked from getting to the stem
vowel in the latter case by the voiced obstruent in stem-initial position).
Despite the similarity between the displacement in the present examples
and the rule of D-S, I will present evidence showing that a third displacement
rule is actually at work. This third rule is distinct from both HTD and D-S,
but it shares with D-S the restriction that the H cannot be associated with
the designated vowel (i.e. the vowel to which it is trying to displace) if
that vowel is preceded by a voiced obstruent. In such cases, the H must
be associated with the vowel preceding the designated vowel. 5 The details
of the formulation of this third displacement rule (call it Displacement-to-
Neutralized Vowel (D-NV)) will be discussed at the point where the relevant
data arise. The ordering of this rule will also be dealt with at that point.
There is another rule in Digo that has the effect of eliminating an
underlying H. The only situation where this rule has so far been found to
be operative is when a high-toned verb stem follows a high-toned object
prefix.^ Only one of these H's is manifested on the phonetic surface. For
example, the surface manifestation of the underlying structure
$ku-a-paput-a$
is ku-a-puput-a 'to hit them', which is tonally indistinguishable from a verbal
form that has a single H in its underlying structure — cf. ku-puput-a 'to hit'.
This suggests that one of the H's from the underlying representation of
ku-a-pup^t-a should be eliminated.
It is not immediately apparent, of course, which H is lost; nor is it
immediately apparent whether the rule in question should change H to L
or simply delete the H. Fortunately, one piece of crucial evidence exists
bearing directly on these questions. Consider the high-toned verb ku-on-a
'to see' (cf. ku-onan-a 'to see one another'). This is a verb that is subject
to Neutralization. Now, if a high-toned object prefix is inserted, we get
ku-a-^n-a 'to see them'. Notice that in this form a H undergoes High Tone
Displacement, associating with the ultimate vowel and thus triggering Rise-
Fall. But we know that the stem H could not be the one that undergoes HTD,
82
since the H associated with the stem -on-a is exempted from undergoing HTD.
On the other hand, the prefix H could undergo HTD, but only if it were the
rightmost H in the word. Thus we see that the stem H must be lost in order
for the prefix H to undergo HTD,
of the stem is simply deleted,
get the intermediate structure
It would be simplest to assume that the H
If it were changed to L instead, we would
cu-a-on-
$ku-a-6n-a$
and if we attempted to associate the first H with the ultimate vowel we
would cross an association line (a situation that is prohibited in auto-
segmental phonology) . I therefore assume a rule of High Erasure (HE) ,
which erases the H specification on the first vowel of a verb stem when a
H is associated with a preceding object prefix.' Given this rule, the
derivation of ku-a-on-a will proceed as follows.
(3)
$ku-a-on-
ku-a-on-
HE
ku-a-on-a
inapplicable
inapplicable
inapplicable
inapplicable
ku-a-on-a
inapplicable
HTD
Neut
D-S
HD
HTS
Rise-Fall
UFL
In this derivation, HE is ordered before
H renders the prefix H eligible for HTD.
HTD since the erasure of the stem
Below I list the rules that I have suggested and indicate those pairs
of rules for which a definite ordering relationship has been established.
(A)
(
High Erasure (HE)
High Tone Displacement (HTD)
Neutralization (Neut)
Displacement-to-Stem (D-S)
High Doubling (HD)
igh Tone Spreading (HTS)
ise-Fall
tterance-Final Lowering (UFL)
I do not believe that there is any direct evidence that HTD must
precede D-S, though naturally this issue depends upon the details of
the formulation of D-S. I have not included the suggested rule of
Displacement-to-Neutralized Vowel in the list of rules since I have
not yet dealt in any way with its formulation or its ordering.
83
2. Noun tonology.
Since some of the most significant evidence concerning Digo tonology
derives from an examination of phrases consisting of a verb plus a noun,
it is necessary for me to examine and to analyze the tonal behavior exhibited
by nouns outside the context of such phrases so that we can properly
appreciate the results of combining these nouns with verbs into verb
phrases. The study of noun tonology, however, is itself a matter that
requires extended treatment and cannot be dealt with fully here. In any
case, much research is still required for a complete understanding of the
full range of noun tonal patterns; all that I intend to do here is to motivate
the existence of three tonally distinct groups of nouns which represent the
simplest cases (from a tonal perspective) and also the most frequent types
in the language.
The simplest way to begin to get at the tonal structure of nouns in
Digo is to compare the simple noun with its locative form. Locatives in
Digo are derived by suffixing -ni to the simple form of the noun. In (1)
below I illustrate a very large class of nouns which both in their simple
form and in their locative form are pronounced with the intonation typical
of verbal words that lack any high tone."
(5)
n-jira
path(s)
n-jira-ni
mu-hama
millet
mu-hama-ni
tsongo
bird(s)
tsongo-ni
mu-hambo
trap
mu-hambo-ni
ma-ruwa
flowers
ma-ruwa-ni
gari
car
gari-ni
tunda
fruit
tunda-ni
banda
shed
banda -ni
goma
drum
goma-ni
chi-ronda
sore
chi-ronda-ni
Nouns such as those in (5) are, I claim, best regarded as being ones which
do not have a H associated with any of their vowels. I will refer to these
simply as low-toned nouns.
Another very large class of nouns in Digo is illustrated in (6). In
their simple form, these nouns are pronounced with the intonation characteristic
of verb words that lack any high tones. However, in their locative form,
these nouns are pronounced in a fashion entirely analagous to verbal words
that have a single high tone associated with them.
(6)
biga
renje
fumo
tsui
vi-yogwe
mu-kono
ny-umba
ny-ungu
gunguhi
ip-fumbati
biga-ni.
renje-ni
fumo-nl
water pot
pumpkin
spear
leopard (s) tsui-n'i"
sweet potatoes vi-yogwe-ni
hand , arm
house(s)
pot(s)
bed leg
side-bar of
bed
mu-kono -ni
ny-umba-ni
V A
ny-ungu-ni
gunguhi-ni
m-
•fumbati-ni
84
Nouns such as those in (2) are, I claim, best regarded as being nouns which
have a H associated with one of the vowels of the stem, this high tone being
lost from the simple form but maintained in the locative form. This sort of
analysis has, of course, a clear precedent in the verbal system. Recall
that the rule of Neutralization has the effect of changing a H (originating
on a stem) into L. This rule accounts for the fact that a high-toned stem
such as in ku-on-a 'to see' (cf. ku-onan-a 'to see one another') is pronounced
like a low-toned verb.
If it is assumed that the noun stems in (2) have a H in their underlying
representations, the question arises as to whether this H is associated with
any particular vowel of the noun. In section 1 it was suggested that
perhaps Neutralization should be formulated so as to affect penult vowels.
If we follow this suggestion, then the H tone on the nouns in (2) would
be associated with their penult vowel. In the case of noun stems of the
shape -CqVCqV, the penult vowel is the first stem vowel. But in longer
nouns, the penult vowel is not necessarily the first stem vowel. For example,
in gunguhi, the penult vowel is the second stem vowel. Thus if we are
going to claim that it is the rule of Neutralization that affects the H
of the nouns in (2), and if Neutralization affects only penult vowels,
then it follows that in the case of nouns there is no restriction that a
H tone may appear only on the first stem vowel. (Recall that in the case of
verbs, I suggested that only the first stem vowel could have a H associated
with it.) This result would be compatible with what we know of other Bantu
languages — in general, a high tone is not limited to the first stem vowel in nouns.
Because of this disparity between verbs and nouns, there are no verbs of
the shape -CgVCgVCo-a that undergo Neutralization whereas there are noun
stems of this shape that do. Notice that if we did not assume that high
tones were associated with particular vowels, it would be difficult to
explain why a noun such as gunguhi undergoes Neutralization whereas verbs
of the shape -CgVCgVCo-a are not susceptible to Neutralization. Thus one
motivation for assuming that H's are associated with a particular vowel
comes from a desire to account in a systematic fashion for this contrast
in behavior.
Let us at this point turn our attention to a third type of noun,
which is illustrated in (7).
(7)
kund^
kXnd^
beans
food
kunde-ni
kande-ni
^ A
chi-kopwe-ni
chi-kopwe
sweet potato
pweza
leaves
octopus
- V A
pweza-ni
zodo-ni
zodo
cooked mangos
bata
duck
bata-ni
bunda-ni
bunda
bunch, cluster
sahani
kofiya
bada
plate(s)
hat(s)
cassava meal
sahani-ni
kofiya-nx
bada-n^
dona
hard porridge
dona-ni ^
chi-tabu
book
chi-tabu-ni
85
The simple forms of the nouns in (7) indicate quite clearly that these
nouns have a H associated with them in underlying structure. This H under-
goes HTD, thus triggering either Rise-Fall or HD followed by UFL. The fact
that these H's do undergo HTD means either that they are associated with a non-
penult vowel that is therefore subject to HTD or that they are associated with a
penult stem vowel, but are marked as exceptions to the constraint that these
vowels do not undergo HTD. For my present purposes it is not necessary to
determine which explanation holds in any given case, since all that is
crucial is that these are instances of nouns that do undergo the displacement
of their high tone to the ultimate vowel. The locative forms in (7) are
quite problematic and I will not attempt to develop an analysis of them
here. They do demonstrate, however, that the nouns in (7), which have a
H subject to HTD, behave differently from the nouns in (6), which have a
H that escapes HTD (and subsequently undergoes Neut).
Suffixation of the locative element -ni provides clear evidence for
three subtypes of nouns — low-toned nouns like those in (5) , high-toned nouns
that fail to undergo HTD and are subject to Neut as in (6), and high-toned
nouns that undergo HTD as in (7). Much additional support for this three-
way distinction will be adduced in section 4, where I examine verb plus
noun phrases; there is, however, additional support from other sorts of
phrases. Here I will mention just phrases consisting of a noun plus the
first person singular possessive word (which consists of the stem -angu
preceded by a prefix that agrees with the noun class of the noun that
-angu modifies) . Consider first the case of nouns that I am regarding as
low-toned.
(8) banda r-angu my shed
chi-ronda ch-aneu my sore
V A
ny-ama y-angu my meat
mu-hama w-^ngu my millet
podzo z-angu my sp. green beans
ma-zu g-^ngu my bananas
On the basis of these data, I would assume that the possessive stem -angu
has a H associated with it — a H that does undergo HTD, The noun has no H
associated with it. All that happens, then, in these phrases is that the
H of the possessive stem displaces to the ultimate vowel, triggering Rise-Fall.
Consider next the case of nouns that fail to undergo HTD, instead being
subject to Neut.
(9) ny-umba y-angu my house
renje r-angu my pumpkin
mu-kono w-angu my hand, arm
gunguhi r-angu my bed leg
ma-zobe g-angu my crabs
ma-dzungu g-angu my sp. pumpkins
m-barazi z-angu my sp. green beans
vi-yogwe vy-angu my sweet potatoes
86
These examples provide clear evidence that nouns like those in (5) are
tonally different from those in (6) . Certain aspects of the tonal shapes
of these phrases, however, are relevant to the concerns of this paper
(beyond just supporting the claim that there is a contrast between the
nouns in (5) and those in (6)) and thus merit scrutiny.
First, notice that the stem -angu has a H that undergoes HTD when
it follows a low-toned noun, as in (8) , but behaves differently after
the high-toned nouns in (9). The behavior of -angu can be easily suiranarized-
-angu behaves, in (9), as though it has a H associated with the penult stem
vowel that fails to undergo HTD and instead is subject to Neut. In other
words, ny-umba y-angu is akin to a-na-som-a and ma-zobe g-angu is akin to
a-na-vug-a . I assume, therefore, that whereas in (8) -angu has a H that is
subject to HTD, in (9) -angu has a H associated with its penult vowel and
this H is not subject to HTD but undergoes Neut instead. How precisely one
should account for this difference in the behavior of -angu is not clear;
the phenomenon seems to be restricted to these possessive phrases (-angu is
not the only possessive stem; a number of others behave in parallel fashion) ,
and I will not dwell on the matter here.
Second, notice that the phrases in (9) provide very strong support
for the claim that nouns such as those in (6) have a H associated with them
underlyingly, even though this H is lost from the surface through the
application of Neut. The H associated with the noun actually appears in
the phrases in (9). It does not, of course, appear on the penult vowel of
the noun, where I have assumed that it is underlyingly located. In
examples such as ny-umba y-angu, the H of the noun displaces to the
penult vowel of the phrase and in ma-zobe g-angu to the antepenult vowel
of the phrase. In other words, the H of the noun behaves entirely analagously
to the prefix H in a-na-som-a and a-na-vug-a. Recall that in section 1,
I suggested that the displacement involved here is not to be accomplished
by D-S, but rather by a third rule of displacement which will be called
Displacement-to-Neutralized Vowel (D-NV) . Assuming that there is such a
rule, distinct from D-S, I can now make some observations about its role.
The possessive phrases in (9) support the view that D-NV must be a phrase-
level rule, since a H in the noun is being displaced either to the neutralized
vowel in the possessive word or to the vowel that precedes the neutralized
vowel. (9) also gives some indication of the ordering of D-NV. Notice that
in the case of ma-zobe g-angu, the H of the noun associates with the word-
final vowel of the noun but does not induce the application of High Doubling.
This indicates that High Doubling must precede D-NV, since otherwise D-NV
would trigger the application of HD. The suggested derivations for
ny-umba y-^ngu and ma-zobe g-angu are given in (10)
(10)
$ny-umba y-angu$
inapplicable
inapplicable
ny-umba y-angu
inapplicable
H-^^^^L
ny-umba y^ngu
inapplicable
inapplicable
inapplicable
inapplicable
H p
$ma-zobe g-angu$
inapplicable
HE
inapplicable
HTD
H L
ma-z6be g-angu
Neut
inapplicable
HD
H^ L
ma-zobe g-angu
D-NV
inapplicable
D-S
inapplicable
HTS
inapplicable
Rise-Fall
inapplicable
UFL
87
A couple comments on these derivations are in order. First, notice
that Neut must not apply to the H associated with the penult vowel of the
noun in these phrases, although it does apply to the H associated with the
penult vowel of the possessive word, I propose that the explanation for
these facts is that Neut applies to a H associated with a penult vowel
(where the H originates in the stem) in the last word in the phrase. In
other words, Neut is a rule that is phrase-level in the sense that it is
triggered by a phrase boundary. See section 4 for more discussion of this
point. Second, the derivations in (10) show Neut preceding D-NV. This
ordering will be critical if it turns out that D-NV is triggered by the L
that results from Neut. This issue will be discussed in section 4 when
more evidence for D-NV is examined. Third, the derivations in (10) show
D-NV preceding D-S. The motivation for this ordering is that we do not
want to give D-S an opportunity to apply in derivations where D-NV applies.
Again, more on this matter in section 4. Finally, if HD precedes D-NV
(as I argued above) and if D-NV precedes D-S (as suggested above) , we will
have the result that (by transitivity of rule ordering) HD precedes D-S.
In the list of rules in (4), D-S was put before HD; this was an arbitrary
move, however, since there seems to be no direct evidence as to how these
two rules should be sequenced (I leave it to the reader to convince him-
or herself of this point).
Possessive phrases can also be used to establish the existence of
a third class of nouns — namely, those in (7) that have a H associated with
them that does undergo HTD. Consider the examples in (11) .
(11) kunde z-angu my sp. beans
kSnde z-angu my food
chi-kopwe ch-angu my sweet potato leaves
zodo r-angu my mangos
pweza w-angu my octopus
These phrases show that after high-toned nouns which are subject to HTD,
-angu behaves just as it does after low-toned nouns — namely, the H of
-angu undergoes HTD and then Rise-Fall. The H associated with the noun
behaves exactly as it does when these nouns are in phrase-final position.
An analysis of these phrases will not be attempted here (since a full
examination of noun phrase behavior is outside the scope of this paper) ,
but the data clearly establishes that nouns such as those in (7) are
distinct both from those in (5) and those in (6)
3. Some new verb tenses.
K§W examine three verbal tenses: the infinitive, the present tense,
and the -ka- perfective. There are other verbal tenses whose tonal patterns
are entirely analagous to the tenses K^W consider. For example, the
future tense parallels the present tense in terms of its tonal behavior.
The future marker -nda- does not have a H associated with it, while third
person subject prefixes in this tense do have a H associated with them;
the lexical tone of the verb stem remains unmodified, just as in the present
tense (in other words, verb stems that have a lexical H continue to have
this H in the present and the future, while verb stems that do not have a
lexical H do not exhibit a H in the present and the future; in Digo, as in other
Bantu languages, the lexical tone of a verb may be altered in some tenses).
Some examples of the future tense:
88
(12) a. low-toned stems
ni-nda-sagar-a 'I will sit down'
ni-nda-tegur-a 'I will knock s.t. off/over'
ni-nda-chit-a 'I will pierce'
a-nda-chYt-S 'he will pierce'
ni-nda-raBiz-a 'I will insult'
a-nda-raBiz-a 'he will insult'
b. high-toned stems that undergo Neutralization
ni-nda-vug-a 'I will cook'
a-nda-vug-a 'he will cook'
ni-nda-som-a ' I will read '
a-nda-som-a 'he will read'
c. high-toned stems that undergo HID
'I will speak'
a-nda-nen-S 'he will speak'
In (12a) there are no H's except in the case of the forms that have the
subject prefix a^-, which is high-toned in this tense. The H of this prefix
displaces to the ultimate vowel, triggering Rise-Fall in a-nda-chit-& and
triggering HD and UFL in a-nda-raBiz-a. In (12b) , the verb stems have a H
associated with them; this H does not undergo HTD but rather is subject to
Neut. In the case of the forms with the a- prefix, the H associated with
this prefix attempts to displace to the penult vowel — succeeding in the
case of a-nda-som-a but landing on the antepenult vowel in the case of
a-nda-vug-a due to the voiced obstruent in front of the penult vowel.
In (12c), the stem has a H tone that is subject to HTD. In ni-nda-nen-a ,
Rise-Fall has applied to the output of HTD. In the case of a-nda-nen-a,
we have the following derivation:
(13) $a-nda-n6n-a$
inapplicable HE
H H
A-nda-nen-a HTD
inapplicable Neut
inapplicable HD
inapplicable D-NV
"^^-^ t!
a-na-nen-a D-S
a-na^Sfeiv4 HTS
inapplicable Rise-Fall
a-na-^nfen=»a UFL
89
Similarly, the -chi- tense (narrative past, conditional) parallels
the -ka- perfective tense in terms of tone. There is a H associated with
this verb tense, regardless of the person/number/gender of the subject
prefix and regardless of the lexical tone of the verb. I will assume that
the H supplied by this particular tense is associated with the -chi- prefix
(a similar assumption in K5W was made with respect to the -ka- tense) . In
addition to this H, the verb stem may be either high- or low-toned. Some
examples:
(14) a. low-toned stems
'I cultivated/if I cultivate'
ni-chi-riB-a 'I paid/if I pay'
ni-chi-v^y-a 'I begged /if I beg'
ni-chi-vumikiz-a 'I agreed /if I agree'
b. high-toned stems that undergo Neutralization
ni-chi-wad-a '1 roasted/if I roast'
ni-chi-som-a 'I read/if I read'
ni-chf-vug-a 'I cooked/if I cook'
c. high-toned stems that undergo HTD
ni-chi-tongoy-a 'I preceded/if I precede'
ni-chi-Bfrik-a 'I sent/if I send'
ni-chi-ban^ng-a 'I spoiled s.t./if I spoil s.t.'
In (lAa), there is a single H (namely, the H associated with the prefix
-chi-) which displaces to the ultimate vowel, triggering Rise-Fall in cases
like ni-chi-rYm-^ and triggering HD and UFL in ni-chi-vumikf z-a . In (14b) ,
there is a H associated with the -chi- but also one associated with the
first stem vowel. This stem H fails to undergo HTD (since it is a stem H
associated with a penult vowel), subsequently being changed to L by Neut.
The prefix H attempts to associate with the neutralized vowel — succeeding
in the case of ni-chi-s6tn-a but remaining on the prefix in the case of
ni-chl-vug-a due to the voiced obstruent in front of the neutralized vowel.
In (14c), there is a H associated with the -chi- and also with the first
stem vowel. The stem H displaces to the ultimate vowel and the prefix H
undergoes D-S, which attempts to associate it with the first stem vowel —
succeeding in the case of ni-chi-tongoy-a , for example, but failing in the
case of ni-chl-banang-^ due to the voiced obstruent in stem-initial position.
This H associated with the first stem vowel then spreads onto the remaining
vowels in the word in the case of ni-chi-tongoy-^ but is blocked from spreading
in the case of ni-chi-banang-a . UFL changes the last H to L in the former case.
But in addition to tenses like the above which are parallel in behavior
to the tenses examined in KfiW, there are other tenses that show distinct
tonal properties. I cannot deal with all of the tenses here, but I do
examine a few especially interesting ones.
90
3.1. The -a- past tense.
The morphological structure of the past tense in Digo is SP+a+(OP)+VS+a,
where SP=subject prefix, OP=object prefix, and VS=verb stem. The choice of
subject prefix does not alter the tonal shape of past tense forms in any way;
consequently, all of the examples cited will be with the same SP — namely,
the 1 sg. subject prefix, which is ordinarily ni- but assumes the form n-
in front of the -a_- past tense marker. Examples of this construction are
given in (15) .
(15) a. low-toned stems
n-a-tsukur-a 'I carried'
n-a-wocher-'a 'I received'
n-a-rongoz-a 'I led'
n-a-tsor-a 'I picked up'
n-a-pig-a 'I hit'
n-a-gur-a 'I bought'
n-a-vugur-a 'I untied'
n-a-vumikiz-a ' I agreed '
n-a-tog3^r-a 'I praised'
n-a-ragur-a 'I treated'
n-a-dz^ng-a 'I built'
b. high-toned stems that undergo Neut
n-a-meg-a 'I broke off a piece'
n-a-san-a 'I forged'
n-a-rum-a ' I bit '
n-a-Bah-a 'I got'
n-a-hem-a 'I cleared forest'
n-a-vwin-a ' I sang '
c. high-toned stems that undergo HTD
n-a-kumbukir-a 'I remembered'
n-a-turuk-a 'I went out'
n-a-puput-a 'I beat'
n-a-nen-^ ' I spoke '
n-a-vwinir-a 'I sang for'
n-a-dunduriz-a 'I put s.t. aside'
n-a -garagar-a 'I rolled about in pain'
n-a-dung-a" 'T pierced'
n-a-ezek-^ 'I thatched'
Consider first the data in (15a) . The verb stems in question are low-
toned at the lexical level, but nevertheless the tonal shapes that they
exhibit in the past tense are entirely analagous to verbal forms that
contain two high tones. The rightmost of these two H tones is displaced
to the ultimate vowel and the preceding H displaces to the first stem vowel
(if it can, otherwise the vowel preceding the stem), from which it spreads
onto following vowels (provided it is not stopped from doing so by the
presence of a voiced obstruent). The data in (15a) parallel precisely
the behavior of high-toned verbs in the -ka- tense. This is illustrated in (16)
91
(16) a-ka-kumbuklr-a he has remembered' (cf. ku-kurabukir-a)
vs. ^ ^
n-a-tsukur-a 'he carried' (cf. ku-tsukur-a)
a-ka-dunduriz-a
vs.
n-a-vumiki z -a
a-ka-tabik-a
vs.
^ V A
n-a-ragur-a
'he has put s.t. aside' (cf. ku-dunduriz-a)
'he agreed' (cf. ku-vumikiz-a)
'he is troubled' (cf. ku-tabik-a)
'he treated' (cf. ku-ragur-a)
The data in (15a), therefore, present no new problems except for the
following: where are the two H's in this construction located? What we
know for certain about these H's is that the rightmost one is subject to
HTD. This tells us nothing really about its location underlyingly — since
H's displaced to the ultimate vowel could originate on a prefix or on a
stem or even (conceivably) on the final vowel -a. The first H in the
word is subject to D-S, and this provides some clue as to its location
since in all other cases a H that displaces to the stem is located to the
left of the stem. Thus it is reasonable to conclude that the first H in
these words is located on one of the prefixes, either the subject prefix
or the -a- past tense marker. Further insight into the location of the
H's can be derived from the data in (15b).
The verb stems in (15b) underlyingly have a H associated with them. The
data in (15a) suggest that the past tense construction has two H's associated
with it, independent of any lexical tone of the verb. Are there, then, three
H's in the case of the verbs in (15b)? Note that the examples in (15b) are
different from those in (15a) . This suggests that there must indeed be a
tonal difference. Wherein does the difference reside? Clearly one difference
is that in (15a) HTD has been able to apply, whereas there is no sign that
HTD has applied in (15b) . One way to explain this failure of HTD to apply
would be to claim that in an example such as n-a-san-a, the rightmost H in
the underlying form is the H associated with the stem -san-. This means that
any other H's in the word (namely, the two H's associated with the past tense
morphology in general) must be located to the left of the stem. If we make
this assumption, then the failure of HTD to apply in (15b) follows from the
fact that a H associated with a penult vowel (where the H originates in the
stem) is regularly exempted from HTD, subsequently undergoing Neut.
A second difference between (15a) and (15b) is that in the latter case
D-S has not displaced a H all the way to the first stem vowel (a failure
that cannot be attributed to the presence of a voiced obstruent), whereas
in the former case such a displacement has taken place (except for instances
where a voiced obstruent has blocked the displacement). Another fact about (15b)
requires mention — namely, there is no surface manifestation of D-NV. We
need, then, to account both for why D-S has not shifted a H to the first
stem vowel and for why D-NV has no surface manifestation.
If we assume that the underlying structure of n-a-san-a is
H H H
^ i I ' d.
$ni-a-san-a$
92
then we will get the following partial derivation:
(17) jy H
$ni-a-san-a$
inapplicable
HE
inapplicable
HTD
ni-a-san-a
Neut
inapplicable
HD
ni-a-san-a
D-NV
At this point in the derivation, D-S will try to apply. D-S as so far
formulated will associate a H with the first stem vowel, unless it is
prevented from doing so by the presence of a voiced obstruent. Thus we
would expect the H associated with the SP to displace to the stem -san-;
but this would yield an incorrect result, since we want this H to associate
instead with the tense marker -a-. I suggest that D-S be constrained so
that it may not associate a H with a vowel that is associated with a L
on the tonal tier . l-'-This constraint would have the effect of preventing
D-S from ever associating a H with a neutralized vowel. Given, then, the
unavailability of the stem vowel, the farthest vowel to the right that the
prefix H can displace to is the tense marker -a.-. The derivation in (17)
will thus continue as follows:
H H.^
5ni-a-san-
(18) $r
ni-^-san-a D-S
ni-a-^SSn-a HTS
inapplicable Rise-Fall
inapplicable UFL
Although this derivation yields an incorrect surface form (*n-a-san-a) , the
correct surface form can be readily obtained if we assume a rule that eliminates
an association line between a H and vowel if that vowel (a) is associated with
a L and (b) is preceded by a vowel that is associated with a H. Call this
rule Fall Simplification (FS) . It has the effect of changing a falling tone
to a low tone if the preceding vowel is high-toned. (I assume that when FS
eliminates the association line between the vowel and the H, if the H is thus
left unassociated with a vowel on the segmental tier, it will simply be deleted
from the representation. Alternatively, it could be allowed to reassociate with
the preceding vowel. In the present situation, at least, the result would be the
same as far as the surface structure goes.) Given the postulation of FS, the
derivation begun in (17) and continued in (18) will conclude as in (19).
(19) $ni-^^^n-a$
ni-a-sc
n-a-san-a deletion of prefix vowel before -a-
-• - 3in-a FS
In this particular case, the relative ordering of FS and Rise-Fall is
immaterial (since Rise-Fall is inapplicable here), so I will return later to
this particular ordering relationship. FS must follow HTS, however, since
FS eliminates an association line that is supplied by that rule.
93
Although there are perhaps other ways of accounting for n-a-san-a
and the other data in (15b) , the preceding account will be adopted in this
paper, FS in particular appears to me to be a quite natural rule and ought
to be invoked in explaining the absence of any surface manifestation of
D-NV. Let us now turn to the data in (15c). The verb stems here are ones
that have a H associated with their first vowel (at the lexical level) , a
H that would be expected to undergo HTD. The analysis of Digo that I have
elaborated here predicts the following derivation.
(20)
H H H
$ni-a-kumbukir-a$
inapplicable
HE
;I H H
ni-a-kumbukir-a
HTD
inapplicable
Neut
inapplicable
HD
inapplicable
D-NV
H^H^ H
ni-a-kumbukir-A
D-S
ni -a^^^m&hkir^
HTS
inapplicable
FS
inapplicable
Rise-Fall
ni-a-Khmblkir-a
UFL
n-a-ftam&ulcir^
deletion of prefix vowel before -a
This derivation yields the correct surface form: n-a-kumbukf r-a . The most
notable point about this derivation is that D-S will associate both of the
prefix H's with the first stem vowel. At the phonetic level, of course, two
H's associated with a single vowel will be identical to one H.
12
The proposed analysis also predicts the correct surface form for an
example such as n-a-ez^-a.
(21)
H H H
$ni-i-4zek-a$
inapplicable
ezek-a
inapplicable
inapplicable
inapplicable
zek-a
inapplicable
inapplicable
H H LHJ.
ni^^^^^zeK-a
inapplicable
H H If H I
n-a^2
„IL,
„^k
HE
HTD
Neut
HD
D-NV
D-S
HTS
FS
Rise-Fall
UFL
deletion of prefix vowel before -a-
94
In this derivation we can see that FS must precede Rise-Fall. If the
reverse order obtained, FS would be able to apply to the output of Rise-
Fall and eliminate the falling tone on the ultimate vowel in favor of a
low tone (since we would have a H associated with a vowel that (a) is
also associated with a L and (b) is preceded by a vowel that has a H
associated with it). By applying FS first, we keep it from affecting the
output of Rise-Fall.
The derivations of the remaining items in (15c) either parallel those
given above or reflect additional phenomena that have been adequately
discussed elsewhere in this paper. I conclude, then, that the tonal
pattern of the -a- past tense can be accounted for if (a) we assume
that there are two H's supplied by the past tense morphology, one of
these H's presumably being associated with the subject prefix and the
other with the -a- prefix; (b) the stem retains its lexical tone
specification; (c) D-S cannot associate a H with a vowel that is associated
with a L (and thus in such cases must associate with the vowel that precedes
the neutralized vowel); and (d) there is a rule of Fall Simplification
that has, roughly, the form given in (22).
(22) FS H (H) L ..^. ^, .. . .
■■ ' condition: the preceding vowel is
N/
associated with a H
where the symbol % means that the association line in question is to be
deleted. (22) utilizes the parentheses notation and thus abbreviates the
two rules given in (23).
(23) a. H H L
"S^ l-^ condition: as in (22)
^<ut condition: as in (22)
V
Given a representation like
$ni-a^^^San-a$
subrule (23a) will delete the association line between the first H and the
neutralized vowel while subrule (23b) will delete the association line
between the second H and the neutralized vowel.
3.2. The imperative.
In this section, I will examine some aspects of the tonal structure of
the imperative in Digo, This form has neither a subject prefix nor a tense
prefix. The final vowel is -a when there is no object prefix present or when
the OP is the 1 sg. -ni-, otherwise the final vowel is -e. The suffix -ni
is added to indicate a plural subject. Although I will give the pi. imp. forms
when I cite data, these forms are problematic in certain respects and I will
not attempt to give an account of them. The reader will perhaps recall that
the locative suffix -ni also yielded some problematic forms; there may well
be a connection between the two sets of data. In (24) t illustrate the imperative
of a low-toned verb — namely, -tsukur- 'carry'.
95
(24) tsukur-a 'carryl' tsukur-a-ni 'pi. carry!'
ni-tsukur-a 'carry me.'' ni-tsukur-a-ni 'pi. carry mel '
a-tsukuV-e 'carry them.'' a-tsukur-e-ni 'pi. carry them!'
The key to understanding the imperative construction, tonally-speaking, is
provided by the example a-tsukur-e. The 3 pi. OP -a- is high-toned. In order
to account for the high-fall on the penult vowel of this example, we must
assume that the stem contains a H that fails to undergo HTD and instead is
converted to L by Neutralization. In other words, the verb must have a H
associated with its penult vowel (given my suggestion earlier that only
penult vowels undergo Neutralization). The derivation of a-tsukur-e can be
seen in (25).
(25)
y H
$d-tsukAr-e$
inapplicable
HE
inapplicable
4-tsukar-e
HTD
Neut
inapplicable
HD
a-tsukar-e
D-NV
inapplicable
D-S
inapplicable
HTS
inapplicable
FS
inapplicable
Rise-Fall
inapplicable
UFL
vation of a-tsukUr-
-e is accep
in support of our claim that D-NV is a distinct rule from D-S. Notice that
the prefix H in this derivation does not associate with the first stem vowel
(as it does in the case of D-S), but rather with the neutralized vowel. I
assume, therefore, that the rule D-NV must have the form given in (26).
(26) H L H-. ^ L
V X V ===> V X V
(where X = any segmental material whatever)
Of course, this formulation is not fully adequate, since we have seen earlier
(cf. a-na-vug-a 'he is cooking') that the presence of a voiced obstruent in
front of the neutralized vowel causes the H to associate with the antepenult
vowel instead of the neutralized vowel. But before we consider how (26)
should be modified to take this fact into account, more data needs to be
assembled. Since (26) is triggered by the presence of a L, the rule that
creates this L — namely, Neut — must be applied before D-NV. I have assumed
that D-NV precedes D-S. The motivation for this is simply that if a
structure satisfies the conditions for both rules, it is D-NV that applies.
However, it is possible that one might be able to allow D-S to apply first
to associate a H with the first stem vowel, and then allow D-NV to operate
on that output and move the H again, this time to the neutralized vowel.
I prefer the simpler derivation where just D-NV will apply.
96
The preceding analysis suggests that the imperative morphology
supplies a H to the penult vowel (since in the lexicon -tsukur- is low-
toned, the H must come from the morphology rather than from the lexicon).
This H fails to undergo HTD, but does undergo Neut. In a-tsukur-e there
is a preceding H in the word that can displace to the neutralized vowel;
no such H is present in tsukur-a and ni-tsukur-a, consequently the only
rule that applies in these forms is Neut.
V ^
(27) vumikiz-a 'say yes!' vumikiz-a-ni pi. say yes I'
ni-vumikiz-a 'say yes to mel ' ni-vumikiz-a-ni 'pi. say yes to mel '
a-vumikiz-e 'say yes to them I ' a-vumikiz-e-ni 'pi. say yes to them.'
Consider the form a-vumikiz-e. The H associated with the OP -a- is manifested
on the prefix. It does not displace to the neutralized vowel. It would seem
that the failure of D-NV to move the H from the prefix must be attributed to
the fact that a voiced obstruent intervenes. This would suggest that the proper
formulation of D-NV is as in (28),
(28) H L
V ====>
i
(where X=maximal number of segments not containing a
voiced obstruent)
(Note that this rule requires that a L follows the H, but does not specify
which vowel in the segmental tier that L is associated with.) There is a
problem with this formulation of D-NV that will arise in section 4, but
the present data motivate the specifications given in (28) .
The derivation of a-vumikiz-e is shown in (29).
(29) ^ H
$a-vumikiz-e$
inapplicable
HE
inapplicable
HTD
H L
d-vumikiz-e
Neut
inapplicable
HD
inapplicable
D-NV
inapplicable
D-S
inapplicable
HTS
inapplicable
FS
inapplicable
Rise-Fall
inapplicable
UFL
Consider next the case of a low-toned verb stem with a stem-medial
voiced obstruent.
(30) rejezer-a 'soak for." rejezer-a-ni 'pi. soak fori'
ni-rejezer-a 'soak for me!' ni-rejezer-a-ni 'pi. soak for me."
a-rejezer-e 'soak for them!' a-rejezer-e-ni 'pi. soak for them!
97
These data illustrate that D-NV cannot displace a H past the first voiced
obstruent between the vowel with which the H is associated and a neutralized
vowel. Thus in a-rejezer-e, the prefix H displaces to the first stem vowel
since it is blocked from associating any further to the right by the voiced
obstruent 2- The formulation of D-NV given in (28) thus finds additional
confirmation.
(31)
So far I have illustrated only the behavior of low-toned verbs.
I illustrate a high-toned verb.-*--^
In
(31) Birik-a
ni-Birik-a
a-Birik-e
'send it.' '
'send me.' '
' send them.' '
Birik-a-ni
V ■'■■
ni-Birik-a-ni
a-Birik-e-ni
'pi. send it I '
'pi . send me.' '
'pi. send them.' '
These examples parallel precisely the data in (24) — suggesting that perhaps
in the imperative the lexical tone specification is deleted and a H is
assigned to the penult vowel for all verbs. Further support for this
proposed merger between high- and low-toned verbs is found in (32) .
(32) vwinir-a
ni-vwinir-
a-vwinir-£
sing to I
'sing to mel '
'sing to them!
vwinir-a-ni .
ni-vwinir-a-ni
. . / A
a-vwmir-e-ni
'pi. sing to I '
'pi. sing to me.' '
'pi. sing to them.'
These data parallel precisely the low-toned verb -vumikiz- illustrated in
(27).
Although my available data is unfortunately skimpy on this particular
point, there is some evidence that the lexical H associated with the first
vowel of a high-toned stem is lost only in cases where the first stem vowel
is the vowel that precedes the penult vowel. Consider the data in (33).
(33)
chekecher-a
ni-chekecher-a
a-chekecher-e
'sift for."
'sift for mel '
'sift for them.'
chekecher-a-ni
ni-chekecher-a-ni
a-chekecher-e-ni
'pi. sift for."
'pi. sift for me.' '
'pi. sift for them!
I suggest that the underlying representation of a-chekecher-e is
H H H
$a-ch4kecher-e$
and that the lexical H associated with the first vowel of -chekecher- has not
been deleted since there is a vowel between it and the penult H. The correct
surface form can be obtained by means of the rules that we have already posited.
The derivation is given in (3A).
(34)
H H H
$a-chfekech4r-e$
inapplicable
HE
inapplicable
HTD
H H L
a-ch^kecher-e
Neut
inapplicable
HD
A-chekecfter-e
D-NV
a-crtekecner-e
D-S
a-cSfeStcRe r-e
HTS
98
a-ch^lcecher-€
inapplicable
inapplicable
FS
Rise-Fall
UFL
There is one problem with the above derivation that requires mention —
namely, D-S disassociates a H from the object prefix and reassociates it
with the first vowel of the verb stem -chekecher-. In K§W it was shown
that, for purposes of D-S, an OP counts as part of the verb stem. Thus a
H associated with a subject prefix or with a tense prefix will reassociate
with the OP rather than with the first vowel of the verb stem proper. Thus
it would seem that in (34) we do not really expect the H on the OP to leave
the OP at all. But the facts are clear. A H associated with an OP displaces
(by D-S) off that prefix onto the stem proper, whereas a H associated with
the SP or the tense prefix will displace (by D-S) onto the OP. The -ni forms
like a-tsukur-e-ni and a-Birik-e-ni apparently illustrate the same point,
but since I haven't provided a full analysis of these forms they cannot
safely be invoked here.
If the derivation in (34) is valid, then it provides additional
support for the existence of the rule Fall Simplification. It is FS
that accounts for the fact that a-chekecher-e has undergone Neut and
D-S (followed by HTS) , but that there is no surface trace of D-NV.
3.3. The present negative.
The tonal structure of the present negative parallels the imperative
and thus provides further confirmation of observations made in the preceding
section. The morphology is somewhat complicated. I give a sample paradigm
for a low-toned verb in (35) .
(35)
si-tsukur-a
ku-tsukfir-a
ka-tsukur-a
ta-u-tsukur-a
ta-mu-tsukur-a
ta-a-tsukur-a
'I'm not taking,
'you're not. . . '
'he's not. . . '
'we're not. . . '
'you pi. aren't,
'they aren't. . . '
carrying
The crucial point here is the high-falling tone on the penult vowel. Given
the analysis of the imperative, the analysis of this paradigm quickly takes
shape. The present negative must involve a H associated with the penult vowel,
In addition, there must be another H located before the penult H. Where
that other H is located can be seen from the paradigm in (36).
(36)
'I'm not untying'
etc.
si-vugur-a
ku-vugur-a
ka-vugur-a
ta-u-vugur-a
ta-mu-vugur-a
ta-a-vugur-a
Clearly, there is a H before the verb stem. The H could be regarded as being
under lyingly associated either with the negative element or the subject prefix
(in the singular forms, these two constituents are fused together in any case)-
I will arbitrarily assume that it is the SP that bears the H. The analysis
99
that I have developed predicts, correctly, that the H associated with the
SP will not be able to displace to the neutralized vowel in (36) due to
the presence of a voiced obstruent in stem-initial position.
If an OP is inserted before the stem -vugur-, the H of the SP will
displace to the OP, although it can go no further than that due to the
voiced obstruent following the OP. Cf. the paradigm in (37).
(37) ta-a-ni-vugur-a 'they are not untying me'
ta-a-u-vugur-a 'they are not untying us'
That the H of the SP will displace as far toward the neutralized vowel
as possible is shown by the data in (38) :
(38) ka-ni-togor-a 'he's not praising me'
ka-ku-togor-a 'he's not praising you sg. '
ka-mu-togor-a 'he's not praising him'
ka-u-togor-a 'he's not praising us'
etc.
In these examples, there is no voiced obstruent until just before the
neutralized vowel; therefore, the H of the SP is able to associate with
the antepenult vowel. It should be noted, incidentally, that although
ordinarily there are some low-toned object prefixes and some high-toned
object prefixes, in the negative present all object prefixes behave as
though they are low-toned. I will not attempt to account for this behavior
here.
The behavior of lexical high-toned verbs in the present negative is
parallel to the behavior of such verbs in the imperative. If the stem is
of the shape -CgVC^VCg-a, no trace of the lexical H can be found on the
surface.
(39) si-puput-a 'I'm not beating'
ku-puput-a etc.
A
ka-puput-a
ta-u-puput-a
ta-mu-puput-a
ta-a-puput-a
This paradigm parallels precisely the forms in (35) based on the low-tone
verb -tsukur-. But compare (40).
(40) si-tanyiriz-a 'I'm not driving off predators'
ku-tanyfriz-a etc.
ka-tanylriz-a
ta-u-tanylriz-a
ta-mu-tanyiriz-a
ta-a-tanyfriz-a
Here, where the first stem vowel is separated from the neutralized vowel
by another vowel, we find evidence that the lexical H has been retained.
The derivation of these forms will be exactly like the derivation of the
imperative form a-chekecher-e given in (34). Unfortunately, the available
data for high-toned verb stems with three or more vowels is skimpy and thus
it is not possible (at present) to investigate the behavior of such stems
fully.
100
4 . Noun plus verb phrases.
Although a complete study of the tonology of Digo sentences is
absolutely essential to an adequate understanding of Digo tone, I have
sufficient materials to explore in detail just one sort of phrase, namely,
those consisting of a verb plus a single (unmodified) noun. Such phrases,
however, shed considerable light on the workings of tone in Digo.
4.1. Phrases involving verb words with no high tones.
I consider first phrases involving verbal forms that lack any high
tones. Verbal forms of this type are: (a) low-toned verbs in the infinitive
when no high-toned OP is present, and (b) low-toned verbs in the present or
future when the SP is Ip. or 2p. and there is no high-toned OP. In (41)
I illustrate the case when a low-toned noun follows, in (42) the case when
a high-toned noun subject to Neut follows, and in (44) the case when a high-
toned noun that undergoes HTD follows.
(41) ku-gur-a nombe 'to buy a cow'
ku-henz-a mu-ganga 'to look for a doctor'
ku-sag-a mu-hama 'to grind millet'
ni-na-pig-a goma 'I'm beating a drum'
ni-na-guw-a tunda 'I'm peeling a fruit'
ku-ni-gurir-a n-guwo 'to buy clothes for me'
(42) ku-takas-a ny-ungu 'to clean a pot'
ku-rand-a chi-tanda 'to spread a bed'
ni-na-jit-a vi-yogwe 'I'm cooking sweet potatoes'
ni-na-gur-a shoka 'I'm buying an axe'
(43) ku-sag-a ma-pemba 'to grind maize' (cf. ma-pemba 'maize )
ku-vugur-a fundo 'to untie a knot' (cf. fundo 'knot')
ku-andik-a chi-tabu y^'to write a book' (cf . chi-tabu 'book') ^
ni-na-tsor-a chi-dafu 'I'm picking a young coconut' (cf. chi-dafu
'young coconut')
ni-na-vugur-a fundo 'I'm untying the knot'
Examination of the data in (41)-(43) establishes that after a verb which
contains no H's in its underlying structure, a noun is pronounced exactly
as it is in its isolation form. The noun is in no way affected by being
preceded by a low-toned verb.
4.2. Phrases involving verb words with one high tone.
Next I turn to phrases where the verb word contains a single H in its
underlying structure. Examples of this sort include: (a) low-toned verb
stems in the infinitive or in the l/2p. present or future tense provided
there is a high-toned OP present; (b) a high-toned verb stem in the infinitive
or in the l/2p. present or future tense; (c) a low-toned verb stem in
the 3p. present or future tenses, if no high-toned OP is present; and
(d) a low-toned verb stem in the -ka- or the -chi- tense. In (44), I give
examples of low-toned nouns after such verbs.
101
(44) ku-onyes-a n-jira to show the way' (cf. ku-onyes-a, n-jira)
ku-afun-a ny-ama 'to chew meat' (cf . ku-afun-'a, ny-ama)
ku-anik-a n-guwo ^'to put clothes in sun to dry' (cf^ ku-anik-a, n-guwo]
ni-na-ezek-a b^nda 'I'm thatching a shed' (cf. ku-ezek-'a, banda)
ni-na-reh-a chi-gwazo 'I'm bringing a peg' (cf. ku-reh-a, chi-gwazo)
ni-na-chekech-a ^-nga 'I'm sifting flour' (cf. ku-chekech-a, u-nga)
ni-na-adz-a mu-tu 'I'm mentioning by name' (cf. ku-adz-a, mu-tu)
a-na-henz-a mu-ganga 'he's looking for a doctor' (cf. ku-henz-a,
^ ^ mu-ganga)
a-na-jit-a manga 'he's cooking cassava' (cf. ku-jit-a, manga)
a-na-plg-a goma" .'he's beating a drum' (cf. ku-pig-a, goma)
a-na-ragiz-a kalamu 'he's ordering a pen' (cf. ku-ragiz-a, kalamu)
a-ka-gur-a n-guwo 'he has bought clothes' (cf. ku-gur-a, n-guwo)
a-ka-jit-a ma-zu /.^'he has cooked bananas' (cf. ku-jit-a, ma-zu)
a-ka-ragiz-a kalamu 'he has ordered a pen' (cf. ku-ragiz-a, kalamu)
There are two immediately striking facts about the data in (44). First,
the low-toned nouns are pronounced just like nouns that have a high tone that
has been displaced to the ultimate vowel. Second, the verbs are (for the
most part — and herein lies a very interesting story) pronounced as though
they have no H associated with them. In other words, it seems as though
the verb gives up its H to the following low-toned noun.'
The analysis of this phenomenon is fairly obvious. Namely, all that we
need to do is assume that the rule of HTD operates on phrases as well as
words. HTD will say that the rightmost H of the input representation
(whether that representation be a single word or a phrase) is associated
with the ultimate vowel in the representation. Given an input such as
$a-na-henz-a mu-ganga$
HTD will associate the prefix H with the ultimate vowel of the noun. Rise-
Fall will then apply to derive the correct surface form.
There is one significant question that must be raised with respect to
this analysis of the data in (44). Does the rule HTD operate once on the
entire phrase, or does it operate twice — first on the verb word, and then
again to the phrase consisting of the verb plus following noun? Given a
structure such as that cited above, either approach would achieve the correct
results. If HTD applies once, it takes the prefix H directly to the last
vowel of the noun. If HTD applies twice, it first of all takes the prefix
H to the last vowel of the verb and then reapplies to associate it with the
last vowel of the phrase. In either case, the prefix H ends up on the last
vowel of the phrase.
Evidence in favor of saying that HTD applies in some sort of cyclic
fashion (first to the verb, and then again to th^ phrase) is provided by^
examples like a-na-pfg-a goma, a-na-ragfz-a kalamu, and ni-na-adz-a mu-tu.
Although in underlying structure these phrases contain just a single H —
originating on the 3 sg. subject prefix in the first two examples and on
the verb stem in the last example, the surface forms clearly reflect two
102
high tones. In section 1, I suggested a rule that these forms provide
crucial evidence for — namely, the rule of High Doubling. This rule says
that a H on an ultimate vowel that is preceded by a voiced obstruent will
double onto the preceding vowel. The effect of this rule is to yield an
output with two H's from an input with just one H. This, of course, is
just the result required in order to account for their being two high
tones in such phrases as a-na-plg-a goma but just a single high tone in
phrases like a-na-gur-a n-gHvlb.
We must now turn our attention to precisely how a form like a-na-pig-a
goma is to be derived. HD is triggered by a H on an ultimate vowel that
is preceded by a voiced obstruent. Thus, given the structure
H
$a-na-pig-a n-guwo$
the only way that we can get HD to apply is by somehow getting the prefix
H to associate with the ultimate vowel of the verb word. There is a rule
that would do this, of course — namely, HTD. But the fact that HTD must
move the prefix H to the end of the verb means that HTD cannot be a rule
that applies just once, to the whole phrase, rather it must apply first
just to the verb itself and then again to the whole phrase. In other
words, it is essential that HTD be cyclic. Furthermore, it is necessary
that HD be cyclic as well, since HD must apply while there is still a
H associated with the ultimate vowel of the verb, even though that H will
eventually be displaced to the ultimate vowel of the following noun. If
these two rules (at least) are cyclic, we can obtain the correct derivation,
as shown in (45). 15
(45) $[[(
a-na-pig-a#] |
a-na-pig-a
a-na-pig-a
;#goma#]]$
first cycle
inappl:
Lcable
HTD
inappl:
Lcable
HD
H H
a-na-pig-a
goma
$
second cycle
inapplicable
HE
H
H
a-na-plg-a
gomd
HTD
inapplicable
Neut
inapplicable
HD
inapplicable
D-NV
inapplicable
D-S
inapplicable
HIS
inapplicable
FS
a-na-pig-a
goma
Rise-Fall
inapplicable
UFL
I conclude, then, from this sort of example that there is a cyclic
application of at least HTD and HD. The issue of the possible cyclic
application of the other rules will be examined as I proceed through
additional data.
103
In (44) I gave just examples where the H on the verb word would
undergo HTD if the verb word were pronounced in isolation. In (45) I
give examples where the H on the verb word would not undergo HTD (but
instead would be subject to Neut) if the verb word were pronounced in
isolation. The noun that follows is, again, a low-toned noun.
W A ,
(46) ku-heg-a n-guruwe to trap a pig (cf. ku-heg-a, n-guruwe)
ku-heg-a n-jiya 'to trap a pigeon' (cf. n-jiya)
ku-ih-a mu-ganga a 'to call a doctor' (cf. ku-ih-a, mu-ganga)
ni-na-Bah-a ma-tunda 'I'm getting fruit' (cf. ku-Bah-a, ma-tunda)
ni-na-vug-a w-ari 'I'm cooking rice' (cf. ku-vug-a, w-ari)
ni-na-tsun-a nombe 'I'm skinning a cow' (cf. ku-tsun-a, nombe)
The most striking fact about these data is that the rule of Neutraliza-
tion has not applied to change the H of the verb stems to L. The H
remains, eventually getting displaced (via HTD) to the ultimate vowel
of the noun. The most straightforward way of accounting for the fact that
Neut has failed to apply to the vowel of the verb stem in these examples
is to formulate Neut so that it applies not to the penult vowel of any
word, but rather to the penult vowel in certain syntactic constructions.
Thus Neut would affect a verb word when its the final word in a verb
phrase, but not when it is medial in a verb phrase, etc. Much work on
the syntax of Digo will be required before I can state precisely the
environments where Neut applies (though preliminary data does show that
it is not simply applicable to the penult vowel of an utterance) , and
I will have nothing more to say on this topic here.
The attentive reader may have wondered why, in section 1, I chose to
disassociate the failure of stems like -heg-, -ih-, -vug- to undergo HTD
from the fact that their underlying H gets changed to L by Neut. That is,
why did I put a restriction on the rule of HTD (preventing it from affecting
penult H's that originate in the stem) and claim that this restiction is
independent of the fact that these same H's will be changed to L's by the
rule of Neut. It might have appeared more plausible to claim that it is
the fact that the H of these verbs is changed to L that makes HTD inapplicable.
The motivation for making the failure of HTD to apply to these verb stems
independent of whether their H changes to L comes from an example like
ku-heg-a n-guruw^. The underlying structure of this phrase is:
H
$[[#ku-h4g-a#] [#n-guruwe#]]$
Given the analysis that I have suggested, HTD will not be applicable on
the first cycle to the verb word. Since HTD does not apply. High Doubling
cannot apply either. Thus at the end of the first cycle, the verb word will
not have undergone any changes. On the second cycle, HTD will associate the
H on the verb word with the ultimate vowel of the phrase. (Recall, HTD is
prevented from applying to penult H's originating in the stem, but on the
second cycle the verb H is not associated with the penult vowel in the input
representation, thus it will be susceptible to HTD.) Rise-Fall will then
produce the correct surface form.
104
But suppose that instead of the analysis I have adopted we claimed
that HTD fails to apply just in case Neut applies. It would then follow
that since Neut does not apply in the phrase under consideration that
HTD should apply to the verb word on the first cycle. But if HTD applies
on the first cycle, High Doubling will also be triggered, and we will
obtain the incorrect surface form *k.u-heg-a n-guruwe'. In order to avoid
this incorrect form, we must guarantee that HD does not apply to the verb
word; in order to guarantee that, we must guarantee that HTD does not apply
to the verb word. But since Neut is not involved in this derivation, the
failure of HTD to apply to the verb must have nothing directly to do with
the application of Neut. Thus I conclude that while there is obviously
a close connection between the fact that H's which do not undergo HTD
are susceptible to Neut (I have expressed this connection by restricting
Neut to penult H's — and only H's that have escaped HTD can be in this
environment) , the present data show that this connection is not to be
expressed by making the failure of HTD contingent upon the application of
Neut. 16
Having established that there is some motivation for my proposed
treatment of Neutralization (and its relation to HTD) , let me turn to
phrases involving a verbal word with a single H in its underlying
structure followed by a noun that contains a H that is subject to Neut
rather than HTD. Examples are given in (A7).
A , V A
(47) ku-afun-a nazi to chew a coconut (ku-afun-a, nazi)
ku-azim-a shoka 'to borrow an axe' (ku-azim-a, shoka)
ku-ar-a demu 'to take a rag' (ku-ar-a, demu)
ku-ar-a chi-demu 'to take a small rag' (chi-demu)
ku-adz-a dzina 'to call by name' (ku-adz-a, dzina)
a-na-jit-a ma-renje 'he is cooking pumpkins' (ku-jit-a, ma-renje)
a-na-gur-a shoka 'he is buying an axe' (ku-gur-a)
a-na-andik-a chi-tanda 'he is making a bed' (ku-andik-a, chi-tanda)
a-na-andik-a baruwa 'he is writing a letter' (ku-andik-a, baruwa)
a-na-tsong-a gun^hi 'he is cutting a bedleg' (ku-tsong-a, gunguhi)
a-na-angamiz-a pHsa 'he is losing money' (ku-angamiz-a, pesa)
a-na-vuw-a ma-zobe 'he is fishing for crabs' (ku-vuw-a, ma-zobe)
a-ka-jit-a vi-yogwe 'he has cooked sweet potatoes' (vi-yogwe)
a-ka-jit-a m-barazi 'he has cooked sp. beans' (m-barazi)
a-ka-jit-a dzungu 'he has cooked sp. pumpkin' (dzungu)
a-ka-jit-a ma-dzungu 'he has cooked sp. pumpkins' (ma -dzungu)
Each of the verb words in (47) has a H associated with it, and in every case
this H is one that is susceptible to HTD. Assuming that HTD is cyclic, the
H of the verb word will become associated with the last vowel of the verb
word on the first cycle. The noun words also have a H associated with them,
but these H's are restricted from undergoing HTD— thus at the end of the first
cycle this H will still reside on the penult vowel of these nouns. (I am
assuming that this H cannot be affected by Neut on the first cycle due to
the fact that at this point, the location of the noun in the phrase is
not yet determined. Neut affects just penult vowels in certain syntactic
positions.) After the first cycle, thenj we will have structures like the following:
105
H H
$ku-afun-a nazl$
H H
$a-na-jit-a ma-r^nje$
H H
$a-na-tsong-a gungahi$
H H
$a-na-vuw-a ma-zobe$
In order to account for the high-falling tone on the noun in examples
such as ku-afun-a nazi, we must appeal to the rule D-NV. We know that this
rule has the effect of associating a H with a neutralized vowel — and we
know that the nouns in (47) are subject to Neutralization. But if we are
going to appeal to D-NV, then it must be the case that D-NV is a rule that
operates on phrases and not just words. In other words, a H associated
with a verb can displace to a neutralized vowel in the following noun.
It is also clear from these data that D-NV must be a distinct rule from
D-S, since displacement is to the neutralized vowel not to the first stem
vowel in an example like a-na-tsong-a gunguhi or a-na-andik-a baruwa.
There is one significant problem with respect to invoking D-NV to
account for these data — and that has to do with the role of voiced obstruents
in inhibiting this displacement. It is clear from (47) that if the
neutralized vowel is preceded by a voiced obstruent, the displaced H ends
up on the vowel in front of the voiced obstruent. The vowel in front of
the voiced obstruent could be the final vowel of the verb, as in a-ka-jit-a
dzungu, or it could be a prefix vowel, as in a-na-vuw-a ma-zobe, or
presumably it could be another stem vowel (although I do not have a
relevant example at my disposal). The problem is that in section 3,
there was evidence that a voiced obstruent anywhere between the H and the
neutralized vowel stopped the movement of the H. My formulation of D-NV
in (28) reflected this inhibiting factor. But notice that in a-na-tsong-a
gunguhi and a-na-andik-a baruwa the voiced obstruent at the beginning of the
noun stem does not prevent the H on the verb from crossing over to the
neutralized vowel. In the case of nouns, it is only a voiced obstruent
that immediately precedes the neutralized vowel that causes the H to
associate with the vowel in front of the voiced obstruent. Why nouns and
verbs should behave differently in this respect, I have no explanation for,
and I have not attempted to reformulate D-NV so as to formalize this
difference in behavior.
The data in (47) provide additional cases where both HTD and HD apply
on the first cycle to the verb word — cf . ku-adz-a dzina and a-na-angaroiz-a
pesa. In these examples, the doubled H on the penult vowel of the verb
remains attached to that vowel while the H on the ultimate vowel of the
verb is displaced to the neutralized vowel (if it can be).
I turn now to the tonal pattern of phrases consisting of a verb word
with a single H in its underlying structure followed by a noun that contains
a H that undergoes HTD (and thus does not undergo Neut). Examples are
given in (48).
106
(48) ku-anik-a ma-pemba 'to put maize in the sun to dry'
(cf. ku-anik-a, ma-pemba)
ku-banang-a kande 'to spoil food'
(cf. ku-banang-§, k¥nde)
a-na-jit-a chi-kopwe 'he is cooking sweet potato leaves'
(cf. ku-jit-a, chi-kopwe)
a-na-vugur-a fundo 'he is untying a knot'
(cf. ku-vugur-a, fundo)
a-na-jit-a zodo 'he is cooking mangos'
(cf. ku-jit-a, zodo)
a-na-tsor-a chi-dafu 'he is picking up a young coconut'
(cf. ku-tsor-a, chi-dafu)
a-ka-fuj-a kande 'he has squandered food'
(cf. ku-fuj-a, kande)
a-na-ragiz-a chi-tabu 'he is ordering a book'
(cf. ku-ragiz-a, chi-tabu)
a-na-suw-a sahani he is washing plates
(cf. ku-suw-a, sahani)
a-ka-angamiz-a rumu 'he has lost a knife'
(cf. ku-angamiz-a, rumif)
ni-na-azim-a chi-karango 'I am borrowing a frying pan'
(cf. ku-azim-a, chi-karango)
'I am spitting blood'
(cf. ku-tem-a has an underlying H that is neutralized, damu)
ni-na-tsun-a nonzi 'I am skinning a sheep'
(cf. ku-tsun-a has an underlying H that is neutralized, nonzi)
The most striking fact about the data in (48) is that the H associated with
the verb in underlying structure has, for the most part, ended up on the
following noun. In particular, it has ended up on the first vowel of the
noun stem (hopping over the noun prefix in the process). The only cases
where the H of the verb has not reached to the first vowel of the noun stem
is when a voiced obstruent stands in the way. Thus (48) clearly involves
the application of the rule of Displacement-to-Stem; (48) also shows that the
rule must be able to apply to a phrase, not just to words.
Let me now illustrate how some of the items in (48) will be derived
under the analysis proposed in the present paper. Consider, first, cases
where D-S succeeds in displacing the H of the verb to the first stem vowel
of the noun.
107
(49)
$[[#d-na-jit-a#] [#chi-k(ipwe#] ]$
-jit-i
H
a-na
inapplicable
H
a-na-jit-a
chi-kopwe
inapplicable
H
chi-kopwe
inapplicable
inapplicable
inapplicable
inapplicable
inapplicable
""■~~^ T^ 5
a-na-jit-a chi-KBpwe
H^-^,__----.,.H
a-na-jit-a ch?^Tct)pwe
inapplicable
inapplicable
H-
a-na-jit-
first
cycle
HTD
HD
second
cycle
HE
HTD
Neut
HD
D-NV
D-S
HTS
FS
Rise-Fall
UFL
(I have assumed, arbitrarily, that in the case of the noun chi-kopwQ, the
underlying H is associated with the first vowel of the stem; I have no
real evidence for this — all that we know for certain about nouns such as
these is that their underlying H undergoes HTD and does not neutralize.)
The above derivation is quite straightforward and requires little in
the way of comment. A number of other phrases in (48) will undergo a quite
analagous derivation — e.g. ku-anik-a ma-pemba, ku-banang-a kande , a-na-vugur-a
fund^, and ni-na-tsun-a nonz^. In examples such as a-na-suw-a sahan^ and
ni-na-azim-a chi-karang5^ the only difference is that the rule of HTS
does more work, since it spreads the H that is displaced onto the first
stem vowel onto both the second stem vowel and also the final vowel.
Examples such as a-ka-fuj-a kande and a-ka-angamiz-a rumO differ from (49)
just in that the rule of HD is applicable to the verb on the first cycle,
doubling the H that has been displaced to the ultimate vowel onto the
penult vowel.
Consider next an example like a-na-raglz-a chl-tabu. The derivation of
this form is shown in (50) .
H H
(50) $[[#^-na-ragiz-a#] [#chi-t4bu#]]$ first cycle
I
a-na-ragiz- I
chi-tabA
HTD
? 3
a-na-raglz-d
H H
chi-tibA
HD
f 3
a-na-ragiz-a
H \
chi-tibii $
second cycle
inappl
icable
HE
inapplicable
HTD
inappl
icable
Neut
Inappl
icable
HD
inapplicable
D-NV
¥
108
H — — — — _H H
a-na-ragiz-a chi-tabu D-S
inapplicable HTS
inapplicable FS
inapplicable Rise-Fall
H H~-______^H L
a-na-ragiz-a chi-faba UFL
This derivation illustrates the case where, in the noun word, a voiced
obstruent precedes the ultimate vowel and thus triggers HD on the first
cycle. This process, combined with the rule of UFL, causes the noun to
be pronounced with a level high tone on the penult vowel. The penult
vowel here is also the first vowel of the noun stem. This means that
when D-S takes the H on the verb and associates it with the first stem
vowel of the noun, there will be no change in the pronunciation of the
noun since it already has a H on the first stem vowel.
In all of the cases so far illustrated, D-S succeeds in reassociatlng
the H of the verb with the first stem vowel of the noun. In (51), I
illustrate the case where the presence of a voiced obstruent prevents the
verb H from reaching the first stem vowel.
? H
(51)
$[[#a-na-tsor-a#][#chi-dku#]]$
first cycle
H H
a-na-tsor-a chi-dafu
HTD
inapplicable inapplicable
HD
$ a-na-tsor-a chi-dafu $
second cycle
inapplicable
HE
inapplicable
HTD
inapplicable
Neut
inapplicable
HD
inapplicable
D-NV
H--,_,__^ H
a-na-tsor-a cRl-dafu
D-S
inapplicable
HTS
inapplicable
FS
a-na-tsor-a cRl-dara
Rise-Fall
inapplicable
UFL
In this case, the presence of the voiced obstruent at the beginning of the
noun stem prevents the verb H from associating with the first stem vowel —
it must associate instead with the preceding vowel. This voiced obstruent
also prevents HTS from associating the H on the prefix with any following
vowels. Other examples in (48) which have a parallel derivation are
ni-na-tem-a damu and a-na-jit-a zodo (the latter form differing just in
that the noun has a voiced obstruent before the ultimate vowel and thus
triggers High Doubling and UFL rather than Rise-Fall).
109
A. 3. Phrases involving verb words with two high tones.
Verb words that contain two high tones in their underlying structure
include: (a) a high-toned verb stem in a 3p. present or future tense form;
(b) a high-toned verb stem in a -ka- or -chi- tense form; and (c) a low-
toned verb in the past tense. In (52) I illustrate phrases consisting of
a verb with two highs followed by a low-toned noun.
(52) a-na-chekech-a u-nga 'he is sifting flour
(cf. ku-chekech-a, u-nga)
^ V A ,
a-na-adz-a mu-tu he is mentioning s.o. by name
(cf. ku-adz-a, mu-tu)
a-na-tsfndz-a ma-dzogoro 'he is slaughtering roosters
(cf. ku-tsmdz-a, ma-dzogoro)
V A
a-na-ezek-a banda 'he is thatching a shed'
(cf. ku-ezek-a, banda)
a-ka-tsukijts-a chi-rond-a 'he has cleaned a wound'
(cf. ku-tsukuts-a, chi-ronda)
a-ka-tsukuts-a chf-donda 'he has cleaned a wound'
(cf. chi-donda is an alternative pronunciation to chi-ronda)
a-ka-fh-a mu-ganga 'he has called a doctor'
(cf. ku-ih-a is a verb with an underlying H that gets neutralized
when in the relevant environment; mu-ganga)
>/ A
a-na-banang-a n-guwo 'he is spoiling the cloth'
(cf. ku-banang-a, n-guwo)
V A
n-a-gwir-a dzogoro 'I caught a rooster'
(cf. ku-gwir-a, dzogoro)
n-a-henz-a mii-ganga 'I looked for a doctor'
(cf. ku-henz-a, mu-ganga)
n-a-pfg-a goma 'I beat a drum'
(cf. ku-pig-a, goma)
Let us consider first examples such as a-na-chekech-a u-nga and
a-ka-tsukuts-a ch£-r6nda. It seems quite apparent that one of the H's in
the verb word has displaced to the first stem vowel In the verb word and
spread onto all of the following vowels. But, if so, then it must be the
case that HTS is a phrasal rule rather than one that is limited to the word
since this spreading of high tones has gone not only all the way through the
verb but also all the way through the noun.
110
A possible derivation for a-na-chekech-a u-nga is given in (53)
H H
(53) $[[#i-na-ch4kech-a#] [#u-nga#]]$ first cycle
a-na-chekech-a inapplicable HTD
inapplicable inapplicable HD
$ a-na-chekech-a u-nga $ second cycle
inapplicable HE
a-na-chekech-a u-nga HTD
inapplicable Neut
inapplicable HD
inapplicable D-NV
a-na-chekech-a u-nga D-S
a-na-chekech-a u-nga HTS
inapplicable FS
inapplicable Rise-Fall
a-na-chekW!?^ ^-n^ UFL
This derivation yields the correct surface form, but it is not the only
possible way to achieve the correct result. In particular, it could very
well be that D-S is cyclic and applies to the verb word on the first
cycle. It could also be the case that HTS is cyclic. If so, it would
apply both on the first cycle (extending the H that has been displaced
to the first stem vowel onto the remaining vowels in the verb word) and
on the second cycle (extending the H in the verb word onto the vowels in
the noun) .
It seems to me that the simplest assumption is that all of the rules
are cyclic — although some of them may be inapplicable until a certain
phrasal level has been reached. In particular, rules like Rise-Fall,
Utterance-Final Lowering, and Neutralization are all applicable only
relative to the end of certain syntactic constituents (the nature of
which I have not yet fully explored). As a consequence, these rules will
not be able to apply until the point where these constituents are being
processed. One reason to consider Neut, particularly, to be in the cycle
(even though it can only apply once, at a certain phrasal level) is that
it must be ordered before the clearly cyclic rule of High Doubling.
Henceforth, I will assume that all of the rules participate in the
cycle. The derivation of a-na-chekech-a u-nga will now be as in (54).
Ill
(5A) $[[#d-na-chikech-a#] [#u-nga#]]
inapplicable
first cycle
inapplicable
:h-i
inapplicable
inapplicable
a-na-chekec
inapplicable
H H
a-na-chekech-a
H H
a-na-chlk^cTv^
inapplicable
a-na-ch4keclv4
inapplicable
a-na-chekech-a u-nga
inapplicable
u-ng.
inapplicable
;h4KtecR^a "Q-nga
inapplicable
inapplicable
inapplicable
u-nga
HE
HTD
Neut, HD, D-NV
D-S
HTS
FS, Rise-Fall, UFL
second cycle
HE
HTD
Neut, HD, D-NV, D-S
HTS
FS, Rise-Fall
UFL
There is one aspect of the derivation in (54) that requires comment —
namely, D-S must not apply on the second cycle to displace the H of the
verb off the verb and onto the following noun. I assume that the reason
that D-S does not apply on the second cycle to a H that has (on the first
cycle) been displaced to the verb stem is that D-S does not displace
across a stem vowel. More precisely, between the vowel to which the H
is linked and the vowel to which it is displacing, there can be no vowel
which belongs to a stem. This notion of not 'crossing' a stem vowel is
rendered slightly more complicated under the assumption that HTS is cyclic,
since when D-S attempts to apply on the second cycle in (54) , the H of
the verb word has gotten associated with three different vowels by virtue of
HTS. The input to D-S is thus:
tt^ H
$a-na-ch4kech-a u-nga$
Does a stem vowel intervene between the vowel that the H in the verb is
linked to and the following noun stem? Yes, if we consider the H linked
to the first vowel of -chekech-. No, if we consider the H linked to the
final vowel of the verb. Consequently, it is necessary to state that no
stem vowel may intervene between any of the vowels that the H is linked to
and the vowel to which the H is displacing.
Examples such as a-na-chekech-a u-nga are interesting in that they
can be used to provide another argument that HTD must be cyclic. Compare
the preceding example with a-na-suw-a sahani. Their underlying forms are
shown below:
H H H H
$a-na-chekech-a u-nga$ vs. $a-na-suw-a sah4ni$
Notice that both have two H's in their underlying structure. Suppose that
HTD is not cyclic, but rather applies just once to the whole phrase. Then
112
HTD will convert the above representations to:
H H H H
$A-na-chekech-a u-ngA$ $A-na-suw-a sahanl$
But now we have a problem. Why does the H of the prefix in the former case
displace to the verb stem, but in the latter example to the noun stem?
There is no readily available answer to this question. The problem does not
arise, however, if HTD is cyclic. In the case of a-na-chekech-a u-nga,
HTD will apply on the first cycle to the verb word and displace the stem
high to the ultimate vowel. The prefix H will not be affected by HTD.
In the case of a-na-suw-a sahani, HTD will apply on the first cycle to the
verb word and displace the prefix H to the ultimate vowel. But now we
have achieved the desired contrast. For in the former example, since the
H in the verb is on the verbal prefix, D-S can displace it only as far
as the first vowel of the verb stem (since to go any further would involve
crossing a stem vowel); while in the latter example, the H of the verb is
on the ultimate vowel, and thus can displace to the first vowel of the
noun stem (since this does not involve crossing any stem vowels) .
I conclude, then, that this contrasting behavior provides independent
evidence for the cyclicity of HTD.
Let me now return to the other data in (52). The derivation of a
number of these examples is entirely parallel to the derivation of
a-na-chekech-a u-nga except that HTS has been interrupted by a voiced
obstruent (and since HTS has been stopped from reaching the ultimate
vowel, it does not block the application of Rise-Fall). Examples of
this sort: a-na-tsindz-a ma-dzogoro, a-na-ezek-a banda, a-ka-tsukuts-a
chi-donda, a-ka-ih-a mu-g^ngS', and~a-na-adz-a m{f-tQ. In some cases the
H of the prefix has not displaced to the verb stem because of the fact t^at
a voiced obstruent precedes the first stem vowel — e.g.a-na-banang-a n-guwo and
n-a-gwir-a dzogoro. The derivations of these items are, however, entirely
parallel to the examples cited immediately above.
I turn next to phrases where a verb word with two H's is followed by
a noun which has a H that is subject to Neut. Examples appear in (55).
(55) a-na-ezek-a ny-umba 'he is thatching the house'
(cf. ku-ezek-a, ny-umba)
a-na-vwar-a chi-tambi 'he is wearing a piece of cloth'
(cf. ku-vwXr-a, chi-tambi)
a-na-Bah-a chi-tanda 'he is getting a bed'
(cf. ku-Bah-a is a verb which has an underlying H subject to
neutralization, chi-tanda)
a-na-afun-a chx-yogwe 'he is chewing a sweet potato'
(cf. ku-afun-a, chi-yogwe)
113
a-na-Birlk-a salamu 'he is sending greetings'
(cf. ku-Birik-a, salamu)
a-na-ez-a gwanda 'he is washing a shirt'
(cf. ku-ez-a is a verb with an underlying H subject to Neut, gwanda)
ni-ka-vundz-a ru-kuni 'I have broken a piece of firewood'
(cf. ku-vundz-a, ru-kuni)
ni-ka-6n-a simba 'I have seen a lion'
(cf. ku-on-a is a verb with an underlying H subject to
neutralization, simba)
^ A ,
n-a-jit-a ma-renje I cooked pumpkins'
(cf. ku-jit-a, ma-renje)
n-a-jit-a dzungu 'I cooked a sp. pumpkin'
(cf. dzungu)
n-a-tsong-a gunguhi I carved a bed leg
(cf. ku-tsong-a, gunguhi)
Clearly, the second H of the verb word in these phrases attempts to
displace to the neutralized vowel in the following noun. The first H of
the verb word attempts to displace to the verb stem. The derivation of
a-na-ezek-a ny-umba illustrates both of these observations.
H H H
$[ [#A-na-4zek-a#] [#ny-aml
(56) $[[#i-na-6zek-a#][#ny-iimba#]]$ first cycle
inapplicable
inapplicable
HE
a-na-ezek-a
inapplicable
HTD
inapplicable
inapplicable
Neut, HD, D-NV
H H
a-na-4zek-A
inapplicable
D-S
inapplicable
ny-Jmba
HTS
inapplicable
inapplicable
FS, Rise-Fall, UFL
H H
H.
a-na-izek-d
ny-amba $
second cycle
inapplicable
HE, HTD
H H
a-na-ezek-a
ny-umba
Neut
inapplicable
HD
H H^ ,
a-na-4zek-a
L
ny^timba
D-NV
inapplicable
D-S, HTS, FS, Rise-Fall,
UFL
The derivation of some other items in (55) is quite parallel to the above.
For example, a-na-vwar-a chl-taWbi, nl-ka-vundz-a ru-kuni, and n-a-jit-a
ma-renje differ just In that the first H of the verb cannot displace onto
the first vowel of the verb stem due to the presence of a voiced obstruent
in front of that vowel. n-a-ts6ng-a gunguhi differs just in that when the
first H of the verb displaces to the verb stem it is able to spread onto
the next vowel (whereas in a-na-ezek-a ny-umba the voiced obstruent after
the first stem vowel prevents HTS from applying).
114
Consider next the derivation of an item such as a-na-Birik-a salamu.
(57)
a-na-Birik-a#
H
] [#sal^niu#]]$
first cycle
inapplicable
inapplicable
HE
a-na-Birik-c
inapplicable
HTD
inapplicable
inapplicable
Neut, HD, D-NV
a-na-Birik-a
inapplicable
D-S
a-na-Birik-a
salamu
HTS
inapplicable
inapplicable
FS, Rise-Fall, UFL
a-na-Blrik-a
saldmu $
second cycle
ina
pplicable
HE, HTD
a-na-BirlkPa
salamb
Neut
inapplicable
HD
a-na-Birlk^
salamu
D-NV
inapplicable
DS
\l
— __^ HL
a-na-Blrlk-a
salamu
HTS
a-na-Birik-a saiamu
inapplicable
FS
Rise-Fall, UFL
The crucial fact about this derivation is that HTS spreads the first H
of the verb word up to the neutralized vowel, thus setting up the environment
for Fall Simplification. This means that the neutralized vowel will be
pronounced with a level low tone, the preceding vowel having a level high
tone. Similar examples: a-na-Bah-a chi-tanda, a-na-afun-a chi-yogwe,
ni-ka-6n-a simba.
In examples like a-na-ez-a gwanda and n-a-jit-a dzungu, the second
H of the verb word cannot displace to the neutralized vowel because it is
preceded by a voiced obstruent, thus the H remains on the final vowel of
the verb. The first H of the verb displaces to the first stem vowel
in the former example but is barred from leaving the prefix in the latter
example (due to the voiced obstruent at the beginning of the verb stem) .
In neither example can the first H in the verb spread since it is followed
by a voiced obstruent. Neither can the second H of the verb spread, for the
same reason. Since neither of the H's in the verb can associate with the
neutralized vowel, that vowel will be pronounced on a low tone.
I will conclude this section by examining phrases where there is a
verb word with two H's followed by a noun that has a H that undergoes HTD
(and thus is not susceptible to Neut). Examples are given in (58).
115
(58) a-na-an£k-a ma-pemba
(cf. ku-anik-a, ma-pemba)
'he is putting maize in the sun to dry'
a-na-tsun-a nonzi 'he is skinning a sheep'
(cf. ku-tsun-a is a verb with an underlying H subject to neutrali-
zation, nonzi)
a-na-vug-a dJ^na 'he is cooking hard maize porridge'
(cf. ku-vug-a is a verb with an underlying H subject to neutrali-
zation, d^na)
n-a-angamlz-a tana 'I lost the bow'
(cf. ku-angamiz-a, tana)
V A
(cf. ku-vugur-a, fundo)
'I untied the knot'
n-a-tsukur-a ma-pemba 'I carried the maize'
(cf. ku-tsukur-a, ma-pemba)
The derivation of a-na-anik-a ma-pemba is illustrated in (59) .
H H H
(59) $[[#a-na-Anik-a#] [#ma-pemba#]]$ first cycle
inapplicable inapplicable HE
|1 H 1^
a-na-anik-a ma-pemba
inapplicable inapplicable
H H
a-na-anik-a inapplicable
a-na-anik-a inapplicable
inapplicable inapplicable
IL H H
a-na-anik-a ma-pemba $
inapplicable
*? H
ma-pemba
i-fnlk-a
a-na-anlk-a ma-pl
Inapplicable
HTD
Neut, HD, D-NV
D-S
HTS
FS, Rise-Fall, UFL
second cycle
HE, HTD, Neut, HD, D-NV
D-S
HTS
FS, Rise-Fall
UFL
The derivations for a-na-tsun-a nonzi and n-a-tsukur-a ma-pemba are quite
parallel. Consider next the derivation of n-a-angamiz-a tan^.
H H H
(60) $ [ [#ni-A-angamiz-a#] [#tana#] ] $ first cycle
inapplicable inapplicable HE
H H U
nl-a-angamiz-a tanA HTD
inapplicable inapplicable Neut
116
nl-a-ane^mlz-a
inapplicable
inapplicable
H H H
ni-a-angamiz-a inapplicable
inapplicable inapplicable
ni-a-angamiz-a inapplicable
ni-a-angamiz-a
inapplicable
H
ingamlz-a tana
inapplicable
tLH H H
ni-a-an^amlz-a tana
ni-a-Angmnlz-a tdna
inapplicable
ni-a-Sngamiz-a tana
HD
D-NV
D-S
HTS
FS, Rise-Fall, UFL
second cycle
HE, HTD, Neut, HD, D-NV
D-S
HTS
FS, Rise -Fall
UFL
The differences between the derivation in (60) and that in (59) can be
attributed to the presence in the former case of a voiced obstruent before
the ultimate vowel of the verb. This voiced obstruent induces the application
of HD on the first cycle and also constitutes a barrier to the spreading of
high tones.
Next consider the case of a-na-vug-a dona.
(61)
H 1^ U
*a-na-vug-a#] r#dor
$ [ [#A-na-vug-a#] [#d6na# ] ] $
inapplicable inapplicable
H
dona
inapplicable
inapplicable
inapplicable
H H
a-na-vug-a
inapplicable
H H
a-na-vug-a
inapplicable
inapplicable
dona $
inapplicable
H H H
nA-vug-a dona
inapplicable
H H VJL^
na-vug-a dona
inapplicable
first cycle
HE
HTD
Neut, HD, D-NV
D-S
HTS, FS, Rise-Fall, UFL
second cycle
HE, HTD, Neut, HD, D-NV
D-S
HTS, FS
Rise-Fall
UFL
In this derivation, on the first cycle D-S can displace the first H of the
verb only as far as the vowel before the stem, since the stem begins with
a voiced obstruent. This voiced obstruent also prevents the spreading of
the high tone. The verb -vug- is one that is exempt from HTD, thus that
rule fails to apply to the verb word on the first cycle. There is an
important point to note about this derivation. The H on the vowel of
-vug- is able, on the second cycle, to undergo D-S and land on the vowel
that precedes the noun stem (it cannot land on the first vowel of the noun
stem since the initial consonant of the stem is a voiced obstruent). But
in undergoing this displacement to the final vowel of the verb, a voiced
obstruent is crossed. This application of D-S across a voiced obstruent is
117
otherwise unattested. Somehow, the fact that the H of -vug- crosses the
^ of the verb, but not the d^ on the noun, must have to do with the fact
that this H is one that is exempt from HTD and subject to Neut. But how
these facts are connected is not clear. I suspect that an understanding
of this sort of case will have ramifications for the over-all analysis
of Digo, but it must at the present time remain a mystery.
The only example in (58) that I haven't yet discussed is n-a-vugur-a
fundo. This item represents another case where the first H of the verb
word is unable to reach the first stem vowel due to the presence of a
voiced obstruent at the beginning of the stem. The second H of the verb
is, however, able to displace to the following noun stem.
The present section represents just the beginning of an analysis of
the phrasal tonology of Digo. There are doubtless many complexities yet
to be encountered. But the data analyzed here have been quite complex
in themselves; however, I believe that the analysis of Digo tonology
that 1 have presented here makes sense out of these very complicated facts
and represents a real insight into the structure of Digo tone.
5. Conclusion.
The Digo data analyzed in this paper provide, I believe, interesting
confirmation for the view of phonology that is generally referred to as
"generative phonology". According to this view, surface phonetic shapes
are sometimes the consequence of the application of a set of phonological
rules applying (in an ordered fashion) to representations which may differ
in significant respects from the phonetic shapes. Notice that the rules
and representations which I have postulated in this paper are entirely in
accordance with this view. For example, I have argued that there are many
items which have a high associated with their penult vowel underlyingly —
e.g. ku-som-a 'to read' (cf. ku-somer-a 'to read to') and ny-umba 'house'
(cf. nyumb?-ni 'in the house') — where this H is never associated with the
vowel in question at the phonetic level. Indeed, I have shown that where a
H appears on the phonetic surface has little direct connection with where
this H originates In underlying structure — e.g. a H that appears phonetically
on a noun may originate in the preceding verb. I have also argued that a
fairly complex set of rules (operating not just on words, but on phrases
as well) mediates between the underlying forms and the surface structure.
There can be no question about whether these rules (in some form) are
"real", for since they operate on phrases consisting of any verb followed
by any noun, they can readily be shown to be fully productive and
necessarily a part of the native speaker's knowledge of Digo.
In addition to supporting the viewpoint of generative phonology, the
Digo data also provides significant evidence in favor of one particular
approach to generative phonology — namely, the autosegmental approach.
There can be no question but that tones in Digo are quite separate from
the segments that bear them. Tones move from one word to another — quite
unlike the segments that manifest the tones.
118
Finally, Digo provides significant evidence in favor of the cyclic
application of phonological rules. The question of cyclicity has been
discussed for many years in generative phonology, but most of the
crucial evidence for the cyclic application of phonological rules has
come from stress phenomena. Digo tone provides a significant new area of
evidence in favor of cyclic application of rules. I do not believe,
however, that the cyclic nature of the rules in Digo can properly be
understood until much more analysis has been done on the interaction of
tone and syntax in Digo.
NOTES
*I would like to express my gratitude to the University of Illinois
Research Board and to the African Studies Program of the University of
Illinois for providing funds that helped make possible the research on
Digo tonology. I would also like to thank the National Science Foundation,
whose Grant No. BNS-7924523 made the writing of this paper possible.
Professor Chin-Chuan Cheng has assisted me at every stage of my work on
Digo and my debt to him cannot easily be repaid. I have discussed Digo
tonology with many people over the past couple years and would like to
thank them all for the stimulation that they have given me; I would
particularly like to thank Nick Clements, John Goldsmith, and Larry Hyman
for the considerable encouragement and discussion that they have provided
in the past few months. My greatest debt, of course, is to my consultant,
Mr. Mfundo Jumaa Mfundo. I can only hope that I shall have the opportunity
to work with him again on this most fascinating language.
■""It should be noted that mb, nd, and n^ (=[Qg]) are treated in Digo
like pre-nasalized stops and they do not behave as though they contain a
voiced obstruent.
2lt is not clear that in the statement of D-S there is any need to
refer to a H that is followed by another H (even though in all cases where
this rule actually applies, such a following H exists), since — after HID
displaces the rightmost H in the word to the ultimate vowel — the only H's
that could move further to the right are ones that do (or that are blocked
from doing so by a voiced obstruent).
■^There are some apparent counterexamples to this claim. There are a
few verbs of the shape -C^-a; e.g. ku-ty-a 'to obey'. The 3 sg. present
tense form of this verb is a-na-ty-a, which suggests that this stem has
a H associated with it, even though the root -ty- has no underlying vowel
(at least, I have found no evidence for such a vowel). Somehow, this H
must be assigned to the penult vowel of the word at some point in the
derivation before Neut applies. The penult vowel in this case is the tense/
aspect prefix -na-. Thus on the surface we have a prefix vowel apparently
undergoing Neut, but the H actually originates in the stem.
See footnote 3 for some discussion of the exceptional cases where the
stem does not have a vowel.
119
I will show later in the paper that in some cases displacement to the
neutralized vowel is halted by a voiced obstruent anywhere between the H
being displaced and the neutralized vowel, whereas in other cases it is
only a voiced obstruent immediately before the neutralized vowel that halts
the shift of the H toward the neutralized vowel.
°It seems to be a very general fact about Bantu grammatical structure
that object prefixes are closely bound to the verb stem. For example, in
certain tenses in Mak.ua (a Bantu language of Tanzania and Mozambique) , a
high tone is assigned to a particular vowel in the verb stem — and an object
prefix "counts" as part of the stem for the purpose of this tone assign-
ment process.
The phenomenon of a high tone being lowered after a high tone is
quite pervasive in Bantu tonology, but I have found no evidence in Digo
that the rule has any wider scope of application than the context defined
in the text.
^K&W note that in utterance-final position, all-low words are
characterized by a mid-falling pitch on their penult vowel. The situation
is actually somewhat more complicated. When our consultant pronounces words
of this sort in isolation (under elicitation conditions), the nature of the
fall on the penult vowel varies considerably. In particular, in short
words (where the penult vowel is the first or second vowel of the word),
the fall may be quite high. In longer words, the fall is not nearly so
high. Consequently, the investigator might, upon first hearing all-low
words, confuse some of them with words that have a high-fall on
their penult vowel (due to the application of Neutralization and Displacement-
to-Neutralized Vowel). However, the difference between these two types of
words is clear when they are juxtaposed with one another.
^It seems likely that kunde-ni is a reflection of the structure:
kunde-nl
where the L on the ultimate vowel derives from a H via Utterance-Final Lowering.
Similarly, it is quite likely that z6d?^-ni derives from the structure
IW
where the rising-falling sequence on the last two vowels Is the result of
the application of Rise-Fall to an ultimate H. The problem is: given that
the noun has a H associated with it, but the -ni does not (apparently), where
does the second H come from? In section 3 some apparently related data are
cited.
lOxhe examples in (11) suggest strongly that the rules of Rise-Fall and
Utterance-Final Lowering are not actually triggered just at the end of an
utterance, but also in certain other configurations (since the data in (11)
clearly reflect the application of Rise-Fall and UFL to the noun, which is
obviously not at the end of an utterance). The question of the precise range
of contexts in which these two rules apply cannot be answered without a
detailed study of Digo sentence structure.
120
■'-■'■Incorporating this constraint into D-S would require that Neut be
applied before D-S, since Neut is the rule that derives the L on the
penult vowel.
-'•^The alternative to allowing both H's to associate with the first
stem vowel via D-S would be to postulate a rule that would have the effect
of deleting one of the H's. Recall, however, that in the case of high-toned
verb stems that are subject to Neutralization, no deletion of a H occurs
in the past tense. Consequently, I have opted for the analysis presented in
the text, which simply allows both H's to associate with the stem via D-S.
■'-■^The following example illustrates that the bilabial continuant sound
that I have symbolized with the letter B does not function as a voiced
obstruent in Digo, but rather as a sonorant.
l4
This is not precisely true, from a phonetic point of view. Recall
from footnote 8 that all-low words have a falling pitch on their penult
vowel; this fall may be fairly high when the penult vowel is the first
vowel of the word, but noticably less high when several syllables precede
the penult. Thus the isolation form of the nouns in (41) and (42) may
be pronounced with a considerably higher fall than would be present in
cases where the noun is pronounced as part of a phrase.
This derivation illustrates why I have formulated HD so that it
results in two H's in the tonal tier. Suppose that instead I had formulated
the rule so that it added an association line between a H and the penult vowel
provided the H is associated with an ultimate vowel that is preceded by a
voiced obstruent. Given such a formulation of HD, on the second cycle
HTD would apply to that H and associate it with the ultimate vowel of the
phrase (disassociating it from the vowels that it had previously been
associated with) . The result would be the loss of the required high tone
on the penult vowel of the verb.
■'■^I should note here that in the account I have given of the interrelation-
ship between HTD and Neut , I have relied heavily on HTD being formulated so
as to move a H to the ultimate vowel. H's located on the ultimate vowel
will escape Neut, which affects only H's associated with the penult. Thus
if this account is correct, we have a strong motivation for assuming that
HTD reassociates a H with the ultimate vowel rather than the penult vowel
(as assumed in K5W) .
REFERENCES
BECKER, Lee A. and David P B. Massamba. 1980. CiRuri tonology (a pre-
liminary view). Studies in the Linguistic Sciences 10.1, 1-13.
GOLDSMITH, John. 1976. Autosegmental phonology. Bloomington, Indiana:
Indiana University Linguistics Club.
GOLDSMITH, John. 1981. Towards an autosegmental theory of accent: the case
of Tonga. Bloomington, Indianai, Indiana University Linguistics Club.
GUTHRIE, Malcolm. 1967-71. Comparative Bantu. Farnborough: Gregg
International Publishers, Ltd.
KISSEBERTH, Charles W. and Winifred J. Wood. 1980. Displaced tones in
Digo (part 1). Studies in the Linguistic Sciences 10.1, 141-177.
Stuiies in the Linguistic Sciences
Volume 11, Number 1, Spring 1931
A STUDY ON CODS-SWITCHING IN TAIWAN
Yen Ling Lee
This study focuses on the investigation of situational
selection of codes (Mandarin versus S®uthern Min) in Taiwan
using the theoretical construct of 'domain' proposed by
Fishraan. Eight domains are adopted in this study: family,
friendship, neighborhood, internal, employment, religion,
education, and administration. A carefully constructed
survey, comprised of questions on the choice of language
variety in different domains, was distributed to a randomly
selected sample of the Mandarin-Southern Min students from
Taiwan at the University of Illinois. The results show:
1 ) Language choice of Mandarin versus Southern Min is
by no means random. A general pattern can be
captured through domain analysis. The use of
Mandarin is predominant in the friendship,
internal, employment, religion, education, and
administration domains; it also has restricted use
in the family and neighborhood domains. The use of
Southern Min is correspondingly predominant in the
family and neighborhood domains; it has restricted
use in the friendship and religion domains.
2) Within each domain, factors which influence
language choice can be isolated such as age of
participant, seriousness of topic, urban versus
rural setting, and nature of discourse (degree of
authority and distance) .
3) The bilingual situation in Taiwan is an unstable
one. There is a shift toward the predominance of
Mandarin as evidenced by its growing use in the
intimate domain.
0. Introduction
In a speech community having a verbal repertoire of more than one
language variety, the choice of one language variety in a given situation
versus the choice of another is far from random. From a communicative
standpoint, each language is sufficient on its own; therefore, the very
co-existence of more than one language in a society over a period of time
implies that each serves a specific function to a certain extent. The
motivation for such functional allocation may be to maintain culture
identity or to convey certain social values as in the case of Norwegian
dialects Ranamal and Bokmal described by Guraperz and Blom (1972). In the
past several decades, linguists have been increasingly interested in the
functional aspects of linguistic codes and the patterning of usage in
multilingual societies (See e.g. Guraperz and Hymes 1972, Hyraes 1964,
Ferguson 1964, Kachru 1978 a, b) . And such studies are essential toward
the understanding of the use of language in a larger societal context. My
interest in this paper is to investigate the sociolinguistic pattern in a
bilingual society - Taiwan. The phenomenon I am particularly interested in
is that of switching: the alternate use of language on the basis of
122
situation and participants (Gumperz and Blom 1972:415) • Some of the
questions I try to answer are: Is there a pattern governing language
choice? If so, what are the spheres of activities that dictate the use of
one versus the use of another? To what extent does each language variety-
serve a different function? How does the language planning effort of the
government affect the language choice? Is there a stable bilingualism in
Taiwan, a diglossia in the Ferguson sense (1964:4'50), or is it a situation
of unstable bilingualism? If it is in the process shifting, to what extent
and in what direction?
Firstly, I will describe the relevant historical and sociolinguistic
background of Taiwan which led to the present state of bilingualism.
Secondly, I will present my study, which takes the form of a survey, on
code-switching in Taiwan. Finally, I will try to analyze the results and
form generalizations on the pattern and nature of Taiwan bilingualism.
1 . Historical and Sociolinguistic Background
1 . 1 Ethnic and linguistic groups in Taiwan
Taiwan is an island off the coast of the southeastern part of mainland
China. The communication between mainland China and Taiwan can be dated
back to the 7th century. However, the first large influx of mainlanders to
Taiwan occurred in the 14th century. Jhe majority were Southern Min
speakers and the rest were Hakka speakers. In 1685, during the Qing
Djmasty, Taiwan became a province of China. In the following 200 years,
mainlanders, mostly Southern Min speakers, continued to settle in Taiwan.
In 1895, Taiwan was ceded to Japan. The Japanese occupation lasted for 50
years until the end of the Second World War at which time Taiwan was
returned to China. In 1949, at the end of the Chinese civil war, the
Nationalist government moved to Taiwan. Along with it came approximately
one million mainlanders comprised of soldiers, government employees,
teachers, merchants and intellectuals. Today, in addition to the 234
thousand aboriginal Malayo-Polynesian speakers, the population in Taiwan
(17 million) consists of 86.4"? old settlers, those whose ancestors settled
before the 20th century, and 13. 6*? newcomers, those who came around 1949
(Xue 1977:338).
The linguistic situation of Taiwan is a complex one. In addition to
Mandarin, the Standard Language, there are various dialects: Southern
Min, Hakka, as well as other Central and Northern dialects. All of these
are distinct entities originated from one language but evolved into
mutually unintelligible dialects. The majority of the population (old
settlers) speaks the Southern Min dialect. The minority (new settlers)
speaks Hakka, Yue , Northern Min, Wu , Can and Northern dialects. The
interest of this paper is focused on the interplay between the Southern Min
dialect and Mandarin.
1.2 Language planning in Taiwan
The Nationalist government started a language standardization campaign
on the mainland in the 1920's. It promoted Mandarin, a Northern dialect,
as the Standard Language. This decision was based on the facts that
Northern dialect was considered the dominant language spoken by 10% of the
123
Han Chinese (Yuan 1960: 22) , that Peking had historically been the political
center of China and that a large wealth of colloquial Chinese novels were
written in Peking speech.
In 1944, the Ministry of Education formulated 5 guidelines for the
promotion of the Standard Language (Fang 1965:130):
1 ) Implementing the standardization of pronunciation of Mandarin
2) Promoting Mandarin as the National Language and basis for wider
communication, both nationally and internationally
"5) Promoting the National Phonetic Symbols as supplement to the
Chinese characters to eliminate illiteracy
4) Promoting the National Phonetic Symbols for the communication
between Han Chinese and the minority ethnic groups
5) Improving the Mandarin teaching methods in schools
When the Nationalist government moved to Taiwan, the promotion of the
Standard became one of the urgent tasks. Since the general population had
been educated in Japanese, another essential guideline was included: to
recover the status of the Southern Min dialect in order to bridge the way
to the full use of Mandarin (Fang 1965:130).
The campaign on the promotion of Mandarin in Taiwan has been vigorous.
The medium of instruction is Mandarin. The mass communication media, radio
and T. v., include some Southern Min dialect programs but the prime-time
programs are mostly in the Standard Language. Although the folk arts such
as local operas and songs are kept in the dialect fonn and are still
popular, the entertainments in Mandarin enjoy even more widespread success.
1.3 Language attitudes
To understand the pattern in language use, the psychological factors
cannot be ignored. The individual attitudes toward the language may need a
detailed, controlled study. However, the sentiments on a larger scale can
be abstracted from trends in the past. There are two forces relevant to
the discussion of Taiwan bilingualism: nationalism and regionalism.
Depending on which force is predominant, the resulting choice of language
can be different.
Language loyalty is defined by Weinreich as an emotional involvement
with a particular language when one feels that language is threatened and
needs to be defended (Weinreich 1953:99). Fishman (1966:452) also made the
parallel observation that language loyalty is primarily a defensive
mechanism in reaction to the presence of linguistic competitions. The
conscious efforts of different ethnic groups to maintain their cultural and
linguistic identity begin only after facing the danger of displacement as
in the case of different American immigrant groups. Weinreich also
correctly observed that "a group's language loyalty and nationalistic
aspirations do not necessarily have parallel goals" (Weinreich 1953:100).
In the case of Taiwan, at a time right after the Japanese occupation, it is
obvious that language loyalty and national aspirations had congruent goals.
The success of the language standardization measures has to be attributed
to the fact that the population at large was feverantly anti-Japanese, and
was eager to identify with the National language. The people in Taiwan
were more receptive to the teaching and the use of the Standard Language.
124
Now, however, more than 30 years after the launching of language
standardization in Taiwan, one observes a rise of regionalism. There has
been an increase in the number of the Southern Min dialect programs on T.
V. and there have been quite a few writers incorporating dialect features
in their work. I view this again as a manifestation of language loyalty.
This time, it is directed toward the dialect since many feel that its
existence is threatened by the vigorous promotion of the Standard Language
(Wu 1972:398). A recent trend of decrease of percentage of Hindi speakers
in certain Indian regions also exemplifies this universal phenomenon of
regionalism versus nationalism (See e.g. Y. Kachru and Bhatia 1979).
Given the interplay of these two forces, nationalism on the one hand
and regionalism on the other, I have also observed an ambivalent attitude
among the Southern Min speakers. Many express the view that a standard
language is essential for national unity but the dialect cannot and should
not be eradicated. While the importance of Mandarin in wider communication
is recognized, many nonetheless express pride in the use of the dialect
because it is rich in archaisms. All in all, the Standard Language seems
to occupy the primary position.
2. Study on Code-switching
Having seen the historical and social factors which help to explain
the present state of bilingualism in Taiwan, I will now go into the
investigation of locating social variables which are determinants in the
patterning of individual usage. If such variables can be abstracted, one
can determine not only the general pattern of usage, i.e. who speaks what
to whom and to what end (Fishman 1972), but also the nature of Taiwan
bilingualism, i.e. whether it is stable or not.
2.1 Theoretical framework
In the past, linguists have isolated many dimensions to characterize
the alternate use of language varieties. The important consideration here
is to form an analysis that not only captures the social level of
regularity but also takes into account the individual fluctuations in
speech (Fishman 1972:247). Oilman and Brown (i960) use two dimensions to
describe the use of personal pronouns 'tu' and 'vous' in several European
languages. They forroulate in terras of sociopsychological relationship
between the speakers: 'power' and 'solidarity'. Similarly, Rubin
(1953:109) in her study on bilingualism in Paraguay isolates a hierarchy of
sociopsychological factors in determining the appropriate language usage
between Spanish and Guarani. The factors ordered according to importance
are: 1) location, 2) formality of interaction, 3) degree of intimacy of
interaction, and 4) degree of seriousness of discourse. There are also
additional factors: school pressure and estimate of the linguistic
proficiency of the addressee. Hyraes (1964) and Ervin-Tripp (I964) also use
the participant, ecological surroundings, and topic as situational
constraints in bilingual communication.
Along the same line as the linguists above, Fishman (1964, 1971, 1972)
formulated the concept of 'domain' to describe the larger and more abstract
societal-institutional context of bilingual interactions. It is an
abstract construct derived from the three components: I) locale of the
125
interaction, 2) role relationship of the speakers, and "5) topic of the
discourse. The purpose of this theoretical construct is to relate the
specific language choice to specific spheres of activity which are
determined by the sociocultural dynamics of the community (Fishman
1972:441).
Since domain is a sociolinguist's summarization of the relevant
societal-institutional contexts of a given community, the domains of one
community may differ from those of another community in number as well as
kind. Schmidt-Rohr( Fishman 1972:441) came up with 9 domains: l) family,
2) playground and street, 3) school, 4) church, 5) literature, 6) press, 7)
military, 3) courts, 9) government administration and Greenfield (1972:23)
in his study on Puerto Rico bilingualism isolated 5 domains: l) family, 2)
friendship, 3) religion, 4) education, 5) employment. The domains are
related to the tripod components role, locale, topic in that each domain is
a composite of the three. For example, the family domain includes
parent-child relationship (domain appropriate person), talk about family
matters (domain appropriate topic) at home (domain appropriate locale).
Even though the set of domains may vary from community to community,
it should not only capture the relevant relationship between language
choice and institutional contexts, but also reveal the pattern of
bilingualism.
Fishman (1968) postulated that typically, in a diglossic community,
the L variety (local, dialectal) is associated with family and friendship
domains whereas the H variety (superimposed, standard) is associated mainly
with the education, occupation and religion domains. In the case of stable
bilingualism, there is a clear division of labor in choice of language for
different domains whereas in the case of unstable bilingualism, the
alternate use of the two languages can occur even in a more intimate and
stable domain such as that of family and friendship. Thus the domain model
is a useful one both to describe bilingual usage and to characterize the
direction of shift.
2.2 Survey
2.2.1 Format
Having the domain analysis in mind, I tailored my study on Taiwan
code-switching to follow the same framework. My goal is firstly to
determine generally what dialect is spoken in what domain and secondly to
determine whether the bilingualism is a stable one using the domain
analysis.
The study takes the form of a self- report survey including questions,
in multiple choice format, on a range of domains. Since domain is
constructed around the cultural, societal values of the community derived
from active participant observation and on a large part based on the
investigator's intuition, I generalize 8 domains which I feel are relevant
for the Taiwan bilingual society: l) family, 2) friendship, 3) education,
4) neighborhood, 5) employment, 6) administration, 7) religion, 8) internal
speech. Within each domain, T list several typical role relationships
(e.g. in the family domain: parent-child), locales including public as
126
well as private (e.g. on the bus, at home, etc.) and ranges of topics (e.g.
politics, academics, etc.). An example of the questions asked is as
follows :
What language do you use when speaking to your parents at home?
1) All Southern Min dialect 2) Mostly Southern Min dialect 3) Both
4) Mostly Mandarin 5) All Mandarin
The list in the questionnaire is by no means exhaustive. But such an
exhaustive list covering all social interactions is very difficult to
compile. To remedy the possibility of overlooking information, 1 also
include questions and spaces for the subject to supply situations to
supplement the ones I have already given.
The questionnaires also include demographic background questions:
age, sex, educational background, place of birth, etc. This information is
relevant because very possibly the individual use of language can be
correlate with his personal background.
A question on language proficiency is also included. Although the
level of language proficiency for both languages does affect the choice of
which language to use, this is not the main concern of this paper. The
major point is to determine the appropriate social context for the choice
of one dialect versus another. Thus the relative proficiency in the
language only serves as a checkpoint so that a person's avoidance in using
a dialect due to a lack of proficiency in that dialect should not be
mistaken as related to a social variable. For the same reason, I also ask
the subjects to make the assumption that all addressees (other than family
members) know both varieties equally well. So again the level of language
proficiency of the addressee does not come into play as determinant of
dialect choice in a given context.
A total of 40 questionnaires were sent out to a randomly selected
sample of Taiwan students. 21 were returned, 13 of which were from male
respondents and 8 from female respondents. Because all subjects belong to
Taiwan regional club here on the University of Illinois campus, they are of
Southern Min parentage and are bilinguals of varying degrees. They are all
graduate students here, and thus similar in level of education. All of
them had done their undergraduate work in Taiwan.
A final word can be said about the format of this survey. The main
drawback of using surveys of this nature in linguistic study is that many
bilinguals may not consciously know what variety he uses in a given
situation. His linguistic attitudes may affect his report on his own
language usage. Also the data collected in this manner may reflect his
projection or recollection of a given situation which may differ from a
spontaneous social context. For these reasons, it should be followed up
with studies using other methodologies. These drawbacks notwithstanding,
this study provides some promising generalizations.
2.2.2 Results
(See Table 1 )
2.2.3 Discussion of the results
127
A. In the family domain (Questions 1-7), the Southern Min dialect is
the predominant language. However there are some exceptions: to the
spouse (Question 5) and to the children (Question 7). In the case of the
spouse, there is a slight majority favoring the use of Mandarin. And to
the children, the majority indicate the use of both. The age factor seems
to play a role here. The younger the addressee, the more that person is
identified with the modern trend, which is manifested as speaking to him in
Mandarin. This seems to be reconfiimed by the answer to Question 9 where
the majority indicate usage of Mandarin in talking to a younger neighbor.
The effect of wider Mandarin education however may also be a factor here.
B. In the neighborhood domain (Question 9-15), the Southern Min
dialect is used most frequently. However, one can make finer distinctions
in that there is a significant decline from the number of people who would
speak Southern Min dialect to the market salesman, to those who would speak
Mandarin to the department store clerk (See Table 2). One may suspect that
the position of locale in the urban-rural spectrum contributes to the
choice of language. Markets and groceries are usually located a few blocks
away from the house, thus the language of the neighborhood. Southern Min
dialect, is used predominantly. However when one moves farther away from
the home and closer to the urban center where department stores and taxi
drivers are found, the language of preference shifts considerably in the
direction of the Standard.
C. In the education domain (Question 16-18) where the language
standardization effort was found to be the most rigorous, it is not
surprising to see that the language used most is Mandarin. In particular,
when the addressee is the teacher (Question 17), the language used is
overwhelmingly Mandarin, regardless of the locale of interaction. In the
case where the addressee is a schoolmate, the language use fluctuates, with
notably more use of dialect when the locale is the home (Question 1 6b) .
D. In the employment domain (Questions 19-21), the majority of people
indicate Mandarin as the language most often used . It is interesting to
note that at work and outside of work, the use of Mandarin in speaking to
the superior (Question 19) and to the subordinate (Question 21) is slightly
more than that to coworkers (Question 20). This may be evidence indicating
that Mandarin is used to signal authority and distance. However, in the
sphere of employment, the boss-friendship boundary is not clear-cut. This
may explain the even split of usage between the two varieties when one
speaks to the boss outside of work (Question 19b).
E. The hypothesis that Mandarin is used to signal authority and
distance is further supported in the administration domain (Questions 22,
24, 25 and 26) and in the case where the addressee is a stranger (Question
23). In all these cases Mandarin is predominantly used.
F. In the friendship domain (Question 8, 27 and 28), with the
exception of gossiping (Question 27e) and conversation at a regional club
(Question 28), Mandarin is the favorite language of interaction. In both
cases, the use of Southern Min dialect may be to indicate an 'in-group'
feeling. Answers to Question 3 indicate that although one uses Mandarin
throughout the dating process, there is an increase of use of Southern Min
dialect as the couple become closer.3 This is congruent to the observation
128
above that Mandarin is used to signal authority and distance.
Symmetrically, the Southern Min dialect is used to signal intimacy.
Answers to Questions 27a-g indicate that the seriousness of the topic is
directly correlated to the frequency of Mandarin usage. The topics high on
the seriousness scale such as national affairs, academics and work all
induce high usage of Mandarin. For the topics which are less serious such
as joking and anecdotes of what happened at home, there is a fairly even
split of usage slightly favoring Mandarin.
G. In the religion domain (Questions 29-34), there are two tendencies
in language use depending on the religion. In the case of Christianity as
signaled by the use of 'church' in the question (Questions 29-31), Mandarin
is favored. In the case of local religion. Buddhism, the Southern Min
dialect is favored. One explanation for this phenomenon may be that
Christianity is associated with urbanization, westernization and modernism
and the H variety is usually tied with these values. Furthermore, Buddhism
is a local religion deeply rooted for centuries and thus identified with
the L variety. It is interesting to note that in the sphere of religion,
locale seems to have precedence over role relationship : speaking to a
priest, to a friend, or to oneself does not have much effect on the choice
of language.
H. The final domain is that of internal speech and emotion (Questions
35-40). With the exception of dreaming (Question 35), all other contexts
have Mandarin as the predominant language. The use of Mandarin in counting
(Question 36) may be explained by the fact that like all other subjects, it
is taught in elementary school in Mandarin and used throughout the
mathematics class in higher school levels. In the case of extreme emotion
such as sadness, anger and excitement (Questions 37, 39 and 40), contrary
to one's expectation that the first language would be used, one finds the
predominant use of Mandarin.^
2.2.4 Analysis of the results
The choice of language variety in different domains is summarized in
Table 3 where the general pattern of usage with reference to the 8 domains
is captured .
As indicated earlier, domain analysis can be used as a criterion to
determine whether the bilingual society under study is a stable one, and if
not, in what direction it is shifting. In a typical diglossic community,
the relationship between choice of language and domain is quite clear-cut.
H variety is associated with religion, work and education domains whereas L
variety is associated with family and friendship domains. However, if one
finds the domain distinction vanishing in terras of language choice as
indicated by the use of H variety in domains norraally associated with L
variety, the bilingual community under study then is an unstable one.
After examining the 3 relevant domains in the Taiwan bilingual
community, one can make several observations. Firstly, one finds that the
predominant use of Southern Min dialect is restricted to the family and
neighborhood domains with limited use in the friendship and religion
domains. The use of Mandarin is predominant in the friendship, internal,
employment, religion, education and administration domains with limited use
129
in the family and neighborhood domains. It seems there is an encroachment
of Mandarin in the family and neighborhood domains(to spouse, children,
younger neighbor) . Secondly, in addition to this contamination of language
use in the intimate domains, there is a shift in the use of language in the
internal speech and friendship domains which are normally associated with L
variety. This provides further support to the postulation that
bilingualisra in Taiwan is shifting toward the use of Mandarin. Thirdly, if
we contrast the habitual language use of the parents (Questions 11, 13 and
18) to that of this generation (Questions 10, 12 and 17), we find a
significantly different pattern in usage. This generation uses Mandarin
significantly more (See Table 4). Similarly, if we contrast the choice of
language when the addressee's age varies (Questions 4 and 9), again one
finds an increased use of Mandarin when the addressee is younger. All
these point to the conclusion that Taiwan bilingualisra is not in
equilibrium as in the case of Paraguay (Rubin 1968) and Canada (Lieberson
1972). It is in a transitional state moving toward monolingualism.
3. Conclusions and Future Research
In the study, I have not only presented the historical sociolinguistic
factors that led to the present state of bilingualisra in Taiwan, but also
examined the pattern of code- switching using domain analysis. Within each
domain, I have isolated factors which may affect the language choice in
Taiwan situation and made tentative generalizations about the possible
outcome of this bilingualisra. To validate the findings of this study,
follow-up studies that include large samples of population with more
diverse backgrounds are needed. In that case, a raore rigid statistical
analysis will be necessary. Also, studies employing different
methodologies, such as interviewing or participant observation, may obtain
results closer to natural interaction situations.
The code-switching phenomenon in Taiwan is by no means an isolated
one. Cross-cultural studies covering formal, functional, and psychological
aspects of code-switching have been fruitful (See e.g. Paradis 1978,
Williamson and Van Eerde 1980) and the present study provides further
evidence, as well as a detailed account of the language dependency on
social contexts in Taiwan.
As indicated in this study, the functional allocation of language
variety in a multilingual community is by no means random. Moreover, in
Taiwan, a bilingual community in the process of shifting, a new norm is
manifested in the increased use of the Standard Language in the intimate
domains.
Another related and relevant area of study is that of code-mixing, a
phenomenon crucial in the study of language change and repertoire change.
It would be interesting to see code-mixing analyzed in this framework.
Does the phenomenon of code-mixing cut across all domains or is it
limited to certain domains? How is it different functionally and fonnally
from the phenomenon of code-switching? If we can identify the locations,
the extent, the forras, and the functions of code-switching and code-mixing,
then we can undoubtedly contribute to the understanding of the activating
force behind language change and language shift. This study provides a
starting point from the Taiwan perspective.
130
NOTES
I would like to thank Professor C. C. Cheng, Professor B. Kachru,
and Professor H. Kahane for their helpful comments and guidance in the
course of preparing this paper.
-'■ Code-switching here is taken in the strict sense: shifting between
languages or language varieties which are not mixed. Various issues
concerning the distinction of code-switching and code-mixing are not taken
up here. For details, see Kachru (1978 a, b) , Warie(l978), Wentz(l977).
2 The dialect groups in China are: (The percentage is based on Yuan
I960. However, the population is inferred based on the present Han
population of 900 million. A more up-to-date census is not available.)
Han peop]
.e
900
million
Northe
rn
dialect
630
million
70
t
Wu
76
million
8.
.\%
Xiang
45
million
5
t
Can
22
million
2.
A%
Hakka
36
million
4
%
Yue
45
million
5
t
Min
Sou
th
Min
27
million
3
%
Nor
th
Min
11
million
1.
2%
3 A similar finding was reported in Paraguay by Rubin (1968:106).
^ In her study on bilingual attitudes among Canadian academics,
Rabel-Heymann (1978) found all languages are used for dreaming and an even
split of languages used for insult.
5 Code-switching in domain analysis is by no means a homogeneous
division. One can not say in an absolute sense that language A is spoken
in domain X, but only that language A is spoken predominantly in domain X.
By doing so, one does not exclude the possibility that there is code-mixing
going on.
131
Table 1 - Results of the Study
All or
Mostly
Con
tex
b
S. Min
Dialect
1 .
to
grandparents
20
2.
to
parents
a) at home
b) outside
19
18
3.
to
sibling
a) daily
b) argue
c) joking
9
12
11
4.
to
relatives
a) elder
b) same
c) younger
19
14
12
5.
to
spouse
7
6.
to
parents-in-law
10
7.
to
son/daughter
4
8.
to
boy/girl friend
a) initially
b) intimate
3
6
9.
to
neighbors
a) elder
b) same
c) younger
14
11
8
10.
to
neighborhood
grocery owner
18
11.
parents to the
same owner
20
12.
to
market salesman
17
13.
pa
rents to the
same salesman
20
u.
to
department
store salesman
12
15.
to
taxi driver
10
16.
to
schoolmates
a) in school
b) at home
c) fieldtrip
d) on bus
e) on phone
f) lecture
g) concert
1
4
2
2
2
2
2
17.
to
teacher
a) at school
b) at home
c) on street
d) on picnic
e) snack stand
outside
school
1
1
2
2
Both
All or Mostly No
Mandarin Response
1
2
1
1
6
5
3
5
3
6
1
5
1
5
3
2
8
1
5
8
6
5
12
2
12
3
4
5
5
3
10
2
7
7
4
5
14
5
11
4
14
3
16
6
13
3
15
4
14
1
19
2
17
1
19
1
17
16
132
Context
S.
18.
parents to teacher
a) at school
7
b) at home
8
19.
to
boss
a) in office
6
b) outside
8
c) on phone
7
20.
to
coworkers
a) at work
6
b) outside
6
21.
to
subordinate
a) at work
3
b) outside
3
22.
to
police
3
23.
to
people just met
1
24.
to
city government
employee
3
25.
to
doctor
6
26.
to
hospital
employee
5
27.
to
friends about
a) national
affairs
2
b) academics
1
c ) wo rk
2
d) anecdote
about home
7
e) gossip
3
f) at wedding
banquet
7
g) joking
5
28.
at
a Taiwan
regional club
11
29.
to
priest at
church
5
30.
to
friend at
church
5
31.
praying at church
6
32.
to
friend at temple
16
33.
praying at temple
15
34.
in
local religious
festival
16
35.
in
dream
6
36.
counting
1
37.
cursing
3
38.
mumbling to self
2
39.
in
excitement
2
40.
sad
3
All or Mostly
S. Min Dialect
Both
All or Mostly No
Mandarin Response
10
8
10
8
11
9
8
11
10
16
14
12
8
6
13
2
18
3
15
6
8
6
7
7
7
7
9
5
5
2
13
5
10
4
9
3
2
2
4
1
3
8
5
4
15
2
15
5
13
4
14
4
13
133
Table IB - Questions on General Impressions
Questions Often Sometimes Never No Response
41 . occasions when
only S. Min
can express
things well 6 13 2
42. occasions when
only Mandarin
can express
things well 10 8 1 2
S. Min Both Mandarin No Response
43. which language
do you feel
closest to 10
Table 2 - Contrasting the Urban-ness of Locale
All or Mostly All or Mostly
Context S. Min Both Mandarin
10. to neighborhood
grocery owner 18 2 1
12. to market salesman 17 2 2
14. to department
store salesman 12 2 7
15. to taxi driver 10 7 4
134
Table 3 - Summary of Language Usage vrith Regard to Domain
Informal Domains ('X' indicates predominant usage)
Majority
Dialect Family
Friendship Neighborhood Internal
S. Min
X
gossip
X
Mandarin
spouse
X
younger
Both
children
Foimal Domains ('X' indicates predominant usage)
Majority
Dialect
Employment Religion
Education
S. Min
Mandarin X
Both outside
to boss
X (temple)
X (church)
Administration
Table 4 - Contrasting Habitual Language Use between Older and Younger
Generations
Context
4. to relatives
a) elder
b) same
c) younger
9. to neighbors
a) elder
b) same
c) younger
10. to neighborhood
grocery owner
1 1 . parents to the
same owner
12. to market salesman
13. parents to the
same salesman
17. to teacher
a) at school
b) at home
c) on street
d) on picnic
e) snack stand
outside
school
18. parents to teacher
a) at school
b) at home
All or Mostly
S. Min Dialect Both
19
14
12
14
11
8
20
17
20
1
5
1
5
3
3
4
5
5
3
10
All or Mostly No
Mandarin Response
1
19
2
17
1
19
1
17
16
10
8
135
REFERENCES
BROWN, Roger and Albert Gilman. 1960. The pronouns of power and solidarity.
In Thomas A. Sebeok ed . Style in language. N. Y. : The technology press
of M. I. T. and John & Wiley. 253-276.
ERVIN-TRIPP, Susan. 1964. An analysis of the interaction of language, topic
and listener. In John J. Gumperz and Dell Hymes ed . The ethnography of
communication. American Anthropologist. 66.6 pt II: 86-102.
FANG, Shi-Duo. 1965. Wushi nian lai zhongguo guoyu yundong shi . (The history
of Chinese national language movement in the past 50 years). Taipei:
Mandarin Daily News, The Committee on Educational Service.
FERGUSON, Charles A. 1964. Diglossia. In Dell Hymes ed. Language in culture
and society. New York: Harper & Row. 429-437.
FISHMAN, Joshua. 1964. Language maintenance and language shift as fields of
inquiry. Linguistics 9: 32-70.
FISHMAN, Joshua. 1966. Language loyalty in the United States. The Hague:
Mouton.
FISHMAN, Joshua. 1968. Sociolinguistic perspective on the study of
bilingualism. Linguistics 39: 21-49.
FISHMAN, Joshua, Robert L. Cooper, and Roxana Ma. 1971. Bilingualism in the
Barrio. The Hague: Mouton.
FISHMAN, Joshua. 1972. The Sociology of language: an interdisciplinary social
science approach to language in society. Rowley, Mass: Newbury House
Publishers.
GREENFIELD, Lawrence. 1972. Situational measures of normative language views
in relation to person, place and topic among Puerto Rican bilinguals. In
Joshua Fishman ed. Advances in the sociology of language II. The Hague:
Mouton. 17-35.
GUMPERZ, John and Jan-Petter Blom. 1972. Social meaning in linguistic
structure: code-switching in Norway. In John Gumperz and Dell Hymes ed.
Directions in sociolinguistics: the ethnography of communication.
N. Y. : Holt, Rinehart and Winston.
HYMES, Dell. 1964. Introduction: toward ethnographies of communication. In
J. Gumperz and Dell Hymes ed . The ethnography of communication.
American Anthropologist 66.6 pt II: 1-34.
KACHRU, Braj . 1978a. Toward structuring code-mixing: an Indian perspective.
In B. Kachru and S. N. Sridar ed. Aspects of sociolinguistics in South
Asia. Special Issue of International Journal of the Sociology of
Language 16.
KACHRU, Braj. 1978b. Code-mixing as a verbal strategy in India. In J. E.
Alatis ed . international dimensions of bilingual Education. GURT:
Washington, D. C.
KACHRU, Yamuna and Tej K. Bhatia. 1978. The emerging 'dialect' conflict in
Hindi: a case of glottopolitics. International Journal of the Sociology
of Language 16: 47-58.
LIEBERSON, Stanley. 1972. Bilingualism in Montreal: a demographic analysis.
In J. Fishman ed . Advances in the sociology of language. The Hague:
Mouton. 231-254.
PARADIS, Michael, ed . 1978. Aspects of bilingualism. Columbia, SC: Hornbeam
Press, Inc.
RABEL-HEYMANN, Lili. 1978. But how does a bilingual feel? reflections on
linguistic attitudes of immigrant academics. In Michel Paradis ed .
Aspects of bilingualism. S. Carolina: Hornbeam Press.
136
RUBIN, Joan. 1968. National bilingualism in Paraguay. The Hague: Mouton.
WARIE, Pariat. 1978. Some sociolinguiatic aspects of language contact in
Thailand. Ph.D. Dissertation. University of Illinois.
WEINREICH, Uriel. 1953. Languages in contact: findings and problems. N. Y. :
Publications of the Linguistic Circle of N. Y. no 1 .
WENTZ, James. 1977. Some considerations in the development of syntactic
description of code-switching. Ph.D. Dissertation. University of
Illinois.
WILLIAMSON, Robert C. and John A. Van Eerde. ed. 1980. International Journal
of the Sociology of Language 25.
WU, Shou-Li. 1972. Taiwan fangyande lai long qu mai. (The origin and
development of the Taiwan dialect). In Taipei Document Committee ed.
The culture of Yellow River Region and Taiwan. Taipei. 381-389.
XUE, Guang-Qian and Chu Jian-Min. 1977. ed. Taiwan in modern times. Taipei:
Zheng-zhong Publishers.
YUAN, Jia-Hua . 1950. Hangyu fangyan gaiyao . (Outline of Chinese Dialects).
Peking: The Language Reform Publishers.
Studies in the Linguistic Sciences
Volume 11, Number 1, Spring 1981
SOME OBSERVATIONS ON DISCOURSE AND SENTENCE GRAMMAR
Jerry L. Morgan
This paper examines two lines of criticism of sentence
grammar, draivn from the literature of discourse studies. It
distinguishes two related positions: that sentence grammar
is incorrect, and that it should be abandoned in favor of a
theory of discourse that reformulates or entirely subsumes
sentence grammar. Some arguments for the first position are
discussed and shown to be unconvincing. The final section
is a speculative discussion of the second position.
1. Introduction
During this time of remarkable growth in the study of discourse,
an important question has been posed for the generative grammarian and
the discourse theorist alike: what is the relation between, on the
one hand, the mental systems that underlie hvunan ability to understand
and produce connected discourse, and on the other hand, theories of
grammar of Chomskyan coloration that take just sentences as their
domain; what I shall call "sentence grammar."
There are a number of positions one might take on the relation
between discourse competence and sentence grammar, either on a prioristic
or empirical grounds. A couple of these positions are potentially
threatening to theories of sentence grammar, and I would like to
examine some arguments that might be advanced for them, to see whether
the arguments are persuasive as criticisms of sentence grammar.
The literature of discourse studies by linguists, from at least van
Dijk (1972) to de Beaugrande (1980) and Givon (1980), contains a number
of proposals with a common thrust: that close study of problems of
discourse analysis and theory shows that sentence grammar is deficient
in important respects, hence incorrect; therefore it ought to be abandoned
in favor of a general theory of discourse. But in fact such proposals
contain two major claims that are logically independent, since one can
be true and the other false. The weaker of the two is the claim that
certain facts about discourse show that generative sentence grammar
cannot be the correct account for sentence syntax and semantics. The
second, stronger position, is that the right discourse theory (or complex
of theories) will do the work of a generative sentence grammar, making
the latter superfluous as a distinguishable component of over-all lin-
guistic ability, thus debunking such important hypotheses as the autonomy
of syntax, the independence of sentence grammar, perhaps even the existence
of a logically independent "language faculty." These are not trivial
issues, as one can see from the large and combative literature in several
fields that has grown around Chomsky's work. So it is worthwhile to con-
sider whether the arguments against sentence grammar that are so common
in the literature of discourse theory actually are a threat to sentence
grammar theory.
138
In this paper I want to examine both the weak and strong positions,
to determine the strength of the arguments for each. I will try to show
that arguments for the weak position have no force, since they overlook
a central claim of sentence-grammar theory. I will argue that the strong
position, though unconvincing in present forms, is more promising, and
worth pursuing. I will also discuss what it would take to make the strong
position a more persuasive one.
2. The weak position
Arguments for the weak position--that is, against sentence grammar--
are generally taken as arguments for the complete abandonment of sentence
grammar, in favor of some other theory. But is is not always clear what
the crucial property of sentence grammar is that the argument is directed
against. So it might be useful to point out some central properties of
sentence grammar. I believe that the following five properties--that is,
claims about the nature of grammar--are characteristic of more-or-less
standard versions of generative grammar:
1. The independence of sentence grammar: Linguistic competence
contains as a distinguishable component a cognitive system
whose domain is the sentence (and, implicitly, smaller expres-
sions that make up sentences) . The fact that this system
interacts with other cognitive systems in performance is not
an argument against the claim that it is an independent
system. The boundaries of this system--that is, what proper-
ties of sentences fall in its domain--is an open empirical
question, not determinable by a priori means.
2. Language-uniqueness : At least some aspects of this cognitive
system--or, from the linguist's viewpoint, the conceptual
vocabulary for describing it--are unique to language, not
found in other cognitive systems.
3. Grammar is formal : Sentence grammar is organized, and, from
the linguist's viewpoint, can be described, purely as a
matter of form, entirely independent of questions of communi-
cation, speaker's intention, and other matters of language use.
4. Structure is Chomskyan : The proper treatment of the notion
"sentence structure" is in the terms of the familiar phrase
markers in the work of Chomsky and many others before and
since; that is, syntactic structure (at a given level) is
reducible to relations of dominance, precedence and syntactic
category, and other notions defined in these terms.
5. More than one level of structure: An empirically adequate
grammar requires that each sentence be assigned at least two
levels of structure. The exact number of levels is an open
question.
Obviously these give properties do not exhaust the conceptual content of
generative grammar; Chomsky's more recent work on "Extended standard
theory," to pick a single example, is far richer. And in fact there is
139
not unanimity on all five properties even within the generative camp.
Properties 4 and 5 are presently controversial within generative grammar,
under attack by proponents of "arc-pair grammar" (Johnson and Postal
1980) and in recent work on "constituent structure grammar" by Gazdar
and his collaborators (Gazdar Cto appear)). Abandoning or replacing these,
properties would indeed constitute an important theoretical shift; but
the result would still count, as far as I can see, as a sentence grammar
of the sort under attack from discourse theorists. Unless I am mistaken,
it is properties 1 through 3 that are relevant to the criticism of
generative sentence grammar vis-a-vis discourse grammar.
The question is, then: do the arguments for the weak position
succeed as arguments against the first three properties?
There are two lines of argument for the weak position. The first,
apparently intended as an argument against the independence of sentence
grammar, is that the only "naturally occurring unit" for linguistic
analysis is the discourse, not the sentence. This argument is far
from convincing. If it is taken as a rather bizarre a priori method-
ological restriction, then it has no bearing on the question of inde-
pendence. Even following this restriction, there is no reason to
believe that it precludes the possibility of discovering, from close
analysis of discourse, motivation for an independent sentence compo-
nent. Taken more seriously, as leading in some obscure way to an a
priori argument against the independence of sentence grammar, again it
has no force. The question of independence is an empirical one--
either there is an independent sentence grammar, or there isn't--and
cannot be decided a priori.
The second line of argument is to criticize sentence grammar as
inherently incapable of providing an analysis of phenomena of various
sorts. Such observations can be construed either as arguments that
sentence grammar is incomplete, or as arguments against one or more
of the first three properties above, van Dijk (1972) uses them to
show, if I understand correctly, that sentence grammar is incomplete,
that there are discourse phenomena that cannot be accounted for by a
sentence grammar:
As long as S-grammars cannot provide satisfactory, general and
consistent, descriptions of the structures underlying discourses,
by formulating the rules which must be mastered by native speakers
to be able to perform the different tasks, we have to consider
them empirically inadequate (p. 7),
But one can't criticize an owl for not being a partridge. This kind of
criticism ignores a central thesis of sentence grammar theory. Chomskyan
generative grammar and its descendants are based on a hypothesis of great
theoretical and empirical importance: the existence of an independent
language faculty, within which there is a sentence grammar with rather
narrow domain. Insofar as sentence grammar in the Chomskyan spirit can
offer only an incomplete account of some language-related phenomenon, it
amounts to a claim about the nature of the phenomenon--that it is outside
the domain of sentence grammar, perhaps outside the domain of the language
faculty altogether, the produc* of some other cognitive system or systems.
140
The incompleteness is not an incompleteness by default or omission, but
a claim, correct or incorrect, about the facts. Thus arguments of this
type against sentence grammar, on the grounds that it cannot account for
certain discourse phenomena, have no force, unless accompanied by a
demonstration that the phenomena in question must be considered to be
in the same domain as phenomena that are central to sentence grammar
theory. But such a demonstration is a very difficult task- -phenomena
are not pre-sorted by nature in this way. There is no basis from which
to argue one way or the other, save to provide support for what I have
called the "strong position": to present an alternative theory which
treats sentence phenomena and discourse phenomena in a single unified
theory, without a distinguishable sentential sub-component. One could
then attempt to compare such a theory with a sentence grammar as expla-
nations of sentential phenomena. Lacking such an alternative, there
is no ground for comparison. Along these lines, observations taken to
be arguments for the incompleteness of sentence grammar could be construed
as arguments against properties two and three above (hence potentially
as arguments against independence) if it could be shown that the phenomena
in question require a uniform treatment of discourse and sentential properties,
and that t!ie sentential properties concerned are central to sentence grammar.
One such argument is the claim that there are important parallels
between properties of discourse and properties of sentences. In fact
there are at least three parallels one might see between sentences and
discourses. First, they both have structure. The point has been made
again and again, and quite correctly, that one key to the understanding
of discourse is the idea that discourses have structure. Then one might
propose "grammars" of some kind for discourses, to generate texts and
assign them structure. From here it is only a short step to the hypothesis
that one can provide a single grammar that treats both sentence and dis-
course structure.
But the parallelism turns out to be rather tenuous on closer inspec-
tion. The kind of structure commonly atrributed to sentences (and not
just by Chomskyans) is not the same kind of structure commonly attributed
to discourses (see Morgan and Sellner 1980 for more discussion). What's
needed is a demonstration that the system for determining discourse struc-
ture can be extended to give a complete treatment of the syntax of sen-
tences, a demonstration so far lacking, though not inconceivable.
A second apparent parallel is that texts, like sentences, have
"meaning" in some sense of this perniciously vague term. For example,
de Beaugrande (1980 p. 37) points out that meaning relations that can
hold within a sentence can hold between independent sentences in a dis-
course. He offers the following pair from Isenberg (1971, p. 155) as
illustration of this point.
Peter burned the book because he didn't like it.
Peter burned the book. He didn't like it.
But again the observation of the parallelism is misleading, in that it
obscures an important difference between the two cases in the illustration.
In the first, sentential case, we understand the relation to hold because
of the parts of the sentence and their mode of combination. But in the
141
second, discourse case, as in all cases of understood relations between
independent sentences in a discourse, we must infer that the relation
is to be understood to hold. Two different mechanism are involved: in
the sentential case, our knowledge of grammar--of the conventional
meaning of the word because, and of just how the meanings of English
expressions are related to the meanings of their parts; in the second
case, our ability to make common sense inferences. The latter can be
cancelled by contextual factors, in the manner of Grice's (1975) conver-
sational implicature; the former cannot. It would be a mistake to ignore
the difference.
The third parallel --anaphoric relations like some antecedent-anaphor
relations, the interpretation of definite noun phrases, and so on--is
similar to the previous one, in that such relations can hold either
between elements within a sentence or between elements in two separate
sentences in a discourse. In this case, though, the evidence suggests
the necessity for a unified treatment. It is fairly clear that such
matters need to be treated in extra-sentential, perhaps discourse terms.
But a perfectly coherent response to such arguments is available to the
proponent of the independence of sentence grammar: namely, just to yield
the territory--to conclude (correctly, to my mind) that such cases are
outside the domain of sentence grammar. Nothing in sentence grammar
theory entails that everything that can conceivably labelled a property
of sentences must be accounted for in sentence grammar.
Another kind of criticism of sentence grammar is based on the obser-
vation that there are discourse explanations for apparently syntactic
facts. There are a number of interesting attempts in the literature
(see Givon 1980 for some recent examples, especially the papers by Garcia
and Erteschik-Shir) . But generally the sentence grammarian can respond
to such analyses in the same way as to previous case: by yielding the
territory, concluding that the existence of convincing discourse explana-
tions shows that the problem was not a syntactic one to begin with.
A related kind of argument is based on the observation that there
are expressions whose meaning and/or syntactic distribution is clearly
to be given in terms of discourse or functional terms. One might conclude
from the existence of morphemes that function as topic or focus markers,
for example, that a complete theory of sentence grammar must incorporate
a treatment of notions like "focus," "topic" and the like. But there is
no reason to accept this conclusion. It is no more necessary than this
one: since English has pronouns like he^ and she whose meaning properties
(hence use) are determined in part by natural gender, the theory of
sentence grammar must contain a theory of physical gender. More plausibly,
these are instances of the interaction of sentence grammar with other
linguistic or non-linguistic cognitive systems. The existence of such
interactions in no way provides arguments for the identity of the inter-
acting systems.
A similar approach is available for dealing with "optional" rules
or constructions with clear discourse value; constructions like "Y-movement,'
for example, that have different discourse appropriateness conditions from
their un-moved counterparts. Sentence grammar need only specify which
orders are possible, i.e. grammatical. The rest should follow from language-
specific discourse rules, from general principles of communication, or from
the interaction of grammar with other cognitive systems.
142
In short, such empirical arguments against the independence of
sentence grammar are not convincing, taken one by one. Conveivably
they could become persuasive cumulatively, by gradually reducing the
domain of sentence grammar to emptiness; but that day is hardly on the
horizon. In the meantime, the sentence grammarian can fairly comfortably
continue to take such observations as progress in the empirical deter-
mination of the domain of sentence grammar.
3. The strong position
Obviously, one persuasive way to argue for the abandonment of a
theory is to present an alternative that gives a superior treatment
of a significant portion of the domain of the theory under attack; for
example, to show how a fairly well articulated theory of discourse
provides an account of some phenomenon that is central to sentence
grammar. IVhat would be required of the theory of discourse, then, is
that it be complete enough for one to examine its consequences at
the sentence level for naturally occurring or constructed discourse,
with a degree of formal detail that approaches that of existing sentence
grammars. Unfortunately, there is no theory of discourse that is that
well developed. This state of affairs is hardly surprising, given the
almost miraculous complexity of the mental systems that underlie our
ability to produce and understand discourse. At this point such a
comparison is impossible Still, it might be useful to sketch in
hypothetical terms some directions that such an enterprise might take.
To begin I need to narrow down a bit what I mean by "discourse
theory." I mean any theory that attempts to satisfy two minimal
conditions :
1. It attempts at least a partial account of the most striking
aspect of discourse comprehension: how an understanding of
a discourse is so much greater than the logical sum of the
parts (i.e. sentences) that make it up.
2. If offers a definition or explication of indispensable but
elusive notions like "topic," "focus," "given/new,"
relevance," "coherence," and "text structure."
Such a theory might fruitfully be framed in terms of communicative
actions, i.e. rules or strategies for the activity of communication,
rather than rules of well-formedness. Matters of ill-formedness would
be recast either as actions that violate rules of communication, or
as inefficient or self-defeating communicative actions, given principles
of communicative efficiency. Such a theory would also need to include
(or appeal to interaction with) a theory of common-sense reasoning, and
would likely include a component of language-specific conventions of
discourse and of other aspects of language use (see Morgan 1978a, b
for discussion) .
If such a theory were available, then, it would be possible to
attempt to recast central aspects of sentence grammar in terms of the
independently motivated discourse theory. Again, lacking a detailed
theory of this type, discussion is necessarily speculative. But a
143
couple of illustrations will help make clear what kind of attempt I have
in mind. The strategy would be to determine how much of the semantics
and syntax of sentence grammar could be treated by discourse theory,
leaving only matters of morphology and the lexicon to sentence grammar.
The possibilities for semantics are rather dubious, it seems to me.
There must be a discourse system for understanding connected discourse
that is heavily based on inference, though perhaps with a language
specific, conventional component as well (by language specific here 1
mean principles that differ from language community to language community,
and must be learned) . Then it is not out of the question that this
system could provide a parallel treatment for problems of compositionality
within sentences, yielding then a single uniform system for all aspects of
meaning analysis, both at discourse and sentence-internal levels. But
such an attempt faces large obstacles. It would be necessary to show how
the compositionality that is a central tenet of sentence grammar could be
dispensed with. The claim of compositionality is that any adequate theory
of semantics must analyze meaning as depending not only on the elements
that make up expressions, but on their syntax- -on the way they are combined
to make up the expression. It is hard to see how to extend an inference-
based understanding system to deal naturally with the difference between
the dog bit the cat and the cat bit the dog without in the process re-
inventing sentence syntax, let alone how to conquer the well-known problems
of the relation of syntactic properties to scope of logical operators. The
likelihood of success of such an attempt is very implausible, I think, though
it cannot be ruled out a priori.
The attempt to recast syntax in discourse terms is perhaps slightly
less implausible. Given a theory well -developed enough to include treat-
ment of the action of referring, and given that language communities can
differ in their conventional rules of discourse, one could attempt to
recast the syntactic rules for English noun phrases as English strategies
for the act of referring.
The possibility of such reformulation of syntax in terms of communi-
cative function is tantalizing, since there are obvious correlations
between syntactic form and communicative function. For example, in lan-
guage after language, the unmarked position of the restrictive relative
clause is adjacent to the "head noun" that it "modifies," as in The woman
who invented the wheel died in 70,000 B.C. Viewed purely formally, this
seems just an unexplainable (though widespread) quirk. Viewed functionally,
on the other hand, it is hardly surprising, since the head and accompanying
relative are uttered in pursuance of a single purpose--to pick out a refer-
ent by describing its properties. Uttering the head and the relative
consitutes a single communicative act, and the temporal adjacency is
unsurprising, assuming some intuitively obvious principles of efficiency.
From this viewpoint it is the cases where the relative is detached from
the head that are surprising.
But this kind of analysis, though tantalizing, also faces serious
obstacles, insofar as form does not always follow function. For example,
how could such a theory explain cases of apparent functional disunity like
extraposed relatives, as in the woman died in 70,000 B.C. who invented the
wheel or verb-particle constructions like John put the cat out, to say
144
nothing of the numerous apparently purely formal conditions and constraints
proposed by generative grammarians from Ross (1967) to Chomsky (1981)? The
burden is clearly on the discourse theorist to show that at least a signifi-
cant fraction of these problems have explanations in discourse and/or func-
tional terms. Personally, I am skeptical that such explanations will ever
be achieved. But I think the knowledge to be gained in the attempt is worth
the effort.
REFERENCES
DE BEAUGRANDE, R. 1980. Text, discourse, and process. Norwood, New
Jersey: Ablex Publishing Co.
CHOMSKY, N. 1981. Lectures on government and binding. Dordrecht: Foris
Publications.
VAN DIJK, T. 1972. Some aspects of text grammars. The Hague: Mouton.
ERTESCHIK-SHIR, N. 1980. Discourse constraints on dative movement. In
Givon (1980).
GARCIA, E. 1980. Discourse without syntax. In GiMon (1980).
GAZDAR, G. To appear. Phrase structure grammar. In G.K. Pullum and P.
Jacobson, eds.: The nature of syntactic representation. Boston:
Reidel.
GIVON, T. 1980. Syntax and semantics, vol. 12: discourse and syntax.
New York: Academic Press.
GRICE, H.P. 1975. Logic and conversation. In P. Cole and J. Morgan, eds.:
Syntax and semantics, vol. 3: speech acts. New York: Academic Press.
ISENBERG, H. 1971. Ueberlegung zur Texttheorie. In J. Ihwe, ed.:
Literaturwissenschaft und Linguistik: Ergebnisse und Perspektiven.
Frankfurt: Athenaeum.
JOHNSON, D. and P. Postal. 1980. Arc pair grammar. Princeton: Princeton
University Press.
MORGAN, J. 1978a. Two types of convention in indirect speech acts. In
P. Cole, ed. : Syntax and semantics, vol. 9: pragmatics. New York:
Academic Press .
MORGAN, J. 1978b. Toward a rational model of discourse comprehension.
In D. Waltz, ed. : Theoretical issues in natural language processing.
Urbana.
MORGAN, J. and M. Sellner. 1980. Discourse and linguistic theory. In
R. Spiro, B. Bruce and W. Brewer, ,eds.: Theoretical issues in reading
comprehension.
ROSS, J. 1967. Constraints on variables in syntax. MIT PhD dissertation.
Indiana University Linguistics Club. Mimeo.
Studies in the Linguistic Sciences
Volume 11, Number 1, Spring 1981
EVIDENCE FOR THE ELSEWHERE CONDITION IN SHONA*
David Odden
According to the hypothesis advanced in The Sound Pattern
of English (Chomsky and Halle (1968)), disjunctive ordering is
a consequence of certain abbreviatory conventions. Chomsky and
Halle state that subrules of rule schemata abbreviated with
parentheses, Greek letter variables or angled brackets are all
disjunctively ordered. A different hypothesis is proposed by
Kiparsky in the paper Elsewhere in Phonology (Kiparsky (1973))
where he proposes that disjunctive ordering is a consequence of
a certain formal relation between pairs of rules. He proposes
there Elsewhere Condition, which states that ' Two adjacent rules
of the form A -I B / P Q and C •> D / R S are disjunctively
ordered if and only if a) the set of strings that fit PAQ is a
subset of the set of strings that fit RCS and b) the structural
changes of the two rules are either identical or incompatible'
(p. 94). I shall show here that in the grammar of Shona, there
are two pairs of tone rules, both of which must be ordered
disjunctively, and both of which satisfy the requirements of the
Elsewhere Condition. Moreover, neither of these pair of rules
may be collapsed with any abbreviatory devices, and thus these
rule pairs stand as counterexamples to the hypothesis that
disjunctive ordering is a consequence of certain abbreviatory
conventions. In the first section, I demonstrate the disjunctive
ordering relation between Sandhi Raising and Clitic Lowering, and
in the second section, I discuss the disjunctive relation between
Associative Raising and Associative Lowering. In the final section,
I discuss the theoretical implications of these pairs of rules,
and argue that these rules provide evidence for the generalized
version of the Elsewhere Condition argued for by Kiparsky.
I. Sandhi Raising and Clitic Lowering.
The first set of rules which I shall discuss is the pair of rules
Sandhi Raising and Clitic Lowering. Under the provisions of Sandhi
Raising, a L tone which stands before a L tone or at the end of a phrase
is raised to a H tone, provided that it is immediately preceded by a H
tone which stands at the end of the word. Thus, the initial L tone found
in the isolation form of the nouns in (1) is changed to a H tone when it
stands after a H tone, demonstrated in (2).
(1) munhu
akabika
zvirongo
zvina
kumusha
(2) ndakatuma mu^nhu
Chipo a^kabika
ndakatuma zvi^rongo
zvirongo zv^na
ndakamutuma kumusha
person
and he cooked '
water pots'
four'
to the village'
I sent the person'
and Chipo cooked '
I sent water pots'
four water pots'
I sent him to the village'
146
On the other hand, when immediately preceded by a L tone, the word
initial L tone is unchanged, as demonstrated in (3). As an example,
the word initial L tone of the noun munhu is not raised to a H tone when
it is preceded by the L tone of the verb handaona 'I didn't see' in the
phrase handaona munhu 'I didn't see the person'.
(3) handaona mijnhu
munhu a_kabika
ndakapa munhu zvi^rongo
zvinhu zvl^na
ndakatuma munhu kumusha
'I didn't see the person'
'and the person cooked'
'I gave the person water pots'
'four things'
'I sent the person to the village'
When a word initial L tone is immediatelt followed by a H tone, the
Sandhi Raising rule does not apply, as is demonstrated in (4).
(4)
badza
mukomana
mapanga
akatora ba_dza
akatora mukomana
akatora mapanga
'hoe'
'boy'
'knives'
'he took
'he took a boy'
'he took knives'
hoe'
The presence of a H tone on the second syllable of the noun badza 'hoe' blocks
the application of Sandhi Raising to the initial L tone, when preceded by
the H tone of the verb akatora in there phrase akatora badza 'he took a hoe'.
Finally, Sandhi Raising will apply to a monosyllabic L toned word,
provided tha the word is followed by a L tone in the next word, or else
stands at the end of the phrase.
(5) bwe
mbwe
na
akaona bwe
akaona mbwe
akatora na
akatora bwe yandakatenga
akatora mbwe dzandakaona
'stone'
'tsoro pieces'
'four'
'he saw the stone'
'he saw the tsoro pieces'
'he took four'
'he took the stone that I bought'
'he took the tsoro pieces that I saw'
On the other hand, when a monosyllabic L toned word stands before
a H tone, Sandhi Raising is unable to apply. Thus, the L tone of the noun
bwe is unchanged when preceded and followed by a H tone, although it is
raised to H when preceded by H and followed by L or pause in (5).
(6) akaona bw£ gdrd
akaona mbwe hilru
*aka6na bwe guru
*aka6na mbwe huru
'he saw a large stone'
'he saw large tsoro pieces'
In order to account for the alternations of L and H tone discussed here,
I propose the Sandhi Raising rule, (7).
(7) L > H / H #
in
147
This rule is stated with the righthand environmental restriction expressed
as a disjunction of environments. An alternative to this expression is
given in (8), where a negative environmental condition prevents the rule
from applying before a H tone.
(8) L^ H / H # -H
Either formulation of Sandhi Raisinf may be accepted, without any significant
consequences for the argument which follows.
The Clitic Lowering rule is a rule which applies to a set of morpho-
logical clitic morphemes which bear underlying H tones. This rule lowers
the H tone of the clitic to L when it is preceded by a H tone. In order
to motivate the claim that these clitics have underlying H tone, I shall
illustrate a third rule, which provides crucial evidence for the underlying
tone of these clitics.
The grammar contains a Sandhi Lowering rule, according to which a H tone
is lowered to L after a H tone, when the following word begins with a H
tone. Thus, as demonstrated in (9), a final H tone is lowered when it is
followed by the word akafa 'he died' which has an initial H tone, but not
before the word akafa 'who died' which has an initial L tone.
(9)
murume
mukadzl
mururae^ akafa
mukadz_l akafa
murume^ akafa
mukadzi akafa
'man'
' woman '
'the man died'
'the woman died'
'the man who died'
'the woman who died'
Similarly, in (10), it is shown that words which end in the tone
sequence LH do not undergo Sandhi Lowering.
(10) badza
badz£ rakavhunika
mukomana
mukomana^ akafa
zvirongo
zvirongo^ zvakatorwa
*zvirongo^ zvakatorwa
'hoe'
'the hoe broke'
'boy'
'the boy died'
'water pots'
'the water pots were taken'
Finally, in (11), it is shown that when a H toned monosyllable is followed
by a H tone and is preceded by a H tone, the tone of the monosyllable
undergoes Sandhi Lowering and becomes a L tone.
(11) mbwa
tsva
anoda
anoda mbwa
anoda mbwa tsva
anoda tsva^ huru
anoda mbwa tsva huru
'dog'
'new'
'he wants'
'he wants a dog'
'he wants new dogs'
'he wants big new (things) '
'he wants big new dogs'
148
If the monosyllable is preceded or followed by a L tone, the H tone of that
monosyllable remains unchanged.
(12) akatora tsva dzandakatenga
akabika tsv£ huru
akauraya mbwa nebwe
akaurayfsa mbwa huru
'he took the new (things) that I bought'
'he cooked the new big (things)'
'he killed a dog with a stone'
'he butchered a big dog'
In order to account for these alternations, I propose the Sandhi Lowering
rule (13).
(13) H ^ L / H (#)
# H
I turn now to the question of the Clitic Lowering rule. When one of
the clitic morphemes -wo ' too' or -zve 'too' is preceded by a H tone, the
clitic will appear with a L tone on the surface, whereas when these clitics
are preceded by a L tone, they appear on the surface with a H tone.
(14) badza-wo
badza-zve^
bhuku
bhuku-wo
bhuku-zve
'hoe too'
'hoe again'
'book'
'book too'
'book again'
Under the assumption that these clitics bear an underlying H tone, a rule
will be necessary to lower that H tone after a H tone.
Crucial evidence in support of the assumption that these clitics
bear underlying H tone is the fact that these clitics trigger the application
of the Sandhi Lowering rule. Thus, as shown in (14), a word ending in the
tone pattern HH changes that tone sequence to HL before the two clitics
-wo and -zve .
(15) hari
hari^-w6
hari^-zve
murume
murume^-wo
murume-zve
'pot'
'pot again'
'pot too'
'man'
'man again'
'man too'
If we assume that these clitics bear an underlying H tone, the lowering
of the final H tone shown above follows immediately from application of
Sandhi Lowering, as shown in (16).
(16) hari-zve
hari-zve
underlying
Sandhi Lowering
On the other hand, if these clitics are assumed to bear underlying L tones,
then it will be impossible to explain the application of Sandhi Lowering
to the preceding H tone.
Given that the clitics -wo and -zve have underlying H tones, some rule
will be necessary in order to account for the fact that the H tone of the
clitic is lowered after a H tone. The lowering of H tone in these clitic
149
morphemes cannot be due to the application of Sandhl Lowering, since Sandhl
Lowering crucially requires the presence of a H tone after the focal H
tone. The H tone of the clitic -wo and -zve, on the other hand, will be
lowered regardless of the righthand context. As demonstrated in (17),
monosyllabic clitics behave differently from monosyllabic nonclitics, in
that only clitics are lowered unconditionally after H tones.
(17) akapa-wo 'he gave too'
akapa-zve^ 'he gave again'
akapa tsva^ 'he gave a new (thing)'
akapa mbwa 'he gave a dog'
Accordingly, I propose the Clitic Lowering rule (18), which applies
only to clitics, and lowers their H tone after a H tone.
(18) H > L / H
[+clitic]
This rule must be ordered after Sandhl Lowering, since, as demonstrated in
(19), application of Sandhl Lowering to a word final H tone prevents the
subsequent application of Clitic Lowering, yielding the correct surface
form hari-wo, from underlying harl-wo.
(19) hari-wo underlying
hari^-wo Sandhl Lowering
hari-wo (NA) Clitic Lowering
If Sandhl Lowering were applied after Clitic Lowering, Clitic Lowering
would incorrectly apply to the H tone of the clitic -wo, as shown in (20),
and would thus incorrectly block application of Sandhi Lowering.
(20) harl-wo underlying
hari-wo Clitic Lowering
*harf-wo (NA) Sandhi Lowering
Clitic Lowering must also apply before Sandhi Raising. This is
shown by the fact that if the underlying H tone of a clitic is lowered by
application of Clitic Lowering, then it does not trigger application of
Sandhi Raising. Thus, as seen in (21), the Clitic Lowering rule first
lowers the initial tone of the clitic -wo , which prevents the H tone of
that clitic from raising the initial L tone of the noun sadza.
(21) akapa-w6 sadza underlying
akapa-wo sadza Clitic Lowering
NA Sandhl Raising
akapa-wo sadza phoneti output
If Sandhi Raising were applied before Clitic Lowering, the underlying H
tone of the clitic -wo would trigger application of Sandhi Raising to the
noun sadza, and then would be subsequently lowered, giving the incorrect
surface form *akapa-wo sadza.
(22) akapa-wo sadza underlying
akapa-wo sadza Sandhl Raising
*akapa-wo sadza Clitic Lowering
150
To summarize the analysis presented up to this point, I have motivated
three rules, viz. Sandhi Lowering, Clitic Lowering and Sandhi Raising, which
apply in that order. I turn now to the evidence which demonstrates that
the rules Clitic Lowering and Sandhi Raising must be applied disjunctively.
A close inspection of the derivation of the underlying form akapa-wo reveals
that, if the two rules in question are applied in a conjunctive manner.
Clitic Lowering will first apply to the H tone of the clitic -wo, yielding
a L tone. This L tone stands after a H tone, and is at the end of an
utterance, so the structural description of the Sandhi Raising rule is
satisfied. Nevertheless, that rule does not apply to the final L tone of
the form akapa-wo, i.e. the rule does not give the incorrect surface form
*akapa-w6. The Sandhi Raising rule must therefore be constrained so that
it does not apply to a L tone which derives from H tone by the application
of Clitic Lowering. Under the hypothesis advanced by Chomsky and Halle,
disjunctive ordering of rules is a derivative property of certain types of
abbreviatory devices, namely angled brackets, Greek letter variables and
parenthesis notation. An inspection of the two rules Clitic Lowering
and Sandhi Raising shows that there are sufficient dissimilarities in the
two rules to prevent them from being collapsed with any of these devices.
These rules are repeated in (23) for ease of reference.
(23) H -> L / H Clitic Lowering
[+clitic]
L - H / H # JLI Sandhi Raising
i'}
Under the hypothesis of SPE, these two rules should therefore be
ordered conjunctively rather than disjunctively. However, under the
Elsewhere Condition proposed by Kiparsky, the disjunctive ordering relation
between Clitic Lowering and Sandhi Raising is predicted automatically.
The two rules are ordered adjacently as required by the condition, and
the structural changes are incompatible, since the feature H is incompatible
with the feature L. Finally, every H toned clitic which undergoes Clitic
Lowering stands after a H tone, and therefore also stands in the correct
environment to undergo Sandhi Raising. Consequently, the required subset
condition is satisfied, and the Elsewhere Condition will provide the
existant disjunctive ordering between Sandhi Raising and Clitic Lowering,
where the hypothesis of SPE cannot provide that relationship.
The only reasonable alternative to imposing disjunctive ordering on
these rules is to restrict Sandhi Raising so that it does not apply to
clitic morphemes. This restriction would therefore attempt to explain
the failure of Sandhi Raising to apply in the surface form akapa-wo by
restricting Sandhi Raising to apply only in a larger domain, where clitics
appear in a smaller domain. At the same time, however, this restriction
would incorrectly predict that clitics with underlying L tones also do not
undergo Sandhi Raising. The proposed restriction incorrectly predicts that
the L toned clitics such as -po 'at there' and -ko 'to there' will never
bear H tone by the application of Sandhi Raising, whereas in fact, as shown
in (24), underlying L toned clitics do undergo Sandhi Raising.
(24) akablka-po 'he cooked there'
akatora-po 'he took there'
akaenda-ko 'he went there'
akatiza-k5 'he ran there'
151
The hypothesis that Sandhi Raising is restricted in its application fails
to account for these facts, so the original version of Sandhi Raising
stands vindicated, and with it, the argument for the Elsewhere Condition.
II. Associative Raising and Associative Lowering.
I turn now to the second pair of rules which are disjunctively
ordered, again in the environment predicted by the Elsewhere Condition,
namely Associative Raising and Associative Lowering. Again, it will be
necessary to investigate three rules in thr grammar, in order to properly
understand the two rules which are crucially ordered in a disjunctive
fashion.
The first rule to be considered is the Associative Lowering rule, which
is discussed at length in Odden (1980). When a noun is preceded by one
of a specific set of prefixes, which I refer to here as Associative Prefixes,
the word initial H tone or tones of the noun are lowered after that prefix.
The lowering of initial H tone is illustrated below in (25) .
(25) mbwa
ne-mbwa
bhuku
ne-bhuku
harl
ye-hari
mbundudz£
se-mbundudzi
As I have argued elsewhere, the Associative Lowering rule which lowers
these H tones applies to the H tone autosegment associated with a sequence
of vowels, and does not apply to the vowel segments themselves. Thus,
the noun mbundudzf has the representation indicated in (26), with a single
H tone autosegment associated with three vowels.
(26) H
mbundudzi
I have also shown in Odden (1980) that the Associative Lowering rule
applies only after prefixes with the morphological feature +Associative ,
and only when that prefix bears a H tone. Failure of Associative Lowering
to apply after other H toned prefixes is illustrated in (27).
(27) i-mbwa 'it is a dog'
i-bhuku 'it is a book'
Va-Huku 'Mr. Chicken'
Va-Mbundudz£ 'Mr. Army Worm'
Failure of Associative Lowering to apply after L toned Associative Prefixes
is demonstrated in (28).
(28) ra-Mbwa 'of Dog'
ra-Huku 'of Chicken'
sa-Mbundudzi 'like Army Worm' ^
'dog'
'with
a
dog'
'book
'with
a
book'
'pot '
'of the
pot'
'army
worm
'like
an army worm'
152
Accordingly, the Associative Lowering rule (29) is proposed, to
account for these alternations.
(29) H * L / H
[+associative]
A second rule which applies in the environment of an Associative
Prefix is the Associative Raising rule, which raises the initial L tone of
a noun to a H tone, when that L tone is followed either by another L tone
or when that L tone stands at the end of the word.
(30) munhu
ne-munhu
mapadza
ne-mapadza
bwe
se-bwe
murimi
ye-murimi
person
with a person'
hoes'
with hoes'
stone '
like stone'
farmer '
of the farmer'
In (31), nouns are preceded by an Associative Prefix when they bear
an initial L tone which is immediately followed by a H tone. It can be
seen that a L tone followed by a H tone is unaffected by the Associative
Raising rule and retains its initial L tone.
(31) badza 'hoe'
se-badza 'like a hoe'
murume 'man'
no-rauriime 'with a man'
mukomana ' boy
ve-mukoraana 'of the boy'
To account for the raising of initial L tone in (30), I propose the
Associative Raising rule (32).
(32) L ^ H / H fLl
[+associative] l#/
This rule must also be constrained to apply only after a prefix with the
morphological feature +associative, since other H toned prefixes such
as the honorific prefix Va- do not trigger application of Associative
Raising, as demonstrated in the form Va-Bwe 'Mr. Stone'. Furthermore,
this rule must be constrained so that it applies only after a H toned
Associative Prefix, since a L toned Associative Prefix does not trigger
raising of the following L tone, as shown in (33).
(33) Va-Bwe 'Mr. Stone'
na-Bwe 'with Stone'
sa-Mbereko 'like Mbereko'
ya-Munhu 'of man'
It can also be observed that the Associative Prefix itself bears a
L tone on the surface when followed by a noun stem which undergoes
Associative Lowering. I assume that the lowering of the Associative
153
Prefix itself is due to the application of a third rule. Prefix Lowering,
which lowers the tone of an Associative Prefix when it is followed by
a H tone.
(34) H -^L / H
[+associative]
This rule must be restricted to apply only to Associative Prefixes,
since the copular prefix i- and the honorific prefix Va- do not undergo
a parallel lowering process before H tones, viz. i-huku 'it is a chicken',
Va-Huku 'Mr. Chicken'. The Prefix Lowering rule must be ordered after
Associative Raising, since its application crucially depends on the presence
of the H tone derived by appl^^tion of Associative Raising, as shown in
the derivation below. ^f
(35) ne-bwe underlying
ne-bwe Associative Raising
ne^-bwe Prefix Lowering
The Prefix Lowering rule must be ordered after Associative Lowering,
since the underlying H tone of a noun stem will not trigger application of
Prefix Lowering if the stem initial H tone undergoes Associative
Lowering. If Prefix Lowering were ordered before Associative Lowering,
Prefix Lowering would incorrectly apply to the prefix in the form ne-hove,
as shown in (36).
(36) ne-hove underlying
ne^-hove Prefix Lowering
NA Associative Lowering
*ne-h6ve phonetic output
Additional support is available for the Prefix Lowering rule, which
demonstrates that the rule applies before any H tone, and not just before
a H tone which derives from the application of Associative Raising to an
underlying L tone. There are a few morphemes which are lexical exceptions
to Associative Lowering, which have initial H tones, but whose H tones are
not lowered after an Associative Prefix. As seen in (37), when one of
these words is preceded by an Associative Prefix, the underlying H tone
of the Associative Prefix is lowered in the presence of the underlying
word initial H tone, as is predicted by the Prefix Lowering rule.
(37) uno 'this'
ne-uno 'with this'
uya 'that'
che-uya 'of that'
upf 'which?'
se-upf 'like which?'
It is therefore to be concluded that the Prefix Lowering rule is at least
partially independent of the application of Associative Raising.
It can be demonstrated that the Associative Lowering rule iterates
from left to right, since application of Associative Lowering to a following
Associative Prefix will prevent that second Associative Prefix from triggering
154
application of Associative Lowering to the noun stem. Thus, as demonstrated
in (38), Associative Lowering applies to the H tone of the prefix -che-
and the derived L tone of the prefix -che- prevents Associative Lowering
from applying to the H tone of the noun hove.
(38) ne-che-hove underlying
ne-che^-hove Associative Lowering
ne-che-hove (NA) Associative Lowering
Similarly, application of Associative Lowering to the second prefix -ne-
in (39) prevents that prefix from lowering the third prefix, which retains
its H tone, and therefore triggers application of the Associative Lowering
rule to the following noun stem.
(39) se-ne-che-hove underlying
se-n£-che-hove Associative Lowering
se-ne-ch£-h6ve (NA) Associative Lowering
se-ne-che-hove Associative Lowering
There is also evidence which shows that Associative Lowering must
apply before Associative Raising, since, when two Associative Prefixes
stand before a noun with a L tone followed by another L tone or a word
boundary, that initial L tone of the noun will be raised by application
of Associative Raising.
(40) ne-che-munhu 'with (the thing) of the man'
se-ne-bwe 'like with stone'
se-che-mapadza 'like of the hoes'
If Associative Raising is ordered to apply before Associative Lowering,
the former rule will apply to raise the initial L tone, and the latter rule
will subsequently apply to lower the tone of the medial Associative Prefix,
as shown in (41) .
(41) ne-che-munhu underlying
ne-che-munhu Associative Raising
ne-che-munhu Associative Lowering
On the other hand, if Associative Lowering were ordered before Associative
Raising, the H tone of the medial Associative Prefix would be lowered,
and would therefore not be capable of triggering application of Associative
Raising to the following noun.
(42) ne-che-munhu underlying
ne-ch£-munhu Associative Lowering
*ne-che-munhu (NA) Associative Raising
It must therefore be concluded that Associative Raising applies before
Associative Lowering.
Given that Associative Raising applies beofre Associative Lowering,
it can be seen in (43) that conjunctive application of these two rules to
the underlying form ne-bwe will incorrectly yield the surface form *ne-bwe.
This derivation yields an incorrect surface form, because Associative
155
Lowering applies to the output of Associative Raising.
(43) ne-bwe underlying
ne-bwe Associative Raising
*ne-bwe Associative Lowering
It is therefore apparent that Associative Lowering must be constrained
in some way to prevent it from applying to the output of Associative Raising.
One solution would be to reorder the rules Associative Raising and Associative
Lowering. However, this solution ca-not be correct for two reasons.
First, it has just been demonstrated in (38) that applying Associative
Lowering before Associative Raising will incorrectly prevent Associative
Raising from applying in the underlying string ne-che-bwe. Moreover,
reordering these two rules will incorrectly allow the output of Associative
Lowering to serve as the input to Associative Raising. As shown in (A4),
application of Associative Lowering to the H tone of the noun mbwa will
create the structure required for application of Associative Raising, and
would then yield the incorrect surface string *ne-mbwa instead of ne-mbwa .
(44) ne-mbwa underlying
ne-mbwa Associative Lowering
ne-mbwa^ Associative Raising
*ne-mbwa Prefix Lowering
Therefore, Associative Lowering and Associative Raising cannot be reordered
in order to resolve the paradox of applying one rule incorrectly to the
output of the other rule.
The two rules cannot be collapsed into a single rule schema employing
angled brackets, parenthesis or Greek letter variables, due to the formal
dissimilarities in the two rules. The rules Associative Lowering and
Associative Raising are repeated in (45).
(45) H ■> L / H
[+associative]
L ■) H H fL>
[+associative] I "J
Therefore, according to the hypothesis of The Sound Pattern of English,
these rules must incorrectly be applied in a conjunctive manner . However,
under the Elsewhere Condition, these two rules will be applied disjunctively.
The two rules are adjacently ordered, and the structural changes of the
rules are incompatible. Furthermore, the subset relationship is satisfied,
since the lefthand environment of both rules is identical, and the right-
hand condition on Associative Raising entails a more restricted environment
of application for that rule than the null righthand condition on the
Associative Lowering rule. Thus, any form which undergoes Associative
Raising will also satisfy the structural requirements of Associative Lowering.
Consequently, the Elsewhere Condition will correctly impose disjunctive
ordering of the rules Associative Lowering and Associative Raising.
156
III. Implications
I have argued here that the Elsewhere Conditions allows us to account
for the disjunctive ordering relationship between the pairs of rules
Associative Lowering - Associative Raising and Clitic Lowering - Sandhi
Raising. One question which might be raised at this point is whether the
Elsewhere Condition does actually cover this relationship, since the
required subset condition may or may not be satisfied by these rules,
depending on what interpretation is given to the subset condition. Under
one interpretation of the subset condition, the two pairs of rules do not
satisfy the subset condition, since the focal segments in both sets of rules
are distinct. Thus, the focal segment for Clitic Lowering must be H, but
the focal segment for Sandhi Raising must be L. The underlying string
akapa-w6 satisfies the structural description of the Clitic Lowering rule,
but not the structural description of the Sandhi Raising rule, since the latter
rule applies to L tones.
One possibility would simply to reformulate the Sandhi Raising and
Associative Lowering rules, so that both rules apply to both H and L tones.
Accordingly, the following revisions mught be proposed.
(46) T ^ H / H# (l) Sandhi Raising
«
T -V L / H Associative Lowering
[+associative]
However, it would appear to be an adhoc move to reformulate the second of
the two rules, just in order to guarantee that the subset requirement on
the Elsewhere Condition will be satisfied.
As stated by Kiparsky, the Elsewhere Condition requires that the pairs
of rules meet a certain subset relationship, but the condition does not
state explicitly at what stage of the grammar this relationship must be
satisfied. There are numerous logical possibilities, the most plausible
of them being the following.
(47) a) The subset relationship is satisfied at the underlying level.
b) The subset relationship is satisfied immediately prior to the
application of both rules.
c) The subset relationship is satisfied after the application
of the first rule, but prior to the application of the second
rule.
Under the third interpretation of the subset requirement (47(c)),
the subset relationship is in fact satisfied in the case of Clitic Lowering -
Sandhi Raising and Associative Raising - Associative Lowering. The structural
change of the first rule in both cases yields the focal segment required by
the second rule. Thus, the data from Shona provide evidence in support for
the third interpretation of the subset requirement.
The examples cited by Kiparsky in support of the Elsewhere Condition
do not decisively support one interpretation of the subset condition to the
exclusion of the other possible interpretations. He cites analyses of
Finnish, Diola-Fogny, Vedic Sanskrit and Classical Sanskrit which support
157
the Elsewhere Condition. In the case of Finnish, the subset relationship
is satisfied only prior to the application of both rules, and thus argue
for the second interpretation. The Diola-Fogny example allow the subset
condition to be evaluated either according to (47b) or (47c), i.e. either
before both rules apply, or before the second rule applies and after the
first rule applies. A similar situation obtains for the Vedic Sanskrit
example, where the application of the first rule does not in any way
affect the applicability of the second rule. Finally, in the case of the
Classical Sanskrit example, which I shall discuss at greater length, the
subset relationship is not observed for the input segments, under any
of the three interpretations of the subset condition (47a-c).
Kiparsky first cites two rules in Finnish, one assimilating final
k to the following consonant, and the second deleting final k. He proposes
that certain words terminate in underlying k, for example /menek/ 'go'.
The following data are offered in support of the assimilation and deletion
rules.
(48) menek#pois -V menep pois 'go away'
menek#alas -i mene alas 'go down'
menek# ■» mene 'go'
The following rules are proposed to account for these altternations.
(49) k 4 Ci / #Ci
k ♦ 0 / #
Final k is not deleted in the surface form menek kotlln 'go home', and
Kiparsky argues that the Elsewhere Condition predicts this state of affairs.
According to his analysis, the underlying form menek kotiin vacuously undergoes
assimilation. Since every k in the environment ^C^ alao stands in the
environment #, the subset requirement is satisfied, and the two rules are
applied disjunctively. It should be noticed, however, that the subset
relationship is satisfied only prior to the application of both rules, and
thus argues for the interpretation (47b) discussed above. The analysis
which Kiparsky gives cannot be made consistent with the third interpretation
of the subset condition, since if the subset condition is evaluated after
the application of the assimilation rule, there will be numerous forms which
do not satisfy the structural description of the second rule (deletion)
but which do undergo the first rule (assimilation). For example, the
surface form menep pois, from underlying menek pois, does not meet the
structural description of the deletion rule. Consequently, the subset
relationship must be evaluated before application of either deletion or
assimilation for the data in Finnish.
Kiparsky suggests the possibility that his purported underlying k
might be reanalyzed as an underlying h and states that 'this is of no
importance to the point at issue' (p. 95). However, the choice of under-
lying consonant is in fact crucial to his analysis, since if the consonant
in question were underlying h, and assimilation and deletion were
reformulated with the consonant h as the focus for both of these rules, the
surface form menek kotiin would not even satisfy the structural description
of deletion, and therefore the Elsewhere Condition would not be relevant
158
in explain that surface form.
In fact, Guerssel (1978) has argued that the Elsewhere Condition is not
needed to explain the nonapplication of deletion in the form menek kotiin.
Guerssel proposes two general conditions on phonological rules, which may
be roughly paraphrased as follows.
(50) i) a rule may not affect the identity of a geminate consonant cluster
unless it also affects the adjacency of the members of the
cluster and
ii) any assimilation rule deletes the boundary symbol between focus
and determinant which otherwise prevents adjacent identical
consonants from being interpreted as geminates, and thus subject
to the preceding condition.
With these two general conditions, Guerssel argues that the nonapplication
of deletion in the form menek kotiin is due to the fact that the application
of the assimilation rule deletes the boundary btween the words, thus rendering
the deletion rule inapplicable.
Guerssel also argues that the analysis of Diola-Fogny proposed by
Kiparsky does not in fact constitute an argument for the Elsewhere Condition,
and can instead be taken as support for the two conditions on geminates
(501) and (50ii). Under Kiparsky';S analysis, Diola-Fogny has two rules,
a deletion rule and an assimilation rule. The assimilation rule assimilates
a nasal consonant to the place of articulation of the following consonant,
changing underlying najum#to to najunto 'he stopped there'. According to
the deletion rule proposed by Kiparsky, a consonant is deleted in certain
types of consonant clusters, including +CC clusters, as well as +C and +#C
clusters. Thus, deletion changes underlying £k£t#bo to eke bo 'death there'.
Kiparsky invokes the Elsewhere Condition to explain the fact that the final
nasal in the surface form najun#to is not deleted, although the underlying
nasal of the phrase ban#na is deleted in the surface form ba#na 'finish
now'. Since, in the former example, the underlying nasal consonant under-
goes the assimilation rule, and since under Kiparsky' s analysis, the subset
condition is satisfied ( and is satisfied at all stages of the derivation,
rendering this example consistent with both interpretation (47(b)) and
(47(c))) , the deletion rule will be unable to apply to any nasal which
undergoes assimilation. Nasals which do not undergo assimialtion, on the
other hand, are subject to deletion, as in the example ba_na. Under the
reanalysis proposed by Guerssel, application of assimilation also entails
deletion of the word boundary symbol between the word najum and the word
to, rendering deletion inapplicable since the structural description of
the deletion rule is not satisfied by the form najunto. Thus, the data of
Diola-Fogny do not force the conclusion that the Elsewhere Condition is
necessary, in order to account for these data. Moreover, the analysis proposed
by Guerssel is simpler than that proposed by the Elsewhere condition, since,
as Guerssel points out, much of the material included in Kiparsky 's
version of deletion and assimilation is necessary only to ensure that the
subset relationship is satisfied.
Thus, out of the original four analyses proposed by Kiparsky in support
of the Elsewhere Condition, two of these analyses are shown not to support
the condition very strongly, since they have an alternative explanation in terms
159
Guerssel's constraint on assimilation rules.
The final and most problematic case for the Elsewhere Condition is the
interaction between certain assimilation rules in Classical Sanskrit and
a rule changing word final s^ to h. As observed by Kiparsky, final coronal
consonants assimilate in place of articulation to a following consonant.
He states that the following assimilations are encountered in Sanskrit.
(51)
t#t
->
an
t#c
■♦
c#c
s#t
■%
?n
s#c
-•
s#c
s#p
■»
<l.#p
or
b#p
s#k
■♦
x#k
or
h#k
s#s
■♦
s#s
or
lj#s
s#s
-%
$#$
or
h#s
s#
■♦
h#
Kiparsky proposes the following two rules to account for these alternations.
(52) [+coronal] -f [aplace] / #/
H-coronal ~j
l+continuantj ->
*rb,piace '~y
I +coronal |
[^-continuant/
Obligatory
[aplace]/ # [aplace] Optional
The argument for the Elsewhere Condition in the case of these rules is
that, just in case s^ is (vacuously) assimilated to a following dental stop,
as in the form devas tisthati 'god stands', the following rule will not apply.
(53) s •» h / #.
Under the analysis given by Kiparsky, the final consonant s^ of the underlying
form devas tisthati undergoes obligatory assimilation to the initial dental
stop _t. Application of this assimilation rule to the final s prevents that
consonant from being weakened to h by (53). In contrast, final s^ is weakened
to h before the consonant kh, as in the form devah khalu, from devas khalu,
since the final consonant s^ is not assimilated to the following velar.
In order for the obligatory assimilation rule to be ordered disjunctively
with final s^-weakening (53), it is necessary to revise the Elsewhere Condition,
with respect to the subset condition. The original subset condition is not
satisfied at the stage in the derivation before assimilation applies, since
there are forms which satisfy the structural description of assimilation but
not s-weakening, viz devat ca, which becomes devac ca 'and from the god'.
Moreover, there are also forms which undergo the assimilation rule which
do not satisfy the structural description of s^-weakening after application
of assimilation, viz. devac ca (and in general, application of assimialtion
does not create any new potential input for s^-weakening) . Therefore, neither
interpretation (47ii) or (47iii) is correct in determining whether the subset
condition in the Elsewhere Condition is satisfied.
The solution proposed by Kiparsky is that the subset requirement should
be modified, so that rules are disjunctively ordered if the environments of
the rules are in a subset relationship, although the foci need not be in a
160
subset relationship. Since the environment of the assimilation rule is in
a subset relationship to the s^-weakening rule, the revised Elsewhere Condition
will guarantee the disjunctive ordering relationship which is observed between
assimilation and weakening.
The optional assimilation rule also has bearing on the Elsewhere
Condition, although Kiparsky does not discuss these data. The consonant
s may optionally be assimilated to the following consonant, according to
the rule proposed in (52). If the consonant s^ is not assimilated, it will
be weakened to _h by application of the s^-weakening rule, (53).
(54) devah pita 'god the father'
deva'l' pita id.
devalj khalu 'god indeed'
devax khalu id.
devab kiip 'god what'
devax kirn id .
devalj sat 'six gods'
devas sat id.
devab sete 'the god slept'
devas sete id.
A fact not mentioned by Kiparsky is that when £ stands before s, it may
optionally assimilate to that consonant, and thus by the Elsewhere Condition,
will not undergo s-weakening. On the other hand, if final £ does not undergo
the optional assimilation rule, it will be weakened to h, since the Elsewhere
Condition will therefore not impose disjunctive ordering on the two rules.
(55) devas steghisyate 'the god will mount'
devah steghisyate id.
muktas syat 'let him be free'
muktah syat id.
In order to restrict s^-weakening without recourse to either the Elsewhere
Condition or Guerssel's constraints on assimilation rules, it is necessary
to reformulate weakening so that it is obligatory before any consonant except
_t, and optional before b, as in (56).
(56) s ■> h / #n-coronal] -s
[-anterior] I
|[-contin.] Oblig 1 |
([+contin.] Optional J y
It would clearly miss a generalization to attempt to build the conditions
into s^-weakening which are necesssary if that rule is to be directly restricted
rather than restricted derivatively by the application of the assimilation
rules in (52) .
It would therefore appear that the surface optionality of weakening
before ^ tremendously strengthens the case for the Elsewhere Condition. Since
the environment of optional assimilation is a subset of the environment of
weakening (#[aplace] versus # ), the rules are adjacently ordered and the
structural changes are incompatible, the rules Optional Assimilation and Weakening
will be disjunctively applied. This entails that if the first rule applies.
161
the second rule cannot apply to the output of the first rule. Therefore, when
Optional Assimilation applies to the form muktas syat. Weakening will not be
able to apply to that output. On the other hand, if Optional Assimilation
does not apply to that form, then Weakening will apply, yielding the form
muktah syat. In the latter case. Weakening is able to apply since the string
muktas syat is not an output of the application of Assimilation.
Needless to say, these data may also be accounted for by Guerssel's
constraint on assimilation rules. Under that constraint, application of
the optional assimilation rule entails deletion of the boundary which stands_
at the end of the word in the form muktas#syat, yielding the string muktassyat.
Weakening is no longer able to apply, since the word boundary required to
trigger that rule is not encountered in the string muktassyat. Therefore, the
Sanskrit assimilation examples do not provide unambiguous evidence for the
Elsewhere Condition , since the same data are automatically predicted by
Guerssel's conditions on assimilation rules.
I have demonstrated here that the evidence originally cited by Kiparsky
in support of the Elsewhere Condition can in fact be accounted for by another
principle, Guerssel's Constraint on Assimilation Rules. However, the Elsewhere
Condition is still motivated to account for the Shona data discussed in the
first two sections of this paper. Guerssel's constraint will not be applicable
to account for the disjunctive ordering relationship observed between those
pairs of rules. The constraint proposed by Guerssel states that:
Given a feature-changing rule of the form X •> Y / Z, if the focus
X takes on some or all of the features of the determinant Z, then
any boundary which intervenes between Y and Z in the output of
that rule will be deleted by convention, (p. 233)
This constraint will not be applicable to the Shona rules Clitic Lowering
(18) , since that rule changes a H tone to a L tone after a H tone, and
therefore does not constitute an assimilation rule. The constraint will be
applicable in the case of Associative Raising, since that rule changes a L
tone to a H tone after a H tone, and will therefore delete the morpheme
boundary which stands between the Associative Prefix and the noun stem,
changing underlying ne+bwe to nebwe . However, the deletion of the morpheme
boundary is irrelevant in guaranteeing the disjunctive relationship between
Associative Lowering and Associative Raising, since neither rule specifies
the presence of a word boundary as a condition for application of that rule.
Since the Constraint on Assimilation Rules cannot account for the disjunctive
ordering relationship between the pairs of Shona rules discussed here, and since
the hypothesis of Chomsky and Halle also cannot account for that relationship,
the Elsewhere Condition is supported as a principle of grammar.
Summary
I have argued here that two separate pairs of tone rules in Shona must
be applied disjunctively, and that this disjunctive relationship can be
predicted only by the Elsewhere Condition. These rules thus strengthen the
empirical support for the Elsewhere Condition as a principle of grammar.
The significance of the Shona rules for the Elsewhere Condition, and therefore
for phonological theory, lies not just in the fact that the Elsewhere Condition
makes the correct predictions, but rather in the fact that it makes these
162
predictions where other theories fail to make the correct predictions. It
is not merely a question of simplicity which argues for the Elsewhere Condition
in Shona, but reduces to the essential question of statability.
NOTES
*An earlier version of this paper was presented at the LSA Winter
Meeting in San Antonio, Texas on Dec. 28, 1980. Data for this paper has
been collected over the past four years from Kokerai Rugara, a native
speaker of the Karanga dialect of Shona, which is a Bantu language spoken
in Zimbabwe. This study has been supported in part by a University of Illinois
Fellowship in Linguistics provided by the Department of Linguistics, in part
by a FLAS Fellowship in African Languages provided by the African Studies
Center at the University of Illinois, and partially by NSF Grant BNS-7924523.
Additional information on the tonal structure of Shona may be obtained in
Odden (1981).
The principles governing the appearance of L tone before proper names
in the Associative construction are discussed in Odden (1981).
REFERENCES
CHOMSKY, N. and M. Halle. 1968. The sound pattern of English. New York:
Harper and Row.
GUERSSEL, M. 1978. A condition on assimilation rules. Linguistic analysis
4:225-254.
KIPARSKY, P. 1973. Elsewhere in phonology. in Festschfirt for Morris
Halle, S. Anderson and P. Kiparsky, eds. New York: Harper and Row.
ODDEN, D. 1980. Associative tone in Shona. Journal of linguistic research
vol.1, no. 2:37-51.
ODDEN, D. 1981. Problems of tone assignment in Shona. PhD dissertation.
University of Illinois (forthcoming) .
Studies in the Linguistic Sciences
Volume 11, Number 1, Spring 1981
TOPIC MARKER ELLIPSIS IN JAPANESE*
Michio Tsutsui
This is an attempt to account for the phenomenon of the
ellipsis of wa, a topic marker in Japanese. First, the func-
tion of wa is reexamined, and it is demonstrated that wa is
not an old information marker but a registered information
marker. Then, two conditions which govern the ellipsis of
wa are proposed. They are New Information Condition (NIC)
and Psychological Closeness Condition (PCC) . The claim of
NIC is that wa which marks new information is not ellipsed.
PCC, on the other hand, claims that the ellipsis of wa which
marks old information has to do with the psychological dis-
tance between the interlocutors and the topic. The concept
of psychological closeness is elucidated through the discus-
sion of PCC. PCC is based on a hypothesis with regard to
the question: when does the topic become psychologically
close to the interlocutor? The hypothesis which is called
Established Topic Hypothesis (ETH) , hypothesizes that the
establishedness of the topic determines the psychological
distance between the interlocutor and the topic, unless his
attitude toward the topic is objective.
1. Introduction
Mastering Japanese particles is probably one of the most difficult
tasks for Japanese learners. It is difficult not only because their
uses are peculiar, but also because there is the problem of ellipsis.
As a matter of fact, some particles are frequently ellipsed in natural
conversation. To provide an example,
(1) Kimi (wa) nani (o) taberu?
you Topic what Ace eat
'What will you eat?'
It is ironical that, on the one hand, the Japanese teacher repeatedly has
to emphasize that students should not drop any particles so that they
learn the correct uses of particles, and, on the other hand, he himself
has to drop them occasionally to make his talk natural. Unfortunately,
in spite of its evident importance, the problem of particle ellipsis has
barely been investigated. Naturally, no Japanese textbooks explain this
phenomenon satisfactorily. The explanation which most of the textbooks
provide is that particles drop in rapid and informal speech. However,
this is not the case. There are cases in which particles drop in formal,
fairly slow speech, and there are also cases in which particles never drop,
no matter how rapid and informal the speech may be. To take a topic
marker wa, for example, it often drops in (2), which is formal, slower
speech; whereas in (3) it cannot drop, even if it is uttered rapidly in
164
an Informal situation.
(2) Watakusi (wa) Smith to moosimasu.
I that say
'I call myself Smith. (= I am Smith.)'
(3) Tom(wa) ii kedo John wa dame da.
good but no good is
'Tom is okay, but John Is no good. '
In this paper, I specifically take wa as the object of study and examine
what rules govern the phenomenon of its ellipsis. In Section 2, I reexam-
ine the function of wa. Then, in Section 3, I present two conditions
which seem to govern the phenomenon in question. Section 4 summarizes
the whole discussion and concludes it.
2. Function of Wa
Before starting discussions about wa ellipsis, it is necessary to
clarify the uses of wa. This section is devoted to this purpose.
According to Kuno (1973a), wa has two uses, namely,
(4) Wa is either thematic or contrastive.
a. Thematic:
John wa gakusei desu.
student is
'John is a student. '
b. Contrastive:
Ame wa hutte imasu ga. . .
rain falling is but
'Rain is falling, but...' (Kuno 1973:59)
Chafe (1970), on the other hand, remarked that wa reflects old information
in contrast with £a, which reflects new information.^ However, as Kuno
(1973b) pointed out, this is not exactly correct. To cite his counter-
example,
(5) Watasi ni wa san-nin no musuko ga iru.
I with three persons of son Nom exist
Tyoonan wa kaisyain, zinan wa
first son company employee second son
daigaku-kyoosi, sannan wa ginkooraan de aru.
university professor third son bank employee is
'I have three sons. The first son is a company employee, the
second son is a university professor and the third son is a
bank employee. '
(Kuno 1973b: 217, n.5)
165
In (5) tvoonan wa, zinan wa and sannan wa all represent new information,
in spite of the fact that they are marked by wa. Alfonso (1966) also
implied an idea similar to Chafe's: "GA draws attention to the word that
precedes it, whereas WA directs attention to what follows after (p. 978)."
Although he did not use the terms old information and new information, the
implication of the above remark is the same as Chafe's idea mentioned above.
To see that wa is not simply an old information marker, let me show another
example:
(6) a. Kimi-tati no dare ga gakusei desu ka.
you (pi.) of who Norn student is Question
'Who among you is a student?
b. Watasi ga gakusei desu.
I Nom student am
' I am a student. '
c. Watasi wa gakusei desu.
I student am
' I am a student. '
To the question of (6a), both (6b) and (6c) are possible answers. The
difference between (6b) and (6c) is that (6b) means 'I and only I am a
student.', while (6c) implies 'As far as I am concerned, I am a student.'
In other words, in (6b) the speaker knows that the other people are not
students, while in (6c) the speaker does not know whether there is another
among them besides him. The important thing is that watasi wa in (6c)
carries new information just as watasi ga in (6b) does. The rest of the
sentence gakusei desu is old information. Generally speaking, the contras-
tive wa reflects new information. To provide some more examples,
(7) a. Explicit contrastive:
A: Soto wa doo desu ka.
outside how is Question
'How is it outside?'
B: Ame wa hutte imasu ga kaze wa arimasen.
rain falling is but wind don't exist
'It is raining, but there is no wind. '
b. Implicit contrastive:
John wa tenisu wa zyoozu desu.
tennis skillful is
'John (as far as tennis is concerned) is good at it.'
The thematic wa, on the other hand, usually reflects old information when it
marks the main topic of the sentence. For example,
(8) a. Anata wa gakusei desu ka.
you student are Question
'Are you a student?'
166
b. Kore wa waCasl no hon desu.
this I of book is
'This is my book. '
Thus, wa reflects both old information and new information.
Then, what is the basic function of wa which is common to the two
uses? In my understanding, the basic function of wa is that of marking
registered information. What I mean by registered information is the
information which is commonly registered in the interlocutor's mind.
There are two kinds of sources of registered information: (1) linguistic
contexts and (2) non-linguistic contexts. The information from linguistic
contexts (henceforth, linguistic information) consists of (1) anaphoric
information, which is the one stated explicitly in the previous sentences,
and (2) non-anaphoric information, which is the one registered in connec-
tion with the anaphoric information. The information from non-linguistic
contexts, or situational information (henceforth, situational information),
on the other hand, comprises (1) the information obtained through the
sensory organs, (2) the information related to the time at which the con-
versation takes place, (3) the information related to the place in which
the conversation takes place, and (4) the information registered in con-
nection with (1) , (2) and (3) . It should be stressed here that regis-
tered information is not the same as old information. To make this point
clear, let me use the sentences in (5) once again. In the second sentence
in (5), tyoonan 'first son', zinan 'second son' and sannan 'third son'
are registered information, because they were registered in the interloc-
utor's mind as nonanaphoric information when the noun phrase san-nln no
musuko 'three sons' was introduced in the preceding sentence. Neverthe-
less, these three nouns carry new information as discussed above.
Generally, a concept X, which is usually expressed by a word or
group of words, goes through three stages until it is established as a
topic, as seen in (9).
(9) Stage 1 (Pre-registered stage)
1 1 X is not qualified for a topic.
X is introduced.
Stage 2 (Post-registered stage)
X is qualified for a topic. Other concepts
which are related to X are also registered
I and ready to be topics.
il
X is presented as a topic.
Stage 3 (Established stage)
X is established as a topic.
Needless to say, there might be cases in which X is registered as the
result of the registration of another concept Y to which X is related.
167
The generalization in (9) applies to both the linguistic information
and the situational information. In the former case, X is introduced
verbally, while in the latter case, X is introduced situationally . It
is noted that there are some concepts which are always registered in the
interlocutor's mind, and therefore do not experience the pre-regis tered
stage. Among them are those of the speaker, the hearer, the present
moment and the place where the conversation is taking place. Words like
watasi 'I', anata 'you', ima 'now' and koko 'here' can therefore be pre-
sented as a topic anytime. Some generic nouns are also pre-registered.
Thus, hito 'person(s)' and kuruma 'car(s)', for example, can be the topic
of the sentence without being registered in advance. If the speaker,
believing that a concept X is registered in the hearer's mind, presents
it as a topic, but it is not registered in the hearer's mind, then the
communication will break down, or the speaker will sound more or less
2
egotistic at least.''
Now, getting back to the claim that wa is a marker of registered
information, let us examine some cases to see if this is true or not.
First, let us examine some examples of thematic wa.
(10) a. John wa gakusei desu.
student is
'John is a student. '
b. Kuzira wa honyuu-doobutu desu.
whale mammal is
'A whale is a mammal. '
c. Boku wa ima san-nen de senkoo wa keizai desu ga. . .
I now junior am and major economics is but
' I am a junior now and my major is economics but...'
(10a) is acceptable if John has been previously mentioned and, as a result,
registered in the interlocutor's mind. It is not acceptable if John has
not been introduced yet. Thus, wa in (10a) is necessarily marking regis-
tered information. (10b) , on the other hand, is the case of pre-registered
information. Since kuzira 'whale' is a generic noun, it is pre-registered
in the interlocutor's mind if he knows it. (10c) has two wa's. The first
one is straightforward. It is marking boku 'l', a word which is associ-
ated with one of the permanently-registered concepts, the speaker. The
problem is the second wa marking senkoo 'major'. According to Kuno (1972),
"The theme must be either anaphoric (i.e. previously mentioned) or generic
(p. 270)." However, the phrase senkoo wa in (10c) is neither anaphoric nor
generic.-^ As a matter of fact, it is new information, presenting a "sub-
topic" of the sentence. If we understand wa as a registered information
marker, however, the problem is solved. Since it is common knowledge that
someone's being a junior of a university normally means he is majoring in
something, the concept of major is registered when the fact that the
speaker is a junior is Introduced in the first clause in (10c). If senkoo
wa is introduced in the first clause, the sentence is unacceptable as
a conversation opener.
168
(11) *Senkoo wa keizai de
major economics is and
boku wa ima san-nen desu ga. . .
I now junior am but
'My major is economics and I am a junior now but...'
Next, let us examine the case of contrastive wa.
(12) a. Ame wa hutte imasu ga kaze wa arimasen.
rain falling is but wind don't exist
'It is raining, but there is no wind.'
b. John wa tenisu wa zyoozu desu.
tennis skillful is
'John (as far as tennis is concerned) is good at tennis.'
Kuno (1973a) states that "the contrastive wa can place nonanaphoric noun
phrases in contrast (p. 46)." If this is the case, (12a) should be able
to be uttered as an opening of the conversation. In my understanding,
however, it is possible only if there is some non-linguistic context which
registers ame 'rain' and/or kaze 'wind' in the interlocutor's mind. In
other words, (12a) is unacceptable as an opening of conversation if there
is no such non-linguistic context. (12a) is acceptable, of course, if it
is preceded by some statement or question which registers ame and/or kaze.
as the dialog in (7a) shows. Generally, when A and B are contrasted, it
is not always necessary that both A and B should be registered in advance,
because they are usually related concepts and if one of them is registered,
the other will be automatically registered. It does not make sense to con-
trast two concepts which are totally unrelated to each other. Likewise,
(12b) is acceptable only if John is under discussion and the preceding
conversation or the situation has to do with his tennis skill. Otherwise,
(12b) will be an egocentric and inconsiderate statement to the hearer, if
not unacceptable.
Finally, let me introduce an interesting sentence in Kuno (1973a):
(13) Ame wa hutte imasu ga, kasa wa motte ikimasen.
rain falling is but umbrella having don't go
'The rain is falling, but I am not taking my umbrella with me. '
(Kuno 1973a:46)
Assume that no mention has been made in the previous conversation of the
rain that is falling now. In this case, ame is neither generic nor
anaphoric. Neither is it the case that it is in contrast with kasa
'umbrella'. To account for this sentence, Kuno assumes that Japanese
allows transpositions of contrastive wa. According to him, (13) is de-
rived from (14) by the wa transposition. ^
(14) Ame ga hutte wa imasu ga, kasa o motte wa
rain Norn falling is but umbrella Ace taking
ikimasen.
go-not
'The rain is falling, but I am not taking my umbrella with me. '
169
According to my analysis, the two wa^' s in (14) are also marking registered
information. Ame and kasa are marked by wa simply because ame is situation-
ally registered, and kasa is registered in connection with the first clause
when it is introduced.
To summarize the uses of wa in terms of the newness of the information
it marks:
(15)
registered information
new information
old information
contrastive
Ex. (7a). (7b)
thematic
Ex. (10c), (13)
Ex. (10a), (10b)
* There is no case in which the contrastive wa marks old information.
3. Conditions of Wa Ellipsis
In the previous section, I examined the function of wa as the prelimi-
nary for the discussions on wa ellipsis. In this section, I discuss two
conditions which seem to govern the phenomenon in question. Section 3.1
has a brief introduction of the two conditions and the discussion of one
of them. New Information Condition; Section 3.2, the other condition,
Psychological Closeness Condition.
3.1 To account for the phenomenon of wa ellipsis introduced in Section 1,
I propose the following two conditions:
(16) a. W£ is not ellipsed when it marks new information.
b. Wa which marks old information is ellipsed when the topic
is psychologically close to the speaker and he believes
that it is psychologically close to the hearer too, at
the moment of utterance.
I call the first condition New Information Condition (NIC) and the second one.
Psychological Closeness Condition (PCC) . NIC is a negative condition, that
is, wa is not ellipsed when it is met, while PCC is a positive condition,
that is, wa IS ellipsed when it is met. In what follows, I will discuss
NIC more in detail.
As I discussed in Section 2, there are two cases in which wa marks
new information: (1) the case in which wa is contrastive, and (2) the
case in which wa is thematic. Now, let us see some examples of the first
case. Look at the following sentence;
(17) Ano hon wa omosiroi ga kono hon wa tumaranai.
that book interesting but this book boring
'That book is interesting, but this book is boring. '
170
The two wa's in this sentence are obviously contrastive and mark new
information. Therefore, they cannot drop.
(18) *Ano hon 0 omosiroi ga kono hon 0 tumaranai.
The following sentence shows an example of the implicit contrastive:
(19) Boku wa kono wain wa suki desu.
I this wine fond am
'I (as far as this wine is concerned) like it.'
Here again, wain wa. is new information and this wa cannot drop to mean
(19) as seen in (20) .
(20) *Boku wa kono wain 0 suki desu.
'I (as far as this wine is concerned) like it.'
The next example involves phrase ellipsis.
(21) Boku wa koohii o nomimasu. Kimi wa?
I coffee Ace drink you
'I'll drink coffee. (What will) you (drink)?'
Notice that the second sentence in (21) is an ellipsed sentence. The com-
plete form would be something like:
(22) Kimi wa nani o nomimasu ka.
you what Ace drink Question
'What will you drink?'
But, in this case, the ellipsed part is clear to the hearer from the context
and is usually omitted in conversation. In other words, nani £ nomimasu ka
was ellipsed because it is recoverable old information, whereas kimi wa is
new information and unrecoverable if it is omitted. Thus, if wa is omitted
in the second sentence, it becomes unacceptable.
(23) *Boku wa koohii o nomimasu. Kimi 0?
Ellipsed parts like nani o nomimasu ka in (21) are sometimes recoverable
from the non-linguistic context. For example, when a teacher comes in the
classroom and finds a student is missing, he may ask, "John wa?" The com-
plete form of this sentence would be something like:
(24) John wa doo sita?
how did
'How did John do? (What's the matter with John?)'
Here again, wa marks the new information John and cannot drop in any case.
(25) *John 0?
171
Now, leC us Cum to the thematic wa. Here I use the sentence in
(10c) once again under a new number.
(26) Boku wa ima san-nen de senkoo i wa> keizai desu ga. .
i:;i
I now junior is and major economics is but
'I am a junior now and my major is economics, but...'
As is discussed in Section 2, the wa marking senkoo. which is a sub-topic
of the sentence, is thematic, but it reflects new information. The second
wa is therefore not omissible. As a matter of fact, if senkoo is the main
topic of the sentence, and old information, it can drop in some situations.
(27) (Boku no) senkoo 0 keizai desu.
I of major economics is
'My major is economics.'
Finally, let us examine the sentence in (28) (« (13)):
(28) Ame fwa I hutte imasu ga, kasa / "^ ) motte ikimasen.
rain falling umbrella having don't go
'The rain is falling, but I am not taking my umbrella
with me. '
Although the two wa's in (28) are thematic, ame wa and kasa wa are both
new information. That is the reason why they cannot drop.
3.2 In Section 3.1, I argued that New Information Condition is a condition
which governs wa ellipsis. In this section, I argue that there is another
condition which determines wa ellipsis, namely. Psychological Closeness
Condition.
Before going into the discussion, it is necessary to elucidate the
concept of psychological closeness first. In this paper I use the term
psychological closeness in the following sense: The referent of a topic
X is psychologically close to a person P when P is not conscious of X as a
being separate from himself. (Henceforth, I sometimes simply say, "the
topic is close to the speaker " to mean "the referent of the topic is
psychologically close to the speaker.") There is another term which repre-
sents the same concept, that is, empathy. In defining the concept of
empathy, Kuno (1976) states that: "I use the term "empathy" to charac-
terize the speaker's identification; in varying degrees, with a partici-
pant in an event, (p. 431)" A slight terminological difference between
empathy and psychological closeness is that the former is a speaker-
oriented term, while the latter is not necessarily so. I use the term
psychological closeness in this paper (1) because I'd like to deal with
anything which can be a topic (e.g., abstract ideas, events, actions),
rather than just "a participant in an event" and thus it seems to be a
good idea to use another term, and (2) because the term psychological
closeness is more convenient to describe the situation from various view-
points. For example, the idea (29a) is expressed as in (29b), (29c) and
(29d) using this term.
172
(29) a. The speaker is empathetlc with the referent of the topic.
b. The referent of the topic is psychologically close to the
speaker.
c. The speaker is psychologically close to the referent of
the topic.
d. The psychological distance between the speaker and the
referent of the topic is small.
We have seen in the above what psychological closeness is. The next
question is: Then, when does something become psychologically close to
a person? With regard to this question, I'd like to present the following
hypothesis :
(30) A concept becomes psychologically close to the interlocutor
when it is well established as a topic in his mind, unless
his attitude toward the topic is objective.
I call this hypothesis Established Topic Hypothesis (ETH) . ETH will need
more explanation. In Section 2, I discussed how a concept is established
as a topic in the interlocutor's mind. ETH is the hypothesis concerning
the further stage beyond the stage 3 (Established stage). Generally speak-
ing, a topic is established when it is presented as a topic. Then, the
longer it remains as a topic without being replaced by another, the more
firmly it is established in the interlocutor's mind. Then, what happens
in the interlocutor's mind is that he becomes unaware of the exis'tance
of the concept the topic represents, as a different being from himself.
In other words, the psychological distance between the topic and the inter-
locutor becomes small.
The same psychological processes as mentioned above seem to take
place in those concepts which have been pre-registered if the interlocutor
reinforces them by continuously doing things like seeing them, hearing
them, feeling them. Thus, such concepts as 'I', 'you', 'now' and 'here'
are well established as a topic in the interlocutor's mind even though
they are not explicitly presented as a topic. Likewise, a concept which
is reinforced by continuous thinking becomes well established as a topic
even if it is not explicitly stated. To put it in another way, a concept
which occupies a person's mind is psychologically close to him. On the
other hand, even though they are pre-registered, generic nouns are usually
not well established as topics because they are not reinforced by contin-
uous thinking or observation. Specific examples will be shown later.
Now, let us examine some data and see if what has been discussed in
the above is the case. First, look at the following sentences:
(31) a. Watasi j wai sensei ni narimasu.
(0 i
I teacher to become
'I will become a teacher.'
173
b. Smith-san. anata ( wal sensei ni narimasu ka.
{rl
Mr. Smith you Question
'Mr. Smith, will you be a teacher?'
Kanai I wa) sensei ni narimasu,
my wife
'My wife will become a teacher. '
Smith-san J wa ) sensei ni narimasu.
(?01
'Mr. Smith will become a teacher. '
Smith to iu hito (wa) sensei ni narimasu.
1*0)
Quotative say person
'A person called Smith will become a teacher. '
In (31a) and (31b), watasi 'I' and anata 'you' are the existances whose con-
cepts are pre-registered and well established in anybody's mind. Therefore,
wa ellipsis is very natural in (31a) and (31b). Similarly, kanai 'my wife',
whom the speaker constantly thinks of and sees, usually has a psychological
closeness to him. Thus, if the speaker believes she is close to the hearer
too, wa drops. If she is actually not so close to the hearer, contrary to
the speaker's belief, the statement without wa sounds somewhat egocentric
to the hearer. On the other hand, the acceptability of" wa ellipsis in
(31d) depends on the situation. If Mr. Smith is a familiar person whom
the interlocutors often see, talk about and/or think of, or have been talk-
ing about, then, wa ellipsis is natural. However, if he is only one of
remote acquaintances of the interlocutors, wa ellipsis is odd. Finally,
in (31e), Smith to iu hito 'a person called Smith' is the expression which
is used when the interlocutors are not very familiar with him, and when
Smith to iii hito is presented as a topic for the first time. Thus, wa
ellipsis in (31e) is totally unacceptable.
The next set of sentences has to do with demonstratives.
(32) a. Kore ( wa) boku no kamera desu.
10 t
this I of camera is
'This is my camera. '
b. Ano hito (wa) watasi no sensei desu.
(wa) watasi
10 \
that person I of teacher is
'That person is my teacher. '
c. Sono hon )wa) omosiroi desu ka.
10 I
that book interesting is Question
'Is that book interesting?'
174
Since kore 'this' in (32a) indicates what the interlocutors have been
looking at, it has been situationally registered and reinforced in their
minds. Thus, it is well established as a topic and wa can drop. Likewise,
ano hito 'that person' and sono hon 'that book' in (32b) and (32c) have
been seen by the interlocutors for some time and well established as a
topic; therefore, wa ellipsis is quite natural in both sentences. The
sentence in (32b) is also used when ano hito 'that person' is not in sight.
Even in this case, wa ellipsis is possible, if that person is well estab-
lished for some reason. (For example, the person was with them a minute
ago, or the person has been a topic of the conversation.) The sentence
in (32a) is also used when sono hon 'that/the book' is not in sight. This
sentence, in this case, is usually used when the speaker is not very famil-
iar with the book. Thus, wa often remains in this case.
Next, let me examine some sentences in which pre-registered information
is involved.
(33) a. Kyoo j wal watasi no tanzyoobi desu.
1(9 ]
today I of birthday is
'Today is my birthday. '
b. Sengetu no itu-ka j^al watasi no tanzyoobi desita.
1*0)
last month of fifth day was
'The fifth of the last month was my birthday. '
c. Koko Wa) samul desu ne.
ID
this place cold is Tag Question
'This place is cold, isn't it?'
d. New York (wal samui desu ne.
1*0]
'New York is cold, isn't it?'
e. Kuzira ( waihonyuu-doobutu desu.
1*0]
whale mammal is
'A whale is a mammal. '
As is discussed in Section 2 and earlier in this section, some concepts
such as this place and this moment are pre-registered and already well
established as a topic. Therefore, in (33a) and (33c), wa ellipses are
very natural, even if those sentences are openings of conversation. How-
ever, in normal circumstances, sengetu no itu-ka 'fifth day of the last
month' and New York are not pre-registered in the interlocutor's mind;
therefore, there is no way for them to be well established at the begin-
ning of conversation. Thus, in (33b) and (33d), wa ellipsis is impossi-
ble.^ Next, look at (33e) . Here, kuzira 'whale' is pre-registered in
the interlocutor's mind as a generic topic. Nevertheless, wa ellipsis
is normally unacceptable. This is because kuzira is usually not rein-
forced in the interlocutor's mind as a topic by constant thinking and/or
observation. ^
175
The next examples show how the situational information affects the
interlocutor's psychology in terms of wa ellipsis.
(34) a. Shuzvutu-situ ; wa'/ doko desu ka.
( 0 S
operating room where is Question
'Where is the operating room? '
b. Ginkoo fwai doko desu ka.
f wa i doko
1*01
bank where is Question
'Where is a bank? '
Suppose that the speaker is in a hospital and utters (34a) to a nurse.
In this case, wa ellipsis is quite natural. This is because shuzyutu-
situ 'operating room' is registered and well established in the interloc-
utor's mind in the hospital situation. On the other hand, if (34b) is
uttered to a stranger on the street, wa ellipsis is quite odd, because
it sounds as if a bank were a necessary existance on the street and every-
body should be feimiliar with it.
As the final piece of evidence for ETH, let me provide another set
of sentences. They are supposed to be uttered to somebody who is looking
for something.
(35) a. Pen wa koko desu.
pen here is
'The/A pen is here. '
b. Pen 0 koko desu.
'The/A pen is here. '
Both sentences mean 'The/A pen is here. ' But the speaker's psychology at
the moment of utterance is different. That is, (35a) is used when the
speaker does not know exactly what the person is looking for. Thus, the
implication is something like 'I'm not sure what you are looking for, but
if what you are looking for is the/a pen, it is here. ' On the other hand,
(35b) is used when the speaker is sure what the person is looking for. The
implication is thus 'I know you are looking for the/a pen, and it is here.'
Thus, if the/a pen is the main concern of the interlocutors and well estab-
lished as a topic at the moment of speech, wa drops; whereas, if the/a
pen is not certainly the main concern of both interlocutors, and therefore
not well established as a topic, it does not drop.
So far, I have examined several examples of wa ellipsis to demonstrate
that ETH is valid as a basis of PCC, supposing that the speaker's attitude
is not objective. However, there is another condition, as is stated in
(30), which supersedes the condition of ETH. That is, if the speaker's
attitude toward the topic is objective, the topic is necessarily distant
from the speaker, because being objective about the topic and maintaining
a psychological distance from the topic are synonymous. In what follows,
I will examine some cases which reflect this condition.
176
First of all, let me remind the reader of the fact that wa ellipsis
is the phenomenon which is usually observed only in spoken language. 9
What does this mean? In general, written language requires a more formal
and objective attitude than spoken language does. In other words, when
one writes, he usually maintains a certain amount of psychological distance
between the topic and himself. This is why wa ellipsis is not observed in
written language.
Next, let us examine the case which has to do with the formality
level of speech,
(36) a. Titi f wa 1 ima byooki desu.
^ wai
1*0 J
father now ill is
'My father is ill now. '
Too-san fwa) ima byooki desu.
{T\
Daddy now ill is
'Daddy is ill now. '
The difference between the two sentences in (36) is that in (36a) a formal
word titi is used to refer to the speaker's father, whereas in (36b) a
casual word too-san is used to refer to the same person. Generally, when
the speaker talks in a formal style, his attitude is objective and the
psychological distance between the topic and himself is large. Thus, in
(36a) wa cannot drop, whereas in (36b) it can drop. Similarly, when one
refers to someone using his full name, the statement becomes formal and
objective; when one refers to someone using his nickname, the statement
becomes informal and empathetic. Thus, in (37a) wa ellipsis is unaccept-
able, while in (37b) it is possible.
(37) a. Susan Smith f wai kono gakkoo no gakusei desu.
[h]
this school of student is
'Sussm Smith is a student of this school.'
b. Susie ( wa) kono gakkoo no gakusei desu.
\0 )
'Susie is a student of this school. '
The next example is a general statement.
(38) a. Tuki / wa"i totemo kirei desu.
[h]
moon very pretty is
'The moon is very pretty.'
b. Kyoo no tuki fwaj totemo kirei desu.
U ]
today of moon very pretty is
'Today's moon is very pretty.'
177
When the speaker talks about something in general, it is an objective
statement. Therefore, the topic is psychologically distant from the
speaker in this case and wa cannot drop, as seen in (38a). However, if
the statement is specific as in (38b), wa ellipsis is possible. The
following is another example of this sort:
(39) a. Sono shu no koto ( wa ) dare-demo sitte-iraasu.
{lis
that kind of thing anybody know
'Anybody knows things of that kind. '
b. Sonna koto ( wa I dare-demo sitte-imasu.
[T]
such thing anybody know
'Anybody knows such a thing.'
In (39a) sono shu no koto represents a group of unspecific things in
general. On the other hand, in (39b) sonna koto is actually an emphatic
indication of sono koto 'the thing' and very specific. Therefore, wa
ellipsis is possible in (39b), but not in (39a).
Finally, let me present the case of recollection.
(40) Boku ] wal damatte gurasu o mitumete-ita.
[%]
I silently glass Ace looking at was
Sinda tuma no koto ga omoidas-areta.
dead wife of thing Nom recalled was
'I was looking at my glass silently. Things about my dead
wife came to my mind. '
As (40) shows, wa cannot drop in the context of recollection, even though
the topic is boku 'I', which is pre-registered and well established as a
topic in the speaker's mind. This is because in this context, the speaker's
attitude is objective and he maintains a certain amount of psychological
distance between the topic and himself. Notice that in the context of
recollection, even the speaker himself is psychologically distant; there-
fore, wa is never ellipsed.
In this section (Section 3) , it has been demonstrated that two con-
cepts have relevance to the phenomenon of wa ellipsis: (1) the newness of
the topic as information, and (2) the interlocutor's psychology toward the
topic at the moment of speech. Clearly, neither of them is in the domain
of syntax or semantics. Rather, the former is basically a concept of dis-
course, and the latter that of pragmatics. The questions are therefore
meaningful only if the data is examined in a discourse, given a situation
at the moment of speech. In other words, the phenomenon of wa ellipsis has
to be treated with the linguistic context and non-linguistic context.
4. Summary and Conclusion
To sum up what has been discussed in the previous sections, wa has
two uses: thematic and contrastive, and marks both old information and new
information. Wa is thus not an old information marker, but a registered
178
information marker. There are two kinds of registered information: lin-
guistic information and non-linguistic (or situational) information.
Linguistic information is further grouped into two kinds: anaphoric infor-
mation and nonanaphoric information. Such concepts as those of the speaker,
the hearer and the present time and the place where the conversation is
taking place are always registered as non-linguistic information. Some
generic nouns are also pre-registered.
To account for the phenomenon of wa ellipsis, I proposed two condi-
tions: (1) New Information Condition (NIC) and (2) Psychological Closeness
Condition (PCC) . NIC claims that wa is not ellipsed when it marks new
information. The claim of PCC, on the other hand, is that wa marking
old information is ellipsed when the topic is psychologically close to
the speaker and he believes that it is psychologically close to the hearer
too, at the moment of utterance. The concept of psychological closeness
is essentially the same as that of empathy. That is, the referent of a
topic X is psychologically close to a person P when P is not conscious of
X as a being separate from himself.
PCC is based on Established Topic Hypothesis (ETH) , which hypothesizes
that a concept becomes psychologically close to the interlocutor when it
is well established as a topic in his mind, unless his attitude toward the
topic is objective. When the speaker's attitude toward the topic is objec-
tive, the topic is psychologically distant from him no matter how firmly
the topic may be established in his mind, and wa ellipsis does not take
place. To locate NIC and PCC, NIC belongs to discourse grammar, while
PCC belongs to pragmatics.
To conclude the discussions, the phenomenon of wa ellipsis seems to
be the reflection of two things: (1) the newness of the topic as infor-
mation, and (2) the speaker's psychology toward the topic at the moment of
utterance.
NOTES
*I am indebted to Seiichi Makino for a number of comments and sugges-
tions throughout the study of this problem. Needless to say, an error or
mistake that this paper might contain is mine.
^Chafe (1970, p. 233).
This is a case in which the speaker violates the maxim ''Be relevant'
of Grice's (1975) Cooperative Principle.
■^Kuno (1973a) states that: "The contrastive wa can place non-
anaphoric noun phrases in contrast, (p. 46)" However, wa marking senkoo
'major' in (10c) is not contrastive either.
Generally speaking, sub-topics seem to carry new information.
^Kuno (1973a, pp. 46-47 n.7).
In (26), boku no 'my' is usually ellipsed as redundant information
by some discourse rule.
179
7
If the speaker is in New York at the moment of utterance, wa
ellipsis in (33d) is acceptable. This is because New York is this
place in this situation.
Q
There are situations in which wa can drop even in the case of
generic nouns. For example,
(i) Eh, kuzira 0 sakana zya nai no?
What? whale fish is not Question
'What? Isn't a whale a fish?'
In this case, the speaker's mind is occupied by the surprising idea that
a whale is not a fish, which was mentioned by someone right before, and
the concept of whale is psychologically very close to him at the moment
of speech.
9
The only exception I have found in modern literature is the stage
direction of a scenario. To the question why particle ellipses take
place in stage directions, I have no answer at the moment.
REFERENCES
ALFONSO, Anthony. 1966. Japanese language patterns. Tokyo: Sophia
University.
CHAFE, Wallace L. 1970. Meaning and the structure of language. Chicago,
IL: The University of Chicago Press.
GRICE, H. Paul. 1975. Logic and conversation. In P. Cole and J.L. Morgan,
eds . : Syntax and semantics: Speech acts, vol. 3. New York, NY:
Academic Press. 41-58.
KUNO, Susumu. 1972. Functional sentence perspective: A case study from
Japanese and English. Linguistic Inquiry 3.269-320.
KUNO, Susumu. 1973a. The structure of the Japanese language. Cambridge,
MA: MIT Press.
KUNO, Susumu. 1973b. Nihon Bumpoo Kenkyuu. Tokyo: Taishuukan.
KUNO, Susumu. 1976. Subject, theme, and the speaker's empathy. In C.N.
Li, ed.: Subject and topic. New York, NY: Academic Press. 417-444.
Studies in the Linguistic Sciences
Volume 11, Number 1, Spring 1981
IKOROVERE MAKUA TONOLOGY (PART 3)*
Chin-Chuan Cheng
and
Charles W. Klsseberth
This paper concludes the description and analysis of Ikorovere
(Makua) verbal tonology begun in Cheng and Klsseberth (1979, 1980).
In particular, we deal with two major phenomena — (a) the shortening
of long vowels before certain consonant clusters (this process was
first introduced in Cheng and Klsseberth (1980), where it is referred
to as VOWEL SHORTENING) and (b) the various fates met by primary high
tones located (or expected to be located) on word-final vowels.
Evidence is given for a rule of HIGH TONE RETRACTION as well as for
a rule proposed in Cheng and Klsseberth (1979) called WORD FINAL
LOWERING. HIGH TONE RETRACTION has the effect of retracting a
primary high tone from the final vowel of a word and placing that
high tone on the penult vowel of the word. WORD FINAL LOWERING has
the effect of lowering a primary high tone on a word-final vowel.
We conclude the paper with some general observations about the
analysis of Makua tonology that has been developed in this and in
the preceding papers in the series, and we summarize briefly what we
think Makua suggests about the nature of phonological systems.
The present paper (which continues the section numbering and example
numbering of Cheng and Klsseberth (1979, 1980)) concludes our description
and analysis of the major tonal features of the Ikorovere (Makua) verb, as
well as segmental phenomena that interconnect in a significant manner with
these tonal features. We concentrate here on two major topics: the inter-
play between the rule of VOWEL SHORTENING (VS) , which we introduced in
section 7, and various segmental and tonal processes operative in the language;
and the behavior of primary (i.e. non-doubled) high tones on word-final vowels.
We conclude the paper with some general observations about the analysis of
Makua tonology that we have developed in this and the preceding papers in
the series, and we summarize briefly what we think Makua suggests about the
nature of phonological systems.
8. MORE ON VOWEL SHORTENING
In section 7 we introduced a rule of VOWEL SHORTENING (VS) which deletes
one mora of a long vowel when that vowel stands before either a geminate
consonant or a cluster consisting of a nasal plus a true consonant. This
rule, however, does not operate in the case of a long vowel that precedes
a nasal plus true consonant when this sequence is in what Kiparsky (1976)
analyzes as a non-derived environment (i.e. when the sequence is entirely
within the same morpheme, and thus has not arisen through the combination of
182
morphemes; and, furthermore, the sequence has not been created by the
application of a phonological rule). VS must, therefore, be restricted
to applying just to inputs that have been derived, either through morpho-
logical processes or phonological processes. The only context cited in
section 7 for the application of VS involved cases where a long vowel
precedes geminates or nasal clusters that have their origin in verb stems
of the structure -VCiCi... or -VNC. . . There are, however, several additional
contexts where VS comes into play. In this section we will examine some of
these other contexts.
One environment where VS applies is provided by a verbal tense that
involves a sequence of tense/aspect (TA) markers: /-aa-mu-/. This tense is
illustrated in (127) below.
(127) k-aa-mw-aap-a 'I was whispering'
y-aa-mw-eet-a nth^ukwa 'he was walking with his stomach protruding'
k-aa-mw-eet-a-ru 'although I was walking'
k-aa-mw-eetet^-a 'I was threshing'
k-a-9-kwec-a 'I was sweeping'
y-a-n-kwec-a vyaa vyaal 'he was sweeping'
y-a-n-tii^g-a 'he was expressing shyness'
k-a-n-lupat^-a 'I was hunting'
y-a-n-tutum-a tu tu tul 'it was foaming'
w-a-y-kukut-a ku ku kul 'it was thundering'
k-a-n-lokotanlh-a ' I was picking up '
y-a-in-pitxkuw-a pitiku pitlkul 'it was rolling'
y-a-m-p ameelac-a phame phamel 'he was sticking things on'
k-aa-mu-n-limel-a 'I was weeding it (noun class 1)'
y-aa-mu-m-purukutul-a purukutui 'he was chewing it (noun class 1)'
We analyze the forms in (127) as having the segmental structure
SP-aa-mu-(OP)-VS-a (where SP=subject prefix, OP=object prefix, and VS=verb
stem). The subject prefixes have their pre-vocalic allomorphs (cf. section A)
before the /-aa-/ prefix. The /-aa-/ TA prefix is extensively used in Makua
(cf. sections 5 and 6 for additional examples of tenses utilizing this prefix).
In the present case the /-aa-/ is used in conjunction with another prefix that
we regard as having the basic representation /-mu-/. Just like the object
prefix /-mu-/ discussed in section 3, this TA prefix /-mu-/ is systematically
modified as the consequence of two rules.
One of these rules is li-GLIDING (u^-GL) , which converts the u vowel to
w in front of another vowel, compensatorily lengthening that vowel in the
process. u-GL has been independently motivated in sections 2 and 3 of this
paper, where we show that it affects the infinitive prefix 11- and the 2 p.
object prefix -u-. The rule of u-GL will take an input structure such as
/k-aa-mu-ap-a/ and convert it to /k-aa-mw-aap-a/. (For the time being, we
are ignoring matters of tone.)
183
The second rule elides the u of /-mu-/ when it is followed by a
consonant (actually, this consonant must in turn be followed by a vowel
or glide — see below). Call this rule u-ELISION (u-EL) . This rule, while
unnamed at the time, was discussed briefly in section 3 in connection with
the object prefix /-mu-/. When u-EL applies, a cluster of m plus a consonant
is created; the ra then assimilates the point of articulation of the following
consonant. Call this rule Nasal Assimilation (NA) . NA is not, however,
restricted just to clusters of m plus consonant. It also affects clusters
of n plus a consonant. For example, there is a present tense morpheme -no-
which optionally elides its vowel before consonants. When the £ is elided,
the n will assimilate to the following consonant. For example, ki-no-hak-a
'I am grinding' has an alternant pronunciation kl-n-hak-a.^ Given the rules
of u-EL and NA, an underlying form such as /k-aa-mu-kwec-a/ will first under-
go u-EL to yield /k-aa-m-kwec-a/ and this representation will in turn undergo
NA to yield /k-aa-n-kwec-a/ . This is not, of course, the correct surface
form — the change of aa to a will be discussed below.
As a consequence of u-GL (which applies when /-mu-/ precedes a vowel) and
u-EL (which applies when /-mu-/ precedes a consonant), the prefix /-mu-/
rarely shows up unaltered on the phonetic surface. The only situation where
/-mu-/ remains unaltered is when it is followed by a (true) consonant cluster,
as in the examples k-aa-mu-n-llmel-a and y-aa-mu-in-purukutul-a . However,
there is only one situation where the TA prefix /-mu-/ is followed by a
(true) consonant cluster — namely, when between the /-mu-/ and the verb stem
there appears a "reduced" form of the object prefix /-mu-/. Both of the
examples cited above fit this description. The underlying form of
k-aa-mu-n-limel-a , for instance, is /k-aa-mu-rau-limel-a/. Such an example
illustrates that the rule of u-EL must be applied iteratively, working from
the end of the word forwards, since the object prefix /-mu-/ must undergo the
rule before the rule tries to apply to the TA prefix /-mu-/. It is only as
a result of the application of u-EL to the object prefix that a consonant
cluster arises and blocks the application of u-EL to the TA prefix. This
right-to-left iterative application of u-EL follows the general pattern
observed in earlier sections of this study where such rules as VOWEL ASSIMILATION,
TONE LOWERING, and u-GL were shown to operate in similar fashion.^
As noted above, application of u-GL and NA to an underlying form such
as /k-aa-mu-kwec-a/ does not account for all of the phonetic changes that
occur. In particular, the prefix /-aa-/ must shorten to -a- when it is
followed by the consonant cluster that arises from u-EL. But, of course,
this is just the rule of VS that we motivated in section 7 1 The derivation of
k-a-n-kwec-a (as far as the segmental aspects of the matter are concerned)
can now be accounted for.
(128) k-aa-mu-kwec-a
k-aa-m-kwec-a u^-EL
k-aa-n-kwec-a NA
k-a-n-kwec-a VS
u-EL must precede NA because the former rule creates the cluster to which the
latter rule applies. Similarly, u-EL must precede VS because the former rule
creates the cluster that triggers the application of VS. There is no ordering
needed between NA and VS.
184
With this account of the segmental structure of the /-aa-mu-/ tense
behind us, we can now turn our attention to the tonal aspect of the data
in (127). When the prefix /-aa-/ is pronounced as a long vowel, it is
pronounced with a rising tone. We can explain this fact by assuming that
the prefix has a high tone associated with the second mora of its long
vowel in this tense. ^ Since the first mora is low-toned, the prefix will
consequently be pronounced with a rising tone. If there is indeed a
high tone on the second mora of /-aa-/, then we would expect this high
tone to double onto the following tone-bearing element by virtue of the
rule of HIGH DOUBLING (which has been amply motivated in preceding sections).
The example k-aa-mu-n-limel-a shows clearly that when the TA prefix /-mu-/
appears in its unaltered form, its vowel does exhibit the high tone that we
predict will appear on it as a result of the doubling of the high tone on
the second mora of /-aa-/. Thus it appears quite safe to conclude that
the tonal structure of the first TA prefix in this construction is /-aa-/
and that /-mu-/ is basically low-toned.
Let us next consider the stem tonology in (127). We claim that the
tone assignment rule given in section 2 as ITA' — namely, assign a high tone
to the first (and third, if there is one) stem vowel — is applicable in the
/-aa-mu-/ tense. Notice, for example, that in every case there is a
high tone on the first stem vowel (cf . k-a-n-kwec -a ) and that the second
stem vowel will also be high-toned, presumably as a result of HIGH DOUBLING,
unless it is at the end of a phrase (cf. k-a-n-kwec-a vyaa vyaa.'). Also,
the third stem vowel is high-toned (cf . k-a-n-l6k6tanih-a and y-a -m-p i t ikuw-a
pitiku pitlkul) unless it happens to be the word-final vowel, in which case
it is subject to the rule of WORD FINAL LOWERING discussed in section 2
(cf. y-a-n-tutum-a tu tu tu!). The vowel that follows the third stem vowel
will be high-toned as a result of HIGH DOUBLING, except when it appears in
phrase-final position (see examples above). The fifth stem vowel never
bears a high tone (cf. y-a-m-p^ameelac-a phame p"ame!).
Recall that in those tenses where ITA' applies, object prefixes bear
an underlying high tone. This high tone causes the high tone assigned
by ITA' to the first vowel of the stem to be lowered (due to the rule of
TONE LOWERING). The high tone assigned by ITA' to the third vowel of the
stem does not get affected by TONE LOWERING, however. The data in (127)
show that these observations are applicable in the /-aa-mu-/ tense. Consider
the example y-aa-mu-m-purukutul-a purukutu.' The object prefix /-mu-/ (reduced
to simply -m- through u-EL) bears a high tone. This high tone has caused
the high of the first stem vowel to be lost (which explains why there is no
doubled high on the second stem vowel). Of course, the high on the object
prefix doubles onto the first stem vowel. The high tone assigned to the
third stem vowel by ITA' remains unaffected by TONE LOWERING and is able to
double onto the fourth vowel of the stem.
The stem tonology in (127) thus provides no difficulties at all — it
simply reflects the ordinary processes that come into play when ITA' is
applicable. We can therefore return our attention to the prefix tonology —
in particular, to the cases where the TA /-mu-/ appears in one of its
altered forms. Consider first the case where it undergoes u-GL. In (129)
we illustrate the derivation that is required for an example like k-aa-mw-eetet
185
(129) k-aa-mu-etet -a%
k-aa-mu-etet -a% ITA'
k-aa-mu-etet -a% WFL
k-aa-mu-etet -a% HD
k-aa-mw-eetet -a% li-GL
The most crucial aspect of this derivation is that since /-mu-/ has received
a high tone through the process of HIGH DOUBLING, when the u glides to w
(and thus ceases to be tone-bearing) , the high remains on the tonal tier and
associates with the compensatorily introduced vowel mora produced by u-GL.
This account of the data requires that we insert the compensatory vowel mora
between the object prefix and the stem vowel. (130) illustrates these points
a little more precisely than does (129).
(130) H H H H
k-aa-mii-etit -a% (after HD)
H H H H j^
k-aA-mw-eetet -a% u-GL
k-ai-mw-e6tet -.
a% reassociation of "stranded" tone
One could avoid this reassociation of a stranded tone by ordering
li-GL before HD, as we illustrate in (131).
(131) k-aa-mu-etet -a%
k-aa-mw-eetet -a%
k-aa-mw-eetet -a%
(after ITA' and WFL)
u-GL
HD
However, we showed in section 3 that when a ij vowel bearing a primary high
tone glides, its high tone must reassociate with the compensatorily introduced
vowel mora that follows. Given that this reassociation of a primary high must
take place, subsequent to u-GL, there is no reason not to believe that a
similar reassociation of a doubled high would take place. We will therefore
continue to assume that the tonal rules precede the segmental rules in their
relative ordering.
We consider next the tonology of forms where the TA /-mu-/ undergoes
^-EL rather than _u-GL. Consider the example y-a-n-kwec-a vyaa vyaa! The
natic. The syllabe yafi. . . \
stem tone is not problemat
has been represented as
having a rising tone — i.e. as having the tonal sequence LH. Do our rules
predict this? In (132) we give the derivation that is expected according
to the analysis we have developed.
(132)
y-aa-mu-kwec-a. . .
y-aa-mu-kwec-a. . .
ITA'
y-aa-mu-kwec-a. . .
HD
y-aa-m-kwec-a. . . .
u-EL
y-aa-n-kwec-a. . . .
NA
y-a-n-kwec-a. . . .
VS
186
In (132) we have assumed that the high tone associated with the vowel of
/-mu-/ as a consequence of HD will remain on the tonal tier when the ij
elides, reassociating with the nasal (which becomes a tone-bearing element
as a result of the loss of the jj ) . Recall from section 7 that when a long
vowel with the tonal melody LH shortens via VS, the resulting short vowel
bears a rising tone. In (132) we have therefore assumed that when /-aa-/
reduces via VS, the result is a rising-toned short vowel. Presumably
the difference between the representation yan. . , predicted in (132) and
the representation yan. . . given in the citation of the data in (127)
would be in terms of the onset of the rise in pitch. We have no instrumental
evidence that would indicate which representation is better in accord with
the facts. If it should turn out that yan. . . is indeed the more correct
representation, all that we would need to do would be to invoke a rule that
would simplify LH to H when it precedes a H in the same syllable. Thus
the interaction of u-EL and the tonal structure of the prefixes in the
/-aa-mu-/ tense presents no great difficulties for our analysis of Makua
tone.
There are other contexts where VS applies in addition to the /-aa-mu-/
tense. For example, when a long vowel in a prefix (or series of prefixes)
stands before a nasal plus consonant arising from the reduction of the
object prefix /-mu-/ via u-EL, that long vowel shortens. (133) provides
just one situation illustrating this point.
(133) a-noo-tipur-a 'he is hoeing deeply/he will hoe deeply'
a-no6-lim-a% 'he is cultivating/he will cultivate'
a-n66-huweel-a% 'he is covering with earth etc'
a-noo-lokotanlh-a 'he is picking up etc'
a-no6-ki-tiikih-a 'he is causing me to be imprisoned etc'
a-n6o-ki-hukulel-a% 'he is sieving beer for me etc'
a-no-n-16wolel-a% 'he is transporting s.t. for him (ch.) etc'
a-no-n-tupulel-a% 'he is cutting s.t. for him (ch.) etc'
a-no-n-terek el-a% 'he is cooking for him (ch.) etc'
(133) illustrates the /-noo-/ tense, which has the structure SP-n6o-(0P)-VS-a,
where the verb stem is subject to ITA'. The high tone on the first mora of
/-noo-/ doubles onto the second mora. Thus this prefix is ordinarily rendered
as a long level high-toned vowel. But in the last three examples in (133)
we see that this long vowel is reduced to a short vowel when It stands before
a cluster resulting from the reduction of the object prefix /-mu-/ to a nasal
consonant in front of a stem beginning with a consonant. The derivation of
a-no-n-terekhel-a% is shown in (134).
187
(134) a-noo-mu-terek el-a%
a-noo-rau-terek el-a% ITA'
a-noo-mu-terek el-a% TL
a-noo-mu-terek el-a% HD
a-noo-mu-terek el-a% PFL
a-noo-m-terek el-a% u-EL
a-noo-n-terek el-a% NA
a-no-n-terek el-a% VS
Notice that VS reduces the sequence 66^ to o^, in effect losing a high tone
that is present at an intermediate structure but absent on the surface.
It is this missing high tone that explains why in a-n6-n-terek"el-a% a
sequence occurs with just three high-toned elements, whereas ordinarily
one only gets either a sequence of two highs or four highs (when one excludes
the effects of PHRASE-FINAL LOWERING and LONG FALL — rules that were motivated
in section 2) .
Yet another context for VS is provided by the -al-e tenses described in
section 5. Recall that these tenses have two forms — one where -al-e is
suffixed to the verb stem, and another where a nasal consonant is infixed be-
fore a stem-final consonant and -e^ is suffixed to the stem. The reader
is referred to section 5 for many examples of this construction. For our
present purposes we just need to point out that when a long vowel precedes
a stem-final consonant, the Infixation of a nasal in front of that consonant
will create a consonant cluster that induces the shortening of this long
vowel. Two distinct cases of this shortening are illustrated in (135).
(135) k a-y-aa-wiih-Sl-e 'he hasn't brought s.t.'
or
k a-y-aa-winh-e ibid.
a-k-e-et-al-e% 'I haven't traveled'
or
a-k-ent-e% ibid.
In the case of k a-y-aa-winh-e, it is the long vowel of the stem /-wiih-/ that
gets shortened before the Cluster arising from nasal-inf ixation. In the case
of a-k-ent-e, the long vowel that shortens has a more complex derivation, which
we illustrate in (136) .
(136) a-k-aa-et-al-e% or; a-k-aa-ent-e%
a-k-aa-et-al-e% a-k-aa-ent-e% AL-TA (see section 5)
a-k-aa-et-al-e% a-k-aa-ent-e% HD
a-k-aa-et-al-e% a-k-aa-ent-e% PFL
a-k-ee-et-al-e% a-k-ee-ent-e% VA
a-k-e-et-al-e% a-k-e-ent-e% VR
inapplicable a-k-ent-e% VS
188
Notice that the infixed nasal is a tone-bearing element. In (136),
it gets the high tone assigned by AL-TA (recall from section 5 that this
tone assignment rule puts a high tone on the second tone-bearing element
in the verb stem). The long vowel that shortens in (136) is low-toned.
Thus the result is a rising tone on the syllable . .ken. . . The derivation
of k^a-y-aa-winh-e is somewhat different.
(137) k a-y-aa-wiih-al-e or k a-y-aa-wiinh-e
k a-y-aa-wiih-al-e k a-y-aa-wifnh-e AL-TA
k a-y-aa-wiih-al-e k a-y-aa-wiinh-e HD
inapplicable k a-y-aa-wii^h-e
inapplicable k a-y-aa-wi^h-e
NA
VS
(We have assumed here that the infixed nasal is n and that it assimilates
to the following consonant via the rule NA, which is independently motivated.
But this point is not crucial to our analysis.) Notice that once again
our analysis has produced an output with a rising tone on a short vowel
followed by a high-toned nasal in the same syllable. But as far as we
can tell, the rising tone in the syllable ...win. .in (137) is not distinct
from the rising tone in the syllable ...ken- -in (136). If there is no
distinction, then we will require the rule mentioned earlier whereby a rising
tone is reduced to low before a high tone in the same syllable.
9. WORD-FINAL PRIMARY HIGH TONES.
In this section we turn our attention to the fate of various primary
high tones that are located (or would be expected to be located) on the
final vowel of a word. This discussion will be somewhat incomplete since
there is some evidence to suggest that the phenomena dealt with here may
be sensitive to non-phonological factors that we have not yet investigated
in detail. However, in the interest of completeness, we have attempted to
bring to the reader's attention a variety of problems regarding word-final
primary high tones.
In our discussion of ITA' in section 2, we omitted from consideration
one class of infinitive forms, a class that has only a handful of members.
The verbs in question are ones which have the stem structure -C,-a (i.e.
there is no evidence that the stem has any vowel other than the final vowel
that is always present at the end of a verb) . We shall refer to such verbs
as "monosyllabic" stems. The tonal pattern of such stems is illustrated
in (138).
(138)
^ , h ^
u-k w-a . . .
u-k^w-a%
'to
die'
-. V ^
u-s-a. . .
u-s-a/
'to
dawn'
u-ly-a. . .
u-ly-a%
'to
eat'
u-w-a . . .
u-w-a %
'to
come'
189
The left-hand column in (138) indicates the phrase-medial pronunciation of
these items, whereas the right-hand column indicates the phrase-final
pronunciation. It will be recalled that in other cases, the infinitive prefix
u- is low-toned and ITA' assigns a high tone to the first stem vowel. Thus
we would expect that ITA' would take an input such as /u-k^w-a/ and give the
output /u-k"w-a/, since the first stem vowel in the case of monosyllabic stems
is the final vowel -a. The items in (138) do sometimes have a high tone on
the final vowel, but this high tone appears only in phrase-final position
and seems best regarded as the double of the high on the infinitive prefix.
(Doubled highs do not appear on phrase-final vowels, due to the application
of the rule PFL.) Thus, although ITA' predicts that a primary high should
be located on the final vowel in (138) , in fact no primary high appears on
that vowel. However, an unexpected primary high does appear on the infinitive
prefix.
We suggest that the best way to treat these facts is to allow ITA' to
apply as expected, putting a high on the first vowel of the stem (the final
vowel -a in these cases) , but to posit an additional rule of HIGH TONE
RETRACTION (HTR) . HTR will say that a high tone on the final vowel of a
word must retract onto the preceding vowel. HTR will be ordered so as to
apply before HD, so that the retracted high will be able to double onto the
final vowel. The derivations of u-k"w-a . . . and u-k^w-a% are shown
in (139).
(139) u-k\-a... u-k w-a%
u-k'^w-a... u-k^w-a% ITA'
u-k\-a... u-k\-a% HTR
u-k'^w-a... u-k\-a% HD
inapplicable u-k w-a% PFL
Further support for the rule of HTR is provided by the -ho- tense dis-
cussed in section 4. The structure of this tense is SP-ho-(OP)-VS-a,
where the SP and /-ho-/ are low-toned and the verb stem is subject to ITA'.
Thus we find examples like ki-ho-rup-a. . . 'I have slept', ki-ho-llm-a. . .
'I have cultivated', ki-ho-kavlh-a 'I have helped'. When a monosyllabic
stem appears in this tense, we get examples like those in (140).
(140) ki-ho-ly-a. . . ki-h6-ly-a% 'I have eaten'
a-ho-k w-a... a-ho-k w-a% 'he has died'
ki-ho-w-a... ki-h6-w-a% 'I have come'
Notice that in these examples the /-ho-/, which is ordinarily low-toned,
bears a high tone which doubles onto the next vowel (that doubled high being
lowered by PFL when it resides on a vowel that is in phrase-final position).
Given the proposed rule of HTR, the fact that /-ho-/ is high-toned in the case
190
of monosyllabic stems follows automatically. The derivations in (141)
illustrate.
(141) ki-ho-ly-a . . . ki-ho-ly-a%
ki-ho-ly-a. . . ki-ho-ly-a% ITA'
ki-ho-ly-a... ki-h6-ly-a% HTR
ki-ho-ly-a... ki-ho-ly-a% HD
inapplicable ki-ho-ly-a% PFL
Although examples such as those above motivate a rule of HTR, the actual
conditions under which it operates require further investigation. There are
conditions under which the rule of HTR clearly does not apply. For instance,
we have not found any examples of the application of HTR where the vowel
preceding the word-final vowel is long. This point can be seen by considering
the negative past continuous tense discussed in section 6. This tense has
the structure k"a-SP-aa-VS-a, where ITA' applies to the verb stem and none
of the prefixes has a primary high tone. For example: k^a-y-aa-terekh-a
'he wasn't cooking', k"a-y-aa-l6kotanih-a 'he wasn't picking up'. In the
case of a monosyllabic stem, we get a form like kha-y-aa-ly-a% 'he wasn't
eating'. This item is low-toned throughout. There is no surface evidence
whatever of the high tone that ITA' should have assigned to the first stem
vowel (the final vowel here) . HTR would have predicted the incorrect form
*k^a-y-aa-ly-a% . Clearly, HTR must be blocked from applying in this example —
and the most obvious means of blocking HTR is to say that the vowel preceding
the word-final primary high must be a short vowel. Of course, given this
constraint on HTR, we still have not derived the correct surface form
kha-y-aa-ly-a% . In order to do that, we need another rule that will lower
a high tone that resides on the last vowel of the word. Call this rule
WORD FINAL LOWERING (WFL) . WFL does not affect doubled highs on a word-
final vowel, so either we must formulate WFL so as to limit it to just
primary highs or we must order WFL before doubled highs come into existence.
For convenience we will adopt the latter approach, but see the next section
for some brief remarks on the matter of distinguishing between "primary"
and "doubled" high tones in rules. Given the existence of WFL, we will get
a derivation such as that in (142).
(142) k a-y-aa-ly-a%
k a-y-aa-ly-a% ITA'
inapplicable HTR
k a-y-aa-ly-a% WFL
We have ordered WFL after HTR. If the reverse order were posited, then WFL
would eliminate the high tone before it ever has a chance to retract in
examples like u-k"w-a% and ki-h6-ly-a%.
There is evidence that we are correct in assuming that ITA' does supply
a high tone to the last vowel in the case of monosyllabic stems in the
negative past continuous, even though in kha-y-aa-ly-a% this high tone is
entirely absent on the surface. This evidence comes from cases where the
verb is in a phrase-medial position. There are two possible pronunciations
191
of the verb, depending on the tonal nature of the following word. In certain
verb tenses in Makua (including the negative past continuous) , a noun after
the verb may appear in two distinct tonal forms. One of these forms will
be referred to as the "focused" form of the noun, the other being the "non-
focused" form. (See Stucky (1979) for some discussion of the phenomenon
in the Imit^ipi dialect of Makua.) In the non-focused form, the noun
retains its underlying high tones, while in the focused form the noun loses
the first high tone from its underlying representation. Since many nouns
have just a single underlying high tone, this means that in their focused
form such nouns are pronounced without any high tones (all things being
equal — which is not always the case, as we show immediately below). With
this much background, consider the examples in (143).
(143) k a-y-aa-puput -a ^ikuwo 'he wasn't washing clothes [non-focused] '
V
k a-y-aa-puput a ikuwo 'he wasn't washing clothes [focused] '
e
h v^ ^ ,
k a-y-aa-ly-a masambara he wasn't eating cassava leaves [non-focused]
k a-y-aa-ly-a mSsambara 'he wasn't eating cassava [focused] '
In (143) we have given examples with a polysyllabic stem like /-puput -a/
and a monosyllabic stem like /-ly-a/. Notice that when the verb is followed
by a non-focused noun, no high tone appears on the final vowel of the monosyllabic
verb stem (nor on the final vowel of /-puput -a/, but more on this particular
verb type later). Apparently, WFL is applicable here, assuming that indeed
ITA' has assigned a high tone to the word-final vowel. The case where a
focused noun follows the verb is different. Here a primary high does
appear on the last vowel of the monosyllabic stem (but not on the last
vowel of the polysyllabic stem /-puput -a/ — again, see below). And this
primary high tone doubles onto the first tone-bearing element in the
focused noun. In other words, even though a focused noun loses its first
underlying high tone, it may acquire a phonetic high on its first tone-bearing
element as a result of HD."
The data in (143) show that while WFL applies to /k a-y-aa-ly-a/ when
this form is at the end of a phrase, it also applies when this form is
followed by a non-focused noun. WFL does not apply when a focused noun
follows. Thus it appears that the conditions for the application of WFL
must take into considered the distinction between focused and non-focused
nouns. At this point it is not clear precisely how this contrast should be
incorporated into the formulation of WFL, thus we must leave this matter as
a problem for further research. What is crucial here is that (143) clearly
indicates ITA' does place a high on word-final vowels, since this high tone
does manage to surface just in case the verb is followed by a focused noun.
Further evidence that HTR does not apply to a word-final primary high
where the preceding vowel is long is provided by the consecutive tense
discussed in section 6, which has the structure k^a-SP-aa-VS-a. The prefix
k^a- fails to trigger HIGH DOUBLING (the only exception we know to this rule),
the other prefixes are low-toned, and the verb stem undergoes ITA'. Examples:
192
k a-y-aa-t aw-a... 'and then he ran away', k a-y-aa-lokotanih-a 'and then he
picked up'. When a monosyllabic stem appears in this construction, we get
examples like those in (144) .
(144) k a-w-aa-vy-a rii.' 'and it burned completely'
k a-y-aa-ly-a nek uu.' 'and he ate until he was nearly full'
k a-w-aa-¥-a ngaal 'and it was morning, and you could see every-
where '
The verb words in (144) contain no high tones except the high associated with
the prefix k"a-. Thus the high tone assigned by ITA' to the first stem vowel —
which in these cases would be the final vowel — does not appear at all. Its
absence can be attributed to the application of WFL. But notice that in
order to get WFL to apply it must be the case that HTR fails to apply, and
the reason that it fails to apply apparently is the long vowel that is in
the penult syllable of the verb word. It should be noted that the consecutive
tense is never followed by a focused noun, thus there are no situations where
the high tone assigned by ITA' actually appears in connection with monosyllabic
stems in this construction.'
In the preceding discussion, we have seen that when HTR fails to retract
a high tone from a word-final vowel, WFL will lower that high both in phrase-
final position and before a non-focused noun. The high will remain, however,
when it is followed by a focused noun. But this discussion has been centered
on a high tone assigned by ITA' to the first vowel of a stem. Recall that ITA'
also assigns a high tone to a third stem vowel. We turn now to the case
where the third stem vowel is also the final vowel of the word.
Recall our account in section 2 of an infinitive verb such as u-l6w6l-a.
We proposed the following derivation.
(145)
u-lowol-a
u-lowol-a
ITA'
u-lowol-a
WFL
u-lowol-a
HD
In other words, we suggested that even though items such as u-lowol-a do not
show any trace of a high tone on the word-final vowel (neither in the event
the infinitive is phrase-final nor in the event it is phrase-medial), ITA'
should be permitted to place a high on that vowel. And we posited a rule
of WORD FINAL LOWERING to lower this high tone. We now see that this rule
is, apparently, well-motivated, since it is required on independent grounds
to account for examples such as k"a-y-aa-ly-a ma¥ambara and k"a-y-aa-ly-a nek"uu!
The alternative approach to items like u-lowol-a would be to constrain ITA'
so that it cannot place a high tone on the word-final vowel in such cases.
But such a constraint is not well -motivated, since we have seen that ITA'
can place a high tone on the word-final vowel when it is the first vowel of
the stem.
193
Assuming that it is correct to allow ITA' to apply in the derivation of
u-low6l-a, we must examine how the rule of HTR fits into the picture. This
rule says that a word-final high tone retracts onto the preceding vowel, if
that vowel is short. Recall also that HTR applies before WFL. Thus it
would seem that HTR should apply in the course of the derivation of u-low61-a.
But if HTR does apply, retracting a word-final high onto the penult vowel,
there is no surface manifestation of this retracted high. In particular,
it does not double onto the following vowel. The correct phrase-medial
pronunciation of u-l6w6l-a is with a low-toned word-final vowel; *u-16wol-a. . .
is incorrect. There is a way in which HTR can be allowed to apply in the
derivation of u-16w6l-a without this retracted high actually being manifested
on the surface — namely, by ordering HTR before TONE LOWERING (TL) . TONE
LOWERING was first introduced in section 3 and given additional support in
section A. It has the effect of lowering any number of consecutive highs
after a high (we are speaking here, of course, with respect to primary high
tones, not doubled highs). The derivation of u-low6l-a. . . , under the proposed
analysis, is shown in (146).
(146)
u-lowol-a . . .
u-lowol-a. . .
ITA'
u-lowol-a. . .
HTR
inapplicable
WFL
u-lowol-a. . .
TL
u-low61-a. . .
HD
In the preceding analysis of u-lowol-a, WFL would not actually be
involved in the derivation. An alternative account exists that would make
use of WFL. In this account, HTR would be modified so as to prevent it
from applying in the derivation of u-lowol-a. For example, we could stipulate
that a high tone cannot retract onto a vowel if that vowel is preceded by
a high-toned vowel. Such a constraint would bar HTR from applying to the
representation /u-lowol-a/, thus permitting WFL to come into play to lower
the word-final high. Given the data so far considered, there is no strong
reason to prefer one of the above alternatives over the other.
At this point, let us return to the examples cited in (143) involving
the stem /-puput^-a/. In the negative past continuous, no high tone shows
up on the final vowel of the verb — neither when it is followed by a non-focused
noun (cf. k"a-y-aa-puput -a ikuwo) nor when it is followed by a focused noun
(cf . k^a-y-aa-puput*^-a ikuwo). We know from examples involving monosyllabic
stems that WFL is blocked from applying before a focused noun. Thus it would
seem as though the data in (143) suggests strongly that it is not WFL that
is responsible for the absence of a high tone on word-final vowels in the
case of stems like /-lowol-a/ and /-puput"-a/. For if it were WFL that was
responsible, then the high tone should appear on the word-final vowel before
a focused noun. If WFL is not at work, then the derivation given in (146) —
which combines HTR with TL — must be correct.
The situation, however, is more complex. Consider the data in (147).
194
(147) k a-y-o-orel-a iheera 'he wasn't saving money [non-focused] for
V him (ch.)'
e
k a-y-6-6rel-e eheera 'he wasn't saving money [focused] for him (ch.)'
The underlying structure of the verb in (147) is /k a-y-aa-a-urel-a/. The
essential difference between this example and the one involving /-puput^-a/
in (143) is that here an object prefix is present. Notice the effect that
(somehow) this object prefix has — it leads to the appearance of a high tone
on the word-final vowel when that vowel is followed by a focused noun. The
appearance of a high tone on this vowel strongly supports the view that ITA'
must be allowed to place a high tone on the third stem vowel even when that
vowel is word-final. But now let us consider how the data in (147) will be
dealt with in terms of the two analyses developed above for explaining when
the high tone is missing from the word-final third stem vowel.
Consider first the proposal that HTR retracts a word-final high from
the third stem vowel and places it on the second vowel, thus creating the
environment for TONE LOWERING to apply. HTR and TL will work correctly to
account for the loss of the word-final high before the non-focused noun
in k"a-y-o-orel-a iheera. But these rules would also predict the loss of
the high in the environment before a focused noun — but this is incorrect.
To achieve the correct results, HTR would have to be blocked from applying
before a focused noun. (It would then share this constraint with WFL,
which is also inapplicable before a focused noun.) But if we postulate such
a constraint on HTR, then we are once again faced with explaining why the
word-final high is lost entirely from the surface form of k^a-y-aa-puput^-a
ikuwo. We could not appeal to HTR and TL to explain the disappearance of the
expected high tone on the third stem vowel, since the constraint on HTR
would block the word-final high from retracting before a focused noun and
the constraint on WFL would block that rule from lowering the high in front
of a focused noun. We return to this problem below.
Consider next the proposal that HTR is blocked from affecting a word-final
high on a third stem vowel (perhaps due to a constraint that the high in
question cannot retract onto a vowel that is preceded by a high tone) .
This constraint on HTR will allow us to derive the data in (147) without
any difficulty, as we show in (148)."
(148) k a-y-aa-a-urel-a iheera k a-y-aa-a-urel-a iheera
k a-y-aa-a-urel-a iheera k a-y-aa-a-urel-a iheera ITA'
inapplicable inapplicable HTR
k a-y-aa-a-urel-a iheera k a-y-aa-a-urel-a iheera TL
k a-y-aa-a-urel-a iheera inapplicable WFL
k a-y-aa-a-urel-a iheera k a-y-aa-a-urel-a iheera HD
k a-y-aa-a-urel-a iheera inapplicable LF
k a-y-aa-a-orel-a eheera k a-y-aa-a-orel-a eheera VL
k a-y-oo-6-orel-e eheera k a-y-oo-o-6rel-e eheera VA
k a-y-o-6rel-e eheera k a-y-o-orel-e eheera VR
195
In (148), WFL applies before a non-focused noun but fails to apply before
a focused noun. The rules of LONG FALL (LF) , VOWEL LOWERING (VL) , VOWEL
ASSIMILATION (VA) , and VOWEL REDUCTION (VR) have been extensively discussed
in earlier sections of this study.
But if we assume that WFL is blocked from applying before a focused
noun, then we still have no explanation for the failure of a high tone to
appear on the word-final third stem vowel in kt^a-y-aa-puput"-a ikuwo .
Thus neither of the analyses that we have sketched accounts for this
particular fact.
We must, then, find some explanation for why there is no trace of
the word-final high in the case of kl^a-y-aa-puput^-a ikuwo while this high
appears on the word-final vowel in kha-y-o-orel-e eheera. We suggest (quite
tentatively) that the difference has to do with the fact that in the former
case there is a primary high on the first vowel of the stem while that high
tone is absent in the latter case due to the application of TONE LOWERING.
(It is the high-toned object prefix that induces the loss of the high tone
from the first stem vowel — thus we can begin to see why the presence of an
object prefix has such an impact on the tonal behavior of the word-final
high tone.)
Suppose that the tonal difference cited above is indeed the critical
factor. Now we can return to a consideration of our alternative analyses
of the loss of high tones from word-final third stem vowels. We begin again
with the analysis whereby HTR affects such vowels. We will need to permit
HTR to apply to all high tones of the specified type except when they are
both preceded by two low-toned vowels and followed by a focused noun. With
this stipulation on HTR, the rule will operate in the case of a representation
like /k'^a-y-aa-puput -a ikuwo/ but not in the case of /k"a-y-aa-a-urel-a iheera/.
(We assume here that TONE LOWERING will precede HTR. ) Notice that this
account still requires that WFL be constrained so as to not apply to a
word-final high tone before a focused noun. This constraint on WFL is
required because in the case of a representation like /k^a-y-aa-ly-a
ma¥ambara/. HTR cannot apply due to the preceding long vowel, and in the
case of /k"a-y-aa-a-urel-a iheera/, HTR cannot apply because of the restriction
that HTR is blocked when two low-toned vowels precede and a focused noun
follows; but in both instances WFL also fails to apply. Thus this approach
to the loss of high tones from word-final third stem vowels necessitates
constraints on both HTR and WFL which refer to focused nouns.
On the other hand, if we assume that HTR is not applicable to high tones
on third stem vowels and that it is WFL that is responsible for their loss
(when indeed they are lost) , then it will be WFL that must be constrained
so that it fails to apply to a high that is preceded by two low-toned vowels
and followed by a focused noun. (This account presupposes that TL precedes
WFL.) Under this analysis, both /k^a-y-aa-ly-a ma¥ambara/ and /k^a-y-aa-a-urel-a
iheera/ will fail to undergo WFL since the word-final high is preceded by two
low-toned vowels and is followed by a focused noun. On the other hand, WFL
will apply to /kha-y-aa-puput -a ikuwo/ since while this word-final high is
followed by a focused noun, it is not preceded by two low-toned vowels.
This analysis does not require that HTR be constrained in any way with reference
to focused nouns.
196
Further data bearing on the behavior of high tones associated with a
word-final vowel is provided by the -no- tense. The structure of this tense
is SP-no-(OP)-VS-a. The SP and the TA prefix /-no-/ are low-toned; the verb
stem undergoes ITA' . Examples are given in (149).
(149) ki-no-vah-a . . . 'I'm giving s.t. away'
ki-no-lim-a, . . 'I'm cultivating'
ki-no-thumih-a 'I'm selling s.t.'
ki-no-hukul-a 'I'm sieving s.t.'
ki-no-leeher-a. . . 'I'm ordering s.t.'
ki-no-hiyerer-a. . . 'I'm balancing s.t. (on the head)'
ki-no-lokotanih-a 'I'm picking s.t. up'
This particular verb tense occurs regularly with a focused noun after
it. When we have a monosyllabic stem, we get a form like ki-no-ly-a meele
'I'm eating millet'. Notice that the word-final vowel receives a high tone
through ITA' (since the word-final vowel is the first vowel of the verb
stem) and this high tone escapes both HTR and WFL. It simply remains on
the word-final vowel, doubling onto the first tone-bearing element in the
focused noun that follows.
A form like ki-no-ly-a meele is to be contrasted with an example
such as ki-ho-ly-a yoolya 'I have eaten the food [non-focused]'. In the
former case, HTR has failed to apply. In the latter case, it has applied.
The critical difference seems to be simply that in the former case the
noun is focused, in the latter case it is non-focused. Thus it seems
that regardless of anything else we might say, there must be a constraint
on HTR that restricts its application before a focused noun.
If we assume that analysis of word-final third stem vowels which
Invokes the widespread application of HTR, then we have already claimed
that HTR is blocked after two low-toned vowels in the event that a
focused noun follows.^ If we assume that analysis of word-final third stem
vowels which invokes the widespread application of WFL (and the non-
application of HTR) , then in order to explain the non-application of HTR
to ki-no-ly-a meele we will have to add a constraint blocking HTR before
a focused noun. Consequently, we have arrived at the situation where
in both of our analyses there must be a constraint on both HTR and WFL
that makes reference to focused nouns. Thus we have little grounds for
choosing between these two analyses.
The -no- tense provides more data concerning word-final third stem
vowels. Consider the data in (150).
197
(150) ki-no-tipur-a imata 'I'm hoeing a garden'
e
ki-no-thumih-a meele 'I'm selling millet'
1 u
V
ki-no-hukul-a uth eka 'I'm sieving beer'
ki-na-a-vahac-a meele 'I'm giving them (ch.) millet'
ki-na-a-lihac-a yoolya 'I'm feeding them (ch.) food'
ki-na-a-lowol-a asaana 'I'm moving the children'
ki-na-a-lamih-o oreta 'I'm curing him of sickness'
(cf. ureta 'sickness')
In cases like ki-no-thumfh-a meele, no high tone shows up on the final
vowel of the word (which is the third stem vowel) . In cases where an
object prefix is present, like ki-na-a-lihac-a yoolya, a high tone does
appear on the word-final vowel and this high doubles onto the first tone-
bearing element in the focused noun that follows. These data clearly
are entirely parallel to the examples cited earlier from the negative
past continuous — k"a-y-aa-puput"-a ikuwo but k"a-y-6-6rel-e eheera. Thus
we have further support that either HTR or WFL (depending upon the analysis
selected) must be constrained so as to fail to apply after two low-toned
vowels when a focused noun follows.
The data that we have presented in this section have not been
sufficient to establish clearly the derivation of verb stems such as
/-lowol-a/, /-puput -a/, and /-urel-a/ when they undergo ITA'. In
particular, the evidence is not very strong with respect to whether
HTR should be allowed to apply at all in the foirms where these verbs
are used. We would opt for allowing HTR to affect just the high tones
on the first stem vowel, but cannot mount a very convincing argument in
support of this position. What is very clear, however, from all of the
data discussed in this section is that ITA' must be allowed to place
a high tone on the final vowel of the word.
In the interest of completeness, we should observe that there are
verb tenses where a high tone appears on the final vowel of the verb and
neither retracts by HTR nor lowers by WFL under any condition. For example,
there is one -al-e tense which is characterized by a tone assignment rule
that puts a high on the last vowel of the stem. This high always remains
on that vowel. Some examples:
(151) ki-pupunt^-e... 'I washed...'
i-ki-puput enl-e... 'he washed for me....'
ki-rap-al-e. . . 'I bathed...'
ki-lokotanihanc-e. . . 'I picked up...'
ki-wi-il-e. . . 'I arrived...'
198
There is also a future tense form in Ikorovere that is characterized by
the assignment of a high tone to the second stem vowel; if this vowel
happens to be word-final, the high remains there. Some examples:
(152) ki-nee-lim-e meelo 'I'll hoe tomorrow'
ki-nee-lokot -e meelo 'I'll pick up tomorrow'
ki-nee-lokotanih-e meelo 'I'll pick up tomorrow'
At the present time, we have no explanation to offer as to why the word-final
high tones in these tenses fail to undergo the rules of HTR and WFL. We do
know that these two tenses have some special syntactic properties, and our
guess is that there will be many links between the behavior of word-final
primary high tones and syntactic (or perhaps even semantic or pragmatic)
considerations. But an understanding of these links must await further research.
10. CONCLUSION
In this paper and in the preceding papers in the series, we have proposed
to account for the immensely complex tonal pattern of the Ikorovere (Makua)
verb by invoking a fairly small set of tone rules which, in conjunction with
a fairly small set of segmental rules, produce the required surface forms from
underlying representations which, while not "abstract" in the sense of
containing "imaginary segments", nevertheless do often differ substantially
from their surface realizations.
A few fairly general remarks are in order concerning our analysis. First,
we have in general taken the tack of allowing our rules to apply in the most
general form possible, invoking subsidiary rules to explain apparent instances
of the failure of the general rule to apply. Call this the "overgeneralization
plus patch-up strategy". The alternative is to somehow place special
constraints on the general rule, blocking its application in the required
environments. Call this the "rule constraint strategy".
For example, we posited a very general rule of HIGH DOUBLING that simply
says that a high tone doubles onto the following tone-bearing element. But
then we added the rules of PHRASE FINAL LOWERING and LONG FALL to account
for the absence of the doubled high on phrase-final vowels and on the second
mora of a long vowel in the context CoV%. This represents the "over-
generalization and patch-up strategy". The alternative would have been to
write into HIGH DOUBLING conditions that would prevent the rule's application
in the two contexts mentioned above. This is the "rule constraint strategy".
In the above case, we have given no empirical evidence establishing the
correctness of our strategy. We do not currently have any unambiguous
empirical data bearing on the problem, thus for the time being at least
we must content ourselves with theoretical considerations. We believe that
the overgeneralization plus patch-up strategy is a most reasonable one when
both the general rule and the patch-up rule are highly natural phonological
processes. HIGH DOUBLING is a widespread rule in Bantu, as is PHRASE FINAL
LOWERING. The patch-up rule LONG FALL is perhaps less well-motivated on
cross-linguistic grounds, but there do seem to be analogues where high tones
199
on penult vowels become falling-toned. Thus we conclude that the over-
generalization plus patch-up strategy is an appropriate one in the present
case.
We followed this same strategy when it came to the tone assignment rule
ITA' . We allowed this rule to apply in a totally general fashion, putting
a high tone on the first and third stem vowels regardless of whether these
vowels were word-final or not. We then invoked HIGH TONE RETRACTION and
WORD FINAL LOWERING to explain the absence of the expected high tone on
word-final vowels. The alternative strategy would have involved constraining
ITA' somehow so that in some cases it would assign a high tone to a vowel
other than the first stem vowel (in cases like ki-ho-ly-a 'I have eaten')
and in other cases would fail to assign a high tone at all. However, in
this instance we found empirical evidence to support the overgeneralization
plus patch-up strategy, since under certain conditions the high tone does
in fact appear on the word-final vowel as predicted.
A second general remark about our analysis is in order. Namely,
we have shown that there exists a need to distinguish between primary high
tones and doubled high tones. For example, TONE LOWERING affects a primary
high that is preceded by a primary high. PHRASE FINAL LOWERING and LONG
FALL affect only doubled highs. In the Imit^upi dialect of Makua, the need
to distinguish between primary highs and doubled highs is even more pervasive
(see Cheng and Kisseberth, forthcoming). In the present study, we have
attempted to use rule ordering to distinguish between primary and doubled
highs. Another approach to the problem would be to provide a structural
distinction. This could be achieved if we did not have HIGH DOUBLING convert
the structure
H I|
/u-likotanih-a/
to the following:
H H H U
/u-l(ikotinih-a/
but instead had it yield the output:
/u-lAl^otlnlh-a/
— that is, if HIGH DOUBLING would not add H's to the tonal tier but instead
add association lines between H's on the tonal tier and the tone-bearing
unit that follows the tone-bearing unit to which the H's are underlyingly
associated.
Given this reformulation of HIGH DOUBLING, then we have a structural
difference between a vowel bearing a primary high and a vowel bearing a
doubled high. The vowel bearing the primary high appears in the structure:
A,
while a vowel bearing a doubled high appears in the structure:
200
(V here actually stands for any tone-bearing unit, whether a vowel or a nasal),
We suspect that this structural distinction may in fact be the appropriate
device for distinguishing between primary and doubled highs, rather than
rule ordering, but this point will have to be explored in depth at a later
time. In particular, it will need to be determined whether a formulation
of HIGH DOUBLING that adds association lines (rather than adding H's to
the tonal tier) will permit a satisfactory account of the principle of high
tone preservation discussed in section 6.
A third, very brief, remark about our analysis remains to be made.
Throughout we have had HIGH DOUBLING apply before the segmental rules of
VOWEL REDUCTION and VOWEL SHORTENING. Whether this ordering is by any
means a necessary one depends on the precise formulation of HIGH DOUBLING
and on the precise way in which VOWEL REDUCTION and the high tone preservation
principle of section 6 work. Since these are matters that must still be
resolved, we do not claim to have established the necessity for applying
HIGH DOUBLING before VOWEL REDUCTION and VOWEL SHORTENING~al though this
ordering has permitted the most straightforward illustration of the way
in which the rules of tonology interact with the segmental rules in Makua.
What has Makua told us about the nature of phonology? We believe that
it has told us a great deal. The tonal system of Makua is extremely complex,
but it is a system that operates with amazing regularity (we know of just
one truly exceptional fact in the verbal system — namely, the prefix k"a-
in the consecutive tense does not induce HIGH DOUBLING) . Not only is it
regular, it is fully productive. Every verb stem in the language (whether
of Makua origin or borrowed from Swahili) behaves in accordance with the
set of rules presented here. The system is regular, it is productive, but
it is by no means a "transparent" system. The underlying forms are often
at considerable variance from their surface manifestations — underlying high
tones sometimes do not appear on the surface, underlying low tones may become
high tones, underlying vowels do not always surface — indeed as many as four
vowels in a row may delete, etc. The rules involved are not superficially
true — e.g. TONE LOWERING does not allow a high tone immediately after another
high tone, but of course on the surface there are many high tones that are
immediately preceded by a high tone. Similarly, HIGH DOUBLING says that
a vowel that follows a high-toned vowel will also be high-toned. But there
are many instances (for a whole variety of reasons) where this is not true.
We conclude, then, that phonological systems may be very complex, very
regular, very productive, very abstract (in the sense of containing representa-
tions that are substantially different from the surface forms and containing
rules that are not at all "transparent") .
201
NOTES
*We would like to thank the Univers ity of Illinois Research Board,
the African Studies Program of the University of Illinois, and the National
Science Foundation (Grant No. BNS-7924523) for their support of the research
that we have reported on in this paper. All of the data presented have
been collected from S.A.C. Waane — our gratitude to him is such that we
shall never be able to express it.
Makua nouns follow the usual Bantu pattern whereby they are organized
into sets which are referred to as "noun classes". In most Bantu languages,
each noun class has associated with it a particular object prefix that a
noun of this class triggers on the verb. In Makua, however, there are just
two noun classes that trigger object prefixes on the verb. One of these
noun classes we refer to as "noun class one", and it triggers the object
prefix /-mu-/, which is the same object prefix that is used to refer to
a third person (child) .
2
Some details concerning nasal assimilation remain to be determined.
For example, both n and m occur before the glides w and y in at least some
contexts. We do not have sufficient data at present to determine fully
how the nasals of/-mu-/ and /-no-/ behave when they come to stand in front
of a glide via the dropping out of the li and o^ vowels in these prefixes.
3
We have indicated that u-EL does apply in the environment of a
consonant followed by a glide. This can be seen from example k-a-^-kwec-a.
Thus we have suggested that u-EL is blocked just in case a cluster of nasal
plus true consonant follows.
It is often the case in Makua that the tonal shape of a certain
prefix varies depending on the tense it is being used in. For instance,
in section 4 we saw that -ho- is low-toned in the construction ki-ho-hukul-a
'I have sieved', but high-toned in the construction k-a-ho-hukul-a 'I sieved'.
In section 7 we gave a possible argument for ordering LONG FALL after
the rule of VOWEL SHORTENING. With this ordering we would be able to explain
why in u-kl-w-a 'to kill me' from /u-kl-iv-a/ (this form presupposes the
prior application of ITA', TL, and HD — see derivation (118) in section 7)
there is no surface evidence of LONG FALL having applied. We do not believe
that this argument is entirely conclusive, and the question of the relative
ordering of the tone rules and the segmental rules is by no means a straight-
forward one. See section 10 of this paper for a few additional comments.
It should be pointed out that nouns are like verbs in that a high tone
in a noun induces doubling onto the following tone-bearing unit. The proper
analysis of a noun such as ma^ambara is that there is a primary high on
the syllable . . .¥a. . . and this high has doubled onto the syllable . . .mba. . .
Thus in the focused form of this noun, the high on . . .¥a. . . is lowered.
This lowering precedes HIGH DOUBLING, thus once the primary high is removed
from the noun, no double of that high can appear.
202
In (144) the verbs are followed by items that are referred to in
Bantu studies as ideophones. Ideophones are identified in our citations
by an exclamation mark after them. They have tonal characteristics that
are quite distinct from the tonal characteristics of other word categories.
In some of the examples, the ideophone is entirely low-toned. They should
not be confused with focused nouns (which may become low-toned through
the loss of their first — and possibly only — high tone). All nouns in
Makua have at least one primary high tone associated with them at the
underlying level.
Q
One detail of pronunciation has been omitted in (148). In the left-
hand derivation, we have ended up with two vowel morae at the juncture
between the verb and the noun — but the correct pronunciation is with just
a single vowel mora. In general, across word boundaries, two vowel morae
are reduced to one. This fails to take place just in case there are high
tones on both morae. Since we have not considered segmental processes across
word boundaries in this study, this matter has not been gone into in the
text.
9 - -
There may be a problem, though. In ki-no-ly-a meele, one of the two
preceding low tones is the subject prefix ki-. A third person subject
that is a child would have a 0 form in the SP slot. We do not have any
examples in our data collection, but it seems very likely that if the subject
prefix is 0 we will still fail to get retraction — that is, we would expect
the form no-ly-a meele, even though just a single low-toned vowel precedes.
But since the facts are not known, we cannot pursue this issue here.
REFERENCES
CHENG, Chin-Chuan and Charles W. Kisseberth. 1979. Ikorovere Makua tonology
(Part 1). Studies in the Linguistic Sciences 9.1, 31-63.
CHENG, Chin-Chuan and Charles W. Kisseberth. 1980. Ikorovere Makua tonology
(Part 2). Studies in the Linguistic Sciences 10.1,15-44.
CHENG, Chin-Chuan and Charles W. Kisseberth. Forthcoming. High tone
doubling in two Makua dialects.
KIPARSKY, Paul. 1976. Abstractness, opacity, and global rules. In A. Koutsoudas
ed.: The application and ordering of grammatical rules, pp. 160-184.
The Hague: Mouton.
STUCKY, Susan U. 1979. The interaction of tone and focus in Makua. Journal
of African Languages and Linguistics 1.189-198.
The following special issues are in preparation:
South Asian Linguistics: Syntax
and Semantics
Editor: Yamuna Kachru
Studies in Language Variation:
Nonwestern Case Studies
Editor: Braj B. Kachru
Papers on Diachronic Syntax
Editor: Hans Henrich Hock
STUDIES IN THE LINGUISTIC SCIENCES
The
Spring 1974
Vol. 4, No
Fall 1974
Vol. 4, No
Spring 1975
Vol. 5, No
Fall 1975
Vol. 5, No
Fall 1976
Vol. 6, No
Spring 1977
Vol. 7, No
Fall 1977
Vol. 7, No
Spring 1978
Vol. 8, No
Fall 1978
Vol. 8, No
Spring 1979
Vol. 9, No
Fall 1979
Vol. 9, No
Spring 1980
Vol. 10, No
Fall 1980
Vol. 10, No
Spring 1981
Vol. 11, No
following issues are available:
1, Papers in General Linguistics
2, Papers on Phonetics and Phonology
Editors: Charles W. Kisseberth and
Chin-W. Kim
1, Papers in General Linguistics
2, Papers on Historical Linguistics: Theory
Editors: Ladislav Zgusta and
Hans H. Hock
2, Papers on African Linguistics
Editors: Eyamba G. Bokamba and
Charles W. Kisseberth
1 , Papers in General Linguistics
2, Studies in East Asian Linguistics
Editors: Chin-chuan Cheng and
Chin-W. Kim
1 , Papers in General Linguistics
2, Linguistics in the Seventies:
Directions and Prospects
Editor: Braj B. Kachru
1, Papers in General Linguistics
2, Relational Grammar and Semantics
Editor: Jerry L. Morgan
1 , Papers in General Linguistics
2, Studies in Arabic Linguistics
Editor: Michael J. Kenstowicz
. 1, Papers in General Linguistics
Orders should be sent to:
SLS Subscriptions, Department of Linguistics
4088 Foreign Languages Building
University of Illinois
Urbana, Illinois 61801
Studies in
The Linguistic Sciences
DIMENS/ONS OF SOUTH ASIAN LINGUISTICS
D.N.S. BHAT Physical identification in Kanada * \
HANS HENRICH HOCK: Sanskrit causative syntax: a diachronic
study y
YAMUNA KACHRU On the syntax, semantics, and pragmatics oj the
conjunctive participle in Hindi-Urdu 35
SURESH KUMAR Discourse structure in a Hindi short storv 51
K.P.MOHANAN Pronouns in Malay a lam 67
RAJESHWARI PANDHARIPANDE Interface of lexicon and
grammar: some problems in Hindi grammar 77
K.G. VIJAYKRISHNAN The syllable in phonological
theory: arguments from Tamil \q\
WILLIAM D. WALLACE Object-marking in the history of Nepali:
a case of syntactic diffusion \ 07
TEJ K. BHATIA Transplanted South Asian languages: an overview 129
TEJ K. BHATIA Trinidad Hindi: three generations of a
transplanted variety \ ^5
N ICOLE DOM 1 NGUE Internal change in a transplanted language 1 5 1
RAJESHWARI PANDHARIPANDE Transitivity in Hindi 161
YAMUNA KACHRU Transitivity and volitionality in Hindi-Urdu 1 8 1
I LI K. BHATIA The treatment of transitivity in the
Hindi grammatical tradition / 95
NA RIND A R K. A GGA R WA I. Reference material in Hindi:
slate of the art 209
Department of Linguistics
University of Illinois
STUDIES IN THE LINGUISTIC SCIENCES
PUBLICATION OF THE DEPARTMENT OF LINGUISTICS
UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN
EDITORS: Charles W. Kisseberth, Braj B. Kachru, Jerry L. Morgan
REVIEW EDITORS: Chin-W. Kim and Ladislav Zgusta
EDITORIAL BOARD: Eyamba G. Bokamba, Chin-chuan Cheng, Peter
Cole, Alice Davison, Georgia M. Green, Hans Henrich Hock, Yamuna
Kachru, Henry Kahane, Michael J. Kenstowicz and Howard Maclay.
AIM: SLS is intended as a forum for the presentation of the latest original
research by the faculty and especially students of the Department of
Linguistics, University of Illinois, Urbana-Champaign. Especially invited
papers by scholars not associated with the University of Illinois will also be in-
cluded.
SPECIAL ISSUES: Since its inception SLS has devoted one issue each
year to restricted, specialized topics. A complete list of such special issues is
given on the back cover. The following special issues are under preparation:
Papers on Diachronic Syntax, edited by Hans Henrich Hock; Studies in
Language Variation: Nonwestern Case Studies, edited by Braj B. Kachru.
BOOKS FOR REVIEW: Review copies of books (in duplicate) may be sent to
the Review Editors, Studies in the Linguistic Sciences, Department of
Linguistics, University of Illinois, Urbana, Illinois, 61801.
SUBSCRIPTION: There will be two issues during the academic year. Requests
for subscriptions should be addressed to SLS Subscriptions, Department of
Linguistics, 4088 Foreign Languages Building, University of Illinois, Urbana,
Ilhnois, 61801.
Price: $5.00 (per issue)
STUDIES IN THE LINGUISTIC SCIENCES
DIMENSIONS OF SOUTH ASIAN LINGUISTICS
Edited by
Yamuna Kachru
VOLUME 11, NUMBER 2
FALL, 1981
DEPARTMENT OF LINGUISTICS, UNIVERSITY OF ILLINOIS
URBAN A, ILLINOIS 61801
TABLE OF CONTENTS
PREFACE i
D.N.S. Bhat: Physical identification in Kanada 1
Hans Henrich Hock: Sanskrit causative syntax: a diachronic
study 9
Yamuna Kachru: On the syntax, semantics, and pragmatics of the
conjunctive participle in Hindi-Urdu 35
Suresh Kumar: Discourse structure in a Hindi short story. ... 51
K.P. Mohanan: Pronouns in Malayalam 67
Rajeshwari Pandharipande: Interface of lexicon and grammar: some
problems in Hindi grammar 77
K.G. Vijaykrishnan: The syllable in phonological theory:
arguments from Tamil 101
William D. Wallace: Object-marking in the history of Nepali:
a case of syntactic diffusion 107
Supplement I: Transplanted South Asian languages
Tej K. Bhatia: Transplanted South Asian languages: an overview 129
Tej K. Bhatia: Trinidad Hindi: three generations of a trans-
planted variety 135
Nicole Domingue: Internal change in a transplanted language . . 151
Supplement 2: Transitivity in Hindi
Rajeshwari Pandharipande: Transitivity in Hindi 161
Yamuna Kachru: Transitivity and volitionality in Hindi-Urdu • . 181
Tej K. Bhatia: The treatment of transitivity in the Hindi gram-
matical tradition 195
Narindar K. Aggarwal : Reference material in Hindi: state of
the art 209
PREFACE
This special issue on South Asian Linguistics continues the tradition
of the Studies in Linguistic Sciences of publishing the on-going research
on South Asian languages. The first such issue was SLS 1 : 2 on "Papers
on Hindi Syntax" edited by Yamuna Kachru. It was followed by another
special issue in 1973 (3:2) edited by Braj B. Kachru. In addition to
these special volumes, papers on South Asian languages were included in
several "general issues" of SLS, too. This issue, however, is with a
difference. First, for the first time we have included specially invited
papers from scholars other than those from the University of Illinois.
There are thus three papers from India, one from Canada, and three from
other U.S. institutions. The University of Illinois contribution to
this volume is less than fifty percent. This 'casting a wide net' is
consistent with the new policy of the SLS. Second, this issue includes
two supplements: one on Transitivity in Hindi and the other on Transplanted
South Asian Languages. Third, a number of papers included here were
originally presented at the Third South Asian Languages Analysis Round-
table hosted by the Program in Linguistics, State University of New
York at Stonybrook, New York, in 1981. The revised versions of these
were specially selected by the editorial board for this issue of SLS.
We are grateful to Professors S. N. Sridhar and Mark Aronoff of SUNY,
Stonybrook for their permission, cooperation and help in this matter,
and for organizing a very stimulating Roundtable with literally a
'shoe-string' budget.
A volume of this type is always a cooperative attempt. This could
not have been possible without the support of an active research group
at our University and their comments, criticisms, and suggestions on
the papers included in this issue. I am grateful to Josephine Wilcock,
Fannie Lambert, andTamara Valentine for their contribution in preparation
of the press copy of this volume.
Yamuna Kachru
Department of Linguistics
University of Illinois
Urbana-Champaign, Illinois
Fall 1981
Studies in the Linguistic Sciences
Volume 11, Number 2, Fall 1981
PHYSICAL IDENTIFICATION IN KANNADA
D.N.S. Bhat
International School of Dravidian Linguistics, Trivandrum
In recent linguistic literature, a great deal of attention
has been paid to the problem of reference. A major distinction,
however, has been overlooked. Noun phrases can be distinguished
not only on the basis of referential identifiability of indivi-
duals or objects being referred to, but also on the basis of
their physical identifiability. This paper discusses the notion
physical identification and presents data from Kannada to show
how this notion is crucial for an adequate description of a
set of wh-word constructions in Kannada.
1 . Introduction
Linguists have generally failed to notice that natural languages
distinguish between noun phrases not only on the basis of the
referential identifiability of objects or individuals that are being
referred to (the so-called definite-indefinite distinction), but also
on the basis of their physical identifiability--i .e. , on the basis
of the ability of the speaker to physically identify the object or
individual that he is referring to. This latter semantic feature is
very relevant in contexts such as that of the interrogative in which
the speaker confesses his inability to physically (but not referen-
tial ly) identify an object, or of the imperative in which the speaker
is expected to necessarily provides a physical identification--if at
all he is to make a specific (identifying) use of a given referring
expression, and so on.
Tlie two distinctions differ from one another on two distinct
parameters. First, the person who is to be able or unable to establish
an identificaiion is the addresseee (as presumed by the speaker) in the
former case and the speaker himself in the latter case; and second, as
pointed out above, the identification in question is referential (on
the part of the speaker) in the former case and physical (on the part of
the speaker) in the latter case.
2.0 Tlie notion 'physical identification'
A speaker can referential ly identify any object or entity that he
can think of, and hence the question as to whether he would be in a
position to referentially identify an object that he is referring to or
not simply does not arise, whereas there would be innumerable instances
in which he would not consider himself to be capable of physically
identifying an object or entity that he is referring to.
Consider, for example, the use of a wh-word in an interrogative
sentence. Generally, such a word indicates that the speaker considers
himself to be incapable of physically identifying the object that he is
referring to; he is in fact requesting his addressee, through such a
sentence, to provide sufficient information for him so that he can establish
a physical identity of that object.
This point can be exemplified with the help of the following sentence:
(la) Where did you buy that shirt?
In using (la), the speaker presupposes that the addressee had bought that
shirt somewhere, but the exact location of that event (i.e., the shop in
which the addressee had bought that shirt) is not apparently known to the
speaker; hence he is trying to obtain that information from the addressee
through (la).
Notice, however, that the identification that needs to be established
here is physical and not referential, and the person who needs additional
information for doing so is the speaker and not the addressee. It is
therefore incorrect to regard the use of wh -words as involving indefinite-
ness, as, I think, some linguists tend to do; one can easily paraphrase a
sentence like (la) into a sentence in which the location under considera-
tion is referred to by a definite noun phrase as shown below:
(lb) Which is the place that you bought this shirt from?
Notice further that (la) can be followed by another sentence (Ic) in
which the location is referred to by a definite noun phrase, and is free
of any interrogation:
(Ic) Did you buy those trousers also in the same place?
The possibility of using a definite expression the same place in (Ic),
and of uttering that sentence immediately after uttering (la) in which
a wh-word occurs instead, even before obtaining a reply to this latter
sentence, indicates clearly, I believe, that a wh-word as such does not
indicate indefiniteness (i.e., inability of the addressee to referen-
tially identify an object), but rather the inability of the speaker to
physically identify the object that he is referring to. Notice further
that even though the speaker has uttered (Ic) immediately after (la),
where the wh-word has been replaced by a definite noun phrase, he would
still be expecting his addressee to provide an answer to (la) as well;
that is, the enquiry resulting from his use of the wh-word is unaffected
by his substitution of that word by a definite noun phrase in a following
sentence (Ic) and hence, it cannot be something which lacks "definiteness ."
3 . 0 The Kannada data
A wh-word can be used in other types of sentences also (i.e., in
sentences other than the interrogatives) in order to indicate that the
speaker is unable to physically identify the object or individual that he
is referring to. It would be interesting in this connection to examine
a complex system of wh -word constructions occurring in Kannada, showing a
number of interesting constraints on usage, all of which can be explained
on the basis of the above claim that the use of a wh-word implies the
inability of the speaker to physically identify the referent.
I wish to examine here five distinct usages of the wh-word of this
nature in Kannada, namely (i) with the alternative suffix o_^, (ii) with
the additive suffix uj_, (iii) with the conditional form of the verb
a:gu 'to become' followed by the additive suffix u_^, (iv) with the emphatic
suffix e: , and (v) in exclamatory sentences. It can be claimed, I believe,
that in all these constructions, as in the case of interrogative sentences
discussed above, the use of a wh-word implies that the speaker is unable to
physically identify the object or entity that he is referring to by means
of the expression under consideration.
3.1 The alternative suffix -o^:
First, consider the use of the wh-word with the alternative suffix oj_
in an indefinite noun phrase contrasted with its absence in the same:
(2a) ra:muvige ondu pustaka be:ka:gide
Ramu-to one book want-having-is
Ramu wants a book .
(2b) ra:muvige ya:vudo: ondu pustaka be:ka:gide
Ramu-to which-or one book want-having-is
Ramu wants a book (i.e. an unknown book).
(2a) is ambiguous in that the indefinite noun phrase ondu pustaka 'a book'
can have either a specific or a nonspecific use. In the former case, the
speaker has a particular book in mind (which, he thinks, Ramu wants to have)
but which he (the speaker) is not actually indicating to his addressee;
whereas in the latter case the speaker has no particular book in mind (and
apparently Ramu also does not, as far as the knowledge or belief of the
speaker is concerned) .
This ambiguity of (2a) is absent in (2b) because of the use of the wh-
word ya : vudo : in that noun phrase. It has only a specific use and not a
nonspecific use. Also, there is an additional meaning involved in (2b) in
that the use of the wh^-word indicates further that the speaker himself
does not know the physical identity of the book which is desired by Ramu.
It is only the physical identity of the object which is unknown to
the speaker and not its referential identity in the above usage; this
can be made clear by the fact that (2b) can be followed by (2c) in which
the noun phrase under consideration has been replaced by the pronoun
adu 'it' .
(2c) adu sikkida ku:41e a:ta udupige ho:gutta:ne
it got-when soon he Udupi-to go-will
He will go to Udupi as soon as he gets it.
The examination of the remaining wh-word constructions would make it clear
that the disambiguation of the specific-nonspecific distinction results
from the use of the alternative suffix oj_ after the wh-word, whereas the
addition of the new meaning, namely that the speaker has no knowledge of
the physical identity of the object under consideration results from the
use of the wh-word itself.
3.2 The additive suffix -u:
The disambiguation of the specific-nonspecific distinction can be
effected by using the additive suffix uj_ rather than the alternative
suffix o: after the wh-word. However, the former cannot be used in the
above sentence (2a), which is a nondurative, indicative one, for reasons
to be discussed a little later (see section 4.0). I will illustrate the
possibility of this disambiguation with the help of the following negative
sentences:
(3a) ninne ra:tri a:ta obba vyaktiyannu kandiralilla
yesterday night he one person (ace) seen-not
He had not seen a person yesterday night
(3b) ninne ra:tri a:ta ya:ranno: kandiralilla
yesterday night he who (ace) -or seen-not
He had not seen someone (unknown person) yesterday night.
(3c) ninne ra:tri a:ta ya:rannu: kandiralilla
yesterday night he who (ace) -and seen-not
He had not seen anyone yesterday night.
(3a) is ambiguous in that the noun phrase obba vyaktiyannu 'a person' can
have a specific use or a nonspecific use. In (3b) the wh-word that has
replaced it, namely ya : ranno : 'someone' can only be specific; whereas
in (3c) the wh-word ya : rannu : 'anyone' which occurs in place of it, can
only be nonspecific. Further, both these latter noun phrases clearly
indicate that their referents are physically unknown or unidentifiable
for the speaker.
Thus the difference between the uses of the alternative oj_ and the
additive uj_ after a wh-word is that the former effects disambiguation
by allowing the noun phrase to have a specific use, whereas the latter
does so by allowing it to have a nonspecific use.
There is an additional meaning difference between the above two wh-
word constructions: the one containing the alternative suffix o: has a
restricted reference whereas the one containing the additive suffix uj_
has an unrestricted reference.
Since the former can only have a specific use, the fact that its
reference is restricted may appear to be rather automatic. But this is
not true of the latter construction. Because, one can think of it as
having either a restricted reference or a non-restricted reference.
3.3 The conditional a: dare
In Kannada, the use of the additive suffix u_^ directly after the
wh-word provides the unrestricted reference, whereas the use of the
conditional form a: dare (of the verb a:gu 'to become') in between the
two gives the restricted reference. Example:
(4a) ra:mu elligu: ho:da:nu
Ramu where-and go-may
Ramu may go anywhere.
(4b) ra:mu elliga:daru: ho:da:nu
Ramu where-become-if-and go-may
Ramu may go to some (unknown) place.
In (4a) , the wh-word elligu: has an unrestricted reference regarding the
place to which Ramu may go, whereas in (4b) the wh-word elliga:daru: has
a restricted reference. That is, according to (4b), the place to which
Ramu would go, even though unknown, is a restricted (but nonspecific) one.
3.4 The emphatic suffix -e:
The wh-word can be used along with the emphatic suffix ej_ in order to
indicate that the choice of a specific referent is being left to the addressee;
since in such a situation the speaker cannot consider himself to be capable
of physically identifying the referent (the choice is yet to be made) his use
of the wh-word can rightly be claimed as indicating the absence of such an
ability. Examples:
(6) ya:re: kerlali na:nu koduvudilla
who-emph. ask-may I give-will-not
Whoever may ask, I will not give.
(7) ya:vudanne: ke:lu kodutte:ne
which- emph. ask give -will
\Vhichever you ask, I will give it to you.
(8) este: kastava:daru: ma:dutte:ne
how-much -emph. difficulty-become-if do-will
However difficult it may be, I will do it.
It is interesting that English also use a wh-word construction in situations
of the above nature.
3.5 The excalmatory sentence
The use of wh-words in exclamatory sentences can be exemplified with the
help of the following sentences:
(9) avanu estu be:ga bandidda:ne
He how-much quickly come-has
How quickly he has come!
(10) avaladu entha: svara
her-it what-type-of voice
What a voice she has!
(11) e:nu male
what rain
What a rain!
Notice that only three wh-words estu 'how much', e:nu 'what' and entha:
(or entaha) 'of what type' can occur in exclamatory sentences in Kannada.
It can be claimed, I think, that in these usages also, the speaker is
indicating his inability or unvfillingness to physically identify (or
express) the exact quality or quantity that has given rise to his surprise.
4.0 Constraints on -o: , -u: and a: dare constructions
I would now like to discuss certain interesting syntactic constraints
or restrictions in the use of three of the wh-word constructions discussed
above, namely, the alternative, additive and conditional.
First, the alternative form is not used in an imperative sentence,
whereas the remaining two forms are. Consider (12a)-(12c):
(12a) *elligo: ho:gu
where-or go
*Go to some (unknown specific) place!
(12b) elligu: ho:gu
where-and go
Go to any (unknown nonspecific) place!
(12c) elliga:daru: ho:gu
where-become-if-and go
Go to some (unknown nonspecific) place!
The constraint is evidently due to the occurrence of specific reference
in (12a). If the speaker is asking his addressee to go to a particular
place, he will have to give him the "physical" identity of that place as
well; because the addressee will not otherwise be in a position to carry
out the order. In (12a) even though the location to be reached is specific j
it is indicated to be physically unidentifiable or unknown to the speaker,
as a result of the use of the wh-word. Hence, (12a) is unacceptable.
Since no specific location has been indicated in (12b) or (12c),
the addressee is free to choose his own location in either of them and
carry out the order. Hence both these sentences are acceptable.
It is interesting to note in this connection that indefinite noun
phrases occurring in imperative sentences have only a nonspecific use.
Consider, for example, the following sentences:
(13a) John wants to buy a book.
(13b) Give him a book!
The sentence (13a) is ambiguous in that the noun phrase a_ book
occurring in it may either refer to a particular book, or to no book
in particular; whereas (13b) is not ambiguous since the speaker cannot
have any particular book in mind while uttering (13b) .
The second constraint that I wish to point out here is that in a
yes-no question or in a protasis, only the conditional form of the wh-
word can be used but not the other two. Examples:
(14a) *ra:mu elligo: ho:dano:
Ramu where-or go-he-did
*Did Ramu go to some unknow specific place?
(14b) *ra:mu elligu: ho:dano:
Ramu where-and go-he-did
*Did Ramu go to any unknow nonspecific place?
(14c) ra:mu elliga:daru: ho:dano:
Ramu where-become-if-and go-he-did
Did Ramu go to some unknown nonspecific place?
The sentence (14a) is unacceptable because the speaker makes a specific
reference to a location which is physically unidentifiable for him and
hence he cannot expect his addressee to say whether Ramu has gone to that
place or not; (14b) is unacceptable because even though the location is
nonspecific, it is unrestricted and hence, a yes-no question about it
cannot be answered; the absence of a specific reference and the presence
of a restriction regarding the location in (14c) makes it possible to
answer the question contained in it, and hence (14c) is acceptable.
The following examples indicate that a similar set of constraints
affects the occurrence of these wh- forms in the protasis of conditional
sentences as well:
(15a) *ya:ro: bandare he:lutte:ne
who-or comes-if tell-will-I
*1 will tell you it some unknown specific person comes.
(15b) *ya:ru: bandare he:lutte:ne
who-and comes-if tell-will-I
*I will tell you if any unknown nonspecific person comes.
(15c) ya:ra:daru: bandare he:lutte:ne
who-become-if-and comes-if tell-will-I
I will tell you if some unknown nonspecific person comes.
The reason for the unacceptability of (15a) and (15b) and for the
acceptability of (15c) are the same as the ones given above.
The third constraint that I wish to point out here is that only
the alternative form can occur in nondurative indicative sentences but
not the additive or the conditional ,
(16a) ya:ro: bandaru
who-or came
Some unknown specific person came,
(16b) *ya:ru: bandaru
who-and came
*Any unknown nonspecific person came .
(16c) *ya:ra:daru: bandaru
who-become-if-and came
*Some unknown nonspecific person came.
This constraint, I think, is due to the fact that, in Kannada,
a generic sentence (or a habitual one) must have a durative predicate
in the indicative mood. The additive and the conditional forms can only
have a nonspecific meaning, which means that the sentences containing
them can only be generic. Hence they cannot occur in nondurative
indicative sentences.
The fourth constraint that may be noted here is that the conditional
form cannot be used in a negative sentence, whereas the other two forms
can. Examples:
C17a) ra:muvige e:no: sigalilla
Ramu-to what-or got-not
Ramu did not get some unknown specific thing.
(17c) *ra:muvige e:na:daru: sigalilla
Ramu-to what-become-if-and got-not
*Ramu did not get some unknown nonspecific thing.
One can either negate an event with reference to a specific object or
with reference to all the objects (of a particular type). It would
not be possible, however, to negate an event with reference to an
object to which a restricted reference has been made, without at
the same time making a specific reference to it. This, I think,
is the reason why (17c) is unacceptable.
The fifth constraint to be noted here is that a numeral can be
used after an alternative form or after a conditional form, but not
after an additive form. Examples:
(18a) ya:ro: mu:varu baralilla
who-or three-persons came-not
Three unknown specific persons had not come.
(18b) *ya:ru: murvaru baralilla
who-and three-persons came-not
*Three unknown nonspecific nonrestricted persons had not come.
(18c) ya:ra:daru: mu:varu barali
who- become-if-and three-persons came-let
Let three unknown nonspecific restricted persons come.
I have used an imperative (concessive) form of the predicate in (18c)
because, as shown by the fourth constraint given above, the conditional
form cannot occur in a negative sentence.
The reason for the unacceptability of (18b) is that the numeral
restricts the number of persons that are being referred to by a noun
phrase, and hence it would conflict with the unrestricted meaning of
the simple additive form of the wh-word. Hence, in order to use a
numeral with it, one will have to add the conditional form of the verb
a:gu 'to become' to it, which has the effect of restricting the reference,
5.0 Conclusion
The above discussion suggests that the notion 'physical identifi-
cation' is crucial not only for giving an adequate account of the five
constructions discussed in section 3.0, but also for explaining the
constraints described in section 4.0. It would be interesting to investi-
gate the relevance of this notion for other languages as well. Note that
the translation equivalents for the Kannada examples suggest that this
notion might be equally useful to account for certain distinctions in
English.
Studies in the Linguistic Sciences
Volume 11, Number 2, Fall 1981
SANSKRIT CAUSATIVE SYNTAX: A DIACHRONIC STUDY"*"' ^
Hans Henrich Hock
University of Illinois, Urbana-Champaign
Interest in the origin and development of the Sanskrit causative
has recently been revived by Cardona (1978) . This paper attempts to
supplement and correct Cardona 's findings. Specifically, it shows that
the class of verbs permitting causative formation in the earliest, Rig-
Vedic Sanskrit consists of intransitives, semi- transitive verbs of mo-
tion, and 'affected-agent' transitives. Causees are marked accusative
and are promotable to subjecthood in the passive. Later Samhita-Vedic
shows an expansion of this class, so as to include other transitives,
Causee marking remains accusative. The crucial period of change is
that of the Brahmanas, where instrumental-marked causees begin to ap-
pear. The starting point for this innovation seems to lie in (the re-
Interpretation of) animate instrumentals of means or instrument. It Is
shown how this transitional stage develops into the highly divergent
systems of Pacini on one hand (with causee marking sensitive to lexic-
ally determined verb classes), and the Classical tradition on the other
(with causee marking sensitive to pragmatic factors). In the latter
tradition, it is observed that not only are accusative-marked causees
promotable to subjecthood, they are promoted preferentially, over
against accusative objects. Moreover, even instrumental-marked causees
tend to hold a position 'higher' than such accusative objects.
1 . INTRODUCTORY
1.1. A very important recent paper by Cardona (1978) has revived inter-
est in the origin and development of the Sanskrit causative. As Cardona cor-
rectly observed, even in the earliest Sanskrit attestations, those of the
Rig-Veda, the causative marked by the verbal suffix -aya- is not limited to
intransitives, but is found also with certain transitives. Moreover, compar-
ative evidence shows this pattern to be inherited from Proto-Indo-Iranian and
even from Proto-Indo-European. (Ibid. 35-7, n. 25.) As a consequence, hjrpo-
theses like those of Jamison (1976) and Haudry (1977:381-2, 385, and passim)
which consider the causative of transitives to be a secondary development
within Sanskrit can no longer be accepted.
1.2. At the same time, however, Cardona notes that it is only a certain
subset of the transitives which is open to this causative construction in
early Vedic. This subset he characterizes as consisting of the following
(9-12, 19-20, 38):
(A) Verbs of movement, such as gam- 'go';
(B) Verbs of perception and knowing, such as ik§- 'see', sru-
'hear, listen', cit- 'perceive, know';
(C) Verbs of consuming (eating, drinking), such as pa- 'drink';
(D) Verbs of saying, such as vac- 'speak'.
10
In addition, causatlves can of course be made from intransitives. In
many cases, these causatives function as simple transitives, as in vardhate
'grows (itr.)' : vardhayati 'grows (tr.). makes grow'. In others, they com-
pete with simple, unmarked transitives, as in vartate 'turns (itr.)' :
vartati/vartayati 'turns (tr.), makes turn'. But also more clearly causative
values can be found, as in sidati 'sits down' : sadayati 'makes sit down'.
Finally, Cardona notes a special group of causatives corresponding to intran-
sitive/passives, such as srnve 'is heard/famous' : sravayati 'makes to be
heard/famous' or myje 'is cleansed/cleans himself /mrjyate 'is cleansed' :
margti/mar.jayati 'cleanses' (12-16, 20, 38).
Syntactically, all of these causatives are characterized by the fact
that the subject of the underlying non-causative (the causee) is marked as
accusative. The case of other constituents remains unchanged. As a conse-
quence, also the case of the underlying object or goal (if any) remains: ac-
cusative or genitive for classes (B) and (C) , accusative or dative for class
(D) , and accusative/dative/locative for class (A) (ibid. 9-13).
In the later language as described by Parjini, a different situation ob-
tains. While causatives of intransitives and class (A)-(D) transitives still
show the accusative of the causee, those of the other transitives (hereafter
referred to as (F)) require the instrumental, except for hr- 'take, carry'
and ky- 'do' which offer an option between instrumental and accusative.
1.3. For this later development of a new category of class (F) causat-
ives, Cardona considers important the following early Vedic correlations be-
tween base verb or passive and causative (19-20, 26-7, 38):
Base verb/pass. Causative
(i) itr. sidati 'sits' sadayati
Cii) (A)-(D) itr. /pass. sypve 'is heard' sravayati
(iii) (A)-(D) act. srnotl 'hears' sravayati
(iv) (F) itr. /pass. mtje/mrjyate 'is (marjayati = trans. act.)
cleansed'
(v) (F) act. margti 'cleans'
Of these, the relationship between myje/mrjyate and the active margti /
"caus." marjayati is the most important: While early on, marjayati is simply
equivalent to transitive/active mirgti, the analogy of class (A)-(D) causat-
ives (like sravayati) made it possible to shift the function of forms like
marjayati to that of a true causative. For a while, then, the old and new
values coexisted and competed. Eventually, however, these formations either
became true causatives or were lexicalized as regular transitives. (Thus,
marjayati eventually became a simple transitive.)
Syntactically, in this development the influence of type (ii) above, i.e.
of the equally intransitive/passive base verbs of class (A)-(D), was most im-
portant in Cardona 's view. For in combination with the fact that sets like
myje/myjyate became exclusively passive, they led to the relnterpretation of
the type marjayati as causatives built on the passive. In this latter categ-
ory, however, the underlying subject was realized in the instrumental case,
the underlying object as nominative. Given the rule mentioned earlier, that
in the causative the nominative of the base structure is converted into an
11
accusative, while other case markings remain unchanged, this would result in
structures with instrumental of the causee and accusative of the underlying
object:
Undi
er lying subject
Underlying object
act. :
Nom. .
Ace.
pass.
:
Instr .
Nom.
caus.
I:
Instr.
Ace.
There was, however, another possible option, namely to extend to the
new causatives the pattern of type (iii) above:
ac t . : Nom . Ace .
caus. II: Ace. Ace.
For a while, these two patterns were in competition with each other.
Eventually, however, the caus. I pattern became generalized for all type
(F) verbs, with only hy- and ky- optionally retaining caus. II (ibid. 27).
Finally, Cardona notes a certain 'encroachment' of the caus. I pattern
on (A)-(D) causatives, both in certain Vedic examples and in statements by
later grammarians who propose certain exceptions to Pacini's rules for
(A)-(D) verbs (28-9). 'However, so far as I know, this construction [with
caus. I for (A) -(D) verbs] had not been fully generalized by Pacini's time.
Indeed, verbs of the semantic groups [(A)-(D)] which had construction [caus.
II J in early Indo-Aryan continued to enter into a causative construction
different from that of the other transitive verbs ..." (29).^
1.4. The purpose of this paper is to supplement Cardona 's observations
outlined in 1.2 above and to propose an explanation for the later develop-
ments which is different from Cardona's views sketched in 1.3 above. It
should be noted, however, that these modifications do not affect Cardona's
view that the causative in -aya- is inherited not only for intransitives, but
also for a subclass of the transitives. This view is, I believe, correct be-
yond a reasonable doubt.
Specifically, in section 2 of this paper I will show that at the ear-
liest, Rig-Vedic stage of Sanskrit, causatives can be made not only from in-
transitives, the quasi-transitives of class (A), and what following Masica
(1975:45-7, 57, and esp. 63) we may call the ingestive transitives of clas-
ses (B)-(D), but from yet another set of transitives — class (E) . And what
this class shares with the other (transitive) classes is that it consists of
verbs with affected agents: The verbal action of such transitives affects
not only the object, but also the subject — at least more saliently so than
it does for other, class (E) transitives. It is true, the distinction in
question is not absolute, but gradient; and the dividing line is not always
easy to draw. But this does not take away from the fact that the distinction
can be important in syntax. Moreover, in the specific case of the Sanskrit
causatives, this non-absolute, gradient nature of the notion 'affected agent'
can be seen to have important consequences.
Section 3 examines the evidence of the Atharva-Veda and other post-Rig-
Vedic mantra collections. At least one quite different, class (F) transitive
12
can be observed to have been added to the list of verbs which can form caus-
atives. In their syntactic behavior, however, the causatives do not yet show
any meaningful change.
The change comes in the next stage, that of Vedic prose. (Section A.)
It is at this stage that causees begin to appear in the instrumental case.
Moreover, already at this point they can be used both with affected-agent
verbs and with others. Finally, there seem to be certain semantic conditions
which favor the appearance of such instrumental causees, and these conditions
make it possible to relate these new structures to a pattern found as early
as the Rig-Veda, namely that of instrumental-marked animate 'means' or 'in-
struments' .
Section 5 shows that the situation in Classical Sanskrit essentially
follows the same lines as that of Vedic prose, suggesting an unbroken tradi-
tion of usage. Also the question of passivization of causatives will be
examined .
Section 6 observes that Pacini's system is quite different from the
Vedic prose-Classical Sanskrit tradition and from the earlier (Rig-)Vedic
situation. It is suggested that this system derives from the system of Vedic
prose, through a set of reinterpretations and generalizations divergent from
those made in the Vedic prose-Classical tradition.'
2. CAUSATIVES IN THE RIG-VEDA®
2.1. At the earliest attested stage of Sanskrit, the following verbal
roots are attested with (genuine) causatives:
CA) Quasi-transitive verbs of motion: _£- 'go, reach', gam- 'go',
dru- 'run', ruh- 'ascend', vis- 'enter'.
(B) Verbs of perception and knowing: ikg- 'see', cit- 'notice,
see; understand',^ vid- 'know' (but cf. below), sru- 'hear, listen;
be heard/famous' .
(C) Verbs of consuming: as- 'eat' (only in participle, cf. below),
dha- 'nurse, suckle', p5- 'drink', bha.j - 'partake of, consume'.
(D) Verbs of speaking: says- 'recite, speak' (only with preverb
a).
(E) Other affected-agent verbs: vas- 'wear', spys- 'touch',
jTudh- 'fight', srad-dha- 'believe' (without causee).^^^
The inclusion of the last verb perhaps requires some comment. Taken
by itself, dha- 'put' hardly would seem to belong into class (E) . However,
it seems that the compound of srad 'belief and dha- 'put' has become lexic-
alized in the meaning 'believe': With one exception, srad- only occurs in
combination with dha-. Moreover, srad-dha- can take an accusative or dative
complement of its own; i.e. it behaves like a transitive verb in its own
right.
(F) Other transitives: None.
2.2. Syntactically, the behavior of these verbs is essentially that
described by Cardona: The causee (if attested) appears in the accusative,
other case relations remain unchanged. The following examples may suffice
13
as illustrations. (Here as elsewhere, causees are characterized by double
underlining. A = accusative, I = instrumental, N = nominative, obi. = ob-
lique.)
(1) tasmin a vesaya girah (1.176.2)
Loc. caus. pi. A
'make the praises enter him'
(2) sam yat janan . . . surah iksayat (1.132.5)
pi. A caus.
'when the hero makes the people look [at him]'
(3) devan . . . payaya havih (2.37.6)
pi. A caus. sg.A
'make the gods drink the oblation'
(4) a tu nah Indra saqisaya (1.29.1)
obi. caus.
'0 Indra, make us speak forth = conjure up'
(Note, however, that nah, being a clitic oblique form, is not unambiguously
marked as accusative, or as any other case, for that matter.)
(5) vastrepa iva vasaya manmana sucim (1.140.1)
sg.I caus. sg.I sg.A
'make the pure one wear [clothe him] (by) a prayer as if (by)
a garment'
2.3. There are however certain verbs which do not seem to follow the
regular case marking pattern.
On one hand, there are verbs which show apparent deviations from the ac-
cusative marking of the causee:
(6) acetayat dhiyah imah jaritre (3.34.5)
caus. pi. A sg.Dat.
'he made the thoughts known to/by the singer'
(7) yajnam devebhyah prativedayan (1.162.4)
sg.A pl.Uat. caus.
'making the sacrifice known to the gods'
Haudry (1977:384, 387-8) considers (6) above to show that the syntax of
causatives is not yet settled in early Vedic. However, two other analyses
are possible. According to one, the dative forms of (6) and (7) are not
causees, but simply refer to the beneficiary of the verbal action. (This is
a coimnon, independently attested function of the dative.) The second analy-
sis would see in these sentences forerunners of a pattern more clearly attest-
ed later, in which original causatives like (ni-/a-/prati-)vedaya- 'make
known' have been lexicalized as simple verbs of speaking and thus take the
accusative or dative of the person informed. (Cf. also 3.3 below.) Given
these alternatives, Haudry 's claim cannot be considered cogent.
2.4. There are also some apparent difficulties with the marking of the
underlying object. Thus in example (5) above, the putative underlying object
manma 'prayer'/ vastram 'garment' apears in the instrumental, not in the ac-
cuative normally found in the non-causative; cf. (8) below.
14
(8) yuvam vastrapl plvasa vasathe (1.152.1)
pi. A non-caus.
'you two are wearing garments of ghee'
Haudry (1977:384, 386) here likewise suggests that the sjmtax of causat-
ives is not yet settled in the Rig-Veda. Again, however, his explanation is
not cogent. For we do find at least one non-causative example of vas- with
the instrumental of its complement:
(9) sam abhrepa vasata parvatasalj (5.85.4)
sg.I non-caus.
'the mountains wear the cloud'
Because of the limited nature of the non-causative evidence, it is dif-
ficult to determine the conditions for the selection of the instrumental, as
well as whether this instrumental is in fact an object or has a different
syntactic function. However, the evidence in 3.4 below suggests that verbs
of clothing may regularly mark their complement either as instrumental or as
accusative. Moreover, yudh- 'fight' likewise takes both instrumental and ac-
cusative complements. Finally, as noted earlier, all the verbs of classes
CA)-(D), as well as srad-dha-, offer a choice between different complement
cases. Perhaps, then, this choice of marking is a regular feature of affec-
ted-agent verbs.
Also spys- 'touch', which in the non-causative (or 'simplex') takes loc-
ative or accusative complement marking, has an example of instrumental object
(or causee ?) , cf. (10). However, given that one of the functions of the in-
strumental is to serve as the 'with'-case, it is easy to see how this situa-
tion might have come about: With a verb of this sort, the causation of
touching can be visualized as consisting not in making A touch B, but A and
B touching each other, or A being in touch with B. Moreover, if the situa-
tion is perceived in this manner, it will become difficult to distinguish
between causee and object. It is, I believe, these special semantic condi-
tions which account for (10). In fact, as will be seen in 4.5 below, these
special semantic conditions lead to the appearance of yet other patterns, in-
cluding that of instrumental complements for the non-causative. Given these
circumstances, it seems unnecessary to see in these instrumentals early in-
stances of instrumental causees. (Note that in two other attestations, RV
6.15.18 and 6.49.12, the more 'grammatical' solution is found, with causee
in the accusative and complement in its underlying case marking, which in
both cases happens to be locative.)
(10) varcasa suryasya . . . tanvam sparsayasva (10.112.3)
sg.I sg.A caus.
'let your body come in contact with the splendor of the sun'
2.4. There is yet another syntactic difficulty in the case of vas-:
The simplex always has middle-voice verb inflection. As is well known, one
of the functions of early Vedic middle voice inflection is to encode reflex-
ivity. It is therefore possible to argue that expressions like abhrepa
vasata should be rendered 'they clothe themselves with the cloud', not as
'they wear the cloud'. The relationship between vaste 'clothes himself' and
vasayati 'clothes s.o. else' thus would be not one of simplex to causative
but of reflexive/quasi-intransitive to non-reflexive/transitive, comparable
to vartate 'turns himself = turns (itr.)' vs. vartati/vartayati 'turns s.o.
else = turns (tr.)'. There is, however, one important difference between
verbs like vartate and verbs like vaste: the former can be viewed as quasi-
15
intransitive, since except for the 'hidden' reflexive, the verb has no fur-
ther arguments. In the case of vas-, however, the verb does have such an ad-
ditional argument, namely the accusative/instrumental complement of the gar-
ment worn. That is, it is transitive to begin with. Finally, note that
there is at least one verb comparable to vas-, namely the reflexive yajate
'sacrifices for himself where there is syntactic evidence that the causative
in -aya- is in fact different from the non-reflexive active simplex yajati
'sacrifices for s.o. else'; cf. 4.6.
2.5. There is evidence in the Rig-Veda which may be interpreted as
showing that the accusative-marked causee of causative verbs has object pro-
perties, i.e. that from a synchronic point of view, the accusative is not
just an arbitrary marker of the causee, but a genuine object marker. This
evidence consists in the passive (-like) participles asita- 'sated' and arpi-
ta- 'made to reach'. These are formed from the causative stems asaya- 'cause
to eat' and arpaya- 'cause to reach', which in turn derive from al- 'eat'
and ]^- 'go, reach', asita- is attested four times, arpita- ten times. And
in all of these fourteen attestations, the context makes it clear that they
refer to promoted causees, not promoted objects; cf. e.g. (11) below.
(11) adat pibat urjayamanam asitam (10.37.11)
'eating, drinking, becoming strong, sated'
2.6. The above findings, then, in general agree with Cardona's observ-
ations, except that the class of transitives making causatives at the Rig-
Vedic stage is larger and includes also non-ingestive affected-agent verbs.
This is important for two reasons: (a) It shows that there is no straight-
line relationship between the pattern of this earliest stage of the language
and the ingestive : non-ingestive distinction made by Panini. (b) Since the
notion 'affected agent' (AA) is a gradient one, it can readily be seen that
and how the Rig-Vedic pattern can slowly begin to be extended to other verbs
which less clearly belong to the AA class or which quite clearly are non-AA.
3. CAUSATIVES IN THE LATER MANTRA COLLECTIONS
3.1. The difference between the stage represented by the Atharva-Veda
and other post-Rig-Vedic mantra collections on one hand and the Rig-Veda on
the other to a large degree is a quantitative one. Although exact counts do
not exist, it is my impression that the textual occurrence of causatives has
increased. Moreover, there are a number of roots which now for the first
time appear in causative constructions. (An incomplete list of these roots
appears below.) Finally, and most importantly, causative attestations are
found from one root (da- 'give') which quite clearly belongs to the non-AA
class.
(A) car- 'move' (VS 23.21), ya- 'go' (VS 25.39), syp- 'crawl,
move' (VS 33.6).
(B) drs- 'see' (AV 4.20.6), jna- 'know' (TS 2.1.11.3).
(C) bhuj- 'enjoy, consume' (AV 19.50.6).
(D) (No new forms noted by me.)
(E) pari-dha- 'put on, clothe' (AV 12.3.51, (19.24.1,) 19.46.4)
(cf. below), j_i- 'conquer, win' (VS 9.11, 12 = TS 1.7.8.1, 4 etc.).
(F) da- 'give' (AV 3.20.8, VS 9.24, TS 1.7.10.1.— These are
16
(quasi-) repetitions of a fixed formula.)
3.2. Syntactically, there is no significant change. The normal pattern
for causee marking still is accusative, even for the new class (E) causative
(cf. (12)) and for the (E) verb ji- 'conquer' (cf. (13)).
(12) (a)ditsantam dapayatu (AV 3.20.8)
sg.A (neg. caus.
participle)
'let him make/cause to give the one who is not about to give'
(13) (i)ndram vajam .japayata (VS 9.11 = TS 1.7.8.1)
sg.A caus.
'make Indra win the prize'
3.3. Here again the causative of vid- 'know' constitutes an exception,
in taking dative marking of the "causee"; cf. (14). This suggests that the
dative marking noted in the Rig-Veda (2.3 above) was not just an isolated
aberration, but did in fact constitute an instance of the lexicalization of
the original causative of vid- as a simple verb of speaking.
(14) imati tasmai prati pra vedaya (TS 3.1.4.1 (mantra))
pi. A sg.D caus.
'announce these to him'
3.4. In the case of pari-dha- 'put on, clothe' (lit. 'put around'),
we are, as in the case of RV srad-dha-, dealing with a lexicalized compound
of dha-. And although dha- 'put 'is hardly an AA verb, pari-dha- when used
ref lexively is as comfortable in the AA class as RV srad-dha-.
Like the semantically close vas- 'wear', this verb shows the possibility
of instrumental or accusative marking for the underlying object, both in the
causative and in the simplex; cf. (15) for accusative and (16) for instrumen-
tal.
(15) (a) pari dhatsva vasab (AV 19.24.4)
non-caus. sg.A
middle
'put on a garment'
(b) (a)tmanam pari dhapayathah vasah (AV 12.3.51)
sg.A caus. sg.A
(reflexive)
'you two cause yourselves to put on a garment'
(16) (a) tena enam . . . pari . . . dha t tana (AV 19.24.1)
sg.I sg.A non-caus.
act.
'with that clothe this man'
(b) indrasya tva varmana pari dhapayamati (AV 19.46.4)
sg.A sg.I caus.
'with Indra 's armour we cause you to clothe [yourself]'
There is, however, a syntactic difficulty similar to that which obtains
for RV vas- (cf. 2.4 above). It might be argued that (15a) and (15b) are
equivalent (except for the difference in number marking of the verb): In
(15a) the reflexive is marked on the verb, in (15b) it has surfaced as 5tma-
nam, making reflexive middle-voice marking on the verb unnecessary. Moreover,
17
it may be argued that the translatlonal difference between causative and sim-
plex is artificial, not only for (15a) vs. (15b), but also for (16a) vs.
(16b). In fact, the existence of the active, non-middle voice type (16a),
taking the same number of arguments as causative (?) (16b) may well be taken
to decide the issue. Whatever the decision may be, however, the case for
taking pari-dhapaya- to be a causative formation is much less clear than that
for vasaya-; for in the latter case there seem to be no attestations compar-
able to (16a) (or to (15b), for that matter).
3.5 The most important development of this stage of the language, then,
consists in the fact that the causative construction has begun to be extended
to non-AA verbs. And what is perhaps equally important is the fact that in
these earliest examples of non-AA causatives, the case of the causee is the
same as it had been for the AA verbs, namely accusative. This suggests that
non-AA verbs acquired causative constructions by a relatively smooth process
of extension of class membership, made possible by the gradient nature of the
notion 'affected agent', and not by the rather abrupt reinterpretation and
syntactic reorganization proposed by Cardona.
4. CAUSATIVES IN VEDIC PROSE
12
4.1. As in the preceding stage, so also here there seems to have been
an increase both in frequency of causative occurrences and in the number of
verbs entering into the causative construction. New additions to the list of
verbs with attested causatives include the following.
(A) kgar- 'run, flow' (TS 6.3.10.2).
(B) khya- 'see' (SB 1.3.1.26, etc.).
(C) ad- 'eat' (AB 4.27.6, SB 2.5.1.6), ghra- 'smell' (TS 7.1.6.6,
^B 4.5.8.5, etc.).
(D)
(E)
(F)
vac- 'speak' (AB 8.6.9 etc.), vad- 'speak' (^B 1.8.3.20 etc.).
slig- 'stick to, adhere' (with locative or accusative comple-
ment) (AB 5.32.6);
lih- 'lick' (^B 14.4.3.4);
ya.j- 'sacrifice for oneself' (AB 1.11.5 etc. etc.);
ap- 'obtain, reach' (^B 2.3.3.16 etc.);
rabh-, labh- 'obtain, take' (TS 2.6.2.5 etc.; TS 2.2.2.2 etc.);
duh- 'milk' (^B 1.7.1.18 etc.; no causee);
vid- 'find' (TS 2.2.8.2);
grab- 'seize' (MS 2.5.6, TS 7.2.7.3, etc.);
vah- 'convey' (AB 1.2.6 etc.);
bhy- 'carry' (KS 23.8 (2x); no causee);
he- 'take, carry' (AB 7.12.4 etc.);
k£- 'do, make' (AB 3.46.2);
pac- 'cook' (^B 14.9.1.1; no causee);
bandh- 'fetter, tie' (^B 13.5.4.4; no causee).
(Note that in the continuum from (E) to (F) some, not necessarily entirely suc-
cessful attempt has been made to order roots in terms of their relative AA
status. )
4.2. What is even more important, however, is that it is at this stage
that instrumental-marked causees begin to appear. And what is equally import-
ant is that most of the new, non-AA causatives appear with accusative causees.
18
Moreover, instrumental causees are not limited to non-AA verbs, but can occur
at least as well with quasi-transitive verbs of motion, as well as with AA
transitive verbs. Finally, with only one exception (slig-, rarely attested
in the early language) , all of the verbs with instrumental causees are at
least occasionally attested also with accusative causees in the early langu-
age. The following list of attestations is, I believe, representative. For
each verb I will give under (a) examples for accusative causees (if any) ,
under (b) for instrumentals.
^- (17) (a) dyavapythivi bhuvanegu arplte (TS 4.7.13.2 etc.)
du. R caus.
(promoted) pass. part,
'heaven and earth have been made to reach the worlds'
tan yat samarpayet artim archet yajamanati (SB 13.3.3.7)
pi. A caus.
'if he were to make them go together, the sacrificer
would go to ruin'
(b) amum te suk ychatu / yam eva dvesti tam asya kgudha ca
sg.N non-caus. sg.I
suca ca arpayati (TS 4.4.4.1-2; similarly passim; a fre-
sg.I caus. quently attested formula, with differ-
ent NPs taking causee position)
'your pain should go to him; whom indeed he hates, to
him he makes his hunger and pain go'/'... him he afflicts
with his hunger and pain' (?)
ruh- (18) (a) rudram grhSn anvavarohayet (TS 3.4.10.3 etc.)
sg.A caus.
'he would make Rudra mount the houses'
(b) enam ibhih prajabhih pratyavarohayati . . . ksatriyam
pi. 1 caus.
... imilii prajab pratyavarohanti (SB 3.9.3.7)
pl.N non-caus.
'he makes these people rise up to him ... these people
rise up to a kgatriya'
na it sirasa sirah abhyarohayipi (^B 14.2.2.50, similarly
sg.I caus. ibid. 51, 52)
'lest I make a head overtop [another] head'
dru- (19) (a) [cf. adhvaryo dravaya tvam somam (RV 8.4.11)
caus. sg.A
'Adhvaryu, make the soma run' ]
(b) yoga vai veditt vfga vedatt pascat vai parltya vyga yogam
sg.N
adhidravati pascat eva enam etat parltya vrsna vedena
hon-caus. 5gT T
adhidravayati (^B 1.9.2.4, similarly 1.7.2.12)
'the altar is the female, the veda the male; (having gone)
from behind the male approaches the female; from behind
he has the male, the veda, approach her [=the altar]'
19
khya-(2Q) (a) eke yajamanam avakhyapayantl (^B 1.3.1.26 etc.)
sg. A caus.
'some make the sacrificer look down'
(b) enam somakrayanya samkhyipayati (SB 3.3.1.11 etc.; rare)
sg.I caus.
'he has the soma cow look at her '/'he has her exchange
looks with the soma cow' (?)
sru- (21) (a) [ sravaya it asya karpa (RV 4.29.3) (cf. also 4.5 below)
caus. pi. A
'make his ears listen/hear' ]
(b) tam vai devaiti sravaya (^B 1.8.3.20)
pi. i caus.
'make him be heard by the gods'
vad- (22) (a) etam atra vacam pratyudvadayanti (^B 1.1.4.11 etc.)
sg.A sg. A caus.
'they make this voice speak out'
tanvas samavadyanta (KS 24.9)
pl.N caus. pass,
(promot. )
'the bodies were caused to converse'
(b) samvadaya enam devaiti (^B 1.8.3.20)
caus. pi. i
'make the gods speak to him' /'make him converse with
the gods' (?)
slig-(23) (b) yatha slegmapa carmapyam . . . saipslegayet (AB 5.32.6)
sg^ I caus.
'as if one were to make glue stick together leatherwork' /
'as if one were to make leatherwork stick together with
glue'(?)
grah-(24) (a) osadhlti eva phalam grahayati (KS 26.5)
pi. A caus.
'he causes the plants to take fruit'
(b) varupena eva bhratyvyam grahayitva brahmapa stypute
sg. I caus. sg.I non-caus.
(TS 2.1.8.2, similarly KS 13.4)
'having caused Varuna to seize the enemy, he lays him
low with the sacrificial formula'
tam varupena eva grahayitva vigpuna yajnena (TS 2.1.4.4)
sg. I caus. sg. I sg. i
'having caused Varuna to seize him, Visnu [and] the
sacrifice [to seize him] '
(This, like (19b) is a highly formulaic expression, re-
curring frequently, with only the NP filling the causee
position varying.)
ji- (25) (a) satyena eva enam uj.lapayati (KS 14.7)
sg.A caus.
'with truth he makes him win'
(b) tayor aryepa varpena saudram varpam jyapayanti (JB 2.404)
sg. I caus.
20
'of these two, he has the Aryan defeat the ^udra'
da- C26) (a) sa eva asmai imin lokin visam pra dapayati (TS 2.1.4.8)
pi. A sg.A caus.
'he makes these worlds, the people give to him'
(b) ahna eva asmai ratrim pra dapayati ratrlya ahar
sg. i sg.A caus. sg. I sg.A
(TS 5.4.9.3, similarly KS 21,9, 35.18; a formula)
'he makes day give him night, night [give him] day'
4.3. The interpretation of the underlined instrumentals in the above
examples is by no means always easy. In some cases, for instance, it may be
argued that the instrumental is governed not bv the causative verb but by the
preverb sam- . (Cf. (20b), (22b), and (23b). ^^a^ There is evidence that
sam- can make for the appearance of an instrumental where an accusative might
be expected, even with non-causative verbs. An example would be (27), with
a Janata being a transitive verb, normally governing the accusative. (Its
meaning, if occurring by itself, is 'know'.)
(27) marudbhib rudrib sam-a.ianat(a) (TS 2.1.11.3)
pi. I non-caus.
'the Rudras agreed with the Maruts'
Even if (20b), (22b), and (23b) are therefore excluded, there still re-
main cases like (17b) and especially (24b) in which it could be argued that
the doubly underlined instrumentals are not really causees, but indicate the
instrument or means through which the action is accomplished. Thus, in the
first example under (24b) , varupena can be said to be identical in function
to the parallel instrumental form brahmagl. The latter, however, accompany-
ing a non-causative transitive, can only be an instrumental of means. The
fact that varupena refers to an animate (divine) being, while brahmapa does
not, would not necessarily preclude this interpretation. For as the second
example under (24b) shows, both animate (varupena , vigpuna) and inanimate
(yajnena) instrumentals, with clearly identical function, can accompany a
causative. And as (28) shows, the same parallelism can be found with non-
causatives.
(28) (e)nam brahmapa devatibhiji ud vapati (TS 5.1.7.3)
sg,I pi. I non-caus.
'he digs her up by means of the sacrificial formula and the
deities'
Even exclusively animate instrumentals of means can be found quite free-
ly throughout the whole Vedic period, both with simple transitives and with
causatives from intransitives, where it is not possible to interpret the
instrumental as a causee. Cf. e.g. the following examples.
(29) agnina rayim asnavat (RV 1.1.3)
sg.I non-caus.
'through Agni one will gain wealth'
(30) prajapatina eva enam cinute (KS 21.3)
sg. I non-caus.
'he piles it with the help of Prajapati'
(31) mitrepa eva yajnasya svigtam samayati (TS 6.6.7.3-4)
sg.I caus. from Intr.
'through Mitra he made calm (= appeased) the good performance of
the sacrifice'
21
It is perhaps also striking that in many of the above constructions the
putative instrumental causees are inanimate; cf. the second example under
(18b); (19b), (26b); and especially (17b), a frequently attested formulaic ex-
pression whose instrumentals always are inanimate. Note also that in the for-
mula of (24b) both animate and inanimate causees may appear, sometimes even
side by side .13 One may well ask how agentive such inanimate causees would
be.
Finally, it is possible to argue that there is an important semantic dif-
ference between the (a) and (b) patterns: Wherever the context is clear
enough to permit a judgment, the causee marked by the instrumental seems to
be less saliently the agent and/or beneficiary of the action brought about by
the act of causation. Thus in (17b) , the hunger or pain which afflicts the
victim cannot be said to derive any benefit from its action or to be affected
by it in any other way. Similarly, in (18b), first example, the point is to
get the people to rise up to the deity, not whether they enjoy doing so or de-
rive any benefit from their action. In (25a) the causee stands to benefit
from his victory, whereas in (25b) the Aryan fights the ^udra in a ritualist-
ic combat whose outcome is predetermined and whose benefits go not so much
to the combatants as to the ritual and its (other) participants in general.
Similarly, in (24a) the plants benefit from their action; but in (24b) the
major concern seems to be the seizure of the enemy. Who accomplishes this
seizure seems to be less important. Finally, in (26), the (a) example seems
to involve a real act of giving. As far as (26b) is concerned, however, the
context shows that the act of giving referred to is a metaphorical one in
which black and white cows are identified with night and day: 'He pours liba-
tions with the milk of a black cow and of a white calf. [Thus] he makes
day give night, night give day.' The case is very similar for (19b). — It
should however be noted that of the two types of causee, the instrumental is
marked (for decreased agency or af fectedness) . The accusative, on the other
hand, seems to be usable for any type of causee, including quite inanimate
and non-affected, non-agentive causees. Thus the context shows that the ac-
cusative causee in the second example under (17a) is inanimate, non-affected,
etc.
The semantic marking of the instrumental causee may be taken as further
evidence that causees of this type are really just instrumentals of means.
At the same time, however, there is evidence (as in (17b) suk rchatu beside
suca . . . arpayati) that the instrumentals in question are from nouns which
can and do function as subjects of the corresponding non-causatives. More-
over, in examples (18b), (19b), and (26b) there can be no doubt that the
doubly underlined instrumentals are in fact causees.
4.4. This rather fluid situation, as well as the special connotations
of instrumental causees, would in my opinion be best accounted for by the hy-
pothesis that instrumental causees are an innovation of this period, growing
out of a reinterpretation (mainly) of animate instrumentals of means as under-
lyingly being the subject of the simplex verb. This development was made pos-
sible by the fact that causees frequently are not specified on the surface,
leaving the causee slot 'empty', as it were. (Cf. sa eva asmai pra dapayati
(TS 2.2.8.4) 'he makes [them] bestow [it] on him'.) It is because this devel-
opment was still in progress that in many instances it is difficult to decide
whether a given instrumental is a causee or not. Moreover, it is the instru-
mental-of-means origin which accounts for the less agential functions of in-
strumental causees.
22
4.5. Here as at the earlier stages of the language, compounds of veda-
yati (lit. 'causes to know') are used in the sense of informing and take the
dative of the person informed. Examples can be found at AB 1.10.2; TS 3.1.4.
1, 6.4.3.3, 7.1.6.5; ^B 1.4.2.3, 3.2.1.39, 5.4.5.31. Also sravayati 'make to
hear' = 'inform' and darsayati 'make to see' = 'show' apparently are now
attested in this construction (TS 3.3.7.2 and JB 3.193-4). Looking ahead to
the language of Classical Sanskrit and of the grammarians we can note that
the number of such lexicalized causatives will further increase. Thus, in
Panini's Dhatupafha clt- 'see, realize', jna- 'know', vac- 'speak', vad-
'speak', vid- 'know'are listed as non-causative verbs in -aya-, with meanings
such as 'inform' or 'instruct'.
As far as object (or causee ?) marking is concerned, spys- also here
causes difficulties: As in the Rig-Veda we find the possibility of instrum-
entals (AB 7.2.7 ). Other possibilities include a dual compound in the ac-
cusative (prana-udanau . . . samsparsayati (^B 4.1.2.24) 'he causes outbreath-
ing and inbreathing to touch [each other]'), as well as double accusatives
or double instrumentals connected with each other by ca ... ca 'both ... and'
(TS 6.4.3.4, ^B 3.9.3.29). Moreover, even the simplex can now take an in-
strumental complement (^B 3.9.4.20). These various possibilities confirm, I
believe, the semantic explanation for the behavior of the causative of spys-
suggested in 2.3 above. (Note that the instrumental of (23b) above may very
well owe its existence to similar semantic considerations.)
4.6. As in the preceding stages, so also here there is a reflexive
whose causative creates certain difficulties, namely yajate 'sacrifices for
himself : yajayati 'causes s.o. to sacrifice for himself. For beside
yajayati we find an active, simplex yajati 'sacrifices for s.o. else' which
refers to the same ritual act as yajayati. However, in this case there is
clear syntactic evidence that yijayati is causative, not just "transitive"
or non-reflexive: yajati takes the dative of the beneficiary of the act of
sacrificing, i.e. of the person for whom the sacrifice is being conducted.
yajayati, on the other hand, refers to the same person by means of the accus-
ative of the causee; cf. (32) and (33).
(32) devan devayate yaja (RV 1.5.12; similarly elsewhere)
sg.D non-caus.
'sacrifice to the gods for the one who is going to the gods'
(33) (a)sman eva angirasah yajayan (^B 3.5.1.14)
pi. A
'may the Angirases make us sacrif ice-for-ourselves'
4.7. As in the Rig-Veda, so also in the Brahmapas, the accusative of
the causee is promotable to subjecthood in the passive; cf. (34) below, as
well as (17a) first example, and (22a) second example above. Moreover, there
are examples also of the underlying object promoted. In fact, with verbs
like duh- 'milk', which take two accusative objects (that of the animal mil-
ked, and that of the milk drawn from the animal), either of the two objects
may be promoted; cf. (35) and (36). However, the usual pattern is that only
one of the 'objects' of the verb (including the causee) is specified on the
surface; and that is the 'object' which has undergone promotion. I have
found only one example where two 'objects' are specified; and here it is the
causee, not the underlying object which has been promoted; cf. (37). Perhaps
not too much should be made of this one example. However, it is noteworthy
that in the later. Classical language, promotion of the causee takes preced-
23
ence over promotion of (other) objects; cf . 5.3 below.
(34) Vina asmal vadyate (^B 13.1.5.15, TB 3.9.14.1)
Sg.N caus.pass.
Cpromot.)
'The veena is made to resound (lit. speak) for him'
(35) agnihotrl dohyamana (^B 12.4.1.9, 12)
Sg.N caus.pass.
(prom.) pres. part,
'the Agnihotf-cow, being caused to be milked ...'
(36) agnihotram dohyamanam (ibid. 6)
Sg.N caus.pass.
(prom.) pres. part,
'the Agnihotr-milk, being caused to be milked ...'
(37) agnyupasthSnam vacayitavyab (MS 1.6.10)
sg.A Sg.N (.prom. ;
caus. gerundive
'[he] should be made to recite the agnyupasthana'
(The causative gerundive refers to the promoted causee.)
4.8. The most important development of this stage, then, is the intro-
duction of an alternative marking for the causee, namely the instrumental,
which appears both with AA and with non-AA causatives. This development
clearly seems to be later than the generalization of the causative construc-
tion to non-AA verbs. Its source seems to lie in the reinterpretation of an
established pattern of (animate) instrumentals of means. It is this origin
which explains an important difference between instrumental and accusative
causees, namely that the former are marked for reduced 'agency' etc.
These findings, of course, are quite different from those of Cardona's
and will therefore call for quite a different explanation of the situation
encountered in Paijini's language.
4.9. There is, however, a certain potential difficulty with the above
explanation, a difficulty which requires some discussion, especially since it
will have some bearing on the explanation of Pacini's system: To characterize
the difference between instrumental and accusative causees, it was necessary
to invoke the notion 'affected agent'. It is therefore, one might argue, dif-
ficult to see how in examples like (18b) we can have an 'affected-agent' verb
(ruh-) , but an agent of that verb (the instrumental causee) which is 'not af-
fected'. Similarly, in examples like (26a), how can we have a 'non-affected-
agent' verb (da-) , but an 'affected-agent' (accusative-marked) causee?
This difficulty, however, is only an apparent one. For I believe it is
legitimate to distinguish between what we might call 'lexical' and 'pragmat-
ic' affected-agent status: By generalization from the more typical situa-
tions, verbs such as ruh- 'ascend, climb' whose agent normally can be con-
ceived of as affected by the verbal action, can be entered into the lexicon
as [+ affected agent] . And this marking may then permit or disallow the ap-
plication of certain syntactic processes. However, in actual usage, in part-
icular pragmatic circumstances, such as those obtaining in (18b), the agents
of such verbs might not actually be affected by, or benefit from, the action.
In those contexts, then, they could pragmatically be treated as [ - affected].
24
Mutatis mutandis, the situation will be the same for lexically non-affected-
agent verbs like da-.
From this point of view, then, the original class of causative verbs is
defined by the lexical feature [± affected], while Vedic-prose causee mark-
ing is governed by the pragmatic use of that feature.
Moreover, one could argue that since at this Vedic-prose stage, causat-
ives can in principle be made from all verbs, the lexical marking has lost its
significance, and only the pragmatic marking is relevant. There would thus
be no conflict between these two different classifications.
While this is no doubt the case at some, probably later stage of the lan-
guage, it is to be expected that this later situation did not come about over-
night, but that the old and new systems coexisted for some time. In fact,
given that even at this Vedic-prose stage, causatives are much more freely
attested with accusative causees than with instrumental ones, that it is still
often difficult to distinguish between instrumental-marked causees and means/
instruments, and that non-AA verbs still are comparatively rarely attested in
causative constructions, it is quite probable that the language of Vedic
prose reflects this transitional stage.
Rather than considering this to be an 'unpleasant' difficulty, however,
I would view this not only as natural but also as an important motivation for
the further developments of the causative. The fact that Classical Sanskrit
opted for a pragmatic [± affected] syntax while the language described by
Pacini decided for a more lexical approach can be seen as divergent attempts
to resolve, as it were, the conflict between pragmatic and lexical marking.
(Because the terms 'agentive' / 'agency' are shorter than terms like 'prag-
matically determined [± affected] marking', I will retain them in much of
the following discussion. Similarly, the terms 'affected agent' /AA will be
used to refer to lexically determined [± affected] marking.)
5. CAUSATIVES IN CLASSICAL SANSKRIT
Except for the fact that the number of non-AA verbs with causative con-
structions is on the increase, 16 the situation in Classical Sanskrit seems to
result from the straightforward generalization of the pragmatically determin-
ed [± affected] marking which had arisen in Vedic prose, and the consequent
elimination of the lexically determined use of that feature. As Speijer
(1886:36-7 with reference) states,
'If one wants to say "he causes me to do something, It is by his
impulse I act", there is room for the [accusative causee], but if it be
meant "he gets something done by me, I am only the agent or instrument
through which he acts", the instrumental is in its place ...'
Though stated slightly differently, this description is eminently compat-
ible with my finding for Vedic prose that the instrumental is marked for de-
creased 'agency' as compared to the accusative. In fact, the notion of re-
duced agency seems to be sufficient to characterize also the Classical use of
the instrumental and to account for the difference between instrumental and
accusative causees.
25
As a consequence, also here we find instrumental causees with all the
sub-classes of the affected-agent class, and accusative causees for non-AA
verbs. Some examples are the following. (In each case, I also give an ex-
ample of the opposite marking, either with the same verb, or at least with a
verb which belongs to the same lexical class.)
(A) Verbs: (38)
sa taiti akramayam asa suddhantam
pi. i caus.
'he had them enter the harem'
vs. ramati tam adhyaropayat plavam
sg.A caus.
'Rama made her go on board the ship'
(C) Verbs: (39) tam svabhib khadayet raja
pi. I caus.
'the king should have dogs devour her'
vs. mantraputam carum rajnim prasayat
sg.A caus.
'he made the queen eat a consecrated porridge'
(E) Verbs: (40)
(F) Verbs: (41)
sudena tam mallam yodhayam asa (MBh 7.191)
sg.I caus.
'he had the cook fight an athlete'
saiphatan yodhayet alpan (Manu 7.191)
pi. A caus. pi. A
'he should make a small group fight in close order'
sa rsin karam adapayat
P-L.A caus.
'he made the ?§is give = pay tax'
vs. panlyopadhaukayitypurugamukhena . . . bilvaphalam . . .
T ^^^ TZ
kapalikab . . . ra.jne pradldapat (Vet. 6. 2-4)
'an ascetic had the chief water bearer give the king
a bilva fruit'
(42) yen(a)
sakhi padam karita (Sak.145.7)
sg.N caus.
(prom.) pass. part,
'by whom the friend has been made to place confidence'
vs. na saksyami kimcit karayitum tvaya
caus. sg. i
'I will not be able to get anything done through you'
Especially instructive is the following pair of examples, taken from one
and the same Sanskrit play. In the active (43) , the focus seems to be on get-
ting the letter written, and the agency of ^akatadasa is of secondary import-
ance. Here the causee appears in the instrumental. In (44), however, the fo-
cus has shlf-tfid to ^akafadasa and the fact that he is the unfortunate person
who had been made to write, the letter. Consequently he is marked as an agent-
ive causative, underlyingly in the accusative which is then promoted to sub-
jecthood in the passive.
26
(43) lekham ^akat:adasena lekhayitva (Mudr .1.19.5)
sg. i caus.
'having had Sakafadasa write the letter'
(4A) yena. . . tadr^am kapa^alekham. . .lekhitati tapasvi sakatadasah
sg.I sg.A (Obj . ) caus. sg.N (promot.)
(Sub j .demo t. ) pass. part.
(Mudr. 7.9.5-6)
'by whom the unfortunate Sakafadasa was made to write such a fic-
titious letter'
5.2. A priori, however, (44) could perhaps also be explained as follows:
The alternative causative passive construction (45) , with non-promotable in-
strumental causee (and with consequent promotion of the accusative-marked
underlying object) would be extremely unusual, if not dovmright unacceptable.
For structures like (46) and (47) with both causee and demoted subject marked
by the instrumental are exceedingly rare.-*-^ Moreover, even if no instrumen-
tal-case demoted subject is specified, instrumental causees seem to be per-
missible in the passive only in structures of the type (48) , where the (pro-
moted) underlying object is human or where, put differently, the promoted ob-
ject is not lower on the animacy hierarchy than the (instrumental) causee.
On the other hand, structures like (49), if they occur at all, seem to be even
rarer than those of type (46). 2"
(45) ???/* yena. . .kapatalekhah. . .lekhitati tapasvina sakafadasena
sg.I sg.N (Obj., caus. sg. i
(Subj.dem.) prom.) pass. part.
'by whom, through unfortunate Sakatadasa, the fictitious
letter was caused to be written'
(46) visakanyakaya rSkgasena. . .ghatital;i tapasvi parvatesvarah
77"! sg.I Sub j . caus. „ /„, . X
sg.I /, V -^ sg.N (Obi., prom.)
(dem.) pass. par t.^ \ j > f /
(Mudr. 1.14. 4-5)
'the unfortunate Parvatesvara has been caused to be killed by
Rakgasa through the agency of the poison-girl'
(47) tena eva. . .datiijapisikaih sulena samaropitab (Vet. 148. 13-4)
sg.I pl~l caus.
(Subj.dem.) * pass. part.
'by him [l was] made to be put on the stake through the agency
of the police'
(48) bhagurayapena apavahitati parvatakaputrab (Mudr .1. 14. 7)
sg. I caus. pass. sg.N (Obj. prom.)
'the son of Parvataka was caused to be driven off through
Bhagurayana '
(49) 111/* sakatadisena lekhab lekhitab
sg.i sg.N caus. pass.
(Obj. prom.) part,
'the letter was caused to be written through Sakafadasa'
The above-noted restriction against structures like (49) seems to be
closely linked to the fact that in passives from causatives like (41) , where
27
both accusative-marked causee and underlying object are specified, it is in-
variably the causee, not the object, that gets promoted to subjecthood, as in
C50) below. Only where no causee is present at all can underlying objects
be freely promoted, as in (51) below. Structures of the type (52), however,
do not seem to be permissible.
(50) ghataka^i tvaya. . .bhayasaipjnam grahayitavyati (Mudr.1.19.28)
pl.N sg.I sg.A (Obj . ) caus. gerundive
(prom. ) (Subj .dem.)
'the executioners should be made by you to take fear'
(51) bhartra ... prasada^i ... dapita^i (^ak.223.13)
sg.I sg.A (Obj. caus.
(Subj. dem.) prom.) pass. part.
'the present has been caused to be given by the master'
(52) *ghatakan tvayS... bhayasaipjna grlhayitavya
'^^'"^ (Subjidem.) ^^'promO* ^aus. gerundive
(translation virtually impossible)
This leads to the conclusion that in Sanskrit the accusative of the cau-
see is 'higher' on the promotability hierarchy than the underlying object.
Moreover, even when marked instrumental, the causee seems to retain some of
this 'higher' status, by blocking promotion of underlying objects which are
lower in the animacy hierarchy. This is interesting, considering that con-
ventional wisdom has it that in causative clause union the causee gets demot-
ed to a position lower than that of the underlying object, even where morpho-
logically the two are marked identically.
6. CAUSATIVES IN THE LANGUAGE OF THE GRAMMARIANS
6.1. As noted earlier, the language described by Papini, roughly con-
temporary with the later stages of Vedic prose, has a causative sjrntax which
markedly differs from what we have observed in sections 4 and 5. In some
ways this situation looks remarkably archaic, in that it is based on lexical,
not pragmatic marking. In other ways, however, it is perhaps just as much
removed as the Classical language from what we find in the Rig-Veda and in
the later mantra collections, except that its divergence is in a different
direction.
The major distinction in Papini's grammar is between intransitives and
Class (A) -(D) ingestives on one hand and other verbs on the other. (Cf.
Papini's sutra 1.4.52 and the discussion in Cardona 1978.) The former take
accusative causees , the latter instrumentals. The only exceptions are hy-
'take, carry' and ky- 'do, make' which offer an option between instrumental
and accusative (cf. 1.4.53). Moreover, there is no indication that there
might be a difference between accusative causee and accusative object in
terms of their promotability to subjecthood in the passive. Presumably, both
are equally promotable. (For examples illustrating these various patterns of
causee marking and of passive promotion, see Cardona 1978; 1-9.)
6.2. The question which must of course arise is the following: How can
this system, so radically different from the Vedic prose-Classical language
tradition and also quite different from the early Vedic system(s) , be related
to these other systems?
28
A priori it would be possible to assume that Panini's language derives
from a source different from the type of language represented in the Rig-
Veda, and that in this different source, non-ingestive affected-agent verbs
simply never formed causatives in -aya-. However, considering that Panini's
language hails from the extreme Northwest of Ancient Indo-Aryan and that most
of the Rig-Veda likewise comes from that region, this argument is not very
attractive.
It is also possible to claim that Pacini's very different system is dif-
ferent because it has a very different historical development, more along the
lines of Cardona's hypothesis than along those suggested here for Vedic prose.
True, some special provision would have to be made, so as to narrow the Rig-
Vedic affected-agent class to just the ingestive sub-classes. However, this
could be easily done along more or less the same line of reasoning as that
which will be developed below.
The only argument against this view would be that it is not necessary,
because it is perfectly possible to derive Panini's system from that of Vedic
prose by means of developments diverging from those postulated for the Vedic
prose-Classical Sanskrit tradition, but nevertheless quite straightforward
and motivated:
As noted in 4.9, there is reason to believe that at the time (roughly)
of Vedic prose, two systems were still in competition with each other, one
in which the feature [± affected] was a lexical one and defined the classes
of verbs traditionally accepted in the causative construction, the other in
which the feature was a pragmatic one and determined the case of the causee.
In the Classical language, this competition was resolved in favor of the
pragmatic feature. However, there is no reason against a different variety
of Sanskrit making the opposite choice, by emphasizing the lexical feature.
Let us assume that this is what happened in the (pre-)history of Pacini's
language. What plausible consequences might this have?
For one thing, we might expect the instrumental marking of the relative-
ly non-affected-agent causee to become associated with, and generalized for,
the lexically non-AA verbs, while the AA accusative would become connected
with AA verbs. Some verbs, especially relatively basic ones like hr- and kf-
could then, as an archaism, retain traces of the earlier option. But since
the lexical, not the pragmatic feature [± affected] has become the important
one, this option would lose its old semantic/pragmatic connotations.
However, as noted earlier, the notion 'affected agent' is an inherently
unstable one, since the borderline between AA and non-AA verbs is not clear-
cut, but gradient. Especially the borderline between class (E) and class (F)
verbs is a very difficult one to draw. There seems rather to be a simple
continuum, from very much affected to very little. Moreover, in some of the
old class (E) verbs, most notably in spfs- 'touch' (but presumably also in a
number of other semantically similar verbs; cf. e.g. sli§-) , there was the
difficulty that causee and object are not always clearly distinguishable and
that in addition, instrumental case marking was one of the options for this
indistinguishable causee/object category. Under these circumstances, a good
way out would seem to consist in giving up the idea of drawing a borderline
between class (E) and class (F) verbs and to establish the line further 'up'
(in terms of af fectedness), at the more clearly defined border between 'in-
29
gestive' and 'non-ingestive' transitives. With this development, however,
the stage of Panini's language has been reached.
6.3. Later grammarians proposed a number of additions and modifications
to Panini's rules. (Cf. e.g. Liebich 1886a: 211-12.) The purpose of the ad-
ditions and modifications, however, is not always clear. For instance, if a
varttika states that nl- 'lead' and v'ah- 'convey' take the instrumental cau-
see, not the accusative, does this mean that there has been a change in the
language? And if so, in what language? That is, did there continue to exist
a grammarians' Sanskrit, separate from the Vedic prose-Classical tradition?
Or do these statements perhaps rather reflect attempts to reconcile the dif-
ferences between actual usage (within the Vedlc-Classical tradition) and
the rules laid down in Pipini's grammar which, in principle, had gotten to
be accepted as the model for correct Sanskrit usage? The fact that most of
these statements address relatively minor issues and do not seem to reflect
actual usage in the Vedic-Classical tradition any better than does Pacini's
grammar, suggests that these statements were indeed simply isolated, non-
systematic attempts to reconcile a particular usage noted by a given grammar-
ian with the rules laid dovm by Papini.
For these reasons I believe that this historical account of the Sanskrit
causative can stop with Papini and need not concern itself with post-Papinian
grammarians.
7. CONCLUSIONS
The major conclusions reached in this paper can be summarized as fol-
lows: At its earliest attested stage, Sanskrit already permitted causatives
from intransitives and (certain) affected-agent transitives. If we disre-
gard lexicalization, the marking of the causee is accusative; and this ac-
cusative causee acts like a genuine object in being promo table to subject-
hood in the passive of the causative.
The next stage of the language, that of the post-Rig-Vedic mantra col-
lections, gives evidence that the class of verbs open to the causative con-
struction was being expanded to include all transitives. The case of the
causee remains accusative, even for the new transitive additions.
It is at the next stage, that of Vedic prose, that the next important
step is taken, namely the development of an alternative, instrumental causee
marking, apparently resulting from reinterpretation of (animate) instrument-
als of means /instrument. This origin accounts for the special, less 'agent-
ive' connotations of this new marking. The resulting conflict between a lex-
ically determined notion 'affected agent' and a pragmatic one can be consid-
ered to give rise to divergent developments in the Classical language on one
hand, Pacini's language on the other. The former opts for the pragmatic
approach and thus continues the development begun in Vedic prose. The latter
chooses the lexical approach and thus, in a sense, reverses the Vedic-prose
development. (In Pacini's language, a further change leads to a lexical re-
classification: the major distinction now is between ingestive and non-in-
gestive verbs, not between affected-agent and non-affected agent verbs.)
Both the Classical and Pacini's language continue the phenomenon al-
ready observed in the Vedic language, namely that accusative causees are pro-
30
mo table to subjecthood in the passive. In Classical Sanskrit, there is in
addition a strong tendency to give preference to the accusative causae over
the accusative object for promotion (if both causee and object are specified).
Moreover, even instrumental causees block object promotion if the object is
lower on the animacy hierarchy. This suggests that in causative clause union
the causee is, in Sanskrit, demoted to a position higher than the underlying
object.
NOTES
This work has been in part supported by a 1979/80 grant from the Univer-
sity of Illinois Research Board. It has grown out of, and benefited from,
conversations with my friend George Cardona, during the 1978 Linguistic In-
stitute at the University of Illinois. The conclusions reached in this paper,
however, are clearly different from Cardona' s. I have also benefited from
discussions with participants in my seminar on diachronic syntax, offered
Spring 1980 in the Department of Linguistics, University of Illinois. Any
errors and omissions in this paper, however, must be considered my responsib-
ility.
2
Accent, being irrelevant to the discussion, has been omitted in all
examples given in this paper. In addition, sentences are quoted with words
in their 'pausa' form, without the external sandhi which often obscures
word boundaries. Data from the Rig-Veda, Atharva-Veda, Vajasaneyi-S^hita,
Kathaka-Saijihita, Aitareya Brahm^na, Jaiminlya Brahmapa, Taittiriya Saiphita
and Brahmaua, ^atapatha Brahmaija, Kalidasa's ^akuntala, Visakhadatta' s
Mudraraksasa, and the Vetalapancaviipsati have been collected and/or cross-
checked by myself. (Except for the Atharva-Veda, I have worked through what
I believe are all the relevant attestations in these texts. For the Atharva-
Veda I have relied mainly on cross-checkings of forms quoted elsewhere in the
literature.) Other data come largely from the following sources: For other
Vedic texts, Cardona 1978 and Delbriick 1888, plus Wecker 1906 for the Upani-
§ads; for the Classical literature, Speijer (1886:36-8) and Renou (1961:472-
3); for the grammarians, Pacini's Agfadhyayl and Cardona 1978, Liebich 1886a.
For simplicity of exposition, Cardona' s presentation has been modified.
^Cardona continues (ibid. , with n. 22) by stating that this pattern per-
sists into Modern Indo-Aryan, including Hindi. However, as I will show else-
where in greater detail, the major distinction in Hindi is not between in-
transitives and class (A) -(D) verbs on one hand and other transitives on the
other. Rather, it is between intransitives and 'affected-agent' transitives
on one hand and other transitives on the other. For the time being, see
Kachru 1976 and 1979.
In fact, there must be some question as to whether these are really
transitive in Rig-Vedic Sanskrit: The only passive-like attestation found
from any of the (A) verbs in Rig-Vedic times, at a time when only transitives
can appear in the passive construction, is the example quoted by Cardona (22):
. . . gataji na adhvg 'like a path that has been traversed'. The promoted object
of this sentence, however, seems to be semantically and syntactically compar-
able to the 'cognate' or 'etymological' objects which can appear with many
other, clearly intransitive verbs. (Cf. Engl. I dreamt a dream : a dream
was dreamt.) Moreover, even in later Sanskrit, including that of Panini,
31
verbs of going only optionally have the promotional passives typical of tran-
sitives. They can also take the purely demotional passive of intransitives.
Cf. Cardona's examples C30) and C62) .
This term has recently been the title of a paper by Saksena (1980).
However, Saksena 's evidence consists only of ingestive verbs and thus does
not go substantially beyond Masica 1975. Moreover, as I will show elsewhere,
some of her arguments do not hold even for all of the ingestive verbs, in the
language treated by her. A much better treatment, using different terms, is
that in Kachru 1976:363, 368 and 1979:310, 313.
'It may be speculated that these developments had taken place in the ex-
treme Northwest, from which Panini hailed. The Vedic prose-Classical tradi-
tion, on the other hand, should perhaps be attributed to the central area of
Madhyadesa which had become the center for most religious and literary activ-
ities. CCf. Hock and Pandharipande 1976:125-8 (with n. 9a) for discussion
and references concerning this dialectal distinction.)
Q
°Here as elsewhere my discussion will be limited to the verbs of classes
(A)-(F) and to genuine causatives of these verbs. Moreover, except for a
few relevant casesy I will not discuss what happens to the underlying object
or goal of such verbs. As Cardona correctly has shown, the normal pattern
is for the case of the object or goal to be retained in the causative active.
^Not all instances of the -aya-f ormations of this root have causative
value; but some clearly do.
■'■Of these causatives, two have outside cognates suggesting that also
these affected-agent causatives may be inherited. For vasaya- 'make s.o.
wear s.th.; clothe s.o.', cf. Goth, wasjan, OE werian 'clothe (oneself /s.o.
else)', OHG werian, ON verja 'clothe s.o.'. (The tendency in Germanic is
to use this verb not only transitively, but also reflexively/intransitively.)
For yodhaya- 'make s.o. fight, order (in)to battle', cf. Lat. iubeo, OLat.
ioub^- 'order'. (On the phonological developments and remakings in Latin,
cf . Leumann et al. 1977:541.) Note also that at least for the root yudh-
there is syntactic synchronic evidence for its transitivity, namely ayuddha-
(RV 8.35.3, 10.27.10) 'irrestible, not to be fought (against)', with passive-
like value of the negated past participle. (For vas- and spys- no unambig-
ously passive-like formations seem to occur in the early language, vas- has
no attestations whatsoever of potentially passive-like ta-participles; spfs-
only has a few, but none of these gives unambiguous evidence concerning tran-
sitivity.)
As a consequence, Haudry's claim (1977:383-4 and passim) that the Vedic
accusative-marked causee is not a true object needs to be reconsidered.
12
The discussion of this section will be limited to the evidence of the
older, Brahmana texts (including the brahmapa portions of the various samhi-
tas of the Black Yajur-Veda)^, except that the Brhad-Arapyaka is here included
as part of the ^atapatha Brahmana. The description in Wecker 1906, supple-
mented by Cardonafs examples, suggests that the situation in the later Vedic
prose is not substantially different.
12a
The instrumental in (21b), then, may have been mechanically transfer-
32
red from (22b) which closely precedes in the same passage.
13
However, in all fairness it should be mentioned that in this formulaic
expression, animate 'causees predominate.
Liebich (1886b:271), apparently unaware of the antiquity and special
motivation of this pattern, identifies this as an example of instrumental
causee.
These glosses are intended only for identification. The actual mean-
ings are more complex: In Vedic ritual, the person (A) who 'sacrifices for
himself normally does not actually perform the sacrifice, but has it per-
formed on his own behalf by a priest who is said to 'sacrifice for A' or to
'cause A to sacrifice for himself /be a sacrificer'.
■"^^Cf . the data in Whitney 1885.
In fact, it must be said in all fairness that in addition to Cardona's
paper, it was my much earlier reading of Speijer's account and my realization
both of its general accuracy for Classical Sanskrit and of its considerable
difference from the predictions of the grammarians which got me interested in
further exploring the history of this construction.
1 8
Unless otherwise indicated, the examples in this section are from
Speijer 1886, where the sources of the examples are detailed.
19
Cf. also Speijer 1886:38. — In fact, (47) seems to be an example of a
passive from a double causative. And though double causatives are not fre-
quently attested in the literature, it seems that the 'topmost' causee in such
constructions normally appears in the instrumental, not in the accusative.
That is, the corresponding active would have had the causee in the instrument-
al to begin with: sa . . . dapjapasikaiti mam sulena sam5ropayati 'he has the
police make me climb with = on the stake'. Moreover, in this example the pro-
moted NP is not the underlying object, but the 'lower-level' causee.
20
Structures of this sort are, however, freely permissible if the instru-
mental noun phrase is not the causee, but the demoted subject, as in capaka-
yena prathamalekhltab . . . lekhati (Mudr.5. 1.8-9) 'the letter first caused to
be written by Capakya'. Cf. also examples (50) and (51).
ABBREVIATIONS OF TEXTUAL REFERENCES
AB = Aitareya Brahmapa; AV = Atharva-Veda (Saunaka recension); JB = Jaimi-
nlya Brahmana (Caland's selections); KS = Ka^haka-Saiphiti; MS = Maitra-
yinl-Samhita [non vidi]; Mudr. = MudrSraksasa (references to the 1930 edi-
tion by Dhruva) ; RV = Rig-Veda; ^B = ^atapatha Brahmana (Madhyandina re-
cension); ^ak. = ^akuntala (page and line references to the 1886 Monier-
Williams edition); TB = Taittiriya Brahmapa; TS = Taittiriya-Sa^hita;
VS = Vajasaneyi-Saiphita (Madhyandina recension); Vet. = Vetalapancaviipsati
(Jambhaladatta's version; page and line references to the 1934 Emeneau edi-
tion) .
33
REFERENCES
CARDONA, George. 1978. Relations between causatlves and passives in Indo-
Iranian. Linguistics in the seventies: directions and prospects, ed. by
B. B. Kachru, 1-42. (= Studies in the Linguistic Sciences, 8:2.)
DELBRUCK, Bertold. 1888. Altindische Syntax. Reprinted 1968, Darmstadt:
Wissenschaf tliche Buchgesellschaf t.
HAUDRY, Jean. 1977. L'emploi des cas en vedique. Lyon: L'Hermes.
HOCK, Hans Henrich, and Rajeshwari Pandharipande. 1976. The sociolinguistic
position of Sanskrit in pre-Muslim South Asia. Dimensions of bilingual-
ism: theory and case studies, ed. by B. B. Kachru, 105-38. (= Studies
in Language Learning, 1:2.) Urbana: University of Illinois, Unit for
Foreign Language Study and Research.
JAMISON, Stephanie. 1976. Functional ambiguity and syntactic change: the
Sanskrit accusative. Papers from the Parasession on Diachronic Syntax,
126-35. Chicago: Chicago Linguistic Society.
KACHRU, Yamuna. 1976. On the semantics of the causative construction in
Hindi-Urdu. Syntax and semantics, 6.353-69. New York: Academic Press.
197 9. Notes on participant roles and grammatical categories in Hindi-
Urdu sentences. Linguistics and literary studies in honor of Archibald
A. Hill, 4.309-18, ed. by M. A. Jazayerey et al. Amsterdam: de Ridder.
LEUMANN, Manu, et al. 1977. Lateinische Grammatik, 1. Miinchen: Beck.
LIEBICH, Bruno. 1886a, b. Die casuslehre der indischen grammatiker vergli-
chen mit dem gebrauch der casus im Aitareya-Brahmapa, 1 and 2. Beitrage
zur Kunde der indogermanischen Sprachen 10.205-34 and 11.273-315.
MASICA, Colin. 1975. Defining a linguistic area. Chicago: University of
Chicago Press.
RENOU, Louis. 1961. Grammaire sanscrite. Paris: Adrien-Maisonneuve.
SAKSENA, Anuradha. 1980. The affected agent. Language 56.812-26.
SPEIJER, J. S. 1886. Sanskrit syntax. Reprinted 1973, Delhi: Motilal
Banarsidass.
WECKER, Otto. 1906. Der gebrauch der kasus in der alteren Upanisad-litera-
tur. Beitrage zur Kunde der indogermanischen Sprachen 30.1-61.
WHITNEY, William Dwight. 1885. The roots, verb forms, and primary deriva-
tives of the Sanskrit language. Reprinted 1945, New Haven: American
Oriental Society.
Studies in the Linguistic Sci en ces
Volume 11, Number 2, Fall 1981
ON THE SYNTAX, SEMANTICS AND PRAGMATICS
OF THE CONJUNCTIVE PARTICIPLE IN HINDI -URDU
Yamuna Kachru
University of Illinois, Urbana-Champaign
The conjunctive participial construction in Hindi-Urdu
has several meanings such as consecutive action/process, manner,
cause, etc., and has been the topic of much linguistic discussion.
This paper examines the three major claims made recently
with regard to this construction: (a) Its various meanings
notwithstanding, the conjunctive participle is derived from
one simple subordinate source (Davison 1981); (b) the participle
marker kar has a 'perfective' meaning (Davison 1981); and
(c) the main verb determines whether the adverbial interpretation
is preferred (Abbi 1980, Davison 1981) . A wide range of data
and a number of arguments are presented in this paper to show
that none of the above claims can be supported. Instead,
it is argued that (a) the construction represents a target
structure phenomenon (Green 1974), i.e., several underlying
representations all end up as the participial construction
on the surface; (b) the participle marker kar has at least
two different meanings, 'perfective' and 'simultaneous';
and (c) the semantics of both the participle and the main verb
together determine whether the adverbial interpretation is
possible. It is pointed out in the conclusion that 'a mostly
pragmatic' analysis of the construction (Davison 1981) misses
a number of crucial styntactic and semantic generalizations.
INTRODUCTION
The conjunctive participial construction (hereafter CP) in Hindi-
Urdu has been investigated in a number of recent studies (Abbi 1980,
Bahl 1974, Bhatia 1977, Davison 1981, Dwarikesh 1971, Kachru 1965, 1966,
1978, 1980, to name just a few). The most interesting claim that has
been made so far is that the several meanings associated with the
conjunctive participle are all pragmatically determined (Davison 1981).
In this paper, I shall examine this claim and provide various arguments
to show that the meanings of the construction under focus are to be
accounted for in terms of its syntax and semantics and in most cases,
it is unnecessary to invoke pragmatic principles. I shall do so by
first examining the following three claims of Davison 1981 and showing
how they do not work: one, the conjunctive participle has the following
syntactic and semantic underlying representation (Davison 1981) :
Y V-kar V Aspect Tense
'perfective'
36
Second, the conjunctive participle marker kar has a 'perfective' meaning
(Davison 1981) and, third, the main verb determines whether the adverbial
interpretation is preferred (Abbi 1980, Davison 1981). 2
In my view, none of the above claims can be supported if the total
range of data is taken into consideration. In fact, the evidence supports
the hypothesis that the conjunctive participial construction is a target
structure phenomenon. That is, several underlying representations
are realized on the surface as the conjunctive participial construction.
Second, the syntactic and semantic evidence show that 'perfective'
is not the only meaning associated with the formative kar. Third,
it is not just the properties of the main verb that determine the
adverbial meanings. Rather, the properties of the subordinate verb
that takes the kar and the properties of the main verb jointly determine
if specific verbs could occur in the underlying representations that
are realized as adverbials.
THE DATA
Let us consider a range of data which consists of sentences with
the conjunctive participles and their paraphrases. Examples 2-4 are
clear cases of temporal reading where the main verb expresses process
or action subsequent to the process/action expressed by the participle.
3
2. usne nahaa kar khaanaa khaayaa
he ag. bathe CP meal ate
Having bathed, he ate (his) meal.
2a. usne nahaaya aur tab khaanaa khaayaa
he ag . bathed and then meal ate
He bathed and then ate (his) meal.
3. ghar aakar raaj ne kapRe badle
home come CP Raj ag. clothes changed
Having come home. Raj changed.
3a. raaj ghar aayaa aur tab usne kapRe badle
Raj home came and then he ag. clothes changed
Raj came home and then changed .
4. New York jaa kar mujhe pataa calaa ki shyaam biimaar hai
New York go CP I came to know that Shyam ill is
Having arrived in New York, I came to know that Shyam was ill.
4a. mal New York gayaa aur tab mujhe pataa calaa ki shyaam
I New York went and then I came to know that Shyam
biimaar hai
ill is
I went to New York and then I came to know that Shyam was ill.
Note that in each case, the process or action expressed by the CP
precedes the process (4) or action (2-3) expressed by the main verb.
37
In this case, the CP has a 'perfective' meaning.
In examples 5-8, the CP's express a manner adverbial meaning.
5. (vahaa baRii phislan thii), khuub pSav jamaakar
there much slipperiness was well feet plant CP
calnaa paRaa
had to move
It was very slippery, we had to walk with very steady steps.
4
5a . * . . . khuub pasv jamaanaa paRaa aur tab calnaa paRaa
well feet had to plant and then had to walk
...We had to plant our feet very firmly and then walk.
5b . ... khuub sadhe pSavo se calnaa paRaa
well steady feet with had to move
. . .We had to walk with very steady steps.
6. vah ThaThaa kar hSs paRaa
he explode CP laugh fell
He burst out laughing (in peals of laughter) .
6a . *vah ThaThaayaa aur tab has paRaa
he exploded and then laugh fell
6b . vah baRe zor8 se has paRaa
he very loudly laugh fell
He burst out laughing (loudly) .
7. vah sabse has kar baatS kartaa hai
he all with laugh CP talking does
He talks to everyone pleasantly.
7a. *vah sabse hastaa hai aur tab baate kartaa hai
he all with laughs and then talking does
He laughs with everyone and then talks (to them) .
7b. vah hastaa hai aur tab sabse baatg kartaa hai
he laughs and then all with talking does
He laughs and then talks to everyone.
7c. vah sabse haste hue / prasanna citta se baaTe kartaa hai
he all with laughing happy heart with talking does
He talks to everyone pleasantly/happily.
8 . vah mere saamne se katraa kar nikal gayaa
he my front from avoid CP emerge went
He passed in front me avoiding (me) .
I.e., He shied away from me.
8a. *vah mere saamne se katraayaa aur tab nikal gayaa
he my front from avoided and then emerge went
38
8b. * [vah katraayaa] aur tab mere saamne se nikal gayaa
he avoided and then my front from emerge went
8c. vah mujhse bactaa-huaa-saa mere saamne se nikal gayaa
he me from saving like my front from emerge went
He passed in front of me, in a way avoiding me.
Notice that in cases of 5-8, the (a) versions which are paraphrases
with the conjunction 'and then' are ungrammatical . In cases of 5-6,
the (b3 versions are close paraphrases with manner adverbs. In case
of 7, the (b) version is acceptable, but semantically distinct, the (c)
version is a close paraphrase with manner adverbials. In the case of
8, the first clause in (8b) is ungrammatical, (8c) is a close paraphrase
with a manner adverbial. Note also that the CP's in 5-8 can be conjoined
with other manner adverbials,
9. is tarah paav jamaa kar nahJi, halke se calo
this way feet plant CP not lightly walk
Walk lightly, not so firmly.
10. usne mujhse has kar nahJJ, beman se baate kii
he ag. me with laugh CP not unwillingly talking did
He talked to me unwillingly, not pleasantly.
Such conjoining in cases of 2-4, yields semantically anomolous results.
11. *usne nahaa kar nahtt, xushii se khaanaa khaayaa
he ag. bathe CP not happily meal ate
He ate his meal happily, not after his bath.
12. *raaj ne ghar aakar nahJi, utsuktaa se kapRe badle
Raj ag. home came CP not enthusiastically clothes changed
Raj changed his clothes enthusiastically, not after coming home.
If, however, the conjoined versions of 2-3 contain time adverbials, the
resulting sentences are grammatical.
11a. usne nahaa kar nahJi, Thiik muh dhone ke baad/nahaane se
he ag. bath CP not exactly face washing after bathing from
pahle/roTii bante hii khaanaa khaa liya
before as soon as the bread was made meal eat took
He ate his meal just after washing his face/before bathing/
as soon as the bread was made, not after bathing.
12a. raaj ne ghar aakar nahJi, usse pahle/Tenis khelne ke baad/
Raj ag. home come CP not that before tennis plaving after
das baje kii xabar sunte hue kapRe badle
10 o'clock news listening clothes changed
Raj changed his clothes before coming home/after playing tennis/
while listening to the 10 o'clock news, not after coming home.
39
It is clear that the temporal CP expresses sequential action, the CP
with manner adverbial meaning does not express sequential action/process.
The manner in which such an act is performed or a process takes place has to
be coextensive with it in time. Hence, in these cases, kar could not
mean 'perfective'.^
The next set of sentences has concessive (13-14), antithetical
(15-16), and causal (17-18) meanings. 6
13. turn mere mitra ho kar bhii meraa vishvaas nahii karte
you my friend be CP even my trust not do
though
You don't trust me even though you are my friend.
14. apne dost ramesh se mil kar bhii use xushii nahii huii
self's friend Ramesh with meet CP even him happiness not happened
though
He was not happy even though he met his friend Ramesh.
15. mohan khaanaa na khaa kar TV dekhne lagaa
Mohan meal not eat CP TV see began
Mohan began to watch TV instead of eating.
16. savitaa shaam ko ghar na aakar ajaayabghar calii gaii
Savita evening home not come CP museum went off
Savita went off to the museum instead of coming home in the evening.
7
17. vah raat - din kaam karke biimaar paR gayaa
he night-day work do CP ill fall went
He fell ill (because of) working day and night.
18. vah Duub kar mar gayaa
he drown CP die went
He died by drowning.
Sentences 13-18 have the following exact paraphrases.
13a. yadyapi/haalaaki tum mere mitra ho, to/phir bhii meraa
although you my friend are then even my
vishwaas nahii karte
trust not do
Although you are my friend, you don't trust me.
14a. yadyapi vah apne dost ramesh se milaa phir bhii use xushii
although he his friend Ramesh with meet even so him happiness
nahii huii
not happened
Although he met his friend Ramesh, he didn't feel happy.
40
15a. mohan khaana khaane ke bajaay TV dekhne lagaa
Mohan meal eat instead TV see began
Mohan began to watch TV instead of eating.
15b. mohan ne khaanaa nahfr khaayaa, ulTe TV dekhnee lagaa
Mohan ag. meal not ate instead TV watch began
Mohan didn't eat, instead, he began to watch TV.
16a. savitaa shaam ko ghar nahrf aaii, ulTe ajaayabghar calii gaii
Savita evening home not came instead museum went off
Savita didn't come home in the evening, instead, she went
off to the museum.
16b. savitaa shaam ko ghar aane ke bajaay ajaayabghar calii gaii
Savita evening home come instead museum went off
Savita went off to the museum instead of coming home in the evening.
17a. vah raat - din kaam karne kii vajah se biimaar paR gayaa
he night day work doing because of ill fall went
He fell ill because of working night and day.
18a. vah Duub jaane kii vajah se mar gayaa
he drown going because of die went
He died because of drowning.
It is tempting to suggest that the concessive reading in 13-14 is a
function of the particle bhii and the negation of the main verb, as
has in fact been suggested in Davison 1981. But, this would not work,
since it is not obligatory for the subordinate or the main clause to
be in the negative and for the CP to be followed by the particle bhii.
What is needed is a strong element of unexpectedness and one of the markers,
as in 19-20.
19. unkaa laRkaa caudah saal kaa ho kar bhii akele sone mg
their son fourteen year of be CP even alone sleep in
though
Dartaa hai
afraid is
Their son is afraid of sleeping alone eventhough he is fourteen.
20. usne caar saal landan rah kar (bhii) briTish myuuziam
he ag. four years London live CP even British Museum
though
nahii dekhaa
not saw
He didn't see the British Museum even though he lived in
London for four years .
It should be pointed out that although both the temporal CP as well as
the concessive CP can be followed by the particle bhii, only the former
has an inclusive meaning. Compare sentences 21 and 22, below.
41
21. vah ghar jaakar bhii paRhegaa
he home go CP also study will
He will study also after going home.
21a. [vah din bhar laaibrerii me paRhegaa) iske alaavaa ghar
he day whole library in study will in addition home
jaa kar bhii paRhegaa
go CP also study will
He will study in the library all day, in addition, he will
also study after going home.
22. vah raajpuut ho kar bhii laRaaii se Dartaa hai
he warrior be CP even war from fears
though
He is afraid of wars even though he is a warrior.
22a. vah shantipriya hai, isliye laRaaii se Dartaa hai. *iske
he pacifist is therefore war from fears in
alaava raajpuut ho kar bhii laRaaii se Dartaa hai
addition warrior be CP also war from fears
He is a pacifist, therefore, he is afraid of wars. *In addition,
he is also afriad of wars being a warrior.
The particle bhii 'even' in 22, is an integral part of the concessive
conjunction and contributes to the "contrary to expectation" meaning.
The particle bhii 'also' in 21, on the other hand, is the "inclusive"
particle as in 23.
23. is itvaar ko mere kaii dost aa rahe hai, aur ek
this Sunday my several friends come -ing are and one
rishtedaar bhii .
relative also
This Sunday, several of my friends are visiting me, and also
a relative of mine.
The antithetical meaning in 15 and 16 is not pragmatically determined,
in fact, both 34b and 35b cited in Davison 1981 are grammatical [24-25
in this paper] .
24. raajaa [mere paas] na aakar kitaab paRhne lagaa
Raja [to me] not come book read began
Raja began to read instead of coming [to me].
25. vah patra na paRh kar bolaa [ki . . . ]
he letter not read CP spoke [that...]
Instead of reading the letter, he said [that...]
The na 'not' preceding the CP in 24-25, signals the meaning 'contrast*.
The temporal, manner, or causal adverbial CP's may not have na_ 'not'
in them; the temporal has an alternate construction as in 26 to negate
the CP.
26. *usne na nahaa kar khaanaa khaayaa
he ag. not bathe CP meal ate
42
26a. usne binaa nahaae khaanaa khaayaa
he ag without bathing meal ate
He ate his meal without bathing.
27. *vah sabse na has kar baate kartaa hai
he all with not laugh CP talking does
He talks to everyone not laughing.
pleasantly.
27a. *vah sabse binaa hase baate kartaa hai
he all with without laugh talking does
He talks to everyone not laughing.
pleasantly.
28. *vah raat - din kaam na karke biimaar paR gayaa
he night day work not do CP ill fall went
He fell ill because of not working day and night.
2 a. *vah binaa raat - din kaam kiye biimaar paR gayaa
he without night day work doing ill fall went
He became ill because of not working day and night. [I. e..
His not working day and night was the cause of his illness.]
The reason for this distribution of the negative particle is clear.
In sentences with temporal CP, the negative may have the CP, the finite
verb or the entire predicate in its scope, as is evident from the
following.
29. usne nahaa kar khaanaa nahii khaayaa
he ag . bathe CP meal not ate
He didn't eat his meal after bathing.
29a. usne nahaa kar khaanaa nahii khaayaa, nahaane se pahle
he ag . bathe CP meal not ate bathing from before
khaa liyaa
ate
He didn't eat after bathing, he ate before bathing.
29b. usne nahaa kar khaanaa nahJi khaayaa, siidhe dafter calaa gayaa
he ag. bathe CP meal not ate straight office went off
He didn't eat after bathing, he went off to his office (after
bathing) .
29c. usne nahaa kar khaanaa nahii khaayaa, haath-muh dho kar
he ag. bathe CP meal not ate hand face wash CP
sirf caay pii lii
only tea drink
He didn't eat after bathing, he only had tea after his wash.
In case of manner, cause, etc., as expected, usually the adverbial is
in the scope of negation.
43
30. turn man lagaa kar nahii paRhte
you mind apply CP not study
You don't study diligently.
31. vah tumhaarii baatg sun kar dukhii nahii huaa hai
he your talking listen CP sorrowful not become has
He is not sad because of what you said.
This is true of adverbs of manner, cause, instrument, etc., as is clear
from the following.
32. usne jaldi se/kalcii se/ tumhaare kahne kii vajah
he ag. quickly scissors with your saying of reason
se parde nahli kaaTe
from drapes not cut
He didn't cut the drapes quickly/with scissors/because of your
asking him to.
In 32, the adverbs 'quickly', 'with scissors', 'because of your asking'
are in the scope of negation (Bhatia 1977). Similarly, in 30 and 31,
the CP's are in the scope of negation. It is hard to imagine why one
would have to specify negative manner, cause, or instrument for actions/
processes unless contrastive information is to be given. In that case,
however, the regular strategy illustrated in 9-10 is adopted.
THE MEANINGS OF CP
To sum up the discussion so far, at least five different meanings
of the CP construction are illustrated in the above examples. These
may be labelled as: 'and then', manner, causal, concessive, and anti-
thetical. This does not mean that some CP's are not ambiguous, in
fact, the following is structually ambiguous.
33. usne ghar aakar bhii khaanaa nahii khaayaa
a) He didn't eat even though he came home.
b) He didn't eat after coming home either.
In the sense of 33b, the particle hii 'only' can replace the particle
bhii 'even' in which case the sentence would mean "he didn't eat only
after coming home. "8 In the sense of 33a, hii 'only' can not replace
bhii 'even'. The structural properties that distinguish the five
different meanings are as follows:
(a) The 'and then' temporal sequential CP answers the question
'when', whereas the manner and causal adverbs answer the question 'how'
and 'why';
(b) The particle bhii 'even' is an obligatory constituent of the
concessive, and na 'not' of the antithetical, tab 'then' is an optional
constituent of the 'and then' CP;
44
(c) The 'and then' CP may contain the inclusive particle bhii
'also, even' or the exclusive particle hii 'only' optionally whereas
the concessive cannot contain hii 'only';
(d) The temporal sequential, manner, and causal adverbials may be
conjoined with like-adverbials but not with unlike-adverbials, as
is clear from 34a and 35a below.
34. vah patra likh kar aur use Daak me Daal kar tab soyaa
he letter write CP and it mail in put CP then slept
He slept after writing the letter and mailing it.
34a. *vah mere kamre me aakar aur thak kar so gayaa
he my room in come CP and tire CP sleep went
He went to sleep after coming into my room and because of
being tired.
35. usne merii salaah maan kar aur uske anusaar cal kar
he ag. my advice accept CP and it of according proceed CP
saphaltaa praapt kii
success obtain did
He obtained success by listening to my advice and following it.
35a. *usne merii salaah maan kar aur has kar sabse baatg kii
he ag. my advice accept CP and laugh CP all with talking did
He talked to everyone because of my advice and pleasantly.
The structural properties and paraphrase relations taken together
suggest that the CP-construction must be derived from several under-
lying representations.
UNDERLYING REPRESENTATIONS: A TENTATIVE PROPOSAL
A tentative proposal as to the underlying representations needed
to derive the range of CP's discussed in this paper is as follows.
36. a. SEQUENTIAL: possibly a coordinate source. Note that
the CP and the finite verb can have different time adverbs:
(i) raaj das saal landan rah kar kal hii dillii lauTaa hai
Raj ten years London live CP yesterday only Delhi returned has
Raj came back to Delhi only yesterday after spending ten
years in London.
Note that barring a few exceptions (Kachru 1980), the subject of CP
must be identical to the subject of the finite verb, which may be termed
the like-subject constraint.
b. CAUSAL: a subordinate source common to CP and kii vajah se.
The CP construction is possible only if the like-subject constraint
holds.
45
c. MANNER: a subordinate source common to CP and other manner
adverbs: the like-subject constraint holds for the CP construction.
d. CONCESSIVE: a subordinate source with the markers yadyapi/
haalSaki. . . to/phir bhii; for CP construction, the like-subject constraint
holds.
e. ANTITHETICAL: possibly a coordinate source with the
antithetical conjunction ulTe 'instead'; the ke bajaay phrase is a
reduced version and so is the CP, the CP having an additional constraint
(like-subject) .
That the properties of the finite verb and the CP must be compatible
is evident from examples such as the following.
37 . vah baRii ho kar Daaktar banegii
she big become CP doctor become will
She will be a doctor when she grows up.
38. *vah man lagaa kar biimaar paR gayaa
he mind apply CP ill fall went
*He fell ill attentively.
CONCLUSION
It is clear that more research is needed to come to definite
conclusions. The following observations can, however, be made tentatively
with regard to the clear cases. If both the CP and the finite verb
involve natural processes or verbs of volitional acts, unless the verbs
are semantically related in specific ways, the interpretation is that
of temporal -sequential action.
39. usne acchii tarah nahaa kar tab roTii khaaii
he ag. well bathe CP then bread ate
He ate (his) meal after bathing.
40. usne acchii tarah cabaa kar (*tab) roTii khaaii
he ag. well chew CP then bread ate
He ate the bread, (*after) chewing it well.
In sentence 39, nahaa 'bathe' and khaa 'eat' are both volitional acts
but not related semantically. In 40, cabaa 'chew' expresses a special
action within the domain of the action expressed by khaa 'eat'. The
same is true of hasnaa 'laugh' and baat karnaa 'talk' and dauRnaa 'run'
and aanaa ' come ' (i.e., 'talk' is a general verb of communication,
'laugh' expresses a special kind of communication, and so on). Similarly,
if the CP expresses a process undergone by an experiencer and the finite
verb expresses a volitional act or vice-versa, the combination yields
a causal meaning.
46
41 . vah thak kar so gayaa
he tire CP sleep went
He went to sleep because he was tired.
42. uskii baate sun kar man uub gayaa
his talk listen CP mind bore went
I was bored listening to him.
This, however, does not mean that a causal meaning is completely blocked
if both the CP and the finite verb are volitional. One example of two
volitional verbs yielding a causal reading is as follows.
43,
tumhaaraa patra paRh kar vah rone lagii
your letter read CP she cry began
a. She began to cry after she read your letter.
b. She began to cry because she read your letter.
In this context, it may be useful to note the differences between the
causal adverbial meaning of the CP construction and the following
use of kar.
44. raNaa ne usko zahar dekar maar Daalaa
Rana ag . him poison give CP hit poured
Rana killed him by giving him poison.
Here kar seems to function as a 'linker' joining a cause proposition
with an effect proposition in the sense of Givo'n 1974. This is further
discussed in Kachru (forthcoming). In this type of 'causative' construc-
tion, the main verb (i.e., the finite verb) must be a causative
(morphological or lexical) and the object (i.e., the affected) of the
main verb must be identical and coreferential with the indirect object
(i.e., the recipient) of the participle. If these conditions are not
met, no 'causative' reading is possible. For instance, the following
do not yield a 'causative' reading.
45. ham usko pakaR kar ghar le aaye
we him catch CP home take came
a. We caught him and brought him home.
b. *We brought him home by catching him.
46. raaj ne kabuutar ko pakaR kar maar Daalaa
Raj ag. pigeon DO catch CP hit poured
a. Raj caught the pigeon and killed it.
b. *Raj killed the pigeon by catching it.
In 45 and 46, the affected object of the main verb has identical
grammatical function in the subordinate clause as well, hence, the
'causative' meaning is blocked. In 47, again, since the conditions
specified above are not met, no causative reading is possible.
47
47. vah usko paise de kar mar gaii
she him money give CP die went
a. She died after giving him money.
b. *She died by giving him money.
This discussion, though based on a limited range of data, makes
it clear that 'a mostly pragmatic analysis', as suggested in Davison
(1981), would ignore the syntactic and semantic generalizations presented
above. The following points emerge from the above discussion and illustra-
tions .
One, the surface CP shares a number of properties, both syntactic
and semantic, with a number of other constructions in the language, and
these could not be accounted for by pragmatic principles alone.
Pragmatic principles, however, are necessary to resolve the ambiguity
of, say, 33 and 43. Two, the formative kar has at least two meanings:
perfective and simultaneous. The simultaneous meaning is involved in
the manner adverbial construction, and the perfective in possibly all
the other constructions discussed above. Three, in view of the different
constraints involved, it is unlikely that a single syntactic source
would account for the entire range of the construction. The only
alternative is to further investigate the phenomenon and determine if
the multiple syntactic sources suggested above can be independently
motivated. The two areas that seem promising are the 'causative'
and the manner adverbial constructions. In case of the manner adverbial
construction, a large number of expressions are well on their way to
becoming 'fixed collocations' or 'idioms' in the sense that given a
main verb, the CP construction with which it can occur is almost
predictable. A number of such expressions is given below.
48. phuuT-phuuT kar ronaa 'to cry bitterly'
sisak-sisak kar ronaa 'to sob' (cry sobbingly)
ho-ho- karke hasnaa 'to laugh with a noise 'ho-ho''
khilkhilaa kar hasnaa 'to laugh'
bhakbhakaa kar jalnaa 'to burn furiously'
umaR kar chaanaa 'to swell and cover' (said of clouds)
garaj kar kahnaa 'to say roaringly'
daat piis kar bolnaa 'to speak, gnashing one's teeth'
caaT kar khaanaa 'to eat by licking'
gaTgaTaa kar pii jaanaa 'to drink by swallowing rapidly'
samhaal kar rakh lenaa 'to keep something carefully'
(i.e., to preserve, keep something
safe)
man lagaa kar paRhnaa/sunnaa 'to read/listen to by applying
one's mind to it' (attentively)
There is no such development in case of temporal or other uses of
the construction. This may be because there are limited choices with
regard to the manner in which specific tasks can be said to have been
performed, whereas the range of reasons, or causes, or time expressions
is much wider. One point, however, seems to be beyond controversy and
48
that is that the CP construction in Hindi-Urdu will contribute significantly
to our understanding of the role of pragmatics in linguistic descriptions.^
NOTES
1 am grateful to the Research Board of the Graduate College,
University of Illinois for supporting the research reported in this
paper, and to Tammie Valentine for assisting me with the collection
of data from various sources. An earlier version of this paper was
presented at the Third SALA Roundtable at SONY Stony Brook, on May 1,
1981, under the title 'Toward an Integrated Analysis of Conjunctive
Participle in Hindi-Urdu.'
2
Abbi (19803 suggests this in relation to CP's with causal and
manner adverbial meanings .
The CP in the gloss is an abbreviation for conjunctive participle
marker.
4
The reasons for the ungrammaticality of the (a) versions of
5-8 are as follows: in 5a, planting one's feet firmly and walking
are semantically contradictory; in 6a, ThaThaanaa is no longer an
independent verb in Hindi-Urdu; in 7a-8a, the verbs hasnaa and katraanaa
cannot take the complements, sabse and mere saamne se, respectively.
In 7b, word-order adjustments yield a grammatical but semantically
pointless sentence and in 8b, the first conjoined clause is ungrammatical
in that katraanaa is not a deleteable object verb. Even repairing
that does not lead to a paraphrase of 8: 8d, yah mujhse katraayaa aur
(*tab) mere saamne se nikal gayaa, 'he avoided me and passed in front
of me'. 8d is as anoraolous semantically as 7b.
This is noted in Porizka (1967-69), also. Davison (personal
communication) suggests that nevertheless, the 'perfective' meaning
even in the manner adverbial is justifiable if 'perfective' is interpreted
as 'prior with continued relevance'. This may work for sentence (5)
but would not work for sentences (6) and (7) and the expressions listed
in (48).
For some speakers, 15 and 16 are ambiguous. For example, 15
may mean either 'without eating' or 'instead of eating'. Even these
speakers, however, find the following unambiguously antithetical.
(i) vah duudh na piikar caay piine lagii
'She started drinking tea instead of drinking milk.'
(ii) vah ghar kaa kaam na karke saare din ghuumtii rahtii hai
'She wanders around the whole day instead of doing the
housework. '
^The -ke is a variant of CP, used obligatorily with the verb
kar 'do' and optionally with other verbs.
49
o
Note that with hii also, 33 is ambiguous, but that has no bearing
on this discussion.
9
One major work consulted as a source for data on CP is the following:
Yaadav, Rajendra. 1951. Saaraa aakaash. New Delhi, India: Aksar
Prakashan.
REFERENCES
ABbI, Anvita. 1980. Semantic grammar of Hindi. New Delhi: Bahri
Publications .
BAIiL, Kali C. 1974. A study in the transformational analysis of the
Hindi verb. Chicago: South Asian Language and Area Center,
University of Chicago.
BHATIA, Tej K. 1977. A syntactic and semantic description of negation
in South Asian languages. Ph.D. dissertation: University of
Illionois.
DAVISON, Alice. 1981. Syntactic and semantic indeterminancy resolved:
a mostly pragmatic analysis for the Hindi conjunctive participle.
In Peter Cole, ed.: Radical pragmatics, pp. 101-128. New York:
Academic Press.
DWARIKESH, D. P. S. 1971. The historical syntax of the conjunctive
participial phrase in the new Indo-Aryan dialects of the MadhyadeSa
("Midland") of Northern India. Ph.D. dissertation. University of
Chicago.
GIVON, Talmy. 1974. Cause and control: on the semantics of interpersonal
manipulation. Syntax and semantics IV, pp. 59-89. New York:
Academic Press.
GREEN, Georgia. 1974. Semantics and syntactic regularity. Bloomington,
IN: Indiana University Press.
KACHRU, Y. 1965. A transformational treatment of Hindi verbal syntax.
Ph.D. dissertation: University of London.
. 1966. An introduction to Hindi syntax. Urbana, IL: Department
of Linguistics, University of Illinois.
1978. On relative clause formation in Hindi-Urdu. Linguistics
207.5-26.
1980. Aspects of Hindi grammar. New Delhi: Manohar
Publications .
. Forthcoming. Conjunctive participle in Hindi-Urdu: syntax.
semantics and pragmatics.
POI^tZKA, V. 1967-69. On the perfective verbal aspect in Hindi.
Archiv Orientalni 35.64-88, 208-231, and 37.19-47, 345-364.
Studies in the Linguistic Sciences
Volume 11, Number 2, Fall 1981
DISCOURSE STRUCTURE IN A HINDI SHORT STORY
Suresh Kumar
Central Institute of Hindi, Agra
Structural analysis of a Hindi short story is presented
in the methodological framework of discourse analysis and the
theoretical framework of semiotics. Macro-structural categories
of the narrative are described in correlation with the relevant
micro-structural categories of the linguistic text of the story
as an exercise in total analysis of the narrative art symbol.
I. Introduction
The study presents a discourse analysis of a Hindi short story Liiciijaa
in terms of structural narratology. A short story is understood as a
semiotic literary symbol--a unified object of narrative and linguistic
structures on one hand, and artistic and socio-cultural wholes on the
other. The theoretical assumption of semiotics provides for a total
and comprehensive model of analysis which may be represented as follows:
Total structure NARRATIVE TEXT Total meaning
Components
. f-J ,
Central Peripheral
4^
syntagmatic 4. paradigmatic focus and modes
axis axis of narration
^ i/ / ^
functions attributes / 1
— -, __ __.
stylistic devices
LINGUISTIC TEXT
A narrative text is, from the structuralist point of view, a unity of
total structure and total meaning, related respectively to the constitutive
and communicative aspects of the text object. The text may be broken down
into components falling into two categories--the central and the peripheral.
Participants or characters and elements or constituents of plot are the
central components which can be seen as arranged at two axes- -the syntagmatic
and the paradigmatic. The interaction of components at syntagmatic axis is
functions. The concept of 'function' is in confirmity with the reductionist
technique of deriving abstractions which could be related to the specific
phenomena occurring in a given text. Modes of narration are the various
conventions of presenting the narrative and are related to discourse types
in general. These modes feed the central cmmponents. The whole complex is
anchored in the linguistic text which is related to the narrative text--
the central and the peripheral components --through stylistic devices, which
are also responsible for giving a touch of individuality to the components
and, as such, the narrative text under study.
52
A few remarks on methodology are necessary.
Tlie study follows structuralist technique of IC analysis which
correlates with temporality- -the distinctive feature of the narrative
structure. The related methodological requirement of the analysis is
the derivation of the core structure of the narrative text. The core
structure of the narrative text is arrived at in a manner which appears
consonant with the simple and organic structural composition of the
story as well as the aim of the study. The method employed here assumes
that it is neither absolutely necessary nor fully possible to insits on
strictly deriving the kernel sentence structure of the entire text by
using the methodology rigorously. As our study is applicational, the
reduction of the linguistic text of the short story into even 'near-kernel'
sentences along with the kernel ones will be sufficient to serve our
purpose. Accordingly, the optional elements of the sentence structure have
been distinguished from obligatory ones by being enclosed in parentheses
in the re-written text (see appendix) . This simple technique helps in
setting off the stylistically and compositional ly relevant linguistic
elements of the discourse structure of the short story as distinct from
the rest.
As our study aims at presenting discourse analysis of a narrative
text, the same is seen here as a many- layered contextual ised structure
with an artistic intent and a cultural dimension. Accordingly, it is
analysed in a hierarehical order of levels of analysis which is consis-
tent with the theoretical assumption.
Metal ingual design of the study tries to strike a balance between
formulaic and tabular representation of facts of structural organisation
of the narrative symbol as usually practised in studies in structural
folkloristics, and verbal representation of the aspects of the symbol
which merit interpretative comment as is the normal practice in literary
studies.
2. Analysis
2.1. Context: Liici-yaa is a story of two married women who are
placed in an identical stiuation at the physical level but show contrast
at the psychological level. In terms of theme dynamics, the story begins
at the point of identicalness and closes at the point of contrast, running
through a course of thematic development which is structurally and func-
tionally responsible for imparting a touch of individuality to the story.
The distribution of identicalness and contrast in situations is governed
by their correlates--the physical and the psychological respectively.
2.2. Narrative macro-structure: The narrative macro-structure of
the story is marked by two inter-related principal super-structure
features: (a) it is participant-dominated, (b) it is single-event. It
is chiefly a story of network of relations that obtain between Sudhaa
and her husband on one side, and Saeedaa and her husband on the other,
with their respective servants playing intermediary roles. The ailing
women asking their husbands to send lichees is the kernel event which
releases a chain of reactions, realising the pattern of structural
relationship between Sudhaa and her husband, and Saeedaa and her husband,
showing the feature of complementary distribution.
53
2.2.1. Total structure of the story is realised by the following
formula:
LIICIYAA --) + INTRODUCTION + INCITATION + BODY -->
+Inter-action 1 + Intra-action + Inter-action 2
+ CLOSURE
INTRODUCTION --> appearance of participants + identicalness (1)
INCITATION --^ appearance of stimulus + identicalness (2)
BODY --> Inter-action 1 -•> interdiction/no-directive (3)
reinforced through repetition (3.1) +
harmony 1
-* prescription/yes-directive (4)
reinforced through repetition (4.1) +
disharmony 1
Intra-action --> reflection + self-deprecation (5)
Inter-action 2 --) remorseful regard (6) reinforced
through repetition (6.1) + harmony 2
-) remorseless disregard (7) +
disharmony 2
CLOSURE --) content vs. discontent = contrast (8-9)
Structural details of BODY are considered in the latter section
while the rest are taken up first. (Note: numerals and sub-numerals in
parentheses in the text of the study correspond to the numbers of the
sections and sentences in the text of the story as given in the appendix.)
2.2.1.1. Component: INTRODUCTION (1)
Mode of narration: summary
Function: appearance of participants
Attribute: identicalness
Style markers, thematic reference, and effect:
(i) aggregative, donO 'both the', appearing in relevant
constructions: Sudhaa aur Saeedaa dono 'Sudhaa and
Saeedaa both' (1:1), dono ke pati 'husbands of both'
(1 : 2) --reference is obvious- emphasis.
(ii) temporal adv. phrases, kaii dino se 'for many days'
(1:1), shaam ko nishcit samay par 'at a fixed hour
in the evening' (1 : 2) --reference to the ailing
women and their husbands respectively--emotive.
54
The physically identical situation of Sudhaa and Saeedaa has been
stressed by the aggregatives-- 'both' were lying ill, 'both' were visited by
their husbands. The pitiability of their condition--they had been lying
ill for many days and were visited by their husbands regularly at a
fixed hour is highlighted by the temporal adverbial phrases.
2.2.1.2. Component: INCITATION (2)
Mode of narration: summary
Function: appearance of stimulus
Attribute: identicalness
Style markers, thematic reference, and effect:
(i) aggregative, dono 'both the', appearing in
relevant constructions: dono ne 'both' (2:1),
dono puruSo ne 'both men' (2 :2) -rreference is
obvious--emphasis .
(ii) modifier, ek hii 'the same,' appearing in
relevant construction, ek hii dukaan se' from
the same shop' (2 : 2) --reference for the husbands --
emphasis .
(iii) temporal adv. phrase, ghar pahucte hii 'immediately
on reaching home' (2 :2a) --reference as above- -
emotive.
The identicalness of situation of the women is again emphasized
through aggregatives. The way their husbands are disposed towards their
wives is underlined by the modifier and the temporal adverbial phrase,
which bring out markedly the identicalness in their situation.
The aforesaid functional attribute, the identicalness, is further
evidenced by the graphonomic fact of singleness of paragraph. Though the
participants are two, the description of their condition is handled in a
single paragraph.
2.2.1.3. Component: CLOSURE (8-9)
Mode of narration: summary
Function: content vs. discontent
Attribute: contrast
Style markers, thematic reference, and effect:
(i) idiomatic expression, bhaavnaao me agar battiyaa mahak
'incense sticks perfume in feelings--feel cheerful and
content' (8 :1a) --reference to Sudhaa- -emotive.
(ii) idiomatic expression, bhaavnaao me makkhiyaa bhinbhinaa
'flies buzzing in feelings--feel desolate and
discontent' (9 :1a) --reference to Saeedaa- -emotive.
55
The contrast in the psychological condition of the women is portrayed
against the backdrop of identicalness in their physical condition, expressed
through near-identical references in the micro-context of the text--Sudhaa
akelii thii 'Sudhaa was alone' (8:1a), and, Saeedaa bhii akelii thii
'Saeedaa too was alone' (9:1a). But their psychological conditions differ,
and the fact is brought out effectively by the choice of idiomatic expres-
sions distributed appropriately between Sudhaa and Saeedaa. These expres-
sions realise functions of content vs. discontent in relationship of
binary opposition. Sudhaa is content with her married life and Saeedaa
discontent.
The graphonomic evidence of contrast in psychological conditions of
Sudhaa and Saeedaa is provided by the fact that they have been allotted
different paragraphs for description of their condition. Paragraph
S goes to Sudhaa and 9 to Saeedaa. The graphonomic split at the level
of this component correlates with the phenomenon of change of functional
attributes- -from identicalness in INTRODUCTION and INCITATION to contrast
in CLOSURE.
2.2.2. The component of BODY is the central component and represents
thematic development of the short story. It is identified in terms of
three sub-components:
Inter-action 1, Intra-action, Inter-action 2
2.2.2.1. Sub-component: Inter-action 1, part I
Mode of narration: semi -dialogue (Sudhaa to servant)
Function: interdiction/no-directive (3), with emphasis (3.1)
Attribute: harmony 1
Style markers, thematic reference, and effect:
(i) adversative, the inter-clause connector, par 'but'
(3: 1) --reference to lichees--extension of incitation.
(ii) negative imperative, mat kahnaa 'don't tell' (3:3a)--
reference to Sudhaa- -emotive.
(iii) adv. phrase of manner, daravaaze ke paas se use bulaa
kar 'calling him back from the door' (3.1:1a)--
reference to Sudhaa- -emphasis.
(iv) negative imperative, mat bhuulnaa 'don't forget'
(3: 1 :2) --reference to Sudhaa--emotive.
(v) adjectives used predicatively, dukhii 'sad' (3:4c),
pareshaan 'worried' (3. 1 :3a) --reference to Sudhaa 's
husband- -emotive.
The reference that the lichees looked puffed up and fresh from outside
but there were worms inside them is the extension of incitation which triggers
off the subsequent components of the text. The function of interdiction is
realised by negative imperative 'mat kahnaa,' and the emphasis therein by
the adverb of manner and the repetition of the negative 'mat'. This together
56
with emotive vocabulary items, 'dukhii' and 'pareshaan', appropriately
realise the attribute- -harmony. Sudhaa ind her husband are depicted as
showing the relationship of mutual care and trust.
2.2.2.1.1. Sub -component: Inter-action 1, part II
Mode of narration: semi -dialogue (Saeedaa to servant)
Function: presciption/yes-directive (4), with emphasis (4.1)
Attribute : disharmony 1
Style markers, thematic reference, and effect:
(1) adv. phrases of manner, chilke samet 'alongwith
the peel,' rumaal me baadh kar 'packed in handker-
chief (4:1) — reference to Saeedaa--repulsion.
(ii) nominal item, imrat phal 'nectar fruit' (4:3)--
reference to lichees--sarcasm.
(iii) verbal itmm, nosh pharmaae '(you) taste'
(4:3b) --reference to Saeedaa 's husband--sarcasm.
(iv) adv. phrase of manner, darvaaze ke paas se use bulaa
kar 'calling him back from the door' (4. 1 : 2) --reference
to Saeedaa — emphasis.
(v) nominal phrase, vilaayatii pleT 'foreign made dish'
(4.1 : 3) --reference to lichees--sarcasm.
(vi) nominal phrase, bahut zyaadaa viTaamin 'a lot of
vitamins' (4. 1 :3b) --reference to lichees--sarcasm.
The function of prescription is realised through Saeedaa commanding her
servant to serve her husband with rotten lichees (4:3,3b), and is stylized
by the choice of relevant lexical items in the adverbial phrase of manner
'chilke samet,' etc. The same is reinforced by repetition of instruction
realised by repetition of imperatives, denaa 'give' and kahnaa 'say' (4:3, 3a).
The style markers occurring in Saeedaa 's statements are overwhelmingly of
sarcastic flavour, and realise the functional attribute, disharmony--
disharmony in relationship between Saeedaa and her husband.
2.2.2.2. Sub -component: Intra-action (5)
Mode of narration: monologue
Function: reflection
Attribute: self -deprecation
Style markers, thematic reference, and effect:
(i) verbal item baRbaRaa 'grumble' (5:1) --reference to
Saeedaa- -disaffection,
(ii) pronominals, kisii kaa 'of one' (5: 2) --reference for
Saeedaa- -kiske 'whose' (5: 3) --reference to third
person- -self- pity.
Ciii) the question word, kyaa 'what' (5 :4) --reference to
Saeedaa- -self- pity.
57
(iv) the idiomatic expression, jaan ghol kar pilaa 'be ready
to forsake everything in life' (5:6b); apnii apnii kismat
'it is a matter of luck' (5:7) --reference for Saeedaa-
self-pity.
The component of Intra-action provides a glimpse of working of Saeedaa's
mind; her pitiable condition, her loneliness which evokes pathetic feelings
in the reader's mind, and consequently realises the functional attribute
of self-deprecation. The emotive expressions of sentence size-level, listed
in the column of style markers, speak of expressiveness of the colloquial
variety, and show a kind of convergence on a single stylistic effect, to
which contributes the graphonomic feature of singleness of paragraph.
2.2.2.3.1. Sub-component: Inter-action 2, part I
Mode of narration: semi-dialogue (husband to Sudhaa)
Function: remorseful regard (6), with emphasis (6.1)
Attribute : harmony 2
Style markers, thematic reference and effect:
(i) temporal adv. phrase, samay se kuch pahle 'a little
before the fixed hour' (6: 1) --reference to Sudhaa 's
husband- -regard .
(ii) antithesis through personal pronouns, tum: mai
'you: I' (6:3,4a) --reference to Sudhaa's husband--
remorse.
(iii) adv. phrase of manner, khud chiil kar 'peeling
himself,' apne hii haath se 'with his own hands'
(6: 5) --reference to Sudhaa's husband- -regard.
(iv) nominal phrase, baRii bhuul 'a blunder' (6:6 —
reference to Sudhaa's husband- -remorse.
(v) verbal items, koshish karuugaa 'I will try' (6.1:2a);
uTh aauu 'I leave' (6.1:2b); yakiin hai 'I am sure'
(6.1:3); aa jaauugaa 'I will return' (6.1:3a)--
reference as above--regard.
(vi) adv. of manner repeated, jaldii 'soon' (6.1 :2b, 3a)
reference as above--regard.
The segment provides an interesting dimension of relationship
between Sudhaa and her husband. The husband speaks and wife listens.
He is remorseful (6) and full of regard (6.1). His regard towards his wife
is pointedly expressed by a chain of adverbs of manner relating to hira--
'samay se kuch phale hii (aa gaye) ' ; 'khud chiil kar'; 'apane hii haath
se (khilayee).' His feelings for his wife are further expressed through a
repetition of the lexical item jaldii 'soon' in his statements (6. 1:2b, 3a).
The component, by virtue of its recurrence, underlines again the functional
attribute of harmony between Sudhaa and her husband.
2.2.2.3.2. Sub-component: Inter-action 2, part II (7)
Mode of narration: semi -dialogue (servant to Saeeda)
Function; remorseless disregard
Attribute: disharmony 2
Style markers, thematic reference, and effect:
58
(i) temporal adv. phrase, un ke jaane ke kaafii der
baad 'long after he (Sudhaa's husband) had left'
(7: 1) --reference to Saeedaa's servant — disregard.
(ii) nominal item, saahab 'sir' (7:2) and its corefer-
ential pronoun unho ne 'he' (7 : 3) --reference to
Saeedaa's husband--disregard.
(iii) negative adverb, (viTaamin) nahii 'no (vitamins)'
(7:3b) --reference to lichees--sarcasm.
(iv) nominal phrase, begam saahbaa 'the lady' (7:3c)--
reference to Saeedaa — sarcasm.
(v) adv. phrase of manner, bekhaTke 'without any hitch'
(7 :3c) --reference as above--sarcasm.
This segment gives a glimpse of the relationship between Saeedaa and
her husband. The servant conveys the message and Saeedaa listens.
Saeedaa's husband is far from remorseful and shows scant regard for his
wife as is evident from sarcastic references in his message to her. The
sub-component, by virtue of its recurrence, underlines again the functional
attribute of disharmony between Saeedaa and her husband.
2.2.3. The component of BODY as treated above shows interesting
features of thematic structural relationship between its sub-components.
The change of functional attribute from identicalness in INTRODUCTION
and INCITATION to contrast as culminating in CLOSURE is appropriately
processed in BODY through its sub-components. In other words, the
attribute of identicalness originates in the first two segments, develops
into contrast in the BODY, and comes to a head in CLOSURE. Following
observations may be made about the structural relationship of the sub-
components with regard to the development of the relevant thematic function.
(i) Inter-actions 1 and 2 have two parts each, and both
of them show the functional attribute of contrast.
Sudhaa's conjugal life is marked by harmony and
Saeedaa's by disharmony.
(ii) Parts I and II of Inter-action 1 show similarity
in repetition of style markers for emphasis (3.1 and
4.1). However, this similarity is absent in the
parts of Interaction 2 (6.1, but there is nothing like
7.1 as its counterpart). This again underlines the
attribute of harmony in Sudhaa's conjugal life and
disharmony in Saeedaa's.
(iii) The sub-component of Intra-action is restricted to
Saeedaa. Besides being instrumental in transtition
from Inter-action 1 to Inter-action 2 and thus accounting
for cohesiveness of the story in general and BODY in
particular, it highlights the state of Saeedaa as a
lonely and dejected woman.
59
(iv) Part I of Inter-action 2 shows the feature/+husband/,
i.e. the husband comes to console Sudhaa; while part II
shows the feature/ -husband/, i.e. the husband does not
come to console Saeedaa. This situation justifies the
respective presence and absence of repetitive devices
in parts I and II in Inter-action 2, which accounts for
psychological proximity and remoteness respectively in
the relationship of the couples. The above contrast in
the condition of Sudhaa and Saeedaa and the related
narrative functions of remorseful regard (6) and
remorseless disregard (7) are sequel to those of inter-
diction or no-directive (3) and prescription or yes-
directive (4), which portray the women's attitudes
towards their husbands .
(v) The contrast in their conditions correlates with
graphonomic features in an appropriate manner in
Inter-action 2. Two paragraphs with larger language
content are devoted to part I, while part II is
restricted to one paragraph with relatively small
language content.
2.3. Modes of narration: The modes of narration have been assigned
the status of peripheral component of the narrative text in our scheme.
They are related to the central components of the narrative text on one
hand and to the stylistic devices in the linguistic text on the other.
While they are instrumental in realising the function and the attributes
in the narrative segment in consonance with narrative requirements, they
are themselves realised through appropriate stylistic devices from linguistic
micro- structure of the semiotic symbol, i.e. the short story.
The story has been told in third person mode of narration. The focus
is on the author, who is the omniscient, though rather impersonal, narrator.
It is neatly distributed between two segments of narration--the scence and
the summary. BODY is composed in scene mode and the rest in summary.
Scence mode is conversational and is realised through semi-dialogue and
monologue. In addition, scene shows expositive and reflective features of
discourse, while the summary mode shows the narrative and descriptive
features. The distribution correlates appropriately with the functions
and their attributes as realised in the story.
2.4 Linguistic micro-structure: Tlie treatment of linguistic micro-
structure of the story is the analysis of the cotextual and contextual
elements of the semiotic symbol that the story is. This treatment has
instrumental, rather than terminal, value, which means that the treatment
owes its value in its relevance to narrative macro-structural analysis,
and is therefore subservient to it.
The treatment falls into two parts: (i) relating to cotextual elements,
which account for cohesiveness of the story as a whole; (ii) relating to
the contextual elements, which represent selection and innovation of items
and patterns of language-use in relation to relevant elements in the macro-
structure of the narrative text.
2.4.1 The relevant cotextual elements may be treated under three heads,
in accordance with the laws of structural hierarchical order of the levels of
60
the text, starting from below: (A) Inter-clause connectors (enclosed in
parentheses at the end of clauses in the analysed text in the appendix),
which are restricted to additive aur 'and' and the adversative par 'but',
while the rest are used for reasons structural rather than semantic. (B)
Inter-sentence connectors, falling into the following major categories:
(a) anaphoric reference, employed overtly, e.g. dono ke pati 'husbands of
both' (1:2), dono puruSo ne ' both the men' (2:2), unhe 'them' (3:2),
unho ne 'he' (3:4), etc.; (b) anaphoric reference used covertly, e.g.
imperative sentences, dekh 'look' (3:3), le 'take it' (4:2), dekho 'look'
(6.1:6); (c) adverbials of various categories, e.g. phir 'then' (5:4),
jaane "God knows' (5:3), kal 'yesterday' (6:6), aaj 'today' (6:7), calte
calte 'while leaving (6.1:5). (C) Inter-pragraph connectors, which relate
directly to the contextual consistency of the text as a whole and fall
into two major categories: (a) nominals including proper names, e.g.
Sudhaa, Saeedaa, naukar 'servant', liiciyaa 'lichees'; (b) adverbial
phrases, e.g. ek din 'once', duusre din shaam ko 'next day in the evening,'
naukar ke jaane ke baad 'after the servant had left,' etc. All these
elements occur sentence-initially at the beginning of the paragraph.
2.4.2. The contextual elements of linguistic micro-structure have
stylistic significance. They are identified intuitively as standing out
in the micro-context of the text and are relatable to the macro-context
of the text in the structural framework. It is interesting to note that
the identifications correlate in a fairly systematic manner with optional
elements of the sentences isolated for determining the basic structure
of the sentences of the text (see Introduction) .
Another relevant observation is that the overwhelming majority of
contextual linguistic elements are modifiers--the modifying structures--
either of substantives or verbs, carrying relevant lexical items of
stylistic significance. For example, the temporal modifier samay se
kuch pahle 'a little before the fixed hour' (6:1) is stylistically
significant inasmuch as it underscores the psychological closeness
between Sudhaa and her husband and thereby contributes to the total effect
of the story. Similarly, repetition of jaldii 'soon' (6.1:2c, 3a),
occurring in appropriate grammatical constructions in the statements of
Sudhaa' s husband, is stylistically significant and can be assigned the
same relevance as mentioned above. The column of style markers in the
section of analysis of narrative macro-structure lists the contextual
linguistic elements.
3. The Cultural Dimension
A reference to the cultural dimension of the art object is an integral
part of the total analysis in narrative semiotics as presented above.
Culturally, the message--the toal meaning of the story--is a well known
fact of conjugal relationship in Indian society which has been dealt
with artistically in the story. The moral is obvious: love begets love;
hate begets hate; mutual love and trust make a couple happy while absence
of these cause misery and depression.
4, Conclusion
The study is presented as an attempt at a discourse structure analysis
of an artistic narrative, and is a modest contribution to studies in Hindi
semiotics on one hand and the structuralists' efforts of "spanning the gap"
61
between the narrative macro -structure and the linguistic micro -structure on
the other. It lays more emphasis on methodology in comparison to finding
something new, if any, in the communicative dimension of the thematic aspec-
of the story. Consequently, I would like to say that Liiaiyda is a homo-
geneous logical, rhythmical discourse culminating in a message of moral
which is of social significance. The above characterization represents
the semiotic nature of the study inasmuch as it is couched in terms of the
relationship between an artistic whole and a cultural whole.
NOTES
*I am thankful to Professor Bal G. Misra who read the draft of the
paper closely and gave suggestions.
REFERENCES
DOLEZEL, L. 1971. Toward a structural theory of content in prose fiction.
In S. Chatman, ed.: Literary style: a symposium. OUP.
HENDRICKS, W.O. 1972. The structural study of narration. Poetics 3.100-123.
LIPSKI, John M. 1976. From text to narrative: spanning the gap. Poetics
19.191-205.
PROPP, V. 1958. Morphology of the Folktale (1928). The Hague: Mouton §
Co. (English translation by L. Scott; 2nd rev. ed. by Louis A.
Wagner, 1968, Austin, University of Texas.)
SCHOTTELNDREYER, A. Mareike and Stephen H. Levinsohn. 1976. The catio
folktale as a play in acts and scenes. Poetics 19.247-279.
APPENDIX
(1)
1. Sudhaa aur Saeedaa (dono) (kaii dino se) aspataal me daakhil tfcii
Sudhaa and Saeedaa both had been in the hospital for many days.
2. dono ke pati (shaam ko nishcit samay par) aate the (aur)
The husbands of both used to visit at a fixed hour in the evening and
2a. Khoj-khabar le jaate the
Inquired of their well-being.
(2)
1. (ek din) dono ne apne-apne pati se liiciyaa bhejne ko kahaa
Once both asked their husbands to send some lichees.
2. dono puruSo ne (aspataal se lauTte samay) (ek hii dukaan se)
liiciyaa khariidii (aur)
Both of them bought, while coming back from the hospital, some
lichees from the same shop, and
3a. (ghar pahucte hii) (apne apne naukar ke haath) (aspataal) bhej dii
Had them sent to the hospital through their servants immediately
on reaching home.
62
(3)
1. liiciyaa uupar se phuulii-taajii thii (par)
The lichees looked puffed up and fresh from outside (but)
la. bhiitar un me kiiRe the
There were worms inside them.
2. Sudhaaneunhe kuuRe kii Tokarii me phikvaa diyaa (aur)
Sudhaa had them thrown in the dustbin and
2a. naukar se kahaa
Warned the servant
3. dekh
look
3a. apne baabuuji se mat kahanaa (ki)
Don't tell your master that
3b. liiciyo me kiiRe the
There were worms in the lichees
4. unhone kitnii muhabbat se liiciyaa bhejiT (par)
How affectionately he sent the lichees but
4a. sunege (ki)
If he learns that
4b. maine liiciyaa nahil khayii
1 did not eat the lichees.
4c. ve dukhii hoge
He will feel sorry.
(3.1)
1. naukar cal paRaa (to)
When the servant started
la. (daravaaze ke paas se) (use bulaa kar) Sudhaa ne kahaa
Sudhaa called him back from the door and instructed him.
2. merii baat mat bhuulnaa
Don't forget what I have said to you.
3. (kahil) (laaR yaa laaparvaahii me) agar kuch kah baiThe (aur)
Lest you should say something in indulgence or carelessness, and
3a. ve bekaar pareshaan ho
He may become worried.
(4)
1. Saeedaa ne liiciyo ko (chilke samet) (rumaal me baadh kar)
naukar ko dete hue kahaa,
Saeedaa gave to the servant the lichees, packed in handkerchief
alongwith the peel, and said.
63
2. le
Take it.
3. ye(imrat phal) unhe (hii) de denaa (aur)
Give these nectar fruits to him and
3a. merii taraf se kahnaa (ki)
Tell him on my behalf that
3b. inhe aap hii nosh faramaae
You taste these for yourself.
(4.1)
1. maukar cal paRaa (to)
When the servant started
2. (daravaaze ke paas se) (use bulaakar)
Saeedaa ne hidaayat kii
Saeedaa called him back from the door and instructed him.
3. (us) (vilaayatii) pleT me (rakh kar) denaa (inhe) (aur)
Serve them on the foreign made dish and
3a. kahnaa (ki)
Tell him that
3b. in me (bahut zyaada) viTaamin hai
There are a lot of vitamins in them.
(5)
1. (naukar ke jaane ke baad bhii) Saeedaa baR baRaatii rahii
Even after the servant had left, Saeedaa kept grumbling.
2. dil me kisii kaa khayaal ho (to)
If he had some feelings for me
2a. ciiz dekh kar khariide
He would buy things with care.
3. (jaane) kis ke sapno me Duube Hoge
God knows whom he has in his heart.
4. (phir) dekhne kii zaruurat bhii kyaa hai
And then what is the need of caring for somebody.
5. mai mar jaauu (to)
Even if I die
5a. (das dino baad) nayii-navelii le aayege
He will bring a new bride just after, say, ten days
64
6. Yah to mai hii huu (ki)
It is only me that
6a. zaruurat ho (to)
If the need be
6b. apnii jaan bhii ghol kar pilaane ko taiyaar rahtii huu
I am always ready to forsake everything in life.
7. sab apnii-apnii kismat hai
It is a matter of luck
7a. aur kyaa
What else
(6)
1. (duusare din shaam ko) Sudhaa ke pati (samay se kuch pahle hii)
aa gaye
Next day evening, Sudhaa 's husband came a little before the fixed
hour.
2. bole
He said
3. kal turn liiciyaa na khaa sakii
You could not eat lichees yesterday
4. mujhe raat bhar aisaa lagaa (ki)
I felt the whole night as if
4a. C^aj) bhuukhaa (hii) so rahaa huu
I am sleeping hungry tonight
5. (aur) unhone ek liicii (khud chiilkar) Sudhaa ko (apne hii haath
se) khilaayii
Then he himself peeled a lichee and offered it to Sudhaa with
his own hands.
6. kal baRii bhuul huii (ki)
I made a blunder yesterday that
6a. malne liiciyaa khud (dekh kar) na lil
I did not myself see lichess while buying them.
7. aaj phawaare jaakar laayaa
I myself brought them today from fountain area
8. (khaas) raambaag kii hal
These are right from Ram Bagh.
(6.1)
1. mujhe (ek zaruurii) miiTing me jaanaa hai
I have to attend on urgent meeting
65
2. mai jaa rahaa huu (par)
I am going but
2a. (puurii) koshish karuugaa (ki)
I will try my best that
2b. (jaldii) uTh aauQ (aur)
I leave it early and
2c. (kuch der) tumhaare paas baiTh sakuu--
Be able to sit with you for some time--
2d. (kah kar) Sudhaa ke pati cal paRe (par)
Said Sudhaa 's husband and started but
2e. (daravaaze se) (phir) lauT aae (aur)
Came back from the door and
2f. bole
Said
3. mujhe (pakkaa) yakiin hai (ki)
I am very sure that
3a. mai (jaldii hii) aa jaauugaa
1 will come back rather early
4. (phir bhii) mai Ram Kumaar ko kahtaa jaauugaa (ki)
Even then I will ask Ram Kumar while leaving that
4a. vah (turant) tumhaare paas aa jaaye
He should come to you immediately
5. (calte calte) unho ne kahaa
He said while leaving
6. dekho
Look
7. dukhii na honaa
Don't be depressed
(7)
1. (un ke jaane ke kaafii der baad) Saeedaa kaa naukar (khaana
lekar) aayaa
Long after he had left Saeeda's servant came with the meals
2. Saahab ek zaruurii milling me gaye hai
Sahab has left for an urgent meeting
3. unhS ne liiciyal bhejii hai (aur)
He has sent some lichees and
3a. faramaayaa hai (ki)
Has conveyed that
66
3b. in me viTaamin nahii hai (aur)
There are no vitamins in them and
3c. begam (saahbaa) inhe (bekhaTke) khaa saktii hai
The lady can eat them without any hitch
(8)
Sudhaa akelii thii (par)
Sudhaa was alone but
la. uskii bhaavnaao me (jaane) (kitnii) agarbattiyaa mahak rahii thii
Hundreds of incense sticks were perfuming in her feelings--
she was cheerful and content.
(9)
1. Saeedaa bhii akelii thii (par)
Saeedaa too was alone but
la. uskii bhaavnaao me (jaane) (kitnii) makkhiyaa bhinbhinaa rahii thii
Hundreds of flies were buzzing in her feelings--she was desolate
and discontent.
Studies in the Linguistic Sciences
Volume 11, Number 2, Spring 1982
PRONOUNS IN MALAYALAM
K. P. Mohanan
M.I.T., Cambridge
The purpose of this paper is to sketch a theory that will account
for the relation between pronouns and their antecedents in Malayalam.
In section 1, I shall propose a precedence principle in Malayalam,
corresponding to what has been called the principle of noncoreference
in English. In section 2, I shall examine the facts of 'strong cross-
over', and, in section 3, demonstrate how the approach proposed in
sections 1 and 2 is superior to the treatment of noncoreference and
strong crossover in the Government Binding theory, as developed in
Chomsky(1981).
1. The Noncoreference Principle
1.1. The noncoreference principle in English (Lasnik (1976), Chomsky (1980))
is demonstrated by contrasts such as the following:
(l)a. Oscar thinks he_ is brilliant.
b. *He thinks Oscar is brilliant.
c. Though h£ is a fool, Oscar adores Mary.
(2)a. Oscar loves his mother.
b. *He loves Oscar's mother.
c. His mother loves Oscar.
(3) a. Near him, Oscar saw a snake,
b. *Near Oscar, he saw a snake.
The underlined NP's are intended to refer to the same individual. Thus,
(lb), (2b), and (3b) are ungrammatical only if Oscar and h£ are interpreted
to refer to the same person. What prevents Oscar from being the antecedent
of the pronoun in these examples is the principle of noncoreference. I shall
restate the essential insight behind the various formulations of the nonco-
reference principle in English as (4). (See Lasnik (1076), Reinhart (1976)
and Chomsky (1980).)
(4) The Noncoreference Principle
(English type languages)
Pronouns cannot c-command their antecedents.
The definition of 'c-command', taken from Reinhart (1976), is as fol-
lows :
(5) «v c-commands ^4 iff the first branching node that dominates «< also
dominates ^ , and •< does not dominate fi .
68
(lb), (2b), and (3b) are ungrammatical because the pronoun c-commands
its antecedent in these sentences, violating (4).
1.2. Before we go on to the discussion of the noncoreference principle in
Malayalam, it is necessary to outline some of the essential facts of clause
structure in the language. Malayalam, like most other languages in India,
is a 'nonconfigurational ' or 'free word order' language, in which the order
of the subject, direct object, indirect object , various adjuncts and the verb
is not fixed. What this means is that in Malayalam, grammatical relations
such as 'the subject of and 'the object of are encoded, not in terms of
syntactic configurations, but in terms of morphological case features. I
have provided extensive arguments to show that Malayalam doos not have a VP
node, but has a flat clause structure as illustrated in (6) . (Mohanan(in
press)) 1.
(6)
kut'ti innale aanaye E^Hi
child-nom yesterday elephant-acc pinched
(The child pinched the elephant yesterday)
1.3. The most striking fact about pronouns in Malayalam is that they can
c-command their antecedents, but not precede them. I formulate the noncore-
ference principle for Malayalam as (7) :
(7) The Noncoreference Principle
(Malayalam type languages)
Pronouns cannot precede their antecedents.
I shall use the property of free word order in the language to demons-
trate the truth of (7). The following sentences show that all and only those
word orders in which the pronoun follows its antecedent are grammatical,
whatever be the c-command relation involved:
2
(8)a. moohan [awante bhaafyaye] nulli
Mohan his wife-a pinched
(Mohan pinched his wife.)
b.*awan [moohante bha^yaye] nulli
he Mohan-g wife-a pinched
(9) a. [moohante bhaaTyaye] awan nulli
Mohan-g wife-a he pinched (=8. a)
b. * [awante bhaaFyaye] moohan nujli
his wife-a Mohan-n pinched
69
(10)a, [moohante bhaarya] awane nulli
Mohan-g wife-n him pinched
(Mohan's wife pinched him)
b. * [awante bhaarya] moohane nulli
his wife-n Mohan-a pinched
(11) a. moohane [awante bhaarya ] aulli
Mohan-a his wife-n pinched
b. * awane [moohante bhaarya] nulli
him Mohan's wife-n pinched
(=10. a)
In all these sentences, the version in which the pronoun precedes its
antecedent is bad, which follows from (7). In (9. a) and (10. a), the pronoun
c-commands its antecedent, which demonstrates that (4) is not applicable in
Malayalam:
(12)a. (=9. a)
moohante bhaaryaye aw an nulli
b. (=10. a)
moohante bhaarya awane
nulli
The examples given above illustrate the irrelevance of the notion
c-command for pronominal noncoreference in Malayalam. Given below are exam-
ples that show that the notion of command is equally irrelevant: even when
the pronoun commands its antecedent, the sentence is acceptable, as shown
by (13. a):
(13) a. [kut^i aanaye aulli enn^ ] awan parahnu
child elephant-a pinched that he said
(The child said that he pinched the elephant)
b. * [awan aanaye nu).li eona ] kutti parannu
he elephant-a pinched that child-n said
(14)a. kutti parannu [awan aanaye nulli enns ]
child-n said he elephant-a pinched that (=13. a)
b. *awan parannu [kufti aanaye nulli enn» ]
he said child-n elephant-a pinched that
Examples (l.c), (2,c), and (3. a) show that pronouns can precede their
antecedents in English; the ungrammatical sentences in (8)-(ll) and (13)-
(14) show that they cannot do so in Malayalam. Examples (lb), (2.b) and
(3.b) show that pronouns cannot c-command their antecedents in English;
70
(9. a), (10. a) and (13. a) show that they may do so in Malayalam. These facts
jBoDow from principles (4) and (7) .
2. Strong Cross Over
2.1. The principle of strong cross over in English is illustrated by (15)
and (16):
(15) a. Who said h£ kissed Mary?
b. Who said Mary kissed him,
c .* iVho did h£ say Mary kissed?
(16)a. Everyone said he kissed Mary.
b. Everyone said Mary kissed him
c. *He said everyone kissed Mary.
(15. a) may be paraphrased as, "For which x did x say x kissed Mary?",
and (15. b) as, "For which x did x say Mary kissed x?". The meaning of (16. a)
is, "For all x, x said x kissed Mary", and (16. b) is, "For all x, x said
Mary kissed x". What is interesting for our purposes is the fact that (15. c)
and (16. c) do not mean "For which x did x say Mary kissed x", and "For all
X, X said X kissed Mary". What blocks these readings is the principle of
strong crossover.
1 shall assume that strong crossover is a principle that relates pro-
nouns to quantified antecedents (who, which boy, everyone, every boy etc.).
Borrowing the essential insight, but not the formulation, from Higginbothom
(1980), I shall state the principle as follows:
(17) Strong Cross Over _
Quantifed antecedents must c-command the pronouns.
I shall also assume, following Reimsjidk and Williams (forthcoming)
that strong cross over is a principle that applies at the level of 'NP stru-
cture', which is a level of representation after NP movement but before
wh- movement. The NP structure of (15)a-c is given as (18)a-c respectively:
(18)a. Who said he kissed Mary.
b. Who said Mary kissed him.
c. H£ said Mary kissed who.
The NP structure of (16)a-c , which do not involve wh- movement, is the
same as the surface structure. In (18)a and b, who c-commands the pronoun, but
in (18)c it doesn't, thereby violating (17). The same explanation holds for
(16)c as well.
71
?.2. Though the principles of noncoreference are different for Malayalam
type languages and English type languages, the principle of strong cross over
is identical. Consider, for example, the Malayalam sentences that parallel
(15):
C19)a. aaia parannu [meefi awane umma weccu enna ]?
who said Mary-n him kiss gave that
(Who said Mary kissed him?)
b. aafsparanfiu [awan meefiye umma weccu enna]?
who said he Mary-a kiss gave that
(Who said he kissed Mary?)
c.*[meeri aare umma weccu enn* ] awan paraniTli?
Mary-n whom kiss gave that he said
(Who did he say Mary kissed?)
In (19c), the quantifier does not c-command the pronoun, which is prohi-
:bited,bythe principle of strong crossover. Observe that the pronoun follows
the antecedent in (19), and therefore the ungrammaticality of (19c) cannot
be explained in terms of the noncoreference principle. The same phenomenon
is found in non-wh quantifiers as well:
(20) a. cilaig parannu [meefi awafe umma weccu enna]
some-n Mary-n them
(Some said that Mary kissed them)
b. cilao paraniTu [award meeTiye umma weccu enna ]
some-n they Mary-a
(Some said they kissed Mary)
c.*[meeri cilare umma weccu enna ] a war parannu
Mary-n sorae-a they
(They said Mary kissed some (people))
(21) a. ellaawarum parannu [meeri aware umma weccu ennsi
all-n Mary-n them
(Everyone said Mary kissed him)
b. ellaawarum parannu [awara meeriye umma weccu enna ]
all-n they Mary-a
(Everyone said they kissed Mary)
c* [meeri ellaawareeytun umma weccu enna ] awaia parafiflu
Mary-n all -a they
(They said Mary kissed everyone)
It may be noted that in the grammatical sentences cited above, the
quantifier c-commandsthe pronoun. In the ungrammatical sentences, on the
other hand, the pronoun c-commands the antecedent asymmetrically. It is
legitimate, therefore, to ask why the strong cross over condition should
be (17) rather than, say, (22) :
(22) Pronouns cannot c-command quantified antecedents.
72
Evidence in support of (17) is found in possessive constructions in
Malayalam. Recall that both subjects and objects are dominated by the S
node in the language. As a result, the possessive inside the subject is
c-commanded by the object, and the possessive inside the object is c-com-
manded by the subject. We get the expected bad results when both the pro-
noun and the quantified antecedent are possessives, i.e., when neither
the pronoun nor its antecedent c-conunands the other. Consider the follow-
ing sentences:
4
C23)a. aar-j [awante bhaafyaye] nulli?
who his wife-a pinched
(Who pinched his wife?)
b. aafe [awante bhaarya] nulli?
whom his wife-n pinched
(Whose wife pinched him?)
c* [aarute bhaaryaye] awan nulli?
whose wife-a he pinched
(Whose wife did he pinch?)
d. * [aarute bhaarya] awane nulli?
whose wife-n him pinched
(Whose wife pinched him?)
e.* [aarute bhaaryaye] [awante amma] nulli?
whose wife-a his mother-n pinched
(Whose wife did his mother pinch ?)
f.* [aarute bhaarya] [awante ammaye ] nulli?
whose wife-n his mother-a pinch
(Whose wife pinched his mother?)
Once again, the same phenomenon is found in non-wh quantifiers:
(24) a. cilafe [awafu^e bhaaryamaai*] nulli
some-a their wives-n pinched
(Their wives pinched some people)
b. cilar.i [awarofe bhaal-yamaaTe] nulli
some-n their wives-a pinched
(Some people pinched thier wives)
c. * [cilafu^e bhaafyamaare] awara nulli
some-poss wives-a they-n pinched
(They pinched some people's wives)
d. * [cilafute bhaafyamaais] aware nulli
aome-poss wives-n them pinched
(Some people's wives pinched them)
e. * [cila¥ute bhaaryamaafe] [awarute ammamaafa] nulli
some-poss wives-a their mothers-n pinched
(Their mothers pinched some people's wives)
f . * [cilarute bhaafyamaafa] [awarute ammamaafe] nulli
some-poss wives-n their mothers-a pinced
(Some people's wives pinched their mothers)
73
The reader must have noticed by now an interesting contrast between the
facts of strong cross over in Malayalam and English with respect to their
subject-object asymmetry. The relevant examples are repeated below:
(25) English
a. *His mother pinched everyone .
b. *Who did his mother pinch?
(26) Malayalam
a. ellaawareyum [awanrfe ammamaar ] nujji
all-a their mothers-n pinched
(His mother pinched everyone)
b. aafe [awante amma] nulli?
whom his wife-n pinched
(Who did his mother pinch?
Why are (26)a and b grammatical in Malayalam, while their corresponding
sentences in English are ungrammatical?
The answer lies in the difference between the phrase structure configu-
rations in English and Malayalam. (25a) and (26a), for example, have the fol-
lowing structures:
(27)
NP
His mother
V NP
I I
pinched everyone
(28)
ellaawareyum awarute ammamaar nulli
all-a their mothers-n pinched
Given the fact that subjects and objects c-command each other in Malaya-
lam, but not so in English, the contrast follows from (17): In Malayalam, the
object antecedent c-commands the pronoun inside the subject KP, and hence co-
reference is possible. In English, the object antecednt does not c-command
the pronoun and hence co-reference is disallowed.
3. Malayalam Pronoiins and the Government Binding Theory
The theoretical interest of pronominal coreference in Malayalam comes
from the clear separation of the facts of noncoreference and strong cross
over. Chomsky (1981), for example, argues that strong cross over and non-
coreference can be accounted for by the following binding condition:
74
(29) R-expressions are free.
The term 'R-expression' refers to what Chomsky calls 'names' (John, the
boy etc.) and 'variables' (traces of wh movement and quantifier raising).
'Free' is 'not bound', where 'bound' is defined as in (30):
(30) o< is bound to f3 iff ■\ and fl bear the same index and (» c-commands \.
Consider the following ungrammatical sentences:
(31)a.* H£ said Mary kissed John
b.*Who.did he say Mary kissed t.?
In (31a), the R-expression John is bound to (=is c-commanded by the
coindexed NP) he, which makes it ungrammatical according to (29) . In (31b)
by choosing who as the antecedent of the pronoun, the pronoun and the
wh-trace become coindexed, thereby creating an informed structure iden-
tical to that of (31a). Thus, both (31a) and (31b) are claimed to be accoun-
ted for by the same binding condition.
The two principles that are collapsed in (28) by grouping 'names' and
'variables' as 'R-expressions' can be stated as follows:
(32) a. Names cannot be c-coramanded by coindexed elements.
b. Variables cannot be c-commanded by coindexed elements.
I have shown that pronouns can c-cocommand their antecedents in Malaya-
lam, violating (32a) (see (12)a,bl. Similar facts are observed in languages
like Chinese and Japanese as well . Therefore (32a) cannot be a universal
principle of grammar that applies to all languages.
Another conclusion that must be drawn from the Malayalam facts is that
the principle governing strong cross over and noncoreference cannot be colla-
psed together as a single principle, as (29) does. In English, the latter is
"Pronouns cannot c-command their antecedents", and in Malayalam, "Pronouns
cannot precede their antecedents". Yet, the two languages have the same prin-
ciple of strong cross over: "Quantified antecedents must c-command pronouns".
The subpart of (29), stated as (32b), does not account for the Malayalam facts.
NOTES
One of the arguments against the VP node is that only those consti-
tuents which are directly dominated by the S node (in (6)) (i.e. the order
free constituents) can be clefted. The correlation between clef ting and
scrambling is obscured by the VP analysis. Further, if we adopt the flat
structure, the structures which generate the cleft sentences are independ-
ently required by the grammar, while the VP hypothesis forces us to invent
special ad hoc rules for these structures.
2.
n - nominative; a = accusative; g = genitive.
75
The terminology of 'cross over' Is derived from the transformational
account of the phenomena which Involved the 'crossing' of a wh phrase across
a coreferentlal NP (cf. Postal (1971)). Such a formulation is clearly inade-
quate, as (a) the phenomenon of cross over is found in non-wh quantifiers
such as every and some, and (b) languages without wh movement (e.g. Chinese
and Malayalam) also exhibit the phenomenon (see Higgnibothom(1980) ) . The
formulation in (17) says nothing about wh movement, and covers all the cases
which the transformational account does not cover.
4.
These questions are slightly odd, since the preferred ones in Malar-
yalam are clef ted questions:
(i) aaf aan^ [paraiiflata [meeri awane umma weccu enni ]?
(ii)aaraana parannat.-> I awan meefiye umma weccu ennTs ]?
(ill) * [meeti aare umma weccu enn]aan^ awan parannaE*
Malayalam has no wh movement. Hence the NP structure in the language
is the same as the surface structure.
■ As in the case of (19), the preferred questions involve clefting.
See Mohanan C1981) for a discussion.
REFERENCES
CHOMSKY, N. 1980. On Binding. Linguistic Inquiry 11, 1-46.
. 1981. Lectures on Governement and Binding. For is Publications
Dordrecht.
HIGGINBOTHOM, J. 1980. Pronouns and Bound Variables. Linguistic Inquiry 11,
679-708.
LASNIK, H. 1976. Remarks on Coreference, Linguistic Inquiry 2, 1-22.
MOHANAN, K. P. 1981. On Pronouns and Their Antecedents. Unpublished ms. MIT.
MOHANAN, K. P. Irl press. Grammatical Relations and Clause Structure in Mala-
yalam. Bresnan Ced) Mental Representations of Grammatical Relations, MIT
Press.
POSTAL, P. (1971). Cross Over Phenomena. Holt, Rinehart & Winston.
REINHART,T. 1976. The Syntactic Domain of Anaphora. Ph. D. Thesis, MIT.
RIEMSDIJK, H. van & E. Williams. Forthcoming. NP Structure.
Studies in the Linguistic Sciences
Volume 11, Number 2, Fall 1981
INTERFACE OF LEXICON AND GRAMMAR:
SOME PROBLEMS IN HINDI GRAMMAR
Rajeshwari Pandharipande
University of Illinois, Urb ana -Champaign
The major goals of this paper are the following: (a) to
discuss three topics in Hindi grammar, i.e., the possessive
postpositions, the progressive construction, and the syntactic
process of raising, and to show that the descriptions of
the above given in the grammars (pedagogical and others)
are inadequate to account for the data; (b) to argue that
a pragmatically oriented analysis is essential to correctly
account for the data; and finally (c) to discuss the empiri-
cal as well as the theoretical implications of the above
proposal .
0. Introduction. It has been argued variously that linguistic
forms are related to extralinguistic settings (Donnellan 1978, Givon 1978,
Green 1975, Morgan 1978, etc.). Also, it is shown that native speakers'
competence includes knowledge of the correspondence of linguistic
facts to contexts. Thus, pragmatics is considered to be an essential
component of language analysis. The major goals of this paper are the
following: (a) to discuss three topics in Hindi grammar, i.e., the
possessive postpositions, the progressive construction, and the syntactic
process of raising, and to show that the descriptions of the above
given in the grammars (pedagogical and others) are inadequate to account
for the data; (b) to argue that a pragmatically oriented analysis is
essential to correctly account for the data; and finally (c) to
discuss the implications of the above proposal.
1.0. Possessive postpositions. In Hindi there are five postpositions
used to express the relationship between the possessor and the possession.
These postpositions are as follows: ka, ke, id (genitive postpositions),
ke (invariable genitive postposition) , k£ (directional postposition) ,
me (locational postposition), and ke pas (locational postposition) .1
What Kachru (1970) pointed out eleven years ago is still true
today, "In the existing grammars and pedagogical works on Hindi-Urdu
the category of possessive postpositions has mainly been described
on the analogy of the possessive in English." For example, it is
pointed out in Fairbanks and Misra 1966, and McGregor 1972, that while
English uses one and the same construction in the following sentences,
different postpositions are used in their Hindi counterparts. Consider
the following examples.
78
1. Jan ke_ / ki_ tin betiya hi
John poss . three daughters are
John has three daughters.
2. jan ke pas ek kitab h£,
John poss . one book is
John has a book.
3. Jan ko buxar he.
John poss. fever is
John has a fever.
h. jan me_ bahut dh€ry h£
John poss. a lot patience is
John has a lot of patience.
Notice that different postpositions are used in Cl)-C4).
The above postpositions are generally divided into two groups,
i.e., ke/ka, ke, ki and ke pas belong to one group while ko^ and me
belong to another. This division is based on the general observation
that ke/ka, ke, ki and ke pas are used when the possessor is animate
and the possession is either animate or inanimate. In this case,
the possession is generally (-abstract) .2 In contrast to this, ko
and m£ are used when the possessor is animate while the possession
is abstract, such as headache, fever, courage, etc. However, it is
not clear what exactly the conditions are which determine the choice
of one postposition as opposed to the other (s) within the same
group.
In the following discussion I will present the data, discuss
inadequacies of the hypotheses proposed so far, and finally point out
that a pragmatically-oriented hypothesis is essential to account for
the data.
1.1. ke/ka, ke, ki and ke pas. First, let us consider the group
which contains ke/ka, ke, ki and ke pas. Consider examples (5) -(13).
Notice that (5) -(7) allow only ke/ka, ke, and ki^. Thus, ke pas is
blocked in (5) -(7). In contrast to this, ke/ka, ke, and ki^ are blocked
in (8) and (9) where ke pas is used obligatorily. Also notice that both
ke/ka, ke, ki and ke pas are allowed in (10) -(13).
79
5. Jan )ke/ka ? ek beta h C
/ *ke pas \
John poss. one son is
John has a son.
6. r
\ mere 7
/ *mere pas \
mere / do bhai h£
I-poss. two brothers are
I have two brothers .
un ke 7 kai dost hg.
*un ke pas)
they-poss. many friends are
They have many friends .
ft ^ - :■
0. -, mere pas/ ek kitab hf.
( *meri j
I-poss. one book is
I have a book.
9. bacce jke pas ? ek lalipap he
Q*ke/ka )
child poss. one lollipop is
The child has a lollipop.
10 . un ke pas ( ek kapre kl dukan h £
(ki s
they-poss. one clothes of store is
They have a clothing store.
11. jixi ike pas '. bari jmotar? hC
(kl ) ' (kar )
John poss . big car is
John has a big car.
12. ajkal (unke pas'* do fsekretari / ht
(unke '. /njkar* (
nowadays they-poss. tworsecretaries?are
(servants )_
Mowa days they have twofsecretarieB.
/servants,
'■]
Several attempts have been made to predict the distribution of
the above postpositions. McGregor (1972:52) suggests, "What is denoted
by the use of ka is a permanent, characteristic or non-fortuitous type
of relationship by contrast with the more 'contingent' possession
expressed by the use of ke pas.". This analysis accounts for the
80
sentences (5) -(9). However, it does not explain mutual interchangeability
of the postpositions in (10)-(12). According to McGregor's hypothesis,
a sentence such as (12) with ka expresses a permanent relationship between
the possessor and the possession, e.g., servant, etc., while C12)
with ke pas expresses a 'contingent' type of a relationship. However,
a close examination of the Hindi data shows that it is not the case.
Consider examples (13) and (14) .
(13 ) un ke do ^nokar the Jo unke ghar me do din
they-poss. two servants were who their house in two days
se zyada na rah sake
more than neg. stay could
They had two servants who could not stay in their house for
more than two days .
(lU) un ke pas ek nOkar tha jo zindgi bhar imke sath rsiha
they-poss . one servant was who life entire them with stayed.
They had a servant who stayed with them all (his) life.
In (13) the postposition ke_ is used and it is explicitly mentioned
that the relationship between the possessor v£ 'they' and the possession
njkar 'servants' is not permanent. Also, notice that (13) does not
present any contradiction. In contrast to this, consider (14) where
ke pas is used and it is explicitly mentioned that the relationship
between the possessor and the possession was permanent and yet (14)
does not present any contradiction. If McGregor's hypothesis was correct,
then, we expect both (13) and (14) to present contradictions.
Another hypothesis is presented in Kachru (1970) . It is claimed
that the apparatus of the inherent features of nouns provided by the
transformational theory (Chomsky 1965) accurately predicts the distri-
bution of the postpositions with the help of spelling rules. Thus in her
hypothesis the choice of Ice versus ke pas is determined on the basis
of the following featu
Possessor Noun
15. f+Animate]
16. a. [+Animate]
b. [+Human]
(Kachru 1970).
Possessed Noun
Postposition
a. f-Animate 1 /
/ (-Abstract) ^
V.< parts of bodyV J
b. [+Animate]
ke
a.
T-Animate 7
( -Abstract \ _
ke pas
b.
r+Human }
-Status ^ ^
ke pas
81
The above feature-system faces the following problems: (a) (15b)
predicts that if both the possessor and the possession are animate,
then k£ is used obligatorily. Within this framework, the possessor-noun
will obligatorily take the postpositions ke/ka, ke, ki if the possessed
noun stands for pets such as cats, dogs, etc. This hypothesis blocks
the use of ke pas in such cases. However, example (13) shows that ke pas
can be used in the cases such as (15b) . (b) This system does not
allow the use of ke_ or ka, k£, ki_ if the possessed noun as the features
such as r+Human "1 (recall (16b)). However, example (12) shows that
|_-StatusJ
ke/ka, ke, ki are not blocked in such cases.
1.2. Semantics: Pragmatics. Another possible hypothesis is
as follows: both ke/ka, ke, ki and ke pas express a possessor-possessed
type of a relationship. However, they are not semantically identical
to each other. While ke indicates emotional closeness/intimacy in the
relationship between the possessor and the possessed, ke pas does not
indicate any such relationship, but rather, it indicates owner-owned
type of a relationship.
This hypothesis predicts the use of ke/ka, ke, ki_ in sentences
such as (5), (6), and (7), where intimate relationships such as parent-
child, brother-sister, friend are expressed. In contrast to this, the
use of ke pas in (8) and (9) is also expected since the owner-owned
type of a relationship is expressed in (8) and (9) . However, the choice
of the above postpositions cannot be determined purely in terms of their
semantics. Were it so, then every sentence with k£ would express intimate
relationships while every sentence with ke pas would express owner-owned
types of a relationship. Thus within this hypothesis k£ and ke pas
would never be used interchangeably. This hypothesis fails to account
for (11) -(13). Moreover, the emotional closeness versus distance
expressed by ke/ka, ke, ki and ke pas cannot be attributed purely to
the semantics of the above postpositions. If it is not possible prag-
matically to derive the above-mentioned meanings then the postpositions
merely express the possessor-possessed type of a relationship.
Consider the following example.
Suppose the context for (11) is as follows: A police officer is
being told by his officer that a thief called John who they are looking
for, has a big car. In this case, the officer is merely reporting the
fact that John has a big car (recall (11)). In this context, (11)
with ki^ does not indicate that John has an emotional attachment for
his car. Thus a sentence such as (11), i.e., jan kl barl imotar?hfc
( car )
'John has a big car.', indicates only the possessor-possessed type of
relationship.
Similarly, if it is pragmatically impossible to interpret the
owner-owned type of a relationship, then, ke pas merely indicated possessor-
possessed type of a relationship. Consider the following situation:
If a police officer is making an announcement about a car_thief called
82
John, and if he utters sentence (H)* i.e., jan ke pas baril^? >h£
'John has a big car.', then (11) even with ke pas, does not indicate
that John owns a big car, but rather, it only indicates that John has
a big car in his possession.
The discussion so far indicates that the choice of the postpositions
cannot be determined purely in terms of their semantics since they
fail to express the same meaning in all the contexts.
If the pragmatic factors are taken into consideration, the above
hypothesis correctly predicts the distribution of the postpositions as
follows: (a) ke/ka, ke, ki and ke pas are used mutually exclusively
in C5)-C9) since certain relationships such as in (5) -(7), are conven-
tionally considered to be intimate while certain others, e.g., (8) and
(9) are conventionally considered to be owner-owned types of relationships.
Therefore, while k£ is obligatorily used in (5) -(7), ke pas is obliga-
torily used in (8) and (9) . (b) In such cases where the relationship
is not conventionally determined, the choice of the postpositions is
determined by speaker's intention to indicate a particular type of
a relationship. Consider the following examples.
17. CmerQ. 7 atmiy koi nahl hfc. . . Tmere /* pas o^te kal h^
( *mere pas ) (^meri \
I-poss. intimate any neg. is ... I-poss. women many are
I do not have anyone (intimate to me). ...I have a lot of women.
CNovel: bat ek prat ki "A Woman's Story", written by Krishna
Agnihotri 197^, p. 1^1, I g-O
The context of the above sentence is as follows: A man is talking to
his friend. He is telling her that he has a lot of women as friends
but he is not emotionally attached to anyone.
In this sentence (17) , ke pas is blocked when an intimate relationship
is intended while ki^ is blocked when the speaker's intention is to
indicate that the relationship is not intimate.
Now consider the following example, (18). In this sentence the speaker
wants to convey his intimate feelings to the hearer. In this case the
use of ke pas is totally blocked.
(18) r mera ? to is sansar me tumhare sivay koi nahi
(*mere pas >
I-poss. really this world in you except for anyone neg.
Except for you, I have indeed no one (intimate to me) in
this world. [Novel: Loi ka tana ' Loi's woof,
written by Rangeya Raghav 1970, p. 26]
(c) If the speaker is not aware of the nature of the relationship
between the possessor and the possessed, or if the nature of the
relationship is not important to him, then he can use ke pas and ke/ka.
83
ke, ki interchangeably. This explains the interchangeability of the
postpositions in (12)and (13).
2.0 ko vs. me. Now I will discuss the other set of postpositions
and point out the conditions which determine the choice of one as
opposed to the other. First, let us consider the data where k£ and
me occur in Hindi .
(19) us fko ? khusi h£
(*me )
he-poss. happiness is
He is happy (literally, he has happiness).
(20) us ,'ko / prasannata h£
i*w.e \
he-poss. cheerfulness is
He is cheerful (literally, he has cheerfulness)
( 21 ) \is J ko ? buxar h C
siko?
{*me)
he-poss . fever is
He has a fever.
(22) usfko? cinta hi
he-poss. worry is
He has a vorry.
(22a) us fko? CmuskilC h£.
l*me \ I dikkat )
he-poss. difficulty is
He has a diffic\ilty.
(23) us (" me ? hahut sakti h£
me?
'*ko J
he-poss. alot strength is
He has a lot of strength.
{2k) usCme? caliki khub ht
5 C me ? c
{*ko\
he-poss. cleverness a lot is
He is very clever (literally, he has a lot of cleverness)
(25) usTme? cancalta h£
he-poss. unsteadiness is
He is unsteady (literally, he has imsteadlness ) .
84
(26) \as Jme^ daya h£
he-poss. kindness is
He is kind (literally, he has kindness).
( 27 ) us ^ me / himmat h £
,[ko^
he-^oss . courage is
He has courage.
(28) us f me? dhfry hC
<Lko$
he-posB. patience is
He has patience.
Notice that in (19) -(22) ko is used obligatorily and m6 is totally
blocked. While in (23) -(25) me is used obligatorily and k£ is blocked.
While in (27) and (28) m5 and ko can be used interchangeably.
2.1. Hypotheses. Kachru (1970) proposed a feature system and
additional rules for the distribution of ko^ and me.
(29) Possessor Noun Possessed Noun Postposition
[+Human] [+Abstract] ko/me
According to Kachru (1970) k£ occurs if the possessed noun belongs to
the sub-class of abstract nouns which denote sensations such as 'pleasure' j
'pain', etc., me occurs when the possessed noun denotes states such as
' enthusiasm', etc.
Notice that (22a) with k£ does not denote any sensation. Similarly
(24) with mS does not denote any state. Also, this hypothesis does
not explain mutual interchangeability of k£ and mS in (27) and (28).
The above data can be handled in a straight-forward fashion within
the following hypothesis. Ko denotes abstract possessions which are
temporarily possessed by the possessor. Those abstract possessions can
be temporary physical mental states, such as khusi 'happiness', buxar
'fever', etc., or possessions such as muskil 'difficulty', etraz
'objection', etc. Those possessions are temporary in the sense that
they are not permanent characteristics of the possessor's physical or
mental disposition. In contrast to this, me denotes permanent qualities/
characteristics which are part of the physical or mental disposition
of the possessor. Recall examples (19) -(22) where temporary physical/
psychological states such as khusi 'happiness ' (19) , buxar 'fever'
(21), etc. are expressed and ko is obligatorily used. On the other hand,
in (23) -(25) permanent qualities such as calaki 'cleverness'
(24), daya 'kindness' (26), are denoted and mg is obligatorily used.
85
2.2 Evidence. The above hypothesis is further supported by the
fact that if it is explicitly mentioned that the possession, namely
the physical/psychological state is not part of the disposition of the
possessor, then a sentence with k£ (30) is fine. However, a sentence
with me (31) presents contradiction. Consider examples (30) and (31) .
(30) us ko aj khusi hf par khusi uske svathav ka
he-poss. today happiness is but happiness his disposition of
dham nahi ht
characteristic neg. is
Today he is happy (literally, today he has happiness) but happi-
ness is not a characteristic of his disposition.
(31) ???us me aj calaki b£, par calaki uske svabhav
he-poss. today cleverness is but cleverness his disposition
ka dharm nahi h£,
of characteristic neg. is
He is clever today (literally, he has cleverness today) btit clever-
ness is not a characteristic of his disposition.
2.3 Pragmatics . Although the above evidence shows that k£ and
mg are semantically similar to each other, the choice of one as against
the other is not determined totally on the basis of their semantics.
Were it so, then, sentences such as (27) and (28) where ko and me_
can be used interchangeably can never occur in the language.
A close examination of the data shows that a number of pragmatic
factors play a role in determining the choice of postpositions.
Consider the following few cases: (a) If a speaker does not know whether
or not a particular quality/state is part of a possessor's disposition,
then he can use ko and me interchangeably, (recall examples (27) and
(28)). (b) When a speaker is specifically referring to a state/quality
as being a permanent characteristic of possessor's disposition, then
only m6 is allowed and ko is blocked. Consider the following example
(32).
(32) yah koi ascary ki bat nahi ki usne yah sankat
this any surprise of matter neg. that he-Ag this disaster ^ __
dhixy ke sath jhela. m|^ usko bacpan se janti hu
patience with survived. I him childhood since know aus
us j me ? dhtry bahut hi
(*ko >
he-poss. patience a lot is
It is not a matter of surprise that he survived this disaster
with patience. I know him since he was a child. He has a
lot of patience.
86
The discussion in this section shows that the choice of ko^ vs. me
is not determined purely in terms of their semantics, but rather,
pragmatic factors such as speaker's knowledge/intentions play a role
in determining their choice.
3.0. The progressive construction. The progressive construction
in Hindi is illustrated in the following sentences. Consider examples
(33) and (34).
(33) jan kitab parh raia ht
John book read prog, is
John is reading the book.
(3I*) jan ghar ja raha he.
John home go prog, is
John is going home.
The analyses of the above syntactic construction presented so
far (Kachru 1966, McGregor 1972, Vajpeyi 1959, etc.) are not satisfactory
since they fail to explain why certain verbs in Hindi fail to partici-
piate in this construction. It is observed that verbs, such as, bethna
'to sit', pakarna 'to hold', and marna 'to die', fail to participate
in this construction. Consider the following examples (3S)-(37).
(35) *vah kursi par b£th rahi ht
she chair on sit prog, is
She is sitting on the chair.
(36) *vah hath me kitab paka^- raha ht
he hands in book hold prog, is
He is holding the book in his hand.
(37) *mera dost mar raha h£
my friend die prog, is
My friend is dying. 4
In the following discussion I shall attempt to define the conditions
which determine whether or not a verb will participate in a progressive
construction.
3.1. Hypothesis . First, I will describe the semantic function of
the progressive construction in Hindi; second, it will be pointed
out that certain verbs fail to carry out this function and thereby
fail to participiate in the construction; third, it will be demonstrated
with illustrations that the participiation/non-participiation of the
verbs in the progressive construction is not determined purely in terms
87
of their semantics but rather, the context in which the verb is used,
also plays an important role.
A close examination of the Hindi data shows that a verb in the
progressive construction expresses progression of an act or a process
or a state, such as sleep, etc. Thus, (33) expresses the action of
reading the book in progress, while a sentence, such as (38) expresses
continuation of the state of sleep. Consider (36).
(58) bacca so raha h£.
child sleep prog, is
The child is sleeping/asleep.
A sentence in progressive construction expresses futurity also.
For example, (33) and (34) ambiguously express the meaning "He is going
to read the book." and "John is going to go home.", respectively.
If the above analysis of the function of the progressive construction
is correct, then we can formulate the following hypothesis: A verb
participates in the progressive construction if the act or process
expressed by the verb is gradual. In other words, a verb participates
in the progressive construction if the progression of the act is
gradual from the point of its beginning to its completion. Thus, the
verbs such as, parhna 'to read', likhna 'to write', jina 'to live',
which express gradual progression of the act/process readily participate
in the progressive construction. However, the punctual verbs^
such as btthna 'to sit down', uthna 'to get up', marna 'to die', rukna
'to stop' , 'cup hona 'to become quiet', and tutna 'to be broken' express
processes which are abrupt and not at all gradual. In other words, their
progression from the starting point to the end is abrupt. Therefore,
those verbs do not participate in the progressive construction in Hindi
(recall examples (35) -(37)). Sentences in progressive construction
with the punctual verbs are grammatical only if they are interpreted
as expressing futurity. Consider examples (39)-(41).
\
(39) vah vaha ruk raha h£
he there stop prog, is
Hejwill stop there.
(is going to stop there
(UO) vah mar raha ht
he die prog, is
Hefwill die soon. ')
ris going to die soon.^
(1+1) vah bttb raha ht
he sit down prog, is
Hefvill sit down. ?
(will be sitting down.^
88
In contrast to the above, in English, punctual verbs when used
in the progressive construction, express the resulting state of the
act/process. Thus, a sentence such as, "She is sitting" means "She is
in the state of sitting", i.e., "She is seated". However, the punctual
verbs in Hindi do not express the resulting state of the act/process
expressed by the verb. Therefore, (35) -(37) are not semantically
similar to their respective counterparts in English.
3.3. Pragmatics. An investigation of the data shows that whether
or not a verb will participate in the progressive construction cannot
be determined purely in terms of the semantics of the construction or
of the verbs. The context in which the verb is used plays an important
role in this decision. Consider the following examples (42) and (43)
where the same verb rukna 'to stop' is used. Notice that when it is
used with reference to a man, (42) is not good, while when used with
reference to a train (43) is fine. This is due to the fact that prag-
matically rukna 'to stop' expresses an abrupt action if it is carried
out by a man, while it expresses a gradual action if it is carried out
by a train.
{k2). ???admi ruk raha ht
man stop prog, is
The man is stopping.
(1*3) tren riok rahi ht
train stop prog, is
The train is stopping.
Similarly, the action of falling is abrupt in the context of a
child (44) while it is gradual in the context of a balloon (45) .
Notice that while (44) is ungrammatical (45) is fine.
{hk) *m£ ne dekha bacca zamin par gir raha the.
I Ag. saw child floor on fall prog, was
I saw the child falling on the floor.
(1+5) in£ ne dekha gubbara havai Jaiiaz se gir raha tha
I ag. saw 'balloon airplane from fall prog, was
I saw the balloon falling down from the
3.4. Punctual verbs and repetitive actions. In addition to the
above discussion, one more fact needs to be mentioned in this context.
In Hindi, if a verb expresses an abrupt action but the action is
performed repeatedly (without any interval), then the verb can participate
in the construction. In this case, the native speakers interpret the
repetition of such an action as the continuation of that action.
Consider the following example.
89
(46) vah uchal raha ht
he jump prog, is
He is jumping.
In (46), the verb uchalna 'to jump' expresses that the act of
jumping is performed repeatedly. Similarly, (44) is alright if it means
'the child is repeatedly falling down on the floor'. However, notice
that the actions expressed by the verbs b t.thna 'to sit down' (35),
pakarna 'to hold' (36), marna 'to die' (37j, do not express repeated
action. Therefore, the verbs in (35) -(37) fail to participate in the
progressive construction.
4.0. Raising. In this section I will discuss the role of pragmatics
in determining whether or not the syntactic process of raising applies
to a sentence. In particular, I will discuss the process of B-raising
in Hindi, wherein the subject of the lower clause is made direct object
of the higher clause/main clause. The verb in the lower clause is
transformed into a participle. This process is already discussed in
Kachru (1966) and Subbarao (1974) . This process is exemplified in the
following sentences.
(1*7) m£ ne dekha ki larka kitab parh raiia ht
I Ag. sav that boy book read prog, is
I saw that the boy was reading a book.
(1*8) m t ne larke ko kitab payhte hue dekha
I Ag. boy' obj. book reading saw
I saw the boy reading the book.
Notice that (48) is the raised counterpart of (47). The subject
of the lower clause, i.e., layka 'boy' in (47) is the direct object
of the higher clause in (48), while the verb payh raha 'read + prog.'
in (47) is transformed into the participle payhte hue 'reading' in
(48).
The above transformational rule is discussed in detailed accounts
presented in Akmajian and Heny (1975), Bach (1974), and Rosenbaum (1967).
It was pointed out (Postal 1973:363-74) that the relationship between
the sentences such as (47) and (48) is not purely syntactic. A close
examination of the Hindi data shows that the conditions which determine
whether or not the rule will apply, do not purely depend on the syntactic
structure of the sentence. Consider the following sentences where the
rule of raising fails to apply.
90
{k9) m\ ne dekha ki bacco ne khana khaya
I Ag. sav that children Ag, meal ate
I saw that the children had a meal.
(49a) *m\ ne bacco ko khana khae hue dekha
I Ag. children-obj . meal having eaten saw
I saw the children having eaten the meal.
(50) mt ne dekha ki usko sardard ho raha h£
I Ag. saw that he-dat. headache happening aux.
I saw that he was having a headache.
(50a) * m £ ne usko sardard hote hue dekha
I Ag. he-dat. headache happening saw
I saw him (in the state of) having a headache.
The failure of (49) and (50) to undergo raising is explainable on
the basis of the following: Kachru, et.al. (1976), pointed out that a
sentence with an oblique subject (subject followed by a postposition)
fails to undergo participialization. Since the subject in (50) is
oblique, it is only as expected that (SO) does not undergo the process
of participialization. Also, it is argued in Pandharipande and Kachru
(1977) that a perfective participle does not refer to the subject
if the subject of the verb is not in the resulting state of the act
expressed by the verb. 6 Thus, in (49), the participle khae hue
'eaten' does not refer to the subject bacce 'children' because the
subject bacce 'children' cannot be said to be in the resulting state
of the action of eating.
4.1. More exceptions. There are a number of sentences however,
which fail to undergo raising even though they do not fall into either
of the categories mentioned above. In this section 1 will examine those
sentences and determine the conditions which block the operation of
the rule of raising on those sentences. Those sentences are the
following.
(51) mt ne dekha ki andhera ghir rEiha h£
I Ag. saw that darkness siorrounding aux.
(a) I saw that darkness was siirrounding.
(b) I saw that the night was setting in.
(52) mt ne jandhere?ko ghirte hue dekha
Qandhera^
I Ag. darkness surrounding saw
(a) I saw the darkness was stirroundlng .
(b) *I saw that the night was setting in.
(53) mt ne dekha ki uske both hil rahe ht
I Ag. saw that her lips move prog. aux.
(a) I saw that her lips were moving.
(b) I saw that she was trying to speak.
91
(5^) mt ne xiske hoth3 ko hilte hue dekha
I Ag. her lips obj . moving saw
(a) I saw her lips moving.
(b) *I saw she was trying to speak.
(55) mt ne dekha ki cifiya cahak rahi h£
I Ag. saw that birds chirp prog. aiix.
(a) I saw that the birds were chirping.
(b) I saw that it was morning.
(56) mt ne ciyiyo ko cahakte hue dekha
I Ag. birds obj. chirping saw
(a) I saw the birds chirping.
(b) *I saw it was morning.
Notice that sentences (51), (53), and (55), are structurally similar
to (47). However, they fail to undergo raising. Also, notice that
(51), (53), and (55), are not semantically similar to their raised
counterparts (52), (54), and (56), respectively. (51), (53), and (55),
express two meanings while their raised counterparts express only one
meaning. In other words, the meaning (b) is blocked in the raised sentences
(52), (54), and (56). In (51), (53), and (55), (a) expresses literal
(vacya) meaning while (b) expresses suggested meaning (vyagya) J
Native speakers feel that the suggested meaning (i.e., in (b)), is more
prominent (pradhana) than the literal meaning in (a), while they feel
that the literal meaning is of secondary importance (g.>na) . Thus,
according to the native speakers it is the suggested meaning which
is in focus. Also, they feel that it is (a) which is suggestive of (b) .
In other words, they feel that the suggested meaning in (b) is based
on the literal meaning in (a).
Now let us consider the raised counterparts (52) , (54) , and (56) .
The meaning in (b) which is prominent in the non-raised sentences, is
totally lost in their raised counterparts.
On the basis of the above facts about the semantics of (51)-(56),
we can formulate the following hypothesis: Raising fails to apply under
the following conditions: (a) if the lower clause expresses two meanings,
i.e., literal and suggested, (b) if the suggested meaning is of primary
importance, and (c) if the suggested meaning is to be expressed.
This hypothesis is based on the facts that while the non-raised
sentences (51), (53), and (55), express both literal and suggested
meaning, their raised counterparts fail to do so.
The above hypothesis attributes the failure of the application of
raising to the semantics of the above sentences. However, whether or
not raising should apply does not purely depend on the semantics of
the sentences such as above. Pragmatic conditions play an important role
92
in this decision. For example, the suggested meaning, such as in [b)
(in (51), (53), and (55)) is derivable from the literal meaning in (a)
only if the pragmatic conditions allow for it. If pragmatic conditions
do not allow for the suggested meaning, then sentences (51), (53), and
(55), convey only the literal meaning. Let us consider the following
case: if somebody utters (55) while reporting his experience of watching
birds chirping in the afternoon in a bird sanctuary, then the suggested
meaning, (i.e., "I saw it was morning"), is blocked. In such cases the
process of raising applies and the raised sentences convey the literal
meaning, (i.e., "I saw the birds chirping.").
This property of a sentence to convey suggested meaning (vyagya)
through its literal meaning (vacya) is discussed in detail in the treatises
on Sanskrit poetics in 800 A.C. by Anandavardhana and again in 1100 A.C.
by Manmiata. Two major points discussed in the above treatises are
relevant for the present discussion. (a) Sabdartha (word-meaning)
and vakyartha (sentence-meaning) are the two semantic levels which serve
as the basis for the suggested meaning, and (b) vySgyartha (suggested
meaning) can come about on the basis of the context. It is observed^
that when certain contextual peculiarities (Kunjuni Raja 1963:311)
are already known to the reader then the primary meaning gives rise
to the suggested meaning. For example, a young girl who is fond of
roaming around with men, utters the following sentences, "Mother, there
are no groceries in the house. Tell me what should we do? The day does
not stay like this forever you know." In the context of the character
of the speaker, (i.e., the young girl), the above sentence suggests
that the speaker wants to go out to see her lover. This suggested meaning
is based on the primary meaning of the sentences.
The role of extralinguistic factor(s) in determining the total
meaning of a sentence has been discussed variously in current literature
in linguistics. For example, Morgan (1978) points out that the conventions
of language use determine the meaning of a sentence such as 'Can you
pass the salt?'. In English, the above question conveys a polite request
'Please pass the salt.''. According to Morgan, it is a convention
of the language use that a question such as the above, is used by
native speakers of English to express a polite request.
4.2. Explanation. In the preceding section we have seen that raising
blocks the suggested meaning. Therefore, in order to convey the
suggested meaning raising should be blocked in the sentences such as
(51), (53), and (55). The question arises as to why raising blocks the
suggested meaning. Borkin (1974) and Bach (1977) have already observed
that a sentence which is not raised conveys a wider range of meaning as
compared to its raised counterpart. In what follows I will discuss
two possible hypotheses for the phenomenon.
4.3. Sentence structure and suggested meaning. The first hypothesis
can be described as follows: in Hindi, there is a correlation between
the sentence- structure and the suggested meaning. The loss of suggested
93
meaning in a raised sentence is a consequence of a more general rule
in Hindi, namely, that if a sentence conveys suggested meaning then
any change in the sentence-structure results in the loss of suggested
meaning. Thus, a syntactic process which causes change in the sentence-
structure is blocked when suggested meaning is to be conveyed. Since
raising changes the structure of a sentence, (i.e., it transforms the
sentence into a participial phrase), it is only as expected that it fails
to convey the suggested meaning. The following evidence supports this
hypothesis: syntactic processes such as, nominalization, and passivi-
zation, which change the structure of a sentence are also blocked for
the sentences such as (51), (53), and (55). First, let us consider
the process of nominalization. The primary function of this process
is to transform a sentence into a nominal phrase. The following examples
illustrate this process.
(57) Jan kitab parhta ht
John book read slux.
John reads a book.
Nominalization applies yielding the following sentence.
(58) Jan ka kitab parhna
John of book reading
John's reading the book.
The above process is discussed in details in Kachru (1966) and (1980)
When this process applied the following changes take place: (a)_the_
subject takes the possessive postposition (ka/ke or ki) (i.e., jan ka
'John's' (58)) and (b) the main verb is changed into its infinitive form,
(i.e., parhna 'reading' (58) ) .
When this process applied to (51), (53), and (55) the resulting
phrases (59), (60), and (61) fail to convey the suggested meaning.
(59) andhere ka ghirna mX ne dekha
darkness of surrounding I Ag. saw
(a) I saw the darkness surrotmding.
(b) *I saw the night setting in.
(60) xiske hotho ka hilna mt ne dekha
her lips of moving I Ag. saw
(a) I saw the moving of her lips.
(b) *I saw her trying to speak.
(61) ciriy5 ka cahakna m^ ne dekha
birds of chirping I Ag. saw
(a) I saw the chirping (of) birds.
(b) *I saw it was morning.
94
Examples (59)- (61) point out that (a) the process of nominalization
changes the sentence-structure and (b) the resulting phrases fail to
convey the suggested meaning.
Now let us consider the process of passivization. This process
is discussed in detail in Kachru (1980) and in Pandharipande (1981) .
This process can be illustrated in the following examples.
(62) vah kam nahi karta
he vork neg. does
He does not do the work.
Passive applies yielding the follovd.ng sentence.
(63) us se kam nahi kiya jata
he by vork neg. did go
(a) The work is not done by him.
(b) He cannot do the work.
Notice that passivization causes the following changes in the
sentence-structure: (a) the ex-subject vah 'he' is followed by the
postposition £e 'by', (b) the auxiliary verb jana follows the main verb
and (c) the main verb is in its participial form when the above process
applies to the sentences such as (51), (53), and (55), the resulting
sentences fail to convey the suggested meaning. In what follows I
will present a sentence similar to (51), (53), and (55), in order to
illustrate the effect of passivization. The reason for choosing an
example other than (51), C53) or (55), is that these examples express
non-volitional acts. In other words, (51), (53), and (55), express
acts which are intentionally performed by the agent. In Hindi, a verb
undergoes passive only if it expresses a volitional act (Pandharipande
1978, 1981). Therefore, (51), (53), and (55) are passivizable. Consider
the following examples.
(64) mg ne dekha ki vah nak bh^hg carha raha he
I Ag. saw that he nose eyebrows raise prog. aux.
(a) I saw that he was raising (his) nose and eyebrows.
(b) I saw that he was getting angry.
Notice that (64) conveys both literal (a) and suggested (b) meanings,
Now consider the following example (65) . (65) is the resulting sentence
after the application of passive to (64). Also notice that (65) conveys
only the literal meaning.
(65) mt ne dekha ki us se nak bhJhe carhai ja rahi hi
I Ag. saw that he by nose eyebrows raised go prog aux
(a) I saw that the nose and the eyebrows were being raised
by him.
(b) *I saw that he was getting angry.
95
Now I will present more evidence to support the above hypothesis.
It will be pointed out that if a syntactic process does not change the
structure of a sentence, then the sentence continues to convey the
suggested meaning. In Hindi, the lower clause in sentences such as
(51), (53), and (55), can be fronted. In this case, the structure of the
fronted clause remains intact. Consider the following examples.
{66) mt ne dekha ki bacca ro raha h£
I Ag. saw that child cry prog. aux.
I saw that the child was crying.
(6T) bacca ro raha h£ yah mt ne dekha
child cry prog. aux. this I Ag. saw
I saw it that the child was crying.
When the lower clauses in (51), (53), and (55) are fronted respec-
tively, then the resulting sentences (68), (69), and (70) continue
to convey the suggested meaning.
(68) andhera ghir raha ht yah mt ne dekha
darkness surround prog. aux. this I Ag. saw
(a) I saw it that the darkness was svirrounding.
(b) I saw that the night was setting in.
(69) us ke ho-th hil rahe the yah mt ne dekha
her lips move prog. aux. this I Ag. saw
(a) I saw it that her lips were moving.
(b) I saw that she was trying to speak.
(70) cifiya cahak rahi thi yah m£^ ne dekha
birds chirp prog. aux. this I Ag. saw
(a) I saw it that the birds were chirping.
(b) I saw it was morning.
In this section it was pointed out that if a syntactic process
changes the sentence-structure, then the resulting sentence/phrase
loses the suggested meaning. In contrast to this, if a syntactic
process does not change the sentence- structure then the resulting sentence
continues to convey the suggested meaning.
4.4. More evidence. The relationship between the sentence-structure
and the suggested meaning is not restricted only to Hindi. The following
evidence shows that it is found in the English sentences as well. Let
us consider the example discussed in Morgan (1978 section 4.1). Consider
examples (71) and (72). While (71) is in direct speech, (72) is in the
indirect speech. Notice that the structure of the lower clause is not
96
the same in (71) and (72). Also notice that while the sentence "Can
you pass the salt?" conveys both literal and suggested meanings in
(71), its counterpart in (72) fails to do so.
(71) He said to me, "Can you pass the salt?"
(a) He asked me whether I could pass the salt.
(b) He requested me to pass the salt.
(72) He asked me whether I could pass the salt.
*He requested me to pass the salt.
Examples (71) and (72) point out that the change in sentence-
structure blocks the suggested meaning.
4.5. Topic. Another possible hypothesis for why raising blocks
the suggested meaning is as follows: Davison (1979) suggests that
whether or not an NP (in English) undergoes raising is determined on
the basis of the following condition: 'The raised NP must be a good
topic, and the rest of the complement clause must be a good comment and
express something about the referent of the topic NP. ' Davison (1979)
further claims, 'NPs whose referent is directly known to the speaker
or to which the speaker refers on grounds stronger than just deduction
or supposition about their existence, are NPs which are excellent candi-
dates for being the topic of the utterance they occur. ' Within this
framework the above phenomenon concerning (51), (53), and (55), will
have to be explained as follows: (a) the raised version of (51), (53),
and (55) conveys only the literal meaning because when the literal meaning
is conveyed overt subjects, i.e., andhera 'darkness' (51) , hOth 'lips'
(53), and ciriya 'birds '(55), are topics . Moreover, when (51), (53),
and (55) convey literal meaning, the referent of the subject NP is clear.
(b) When the lower clause conveys the suggested meaning, no NP within
the clause can be said to be the topic but rather, the whole proposition
is the topic. Moreover, the referent of the topics, (i.e., the night
is setting in (51), she was trying to speak (53), and it was morning
(55)) are known on the basis of the conventions of the language use.
Since there is no particular NP which can be labelled as topic, no
NP can be raised out of the lower clause to the higher clause. As a
result, the raised counterpart fails to convey the suggested meaning.
4.6. Problems. The above hypothesis faces the following problems:
(a) it is not clear why the subject NP is the topic when literal
meaning is to be conveyed while it is not when the suggested meaning is
conveyed. Kachru (1980) points out that an informal definition of
topic in Hindi is as follows: if the question 'what?' is asked with
reference to the verb in a Hindi sentence, then the part of the sentence
which answers the question is usually the topic of the sentence.
According to this definition the whole proposition, i.e., the lower
clause will have to be labelled as topic in sentences such as (51),
(53) , and (55) . No particular NP can be treated as the topic. If
this analysis is correct then raising should be blocked irrespective
of whether or not the lower clause conveys the literal/suggested
97
meaning; (b) according to Davison's hypothesis, the NP whose referent
is directly known to the speaker and hearer, constitutes an excellent
candidate for topic which is raised comfortably. However, the following
example shows that even when the referent is directly known to the
speaker/hearer raising is blocked.
(73) . m^ ne dekha ki meri batg usko paresan kar rahi he
I Ag. saw that my talk him bother do prog aux.
I saw that my talk was bothering him.
Raising applies to (73) yielding the following sentence.
(74) *mt ne apni bate ko usko paresan karte hue dekha
I Ag. my (self's) talk him bother doing aux. saw
I saw my talk was bothering him.
In (73) meri bate 'my talk' is the subject NP, the sentence expresses
literal meaning and the referent of the subject NP, i.e., meri bate
'my talk' is clear and yet raising fails to apply, and (c) moreover,
the hypothesis misses the generalization, namely, that suggested meaning
is blocked if the structure of the sentence is altered. In contrast
to this, the former hypothesis explains the phenomenon in (51), (53),
and (55), on the basis of a more general rule which is independently
motivated and is needed in the language.
5.0 Conclusion. In the preceding sections it was pointed out that
(a) in order to provide an adequate analysis of the possessive post-
positions, the progressive construction and Raising, pragmatic conditions
need to be taken into consideration, (b) two major aspects of raising
are observed, (i) raising blocks the suggested meaning and (ii) raising
changes the structure of the sentence ^therefore it fails to convey
the suggested meaning. It was further pointed out that (ii) follows
from a more general rule in Hindi, namely, that any change in sentence-
structure results in blocking the suggested meaning.
The discussion in this paper points out that it is the interaction
of grammar and pragmatics that explains the linguistic facts. However,
at present it is not clear how to formally represent this interaction.
In order to do so we need to devise an empirical criterion to define
'grammatically' determined versus 'pragmatically' determined aspects/
features of language.
The pragmatically oriented hypothesis presented in this paper
is also relevant for language teaching and learning, as Oiler (1969:47)
points out, "The concept of pragmatics has definite implications for
theories of language learning and methods of language teaching with respect
to material construction for instance it indicates that the structures
should be presented in meaningful contexts." The hypothesis discussed
here is also relevant for devising what Wilkins (1977) labels as "notional"
syllabuses" or what is termed as "communicative syllabuses" in Munby
(1978). Such syllabuses are based on the assumption, namely, that
98
"...since those things that are not conveyed by the grammar are also
understood, they too must be governed by 'rules' which are known both
to speaker and hearer. People who speak the same language share not so
much a grammatical competence as a communicative competence" (Wilkins
1977:11). In order to develop communicative competence in students
a pragmatically oriented explanation of linguistic facts is essential.
Finally, this paper points out that the grammar of a language
does not always adequately explain what is conveyed by the speakers
of that language and therefore, ■ pragmatics needs to be incorporated
into the core of the study of language.
NOTES
An earlier version of this paper was presented at the Second South
Asian Language Analysis Roundtable held in April 1979, at the University
of Illinois at Urbana-Champaign.
Locational postpositions are so-called because their primary
function is to indicate location. In English, for example, the locational
prepositions are in_ and at^. The locational postpositions me_ 'in' and
ke pas 'near' in Hindi are functionally similar to the locative case
in Sanskrit, German, etc. In addition to the above, those postpositions
are used to indicate a particular type of possession (cf. section 1.1.).
2
The opposition between abstract versus -abstract can be explained
as follows: the possessions such as son, daughter, book, etc. are
(-abstract), i.e., the above type of possessions have an existence,
independent of the possessor. In constrast to this the abstract posses-
sions are fever, courage, etc. which do not have a existence independent
of the possessor.
An interesting piece of evidence to support this hypothesis comes
from the following: in Hindi, usko khusi h^'he is happy' (literally,
'he has happiness' indicates that he is happy because of some circumstan-
tial reason. If the speaker wants to indicate that he is a happy ty^e
of person, then the speaker uses another construction, i.e., vah khus
mizaz admi h£ 'he is a man of happy disposition'.
4
Example (37) is grammatical in the following two contexts, i.e.,
(a) when the speaker is actually witnessing his friend approaching
death and (b) when marna 'to die' expresses metaphorical meaning
"to work to death" (i.e., to do very hard work).
For more discussion on punctual verbs refer to Lyons (1969) .
99
For example, it is not possible in Hindi for a perfective participle
such as roya hua 'cried' to refer to layka 'boy' (i.e., roya hua larka
'cried-boy ' is not grammatical in Hindi since the boy cannot be said to
be in the resulting state of the act of crying. Similarly, likhi hui
'written' does not refer to the subject, e.g., larki 'girl' (i.e., *citthi likh:
hui layki 'the girl who has/had written a letter'. ). Notice that the girl
in the above sentence cannot be said to be in the resulting state of the
act of writing.
For a detailed discussion on tlie literal (vacya) and
suggested meaning refer to (a) kavyaprakasa(1955) (Chapters 1-3), ed.
by Sukhtankar S. S. (for complete reference. see bibliography) and
(b) Studies in Dhvanyaloka. 1938, by Subbrama Pattar P. S. (for reference,
see bibliography) and (c) Kunjuni Raja K. , 1963, Indian theories
of meaning (for reference. see bibliography).
REFERENCES
AGNIHOTRI, Krishna. 1974. Bat ek Orat ki . Delhi: Indraprasth Prakashan.
AKMAJIAN, A. and Heny, F. 1975. Introduction to the principles of
transformational syntax. Cambridge: MIT Press.
BACH, E. 1974. Syntactic theory. New York: Holt, Rinehart and Winston.
. 1977. Review of Postal: on raising. Language 53.
BORKIN, A. 1974. Raising to object position. Papers in Linguistics.
CHOMSKY, Noam. 1965. Aspects of the theory of syntax. Cambridge:
MIT Press.
DAVISON, A. 1979. Linguistics and the measurement of syntactic complexity:
the case of raising. Reading Center Technical report Metatheory
No. 3. Application of Linguistic Theory in the Human Sciences.
East Lansing: Michigan State University, pp. 1-44.
DONNELLAN, K. S. 1978. Speaker references, descriptions and anaphora.
In P. Cole, ed. : Syntax and semantics. Vol. 9, Pragmatics.
New York: Academic Press.
FAIRBANKS, G. and B. G. Misra. 1966. Spoken and written Hindi.
Ithaca: Cornell University Press.
GIVON, T. 1978. Negation in language, pragmatics, function, ontology.
In P. Cole, ed.: Syntax and semantics Vol. 9, Pragmatics.
New York: Academic Press.
GREEN, G. 1975. How to get people to do things with words: the
whimperative question. In P. Cole and J. Morgan, eds . : Syntax
and semantics. Vol. 3, Speech acts. New York: Academic Press.
KACHRU, Y. 1966. An introduction to Hindi syntax. Urbana, Department
of Linguistics, University of Illinois, mimeographed.
. 1970. A note on possessive constructions in Hindi-Urdu.
Journal of Linguistics 6.1.
. 1980. Aspects of Hindi grammar. New Delhi: Manohar
Publications.
, B. Kachru, and T. Bhatia. 1976. The notion 'subject', a
note on Hindi-Urdu, Kashmiri and Punjabi. In M. Verma, ed.: The
notion of subject in South Asian languages. University of Wisconsin
Publications Series Vol. 2.
100
KUNJUNI Raja, K. 1963. Indian theories of meaning. Madras: Adyai
Library Series 91.
LYONS, John. 1969. Introduction to theoretical linguistics. Cambridge:
Cambridge University Press.
MCGREGOR, R. S. 1972. Outline of Hindi Grammar. Oxford: Clarendon
Press .
MORGAN, J. 1978. Two types of convention in indirect speech acts.
In P. Cole, ed.: Syntax and semantics. Vol. 9, Pragmatics. New
York: Academic Press.
MUNBY, J. 1978. Communicative syllabuses design. Cambridge: Cambridge
University Press.
OLLER, J. W. 1969. Some psycholinguistic controversies. In J. W.
Oiler and J. C. Richards, eds.: Focus on the learner: pragmatic
perspectives for the language teacher. Newbury, MA: Newbury House.
PANDHARIPANDE, R. 1978. Exceptions and rule government: the case
of passive rule in Hindi. Studies in Linguistic Sciences 8.1.
and Y. Kachru. 1977. Relational grammar, ergativity, and Hindi-
Urdu. Lingua 41.
POSTAL, P. 1974. On raising: one rule of English grammar and its
theoretical implications. Cambridge: MIT Press.
RAGHAV, Rangey. 1970. Loi ka tana. Delhi: Rajpal and Sons.
ROSENBAUM, P. 1967. The grammar of English predicate complement con-
structions. Cambridge, MA: MIT Press.
SUBBARAMA, Pattar. 1938. Studies in Dhvanyaloka. Trichur: Mangalo-
dayam Press.
SUBBARAO, K. V. 1974. Noun phrase complementation in Hindi. Urbana,
University of Illinios Ph.D. dissertation. Unpublished.
SUKHTANKAR, S. S. (ed., trans 1., and annot.) 1933. Kavya-prakasa.
Chapters 1-3. Bombay: Bombay Book Depot.
. (ed., transl., and annot.) 1941. Kavya-prakasa. Chapter
10. Bombay: Karnatak Publishing House.
VAJPEYI, K. 1959. Hindi shabdanushasan. Banaras : Rajkamal Prakashan.
WILKINS, D. A. 1977. Notional syllabuses. Oxford: Oxford University
Press.
Studies in the Linguistic Sciences
Volume 11, Number 2, Fall 1982
THE SYLLABLE IN PHONOLOGICAL THEORY:
ARGUMENTS FROM TAMIL
K.G. Vijayakrishnan
Central Institute of English and Foreign Languages, Hyderabad
This paper shows that the syllable plays an important role
in Tamil phonology. It is claimed in Halle and Vergnaud (1980)
that unlike the onset and the rime, the constituents of a
syllable, the syllable itself as a constituent hardly plays any
role in phonological theory, and therefore its relevance to
phonological theory is marginal. This paper shows that the
syllable does play an important role in the description of
morphological and phonological generalizations in Tamil, and
that its role in Tamil phonology is far from marginal.
0. Introduction
This paper is mainly a reaction to the sceptical statement made in
Halle and Vergnaud (henceforth HV) (1980) about the need for the theoreti-
cal construct 'syllable' in phonological theory. They state that unlike
the onset and the rime, the constituents of a syllable, the syllable
itself as a constituent hardly plays any role in phonological theory,
and that therefore, its relevance to phonological theory is marginal. I
will show in this paper that the syllable does play an important role in
the description of morphological and phonological generalisations in
Tamil, and that its role is far from marginal.
I am aware that in the recent literature [Kiparsky (1979), Selkirk
(1980), Prince (1980), etc.], HV's verdict regarding the role of the
syllable in phonological theory has not been unanimously accepted. The
decision to label the onset and rime as W and S respectively makes it
obligatory to recognise the '^ of which these form the binary branches.
I will merely mention in passing that the arguments put forward by
Selkirk (1980) for recognising the construct '" make use of only surface
phonetic phenomena like aspiration in English, and flapping in American
English, thus not refuting the HV stand that *■ is only a surface pheno-
menon, and in that sense, (perhaps) marginal.
1. Preliminaries
Before one can discuss the validity of the construct 'syllable'
in Tamil, one must establish the nature and structure of its components,
the onset, the rime, and the appendix. These constituents have already
been proposed in the literature to date. For instance, English words
like fix and consort have been argued to be syllabified as:
(l)a.
onset rime appendix
f i k s
102
b.
onset rime onset rime
k on sort
I shall assume that the structure of the syllable in Tamil is as given
in (2)/
(2) a. Onset Template for Tamil
Onset
[-syllabic]
C) ( )
b. Rime Template for Tamil
Rime ,,,..,
r+syllabic I
[■^ F J where F is the complete feature
A complex for vowels.
^ ( )
The consequence of (2)a. and b. is that we will have syllabifications
such as illustrated in (3):
(3)a. o r o r
I I I I
p a c i 'hunger'
b. o r 0 r
I /I M
p aa tt i 'grandmother'
c. o r o r
1 I ^ /I
c a mp aa 'a variety of rice'
or or
c e yt
1!
yt 1
I have argued in Vijayakrishnan (in preparation) that lexical entries
in Tamil should specify whether the rightmost segment is an appendix or not.
This information is necessary to capture the systematic differences between
pairs like (4)a. and b.
(4)a. a
k a n 'eye'
b. k a n 'joint of a bamboo'
103
There are several systematic differences between (4a) and (4b) that
motivate the representations given above. (4a), for example, has two
surface variants, kan and kanni, but (4b) has only one, namely, kani.
Similarly, when compounded"]^ (4a) yields forms like katcewi (from /kan/
'eye' and /cewi/ 'ear'), while (4b) gives forms like kantkkaal 'ankle'
(from /kan/ 'joint of a bamboo' and /kaal/ 'leg').
Now that we have established the nature and structure of the
constituents of the Tamil syllable, we are ready to take up the issue
of the role of the syllable in phonological processes in Tamil. In
section 2, I shall show that there is a morpheme structure constraint
in Tamil, the formulation of which has to make crucial use of the con-
struct 'syllable'. In section 3, I shall demonstrate that the process
of gemination in compounds applies only if the first stem obeys certain
conditions which include that it be monosyllabic, and that the relevant
environment for the rule cannot be adequately stated without referring
to the syllable.
2. A Morpheme Structure Condition
It is not clear whether the facts pertaining to possible and
impossible Tamil sequences ought to be stated as a constraint of
possible sequences or as implicational statements. We will not go
into this problem in this paper. We are only concerned with the manner
of stating generalisations pertaining to possible Tamil sequences.
I will look at one morpheme structure condition relating to the
occurrence of branching rimes in Tamil sequences. In Tamil, non-derived
sequences cannot end in non-low branching rimes if the sequence has more
than one syllable. Monosyllables can have branching rimes. Thus, we
have [pu] 'flower', [po] 'go, [pT] 'shit', etc., while [kalapu] , [_tarapo] ,
and [canapT] are not 'possible' Tamil sequences at all.
One might say that this can be stated easily on the string which
projects all and only the rimes of sequences.
•
(5) * [... r r ] Projection: r
[-low]
A
But this is incorrect for two reasons. It would brand the perfectly
acceptable strings in (6)a. and b. as unacceptable, as the rime projection
would ignore the final segments which are not part of the rime.
(6)a. /kareen/ 'a male or female elephant'
b. /kuNpalamooc/ 'fish brought from the Maldives'
c. /titukuut/ 'suddenness'
(7)a.
b.
/ taNpuur/ ' tanpura '
a
/alloon/ 'moon'
104
Therefore, a rime projection would not suffice. We can state this
elegantly using <^ as (8) demonstrates.
(8) M. S. C.
* {...<r- <r-^ ] Projection: <7 '^
[-low] r a
The constraint is now formulated on a projection of the superordinate
ble and the syllable along with the rime,
elements, the starred sequence is unacceptable.
syllable and the syllable along with the rime. In a projection of these
One could argue that it is possible to state the constraint at the
level of onsets and rimes, as * [... r o r]. But it is fairly obvious that
[-l^w]
the sequence o r is used as a notational variant of <»~.
3. Sonorant Onset Gemination
Yet another generalisation that involves the crucial use ofo- is a
process of gemination in compounding. Here, the initial sonorant onset
of the second stem is doubled if and only if the first stem is monosyllabic
and contains a nonbranching rime. Consider (9) and (10):
(9)a. /pu/ /neri/ [punneri]
'conduct'
4
b. /mu/ mati/ [mummadi]
'three' 'lap' 'the three folds on the belly of
beautiful women'
CIO) a. /koo/ /naagam/ [konagam]
'king' 'snake'
b. /tii./ /naattam/ [tTnatram]
•bad' 'smell'
c. /maa/ /nilam/ [manilam]
'big' 'land' 'earth'
d. /puu/ /mati/ [pumadi]
'soft ''lap' 'the soft udders of a cow'
e. /naa/ /makal/ [namagal]
' tongue •' daughter ' 'goddess of learning'
The doubling of the initial onset of the second stem is restricted to
compounds with monosyllabic first stems with nonbranching rimes. If the
initial stem is not monosyllabic, the doubling in question does not take
place.
(ll)a. /ati/ /mati/ [adimadi]
'bottom' ' lap' 'secret place in waist cloth'
b. /koti/ /maram/ [kodimaram]
'flag' 'tree' 'flagstaff
105
We may state the rule of gemination as follows:
5
(12) C — V CC
-TJ
r
As far as I know, rule (12) cannot be formulated without the use of '^
or a notational variant of it.
NOTES
See Vijayakrishnan (in preparation) for a detailed discussion.
2
We can extend this generalisation to low vowels by saying that in
words like [raja] 'king', the final [a] is added later to the stem /raaja/.
3
Onset final sequences will have the structure^, and therefore, when
"' is projected, the final T will be projected without the onset of course.
For instance, /wilayaat/ 'play' will have the structure /a- a- «- _/.
/A /^ /v //
' w 1 1 a y aa t'
mu is related to the form muu as in words like [muunri] 'three'. The
underlying representation for [mummadi] , however, cannot be /muu+mati/,
as there is no general rule shortening the vowel in this environment. I
assume that /mu/ and /muu/ are listed in the lexicon as distinct suppletive
stems. I am grateful to Harold Shiffman for bringing this to my attention.
The rule has a few additional complications which are not relevant
for our purposes.
REFERENCES
HALLE, M. and J.R. Vergnaud. 1978. Metrical structure in phonology.
Unpublished ms. M.I.T.
. 1980. Three dimensional phonology. Unpublished ms . M.I.T.
KIPARSKY, P. 1979. Metrical structure assignment is cyclic. Linguistic
Inquiry 10.421-441.
LIBERMAN, M. and A. Prince. 1977. On stress and linguistic rhythm.
Linguistic Inquiry 8.249-336.
PRINCE, A. 1980. A metrical theory of Estonian quantity. Linguistic
Inquiry 11.511-562.
SELKIRK, E. 1980. The role of prosodic categories in English word
stress. Linguistic Inquiry 11.563-606.
VIJAYAKRISHNAN, K.G. In preparation. Some aspects of the phonology of
Tamil and Tamilian English. Ph.D. thesis. Central Institute of
English and Foreign Languages, Hyderabad.
studies in the Linguistic Sciences
Volume 11, Number 2, Fall I98I
OBJECT-MARKING IN THE HISTORY OF NEPALI:
A CASE OF SYNTACTIC DIFFUSION
William D. Wallace
University of Illinois, Urbana-Champaign
Modern Nepali uses the postposition lai to mark human
direct objects, indirect objects, and dative-subjects. Early
Nepali , however , used a variety of postpositions for each of
these grammatical relations. Therefore, a change occurred in
the encoding strategy for objects and also in the postposition
used. The motivation for this change seems to have been to
treat pairs of objects more alike — dative-subjects and indirect
objects; indirect objects and direct objects. The grammar of
Nepali treats dative-subjects and indirect objects alike as
illustrated by the rule of VOLEX; direct objects and indirect
objects are also treated similarly, there being surface struc-
tures in which these case-roles are ambiguous. The change in
object-marking has diffused through the grammar beginning with
dative-subjects and moving then to indirect objects and then to
human direct objects. The progress of this change shows that
syntactic change diffuses through the grammar gradually, spread-
ing in frequency and number of possible environments.
Introduction
Recent work on syntactic change, such as Chung 1977 and Naro 198I ,
has shown that syntactic change progresses gradually through a grammar;
and usually, syntactic change begins as an adjustment to the grammar to
fit the reinterpretation of surface structures. In this paper, I shall
present evidence on a shift in object-marking in Nepali which supports
this view of syntactic change.
The paper is organized as follows: section 1 presents the object-
marking system of Modern Nepali; and section 2 presents that of Early
Nepali, c. 1350-1700 A.D. ; in section 3, I discuss what motivations there
might be for a shift in object-marking from Early to Modern Nepali; and in
section h, 1 examine possible paths for this change; sections 5 and 6 con-
tain data from the intervening stages of the language; and section 7 ex-
amines what recent grammars have said about object-marking; in section 8,
I discuss the resiilts of this research and "syntactic diffusion."
1. Object-Marking in Modern Nepali
1.1. The object-marking strategy of Modern Nepali is to encode all objects
with the same postposition. This postposition lai is used principally to
108
mark three grammatical relations: definite/human direct object (DO);
indirect object (lO); and dative-subject (KO).-'-
Nepali DOs are unmarked if nonhuman (l); however, if human or emphatic/
definite, DOs are marked with lai (2-3):
(1)
(2)
(3)
mai-le kam gar^
ma gai herchu
mai-le syam-lai here'
ma sipai-lai herchu
mai-le kam-Xai gare
ma gai -lai herchu
lOs are marked with lai whether human or not ;
(U) a. mai-le syam-lai pani die
b. mai-le phul-lai pShi di'?'
'I did the work'
'I see the cow'
'I saw Shyam'
'I see the soldier'
'I did that work'
'I see that cow'
'I gave the water to Shyam'
'I watered the flower'
Essentially then, any object of a transitive verb which is human or
semantically /pragmatically distinctive is marked with lai.
In addition to marking lOs and DOs, lai is used as the marker of
dative-subjects (cf. Masica 1976:159-69; Kachru 1970), NPs which seem to be
the semantic subject of certain sentences but are not the grammatical
subjects as in (5):
(5) a. r'am-lai bhok l*agyo
b. sita-lai dukh cha
c. EfakTar-lai ris uThcha
d. ma-lai apata paryo
e. ram- lai aulo cha
'Ram became hungry'
'Sita is sad'
'the doctor feels angry'
'I had an accident'
'Ram has malaria'
For the purposes of this paper, I shall assume KOs only occur with intransi-
tive verbs; thus, they can be syntactically differentiated from lOs and DOs. 2
(Also, I shall refer to all three grammatical relations as "objects.")
1.2. This system of case-marking nonsubjects is common among South Asian
languages; e.g., Hindi shows the same object-marking strategy with ko: ^
(6) a. r'am-ne phal kh'aya
b . r'am-ne larke-ko dektia
(7)
'Ram ate the fruit'
'Ram saw the boy'
rajn-ne sy'am-ko patra likhS 'Ram wrote a letter to Shyam'
ghore-ko c'ar'a do 'give fodder to the horse.' '
109
(8) a. ram-ko bhuk lagT 'Ram became hungry'
b. pifS-ko krodh aya' 'Father became angry'
c. larkT-ko bukh'ar hai 'the girl has a fever'
So, Nepali is not unusual in having this object-marking strategy.
1.3. The use of this object-marker in Nepali dates back at least a century,
for Grierson (l9l6:lTff-) presents I'ai as the only marker of DOs , lOs , and
KOs , and most of the texts he used to describe the language were written in
the late 19th century. In the passages Grierson quotes as examples from the
Calcutta Auxiliary Bible Society translation (1902), lai appears exclusively
as the object-marker:
(9) a. ani tyes-le tyes-Iai sungur charaunu-l^i "aphna kheta ma paThaiyo
'And he sent him into his fields to feed the swine. '
^^
b. ani tyes-le tiniharu-lai aphnu jivika baDi diyo
'And he divided his possessions between them. '
c. ani tyes-1'ai apugye hunu Xagyo
'And he began to have want. '
Also, in the stories told to R.L. Txu-ner (1921 ) by Gurkha soldiers from the
First World War, I'ai is the only marker of DOs, lOs , and KOs:^
(10) a. tyo turkai-lai paya hunde tesai-1'ai kaTera tes-ko khun
p'ani jhaT khaera tirkha! bujhliune thiyau
'If we had caught that Turk, after killing him we would have
assuaged our thirst in drinking his blood like water. '
b. hamrai topkharfS-le unharu-1'ai dherai goar diyo
'Our guns gave much help to them. '
c. jamadar dalbir-Iai hukum thiyo pacisoTa bakhra kaTnu ko
'Jamadar Dalbir received an order to kill 25 sheep. '
So, Nepali employs a common South Asian object-marking strategy, using lai
to encode DOs, lOs , and KOs, and this strategy has been employed for at
least one hundred years .
2. Object -Marking in Early Nepali
2.1. There is, however, no grammatical reason why in Hindi or Nepali these
three grammatical relations should be identified in surface structures with
a unal case-marking. In Latin, for example, the dative and accusative cases
mark lOs and DOs, respectively, and constructions like KOs in South Asian
languages may appear with a variety of cases including accusative, dative.
110
and genitive. 5 A similar situation exists in Sanskrit: again DOs and lOs
are marked with separate cases , and KO-like structures may be construed
with genitive or dative."
Indeed, Nepali was at one time more like Latin and Sanskrit in its
object-marking strategy, in that DOs, lOs , and KOs were not marked with a
single postposition. Rather, in Old Nepali there were several strategies
used for marking objects, and each of the three grammatical relations had
a different set of markers. Between the time of the oldest NepsLLi records
(c. 13OO-I35O A.D.) and the end of the 19th century, a major change occurred
not only in the surface resilization of the object postposition but also in
the strategy the language uses to mark objects. That is, the marker changed,
and the system of object-marking changed. Essentially, we are dealing with
a change from one arbitrary system to another, although during the progress
of the change these three grammatical relations may have become semantically
grouped together and that may have affected their becoming syntactically
identified.
2.2. Let's look at the earliest examples of object-marking in Nepali. Until
about 1700, the examples are not as complete as we would wish, because
most of the surviving documents are deeds of gift inscribed on metal, and
these texts are very formulaic. For example, human DOs do not appear fre-
quently in these texts. However, it is apparent there are a variety of
strategies available for marking the grammatical relations, and no one
strategy dominates all three.
2.2.1. The earliest example of lai is found marking KOs in the late lUth
century:'^
(11) a. r'amadasa p'gdhy'a tin lay a sudam may'a bhaicha (1398)
'Ramdas Padhya and three others received this gift. '
b. r'amadasa padhya lahi vramavitra may'a bhaicha (l398)
'Ramdas Padhya received this Brahman land gift. '
In the same construction we also find upar instead of lai .
(12) a. sri sri sri sri iri bHanasli-ko agya manupadhya -upar
may a bhaicha (1529)
'As an order of Shri Panch Bhana Shah, Manu Padhya has
received this gift. '
b. hany"! vohora-upra may'a yesi bhaicha (l6olt)
'Hanya Bohora has received such a gift. '
Finally, kana is found very late in this period marking KOs:
(13) jasa vaja-kana laiiu paDa ta tas-ko up'au kaliV (16I+3)
'I shall tell the cure for a hawk which loses blood. '
Ill
2.2.2.' A similar situation, i.e. much variation, exists in the case-
marking strategy for lOs in this period. Very early we find lOs marked
with anusvara , as in (lU):
(lU) a. "apana ucitai kari jayakara paDitai Dhuyan (1336)
'Having made it suitable for himself, may Jayakar Pandit
enjoy this . '
b. meri Janmauti Dhoya akhara padhaya-ka pras'ada golhu .joisY
kanakapatra-ki bhasha pasa kari akry'a chu a(lo) ... (1356)
'Having made the promise of this goldplate to Golhu Joisi
as a favor for teaching me letters and making my horoscope,
I have freed from authority this field ..."
But there are also throughout the period lOs with no surface marking at all:
(15) a. ... siu^arma visnudasa pasa kiy^ chau (lUoH)
'We have made this promise for Shivasharma and Vishnudas.'
b. ark a rlja dinu nahi (1663)
'Don't give it to any other king.'
Inflected pronouns were also used:
(16) ... van pasho inu tin jana-le h'amu diyako cha (l590)
'These three men have given us the forest and hillside.'
And kana is found marking lOs since the l6th century:
(17) a. pHTho mari ragatyaya daicavala lai gau ghara-kana vilo
ra.la surtisahi Thakura-kana saveTTo di paJdca garyako cha (1581)
'Having brought bloodied yogurt-and-rice, having killed a
young he-goat, this (promise) has been sealed having given
a haunch to King Surti Shah Thakur and a share of meat to
the neighborhood. '
b. athav"? mad'S.gni ho ta adda mada-sita cir'ai-ko masu
musa-ko m'asu bhijaunu sikara-kana dinu (16U3)
'If there is indigestion, wet mouse-meat and bird-meat
with liquor ajid give it to the hawk. '
2.2.3. DOs in this period are usually found with no surface marking, and
this includes human DOs :
(18) a. Jo yo kanakapatra bhit"ara-ki bh'ashlC gHala ghalava shosa sos'ava
so "apna dyopitara ekai sai purush"? l^bhinaraka gHala (1356)
'Whoever destroys or has destroyed, steals or has stolen the
promise of this goldplate, he shall put into hell one-
hundred members of his family.'
112
b. apaiia raja mariai becnu (1663)
'Having warned your king, you may sell. '
However, there are also examples of anusvara and kajia for DO marking:
(19) vrahma vishNu ishvara vuddha dharma saTgha eti-lca deva ghale (1356)
'You will destroy such gods as Brahma, Vishnu, Ishvar, Buddha,
Dharma , and Sangha . '
2.2.i*. From these data, we can see there is no unified object-marking
strategy for KOs , lOs , and DOs in this period. DOs appear generally un-
marked, although they may be marked with anusvara or kana. lOs are
usually marked in some way with anusvara, kana, or an inflection, but some-
times lOs remain unmarked. KOs are always marked, and we've seen lai , upar,
and kana used for this purpose. Unfortunately, there is not enough data to
really tell what is the "system" at this stage, nor whether 1350 to I7OO is
a representative period; however, it is apparent that KOs, lOs, and DOs are
not identified syntactically or semantically — they are not considered
"alike."
2.3. Between 1700 and 19OO the syntax of KOs, lOs, and DOs in Nepali con-
verged; and the language developed a unal postposition lai to use with all
three. Nepali object-marking presents a case of syntactic change in which
the encoding strategy of certain grammatical relations has converged and
one postposition has become the dominant marker for encoding all three of
these objects.
In the following sections , I shall give more details on the intermediate
steps in this change and also discuss why and how this change progressed.
The progression of this change is recorded in documents from the l8th sind
19th centuries, so we can examine the chajige in progress. Unlike most ex-
amples of change in progress that have been discussed in the literature
(Chung 1977; Naro 1981), this change has been completed. And so, the
historical evidence can show us how the change progressed and how the vari-
ation which was at one time the norm for object-marking eventually died out.
3. Motivations for Convergence
3.1. We have seen that KOs, lOs, and DOs were at one time syntactically dif-
ferentiated by the use of separate sets of markers, and now they are identi-
fied syntactically by the common marker lai. How these grammatical relations
are alike probably has had some affect on the process of the convergence in
their syntactic encoding. So before looking at the data on this change be-
tween the oldest and modern periods, I'd like to discuss syntactic and
semantic points of similarity and contrast among these three objects to see
whether there are any facts that may explain why Nepali speakers would be
motivated to treat them alike. 9
3.2. For the analysis of syntactic similaj'ities among KOs, lOs , and DOs, I
shall assume there are at least three levels of grammar on which they have
syntactic functions. These I shall refer to as their "underlying function,"
"grammatical relation," and "surface case-role. "10
113
Let us assume that at the underlying level subjects are what sentences
are about, and objects are NPs that verbs associate with subjects. Ignoring
all other constituents, we can have three basic underlying structures:
(20) a. S V b. S 0 V c. S 0' 0 V
0' and 0 are both objects, but they are differentiated because the number
of objects a verb must associate with a subject affects greatly the syntax
of the sentence, and because objects that assume the grammatical relation
indirect object are secondary to direct objects in frequency and sometimes
in syntactic behavior.
On the level of grammatical relations , NPs are defined by functions
like those in Relational Grajnmar (Johnson 197^; Postal and Perlmutter 197^)-
From the structures in (20) we may derive four basic structures:
(21) a. PS VI b'. KO PS VI b. AS DO VT c . AS 10 DO VT
Since the grammatical relations selected depend in part upon the verb,
transitivity has been specified in the verb on this level. (21a) repre-
sents a subject and intransitive verb; (21b') and (21b) represent underlying
SOV combinations. The underlying subject of an intransitive verb becomes
a KO if an 0 is present (excluding equational sentences), and the 0 becomes
the PS. The underlying subject of a transitive verb becomes an AS , an 0
becomes a DO; and in (21c), an underlying 0' becomes an 10. Again, PS and
AS are symbols used mainly for reference.
Finally, the surface case-roles are specified by postpositions, as in
(22) for Modern Nepali.
(22) a. NP-0 VI
b • . NP-ia. NP-0 VI
b. NP-le NP-0/rai VT
c. NP-le NP-lai NP-0/lai VT
3.2.1. First, we shall examine KOs and DOs which seem the least similar
of the three possible pairings. KOs are derived from underlying subjects,
while DOs from underlying objects. DOs remain objects at the grammatical
relation level; KOs are "demoted" to a chomeur relation. KOs are always
marked at the surface , while DOs are marked sometimes , and sometimes they
are not marked. Semantically, KOs are almost always human, but there is no
such restriction on the DO. KOs and DOs don't occur in the same sentence
structures , so the opportunity to perhaps confuse their surface syntactic
roles does not present itself. Also, I know of no rule which operates with
KOs and DOs as its domain. Thus, there are few common points between KOs
and DOs that might motivate Nepali speakers to want to identify them in
surface structure.
3.2.2. lOs and DOs seem fairly similar at all levels. Their underlying
functions are the same — both are objects associated with a verb. They do
have different grammatical relations, and usually they are encoded dif-
ferently, but whatever marking has been possible for DOs also seems to have
114
been possible for lOs , although the opposite is not necessarily true.
Semantically , lOs are usually human, while again there is no such restriction
on DOS.
In Modern Nepali, lOs and DOs are subject to some of the same syntactic
rules, but this may be a consequence of the change in surface marking — I don't
have any evidence to support saying they have always been treated alike
syntactically. However, one structure that is not new can produce surface
ambiguity in the roles of DOs and lOs , that is conjunct verbs of the form
NP + garnu 'to do NP'.^^
In sentences with these conjunct verbs , the noun which appears to be
a direct object of garnu loses its syntactic identity, and so sentences occur
with two objects as in (23):-'-^
(23) a. ram-le ghaxharu has garyo
'Ram destroyed the houses. '
b. das jana-ko sallah-sita naya! kam suru garchau
'We'll begin a new job with the advice of ten people.'
So, in these sentences, the leftmost NP — gharharu or naya kam — should be
considered the direct object.
Such sentences may also occur with human objects:
{2k) a. ma euTi keTi-lai prem garth*?
'I used to love a girl. '
b. ma timi-lai biswas garchu
' I believe you. '
c . un-le buDhe sakal-ma taruni swasni-Xai biha gare
'In his old age he married a young wife- '
Whether these compounds are lexicalized or not, two surface analyses for
these sentences are possible: S DO NP+V or S 10 DO V. The former analysis
is strengthened by the noun conjunct losing its syntactic identity as in
(25), and the latter analysis by structiires in which the separateness of the
noun conjunct is emphasized because the preceding object is not in an
object case as in (26).
(25) dherai gharharu has bhae
many house-pl. destruction occurred-pl.
'Many houses were destroyed. '
(26) ma baDo sukh -ko anubhav gare
I big happiness-of experience did-ls
'I experienced great pleasure.'
115
The two possible analyses would make ambiguous the grammatical rela-
tion of these human NPs. I would not say that this has caused lOs and DOs
to develop similar surface marking, but it is an instance in which the role
of lOs and DOs could be confused, thus weakening their separateness and
strengthening a speaker's hypothesis that they should be treated alike.
3.2.3. As for KOs and lOs , they are very unlike in underlying function.
KOs are derived from subjects, while lOs are derived from objects. Their
grammatical relations are different, and it's only been recently that they
have been commonly marked for surface case-role. Both, however, usually
precede an unmarked NP derived from an underlying object. Semantically ,
both are usually human, and both are involved in the meaning of the sentence
in terms of "acquiring" the underlying object.
In Nepali, there is a rule which appears to treat KOs and lOs alike.
(Actually, "rule" may be too strong a term considering the differences be-
tween these two grammatical relations; perhaps "system of correspondences"
would be better vrntil further research has been done on the data. ) This
rule, O-V lexicalization (VOLEX), makes direct correspondences between
lOs and DOs and KOs and PSs. The first case, with transitive verbs, is
illustrated by the examples in (27-29).
(27) a. ma timi-lai Dh'aT k\ira gardina
b. ma timi-lSi D11B.Tdiria
'I wouldn't lie to you. '
(28) a. us-le aphna pariv'ar-l'a.i rias garyo
b. us-le "Sphiia parivar-lai nSsyo
'He destroyed his family. '
(29) a. b'abu-le fam-lai mar dinubhayo
b. U3,bu-le rain-l'Si m'arnubhayo
'Father beat Ram. '
I won't speculate on the form of this rule, because it's not clear whether
an underlying O-V combination is univerbated, or whether a verb like n'asnu
may be "conjunctivized," splitting into 0 and V. In either case, there is
a surface correspondence between lOs and DOs in sentence pairs related by
VOLEX.
This same rule can also be used to explain corresponding sentence pairs
like those in (30-32):
(30) a. abhagi-lai kHane beTff-ina ris uThcha
b. abh'agi khane bel"a-ma rislal[cha
'The unfortunate man gets angry at meal times.'
116
( 31 ) a • ma-lai sarai bhok l^geko cha
b. ma sarai bhok'aeko chu
'I have become very hungry.'
(32) a. ma-Iai r^t bhari nid parena
b. ma rat bhari nidaena
'I didn't sleep all night.'
Again 0-V combinations are univerbated, or a verb like nidaunu is conjuncti-
vized, but in these data, the verbs are intransitive, so KOs appear with
conjunct verbs and the corresponding NP is a PS with the unitary verb. 13
The similarity between these two sets of data indicates both are
instances of VOLEX. Even more important to our discussion of these gram-
matical relations is the fact that VOLEX operates ergatively: the 0 of
intransitive verbs is the subject, while the 0 of transitive verbs is the
direct object; the "promoted" NP becomes either the intransitive subject
or transitive direct object. And Nepali, like many South Asian languages,
is ergative in the perfect tenses, so the PS and DO would be identified in
surface structure with the same marking — 0. Thus, both KOs and lOs would
correspond to the same surface case-role, the ergative, and so they would
be treated alike by VOLEX.
Nepali, however, is ergative mostly in subject-marking.-'-^ The agent
of the transitive verb controls verb agreement and other subject properties,
and this situation has been present since the earliest records (cf. lUb, 15a).
Older Nepali does have examples in which, like Hindi, the DO controls verb
agreement, as in (33b):
(33) a. c'ar rata p'asKa gayako chan (l590)
'Four red blankets were provided. '
b. ... van p'asho inu tin janS-le hSnu diyako cha (1590)
'These three men have given us the forest and hillside. '
But for the most part, DOs do not control subject properties. On the other
hand, a PS in a dative-subject sentence also does not control subject pro-
perties other than verb agreement as is true of other South Asian languages
(Kachru, Kachru and Bhatia 1976). In (3'+a) the KO controls reflexivization
and in (3^b) it controls conjunction reduction:
(3^) a. un-Iai aphu-mathi daj^ST lagyo
'He felt pity on himself. '
b. masu kliaera ram-Iai ris uThyo
'After eating meat. Ram becajne angry.'
117
While the correspondences between PS and DO are not exact, the domain
of VOLEX still provides an example of how KOs and lOs are treated alike
in the grammar of Nepali.
3.3. In the preceding, I have tried to find explanations in the grammar
of Nepali for why speakers would treat these three dissimilar grammatical
relations alike and so cause the convergence of their syntactic encoding
in the modern language. We have seen that KOs and DCs are not very similar,
except possihly in the area of the semantic feature human; IDs and DCs are
similar syntactically, and there is at least one structure which could lead
to ambiguous aneilyses of their roles in surface structure. KOs and lOs are
similar semantically, and in addition, they are treated alike by the rule
VOLEX. All these points of similarity would be present in earlier stages
of the language, although the conjunct verb structures and VOLEX may have
been developing with the convergence.
These similarities generally represent interpretations based on surface
structures, rather than underlying f\mction. VOLEX would identify KOs and
lOs mainly by marking — in earlier stages, corresponding NPs woiold be marked
with a dative postposition if KO or 10, and 0 if "promoted." Conjunct
verbs are only ambiguous when compared with other surface structures. So,
there are motivations in Nepeai for KOs, lOs , and DOs to be treated alike,
and these depend on their surface syntactic and semantic similarities.
h. Some Predictions
The results of the preceding section tell us that Nepali speakers
associate KOs, lOs , and DOs in the structure of the language, so the con-
vergence in their marking is not just a random change. These results also
make certain predictions about the possible paths object -marking convergence
may follow. We know the strongest bondings of the three objects are KO-IO
and lO-DO. So, the change could spread from lOs to KOs on the one hand and
DOs on the other. Or, we might find a pairing phenomenon, in which KOs and
lOs converge and lOs and DOs converge eventually producing unity among all
three. Or, there could be a chain progression or diffusion of the marking
strategy — KOs shift, lOs follow KOs, and DOs follow lOs. Of all the possible
permutations of convergence patterns, these three are the most likely.
5. Object-Marking 1T00-1TT5
5.1. Now let's look at what actually occurs in the intervening stages. In
this section are presented data from about 1700 to 1775. There is still much
variation in the strategies used to mark KOs, lOs, and DOs; each has a set
of markers which are not identical in content or frequency of usage.
5.2.1. During the l8th century, we find KOs marked with kana (35), lai (36),
upar (37), and tir (38). kana and I'Si are the most common and least re-
stricted in environment; upar emd tir are only found in constructions like
the one illustrated here for each.
118
(35) a. jo tyati na gara us-kana pacamah'ap'gtaka (1712 )
'Whoever doesn't act in this way, he commits the five
great sins. '
b. pviNya-kl garniya-hoi garauNiya-kana puNya adhika hucha (1723)
'He who has merit done has more merit than he who does merit. '
c . daradastura meTi salami aru sabhai-kana parala
so dastura timiharu-le pani cahravanu parala ( 1766a)
'Having done away with the custom, all others receive that
privilege which you had to raise cattle. '
(36) a. papa garauHiya hota ta p'apa garNiya-ISi p'apa adhika h'licha (1723)
'He who sins sins more than he who causes sin to be done. '
b. 25 .fana oD mahal ciThnya-lai ^ta jagyo (l75l)
'25 men were employed to lay bricks. '
c. tesa kurS-ma ta-lai dosha chaina (l770)
'You aren't guilty in this matter. '
(37) ••• abhogya dharala-ka pariyara manikyT jagya dharalT-upar
maya yesi bhaicha (1719)
'Manike and Jage Dharala and the family of Abhoge Dhaxala
have received such a gift.'
(38) 7 rupiy"a agl"! 7 visa rupiyaT 53 Than kapaDa ukildara-tira laTgya (l75l)
"The overseers received 7 rupees each for a total of l^iO rupees
and 53 thans of cloth. '
5.2.2. lOs in this period are foiand marked by kana (39), lai (i*0), and
0 (^1), the first two being the most frequent.
(39) a. tSh"a-mah"a jo visves'vara-ko puj'S gaxa us-kana pacasa
pathi dhana dinu (1712)
'Give 50 pathis of rice to whomever worships Vishveshvar
in there. '
b. vahiNi-kana prasaga garnya (1723)
'to have sex with a younger sister'
c. vaUiya-kana ajnana-maha kshatriya-ko Sdha diNu (1723)
'to give half a kshatriya's share in ignorance to a merchant'
119
d. rogi manushya-kana cilla-ko ajirNa mugi-ka curNa-le harcha (lY52
'The powder of lentils cures indigestion from oily foods
for a sick man. '
e. esfS sri r'Smacandra-kana namaskara gari (l7T3)
'having greeted Ramacandra in this way'
(Uo) a. ... kulo-ko pani sabai-laya vadikana aghika paida-sito
leno (I7li+)
'Take that which was produced first after distributing
to all the water from the canal. '
h. sudra-lai ajnana-ma vaisya-ko adlia ... dinu (lT23)
'to give half the share of a vaishya in ignorance to a shudra'
c. naladuma kaThmaPau-lai dikana ( 175 5a)
'having given Naladum to Kathmandu'
d. tyo hati mahuta-lai lay era (177^)
'having brought the elephant to the Mahout'
(Ul) a. ... suratha saha-le yo resha m'anikya' ra Jagya! dharala
vaks'yo (1719)
'Surath Shah gave this authority to Manike and Jage Dharala. '
b. svaniata prasaga garnu (1723)
'to have sex with one's own mother'
c. hami dheva svaddhachau ( 1766b)
'We shall ask the Dhewas. '
5.2.3. The set of postpositions used to mark DOs is the same as that for
lOs — kana (hz) , lai (1+3), and 0 (hk) . However, unlike lOs , DOs are found
marked by 0 as often as marked by kana or lai .
(U2) a. nagi stri-kana dekhnu (1723)
'to see a naked woman'
b. ... dhanadatyS nevaxa-kana hi Jo Hami-le paJcrikana
visvamitra misra-lai vaksyako ho (1768)
'Having captured Dhanadatya the Newar yesterday, we gave
him to Vishvamitra Mishra. '
c. reTvaMa-kana mSri (1773)
'having killed Havana'
120
(i+3) a. mitra-rSi marnu (1723)
•to kill a friend'
b. kaski-lai choDi gorkKa-le kaKa janu cha (1755^)
'Having left behind Kaski , where can Gorkha go?'
c. testa saidhuva-lai rashya pheri dhamadhuma hola (1767)
'Having placed the rebels there, again there will be a revolt.
(i+U) a. kheta vecano bhaya-dekhi r5.j'a janapano caicha (171^)
'After selling the field, it is necessary to inform the king.'
b. esai-lai chori dinu (1723)
'to give a daughter to someone'
c. yaha-vaTa bhalS manisa paThauchau ( 17 5 5b)
'We shall send a good man from here. '
d. valia-ko mahuta kajnaikana (l77^)
'having hired the Mahout from there'
5.3. The data presented in this section show individual variation in the
object-marking strategies for KOs , lOs , and DOs. All three share kana and
Xai ; lOs and DOs also use 0; KOs occur with upar and tir . lOs and DOs
while similar in postpositions used differ in the frequency of the use of
0. So, we cannot say that any two objects have exactly the same encoding
strategy.
6. Object -Marking 1775-1850
6.1. During the period from 1775 to I85O, we find that KOs, lOs , and DOs
are marked with ISi and kana (DOs are sometimes found with 0); however,
marking strategies vary from text to text. Usually KOs will be marked by
ISi ; lOs by kana and lai in roughly equal proportions ; and DOs are marked
by kana somewhat more frequently than by lai .
This change in the progress of the convergence coincides with a change
in the political and social structure of Nepal, for by 1770, Prithvinarayan
Shah, the king of the city-state of Gorkha, had conquered what is now
western and central Nepal, including the three dominant Newar city-states
of the Nepal Valley. His successors during the next UO years would extend
the boundaries of their kingdom to its present borders. Nepali, as spoken
in Gorkha, would have become the political and dominant social language in
all parts of the country. The effects of this social change may also have
speeded the progress of object-marking convergence.
6.2.1. In the texts from this period we find much variation in the possi-
ble distribution of object-markers. In the Dibyopadesh (c. I78O) of
121
Prithvinarayan Shah, for example, we find IsTi used exclusively for KOs (U5)
and lOs (^^6), while DOs are marked ty lai (U7) and 0 (U8).
(it5) a- ma-lai asarje lagyo
'1 was amazed. '
b. ma-lai yavaTa kura-ko sadhyaha lagi rahecha
'I'm still wondering about this one thing.'
(i+6) a. tina-lai volahaTa-ko ruka! leshana jaisi
'as if to write a summons to them'
b. tin jata-Iai pahaDa am phadarapha pani kasai-le na sholnu
'Don't let anyone open the mountain to those people for
general coming and going. '
(U7) a. usai vel'a-m'a ma pani vimajyau ra bhanu Jaisi
kulanandra Jaisi pujaheri rana-lai DSkya
'At that time I summoned Vimajyau and Bhanu Jaisi,
Kulanandra Jaisi, and Pujaheri Raiaa. '
b. tava raja-le sevaka-lai ghar-ma na m'arnu
'Let not the king beat a servant in the house.'
(^8) teri chori sivar"ajna basnyata-ko choro kehersi vasnyata-lHi deu
'Give your daughter to Kehersi Basnet, the son of Shivaram Basnet.'
6.2.2. In other texts there is more variation in what marking is allowed.
KOs, however, use I'g.i almost exclusively; the only example of a KO marked
with kana that I've found is in the Ramayan of Bhanubhakta from about I85O
(i+9b).
(U9) a. tyo pratyupaltSra-ko ich'a ma-llCl chaina (l823)
'I don't desire the service be returned. '
b. ma-kana ta milyo r'ajya vana-ko
'I found a forest kingdom.' (cited in Bhattarai 19T6:UT3)
lOs (50) and DOs (51) can both be marked by I'ai or kana:
(50) a. kaun taraha-ko cija vastu vanTi diny'a ho
so ma-kana ajna gara (l823)
'Whoever can give me such a thing may command me. '
b. Jo ram-ko vaDhiya katha ma-lai prasna garyathyau tye k»tha
timi-l'Si est'a prakara-le mai-le varNana gary"a (l833)
'I described to you in this way that good story of Ram
about which you asked me. '
122
(51) a. tara ma tes-ISi kahile dekhnu paudainu (1819)
'But I'm not allowed to see her. '
t. aputri putraheru-kana paudacha (l833)
'The childless get sons.'
6.2.3. In the Vetalapancavimsati (Riccardi 1971 )> we can examine KOs , lOs ,
and DOs — from about l825 — in a fairly large corpus. In this work, kana is
used to mark animate objects. 5
(52) a. bikrajnake^ari-kana kasai eka ghora purusha-le hari li
(VF 15-16)
'some terrible being having taken away Bikramakesari '
b. mytaka li ayak'a ra.la-kana dekhi (V25, l)
'having seen the king bringing the corpse'
kana is also used to mark some lOs :
(53) a. kshamti^ila nama aghori-kana praNama gari (VF 151-52)
'having done obeisance to the ascetic named Kshamtishila'
b. bhuta-preta-pisaca ityadi-kana "Snaiada diny'a (VF 125)
'giving pleasure to all sorts of ghosts, spirits, and demons'
On the other hand, lai is used as the marker of other lOs (5^) ajid all
KOs (55).
(5^) a. tasartha timi-lai betala-sadhana-bidya ma dimchu (VF 36)
'Therefore, I shall give you the knowledge of how to
summon vetalas. '
b. esta esta ratna chattisa saya eka 36OI timi-le ma-l^i
caDhayau (VF 70-71 )
'You've offered me 36OI of such jewels. '
(55) a. tasartha tyo bidya ma-lai cahidaina (VF k2)
'Therefore, I don't want that knowledge.'
b. arka-l'Si duhkha parda (VF 75)
'others being sad'
The data from the Vetalapaficavimsati suggest there is some stylistic
variation among texts of this period in the use of lai and kana. Some
texts use I'ai only, others may limit the use of lai or kana to certain
environments. This period represents a stage in the evolution of object-
marking in which speakers recognize the lai /kana variation as a
123
marker of stylistic variation, whereas in the early l8th century, the
distinction appeared to be a lower-level syntactic phenomenon.
6.3. In this section, we have seen that lai is almost dominant for KOs
in this period. lOs and DOs both may be marked with lai or kana, and DOs
may be marginally marked with 0. In this period, it appears there is
stylistic variation in object-marking, because texts vary in their object-
marking systems.
7. kana in Modern Nepali
We have seen that kana was a strong competitor for lai during the l8th
and 19th centuries. Today, it is seldom mentioned as an object-marker,
except in the most complete texts.
16
In recent grammars there is no question that lai is the postposition
to be used with KOs, lOs , and DOs; if kana is mentioned, it is usually
in another context. Turner 1931 glosses kana as being equivalent to lai .
Bhattarai (1976:^473) mentions kana in his discussion of the karma karaka
'object case.' He states that kana is found sometimes only in poetry, but
that it was used in older forms of the language for both poetry and prose.
Verma and Sharma (1979:15) mention kana only in its use as a conjunc-
tive participle:
(56) ma bhat khaikana gae' 'I left after eating."
The conjunctive participle in Nepali is formed by adding -i_ or -era to the
verb stem, but -i+kana is an alternative. It seems to be the fate of kana
to be subsumed in this form, for Sharma (l980b:10U) states that these
participles are "stem + ikana . " When this reanalysis is complete, kana
will have virtually disappeared.
8. A Case of Syntactic Diffusion
From the preceding array of exeimples , we can see that between 1700 and
1900 lai became the dominant object-marker in Nepali, gradually increasing
its frequency and possible environments. We also find that the object-
marking strategy of the language underwent a radical change from a steige
at which each grammatical relation was marked with its own set of postposi-
tions to the current stage in which all three use the sole object postposi-
tion lai. The progress of this change is illustrated in Table A on the
next page. (Table A represents the data as presented here — further research
may suggest a more finely-graded sequence. )
Table A reveals that of the three possible paths of convergence we
predicted for lai -marking the third hypothesis is the correct one. There
is a chain shift pattern in which YKi dominates first KOs, then lOs , and
then DOs. As iSi becajne more dominant, alternative postpositions receded
to marginal usage. Obviously, the change did not complete itself for any
one grammatical relation before moving on to the next; rather, it was
affecting all three in various degrees at any one time.
124
pre-lTOO
1700-1775
1775-1850
post-1850
KO
lai /kana/upar
lai/kana/ (upar ) / (tir ]
lai / ( kana )
lai
10
kana/v/0/(dat)
lai/kana/ (0)
lai/kana
l"ai
DO
kana/V/0
lai/kaiia/0
lai/kana/ (0)
lai / ( kana )
Progress of iSi in Nepali
(Parentheses denote marginal usage)
Table A
Thus, examining the grammar of Nepali at a particular time prior to
1850 should give a picture of much variation in object-marking, like the
views of change in progress presented by Labov 1963 or 1966.
The chronological profile of Table A illustrates the syntactic equiva-
lent of the progress of phonological change in the theory of lexical dif-
fusion. The work of W. S-Y. Wang, C-C. Cheng, M. Chen, and H.I. Hsieh on
Chinese dialects (cf. Chen and Wang 1975 for summary and references) has
shown that a sound change diffuses gradually through the grammar and
gradually expands its domain to a greater number of lexical items, even-
tually applying in all possible environments.
In the history of Nepali, we find "syntactic diffusion" of an object-
marking encoding strategy for KOs , IDs, and DCs. lai is reanalyzed as
the only object-marker rather than a varisuit, and these three grammatical
relations are reanalyzed as being alike syntactically/semantically rather
than being different. These changes diffuse through the grammar gradually
over a period of two or more centuries , as lai expands its domain from all
KOs to all IDs to all hijman DCs. This expansion was motivated by the gram-
mar tending to treat KOs, lOs , and DOs alike causing convergence of all
three. D\iring the period of diffusion, syntactic correspondences like
VOLEX and surface structures like conjunct verbs occur which allow speakers
to identify KOs with lOs and lOs with DOs, thus creating new environments
for the lai -marking rule.
As pointed out by Chung 1977 » syntactic change involves Just such a
series of discrete changes gradually actualized in the grammar. The stages
of Nepali history represented in Table A clearly show that in the l8th and.
19th centuries (and perhaps before) the language was in a state of transi-
tion as the change in object-marking strategy diffused through the grammar.
Similar progressions with historical stages have been shown for the Poly-
nesian shift from accusative to ergative morphology (Chung 1977) and the
loss of verb agreement in Brazilian Portuguese (Naro I981). Also, in
Nepali, Polynesian, and Brazilian Portuguese, the syntactic changes resolve
clashes between surface analyses and grammatical outputs. Thus, syntactic
change begins as a solution to surface reinterpretation and proceeds gradu-
ally, expanding its scope to more and more environments.
125
In summary, I have shown that there has "been a shift in the object-
marking strategy of Nepali, that the motivation for this shift seems to
have been to treat certain grammatical relations identified in surface
structures more alike, and that the change diffused gradually through the
grammar, expanding its domain over a period of several centuries.
NOTES
-'-For further information on the uses of Tai , see Verma and Sharma
(I979:38ff.)
I've also assiomed that one can always tell the difference between
lOs and DOs; especially among speech act verbs, there may be alternate
judgements.
^For further examples, see Central Hindi Directorate (l975:38ff.)
^In presenting his examples , Turner uses the Royal Asiatic Society
transcription plus modifications. The data in (lO) have been normalized
with Roman transcription.
5cf. Hale and Buck (1966): accusative of feeling $352; dative of
agency $373; dative of possession $37^; dative of direction $362-6k.
Cf. Delbriick (1968): genitive/dative of sickness and illness
pp. 5-6; genitive/dative of possession pp. ikk-k^ , 16I-62.
7Most of the Nepali data is from short selections or inscriptions in
Fokharel 1975' These have been referenced by date only ajid are listed in
the primary sources by date also. In each example, the relevant grammatical
relation is underlined. The data is given in strict romanization of
Fokharel 's text.
o
These constructions may also be found with an agent phrase:
a. yiti kshek'S mah"araja-le mahya bhaicha (l59l)
'Such a compound has become a gift through the king. '
b. mxiktisahi silimsahi mahara j a- vaTa maya bhaicha (l59l)
'This is a gift from King Mukti Shah and King Silim Shah.'
-^The grammatical theory in which we examine these objects may influ-
ence how similar or how different we believe these three to be. A graimnar
of Old Nepali wo\ild not be motivated to group KOs , IDs, and DOs together,
while a grammar of Modern Nepali would find such a grouping valuable. I
doubt that my analysis will avoid such problems, but it provides a useful
framework in which to discuss these objects without a dependency upon
surface roles. It is not intended to explain all areas of Nepali syntax.
126
-'-'-'Shibatani 19TT has shown that certain rules of Japanese refer
crucially to grammatical relations, while others refer crucially to surface
case-roles; thus, any grammar must take into account more than one level
of NP-hood. Although I shall not be discussing such rule domain here,
this method of analysis can yield some insights into similarities among
the three Nepali NPs .
l^Cf. Kachru 1978 for conjunct verbs in Hindi.
-'-^The following sentences in this section are taken from Abdulky 197^:
p. 69 {30b); p. 70 (30a); p. 83 {3Ua); p. 85 (2lia), (2l+b), (26); p. 86
C25); p. 88 (31a), (31b), (32a), (32b); p. 121 (23b); p. 122 (2l+c). She,
too, uses a phonetic transcription which I have normalized to conform with
my other data.
-'■^Although Abdulky 197*+, Turner 1931, and Verma and Sharma 1979 gloss
sentences with pairs like bhok lagnu and bhokaunu/ ris uThnu and risaunu
with the same meaning, there is probably some difference in their semantic
interpretations or pragmatic uses or both. As pointed out to me by
Rajeshwari Pandharipande , such pairs in Marathi do show such differences.
This would mean that the correspondences represented by VOLEX only occur
on the syntactic level, and that on a semajitic or pragmatic level these
sentences would not correspond. Further research will reveal what the
relationship between these pairs on all levels really is.
-'-^For discussions of ergativity in Nepali, see Abadie 197^, Kachru
and Pandharipande 1978, and Sharma 1980a.
-'-5selections from the Vetalapaflcavimsati are from the Frame Story (VF)
or Story 25 (V25); the line numbers refer to Riccardi's text.
l^The equivalent of kana is still in use in the Western Pahari lang-
uages of Kumauni and Garhwali as the dative-accusative marker, as reported
by Grierson I916.
REFERENCES
Primary Sources
PSV 3-5
PSV 7-13
PSV 21-22
PSV 23-25
PSV 33
PSV 31+
PSV 35
PSV 36-38
PSV 39-^0
PSV 75-82
PSV 83-8I1
1336
Punyamalla
Copperplate
1356
Prithvimalla
Goldplate
1398
BaliraJ
Copperplate
lUoit
Medini Varma and Baliraj
Copperplate
1529
Bhana Shahi
Edict
1581
Stirti Shahi Thakur
Copperplate
1590
Three Jaisi
Copperplate
1591
Arjanya Buda
Edict
1601+
Sangram Shaiii
Copperplate
I61t3
Anon.
Bajpariksha
1663
Jagatprakash Malla
Edict .
127
1712
Raghau Jaisi
Inscription
PSV 93-9it
lTli+
Bhupatandra Mall a
Copperplate
psv 95-96
1719
Surath Shah
Deed
PSV ioU-105
1723
Premnidhi Pant
Prayascit Pradip
and Nripasloki
PSV 108-23
1751
Sudarshan Shah
Copperplate
PSV 135-38
1752
Gangavishnu Dv:
-Ja
AjirNamangari
PSV 11+0-U5
1755a
Jayaprakash Malla
Duty Letter
PSV 156-57
1755b
Prithvinarayan
Shah
Letter to Pandit
Rajvalocan
PSV 151+-55
1766a
Prithvinarayan
Shah
Order
PSV 158-59
1766b
Prithvinarayan
Shah
Letter to Hari Pandit
PSV 160-61
1767
Prithvinarayan
Shah
Letter to Hari Pandit
PSV 16U-65
1768
Prithvinarayan
Shah
Edict
PSV 166
1770
Prithvinarayan
Shah
Letter to Abhiman
Simha
PSV 167
1773
Anon.
AJirNamanjari
PSV 1U9-52
177i*
Prithvinarayan
Shah
Letter to Abhiman
Simha
PSV 168-69
1780
Prithvinarayan
Shah
Dibyopadesh
PSV 17U-88
1819
Munsi
Three Stories
PSV 298-300
1823
Anon.
Mahabharat Sabhapurva
PSV 3k6
1833
Anon.
Ramashvamedha
PSV 351-52
Secondary Sources
ABADIE, Peggy. 197^. Nepali as an ergative language. Linguistics of the
Tibeto-Burman Area 1.156-77-
ABDULKY, Vicki June. 197^. A formational approach to the semantic
structure of Nepali. Ph.D. dissertation. Cornell University.
BHATTARAI, Rohiniprasad. 1975- Virhud Nepal i-Aryakar ana. Kathmandu:
Nepal Rajkiya Prajha-Pratisthan.
CENTRAL HINDI DIRECTORATE. 1975- A basic grammar of Modern Hindi.
New Delhi: Ministry of Education and Social Welfare.
CHEN, M.Y. , and W. S-Y. Wang. 1975- Sound change: actuation and imple-
mentation. Language 51.255-81.
CHUNG, Sandra. 1977- On the gradual nature of syntactic change.
Mechanisms of syntactic change, ed. by Charles N. Li, pp. 3-55.
Austin: University of Texas Press.
DELBRUCK, Bertold. 1968 (1898). Altindische syntax. Darmstadt:
Wissenschaftliche Buchgesellschaft.
GRIERSON, G.A. 1916 . Linguistic survey of India, Vol. IX: Indo-Aryan
family, central group, Part IV. Delhi: Motilal Banarsidass.
HALE, William Gardner, and Carl Darling Buck. 1966 (1903). A Latin
grammar. University: University of Alabama Press.
JOHNSON, David E. 197^. Toward a theory of relationally based grammar.
Ph.D. dissertation. University of Illinois at Urbana-Champaign.
KACHRU, Yamuna. 1970. The syntax of ko_-sentences in Hindi-Urdu.
Papers in Linguistics 2.299-31'+.
. 1978. Conjunct verbs: verbs or verb phrases? Proceedings of the
Twelfth International Congress of Linguists, ed. by Oskar E. Pfeiffer
and Thomas Herok, pp. 366-70. Innsbruck: Innsbrucker Beitrage
zur Sprachwissenschaft.
128
KACHRU, Y., B.B. Kachru, and T.K. Bhatia. 1976. The notion of 'subject':
a note on Hindi-Urdu, Kashmiri, and Panjabi. The notion of subject
in South Asiaji languages, ed. by M.K. Verma, pp. 79-108. Madison:
University of Wisconsin.
KACHRU, Y., and R. Pandharipande. 1979- On ergativity in selected South
Asian languages. South Asian Language Analysis 1.193-209.
LABOV, William. I963. The social motivation of a sound change. Word 19.
273-309.
. 1966. The social stratification of English in New York City.
Washington, D.C. : Center for Applied Linguistics.
MASICA, Colin. 1976. Defining a linguistic area. Chicago: University
of Chicago Press.
NARO, Anthony J. I98I. The social and structural dimensions of a
syntactic change. Language 57.63-98.
POKHAREL, Balkrishna. 1975- Pafic shaya varsha. Kathmandu: Sajha Prakashan.
POSTAL, P.M., and D. Perlmutter. 197^. Relational grammar. Unpublished.
RICCARDI, Theodore. 1971- A Nepali version of the Vetalapancavimsati .
New Haven: American Oriental Society.
SHARMA, Tara Nath. 1980a. The auxiliary in Nepali. Ph.D. dissertation.
University of Wisconsin at Madison.
. 1980b. Class notes from Beginning Nepali, Summer I98O. University
of Wisconsin at Madison. Unpublished.
SHIBATANI , Masayoshi. 1977. Grammatical relations and siorface cases.
Language 53.789-809.
TURNER, R.L. 1921. Specimens of Nepali. Indian Antiquary 50.8i|-92.
(Reprinted in 1975— R.L. T\irner Collected Papers 1912-1973,
London: Oxford University Press.)
. 1931. Dictionary of the Nepali language. London: Kegan Paul,
Trench, Trubner and Co. , Ltd.
VERMA, M.K., and T.N. Sharma. 1979- Intermediate Nepali structure.
New Delhi: Manohar.
studies In the Linguistic Sciences
Volume 11, Number 2, Spring 1982
TRANSPLANTED SOUTH ASIAN LANGUAGES;
AN OVERVIEW
Tej K. Bhatia
Syracuse University, Syracuse
INTRODUCTION
Research on transplanted languages (languages of the immigrants)
represents one of the newly emerging and rapidly growing areas of socio-
linguistic research. Stage for such research was set by the works of
four scholars, Werner Leopold, Einar Haugen, Uriel Weinreich and William
Mackey in the area of bilingualism and language contact in the late 1940' s
and early 1950' s. Consequently, a large body of sociolinguistic literature
on immigrant languages grew in the sixties and seventies (for references
to works of the above four scholars and works on transplanted languages,
see Haugen 1978). Although a great deal of work has been done on this
topic in the past two decades, the scope of these studies is restricted
to transplanted European languages in the United States alone. As regards
the immigrant South Asian languages in the USA, no serious (socio)-lin-
guistic work has been carried out as yet.
Several South Asian (henceforth, SA) languages have been transplanted
around the world as a result of migration from Asia. The languages which
have gained considerable importance in their new setting are the following:
Tamil in South-East Asia; Gujarati in Africa and North America; Punjabi in
Western Europe and North America; Hindi in Western Europe, South and North
America, the Caribbean and the islands of Fiji and Maritius.
The work on transplanted SA languages did not begin until the seventies.
In India, the work on this topic began as part of a campaign to promote the
image of Hindi and Tamil as 'international languages'. Therefore, world
conferences of Hindi and Tamil were organized in India and sometimes out-
side India. What emerged from these conferences were proceedings which
contained reports on language maintenance and the teaching of these two lan-
guages around the world. Consider, for example, the proceedings entitled
'Visva Hindi Darsan' (1975) which was released on the eve of the first
World Hindi Conference held in Nagpur, India in 1975. "^hese proceedings
have a separate section on Hindi around the world. This section presents
the status of Hindi in South Africa, Fiji, Trinidad, Surinam, Singapore,
and the teaching of Hindi in several countries of Europe, North America,
Asia and Africa. These reports are very brief (ranging from half a page to
two pages) and have missionary overtones. Although these reports did not
initiate any serious linguistic research on transplanted SA languages,
nevertheless, they did generate some interest and awareness about these
varieties, mainly in India, and opened a new channel of communication be-
tween Indian scholars and the speakers of transplanted SA languages. As a
result, some journalistic articles and even one or two dissertations on the
130
non-native literatures of these varieties were written.
As regards other SA languages, namely Gujarati and Punjabi, no work
has been done either in India or abroad. Only some remarks are offered in
the studies on East Indians abroad. Also, to the best of my knowledge, no
work has been done on transplanted Tamil varieties either. Therefore, in
what follows I will restrict myself to the transplanted varieties of Hindi
alone. It should be added that a large part of work on transplanted Hindi
has been carried out primarily in the United States.
RESEARCH ON TRANSPLANTED HINDI: ITS NATURE AND SCOPE
The work on transplanted varieties of Hindi is still in its infant
stage and can be grouped into two classes: secondary and primary studies.
Before I detail the nature and scope of these studies, a note of caution is
in order. In venturing to write on this topic, one is at once made con-
scious of a number of difficulties. In the first place, most of the stud-
ies on this topic are still in their unpublished forms. Secondly, the geo-
graphical area covered by the present study is quite vast. In such cases,
the danger of omissions - especially the omission of localized studies - is
quite great. In spite of the fact that all major and important studies
have been included in this work to the best of my knowledge, I am not claim-
ing a comprehensive survey. Also, this study does not include the papers
presented on transplanted languages during the 2nd South Asian Language
Analysis Roundtable held at the University of Illinois simply to avoid re-
petition since some of those papers have been included in the present volume.
SECONDARY STUDIES
Some of the earlier remarks about the variety of Hindi spoken in Fiji and
the Caribbean appeared in a number of studies on Indian Indentureship by
Verrill (1931), Neihoff and Neihoff (1960), Klass (1961), Burns (1963),
Wood (1968), Tinker (1974), LaGuerre (1974) and Gillion (1977). In addi-
tion, some travelogues and missionary memoirs such as by Moister (1883),
Kinsgley (1871), Grant (1923) contain valuable linguistic information.
These works are not only important for historical reasons but also serve
as excellent sources for sociolinguistic and linguistic information such as
historical background about the number of SA languages and their speakers,
language attitudes, lexical borrowings, language variation, literacy and
the role of Hindi in its transplanted setting.
PRIMARY STUDIES
The primary studies can be classified into three groups: (I) Descrip-
tive Linguistic, (II) Pedagogical and (III) Comparative Studies.
DESCRIPTIVE LINGUISTIC STUDIES
Descriptive studies marked the beginning of linguistic and sociolin-
guistic studies on transplanted Hindi. These studies appeared primarily in
the form of unpublished research papers and unpublished Ph.D dissertations.
Domingue (1971) represents the first systematic and detailed attempt in
this area on Maritius Hindi. To this date, it continues to be the only lin-
guistic description of the Hindi spoken in Maritius. This study, which is
131
primarily restricted to loan phonology, discusses the various manifesta-
tions of Creole English and Standard Hindi in Mauritian Bhojpuri. In
addition to this, the study also presents a brief discussion of gramma-
tical innovations in this variety of Hindi. This work was followed by
Mohan's excellent study (1978) on Trinidad Hindi. Mohan (1978) is pri-
marily a morphological study of the first generation Trinidad Rhojpurl.
The study is intended to serve as a reference grammar of the speech of
fluent Trinidad Bhojpuri. It attempts to account for the most conserva-
tive variety of Trinidad Hindi which is spoken by middle aged and elderly
Indians in rural Trinidad. Twenty five native speakers of Trinidad
Bhojpuri between the ages of sixty five and ninety five years of age
served as subjects. Needless to say, this study represents the only sys-
tematic grammar of Trinidad Hindi. A Ph.D dissertation on Guiana Hindi is
currently in progress, which aims at the language attrition in Guiana.
A small number of research papers are also available on the topic
under consideration. Siegel (1975), Moag (1978) and Tiwari (197 9) deal
with Fiji Hindi. Siegel's study outlines the structure of Fiji Hindi and
brings out some salient characteristics of Fiji Hindi. Moag's paper pre-
sents a sociolinguistic profile of Trinidad Hindi together with the lin-
guistic adaptation of Trinidad Indians. The primary focus of this paper is
'to recount the development of the two outstanding traits of the Fiji In-
dian Community, namely it's linguistic unification and its retention of a
communal language as a badge of cultural identity. In addition, the paper
also presents a discussion on the linguistic maintenance of Fiji Indians
and the indigenization of Fiji Hindi.' The paper points out the develop-
ment of special contact codes (Pidgin Hindi and Pidgin Fijian) by Fiji
Indians for communicating with non-Indians in the new setting. Tiwari
(1979) notes the presence of Bhojpuri, Avadhi, Standard Hindi and 'Sahibi
Hindi' elements in Fiji Hindi and concludes that in spite of some idiosyn-
cratic properties, Fiji Hindi essentially derives its grammar from
Standard Hindi. ^
Durbin's paper (1973) represents the first linguistic study which
brings out the formal syntactic changes in Trinidad Hindi. The explana-
tion of the changes are sought in the socio-cultural changes of the East
Indian Community that have occurred since the East Indians' arrival in
Trinidad. The paper demonstrates that the socio-cultural changes in the
community have influenced the range of functions of the language which in
turn has determined the direction of structural changes in the language.
The only work of linguistic nature on Surinam Hindi has been done
by Bosch. Bosch (1978) addresses itself to the problem of tense marking in
the embedded sentence in Surinam Hindi. The other two studies (1978a and
1978b) focus on syntactic aspects such as verb phrases and participant
reference in that language. The only account of transplanted SA languages
in Africa is presented in Neale (1974) .
PEDAGOGICAL STUDIES
Two important works of pedagogical nature are the texts on Fiji and
Surinam Hindi by Moag (1977) and Huiskamp (1978), respectively. Besides
the exhaustive nature of the Moag text, the other important features of
this text are: the use of the Devanagari script together with the Roman
132
transliteration and detailed grammatical notes. The planning of lessons
utilizes the concept of sequencing, which involves selection, grading,
presentation and testing. The text compiled by Huiskamp constitutes a
second volume of the series 'Languages of the Guianas'. This text is a
part of an audio-visual course in Surinami Hindi which is modelled after
"First Things First", an English audio-visual course for beginners by L.G.
Alexander (London: Longsman Group Ltd, 1967). The course contains twenty
lessons which are followed by word index, grammatical 'comments' and index.
Following the preface is a set of instructions for students and instruct-
ors regarding 'how to make the best use of this course'. The value of the
course as a pedogogical text is not much. The text emphasizes the ac-
quisition of lexical items. The grammatical notes are very sketchy and the
transliteration scheme is based on Dutch. There is no text written to
teach Trinidad Hindi as a second language. Recently, the Sanatan Dharma
Sabha of Trinidad has introduced some texts for school-age children. The
texts are modelled after the children's textbooks of the National Council
of Research and Teaching with two notable changes: One, more indigenous
vocabulary is introduced in the texts and two, the English transliteration
of the vocabulary presented in the Devanagari Script is also given.
COMPARATIVE STUDIES
The only work of the comparative nature has been done by Durbin (1977).
This work addresses itself to the problem of 'identification of a language'
in general and the language of Trinidad East Indians in particular. The
study notes that the identification of the specific languages brought by the
Indian ancestors of the contemporary East Indian population remains
'guess work'. Employing objective comparative methods, Durbin attempts to
identify the speech of Trinidad East Indians.
CONCLUSION
The transplanted varieties of Hindi grew in a linguistically alien,
socially unfavorable and sociolinguistically complex environment. Neither
its speakers nor the speakers of dominant languages in the area paid any
serious attention to these varieties. In spite of the growing interest in
the investigation of immigrant languages, transplanted varieties of Hindi
and other SA languages still continue to be a neglected area of investiga-
tion. As such, the primary aim of the investigations carried out in the
area of transplanted varieties of Hindi has naturally been to present some
important demographic information about these varieties and to address them-
selves to the question of language identification. In addition to this,
these studies devote considerable attention to the adequate linguistic de-
scription of these varieties and to the indigenization of Hindi in its new
settings. The other concerns of these studies have been language mainten-
ance and problems of language pedagogy. In short, the primary concerns of
these studies have been four fold: (I) demographic information, (II) lan-
guage identification and description, (III) language contact and change,
and (IV) language pedagogy.
Needless to say, there are several areas open for future research in
this field of investigation. A great deal of work remains to be done on
topics such as linguistic variation, contiguous generational speech, Ian-
133
2
guage attrition and language death, language maintenance, functional range
and typology of the transplanted varieties of Hindi.
NOTES
This point is open to further investigation.
2
Currently a project dealing with Trinidad Hindi: a dying language
is in progress.
REFERENCES
BOSCH, Tinke. 1978. The distribution of verb forms in Saranami Hindustanie
narratives. Lenguas de Penama.
. 1978a. Sentence introducer in Saranami Hindustanie. Lenguas
de Penama.
. 1978b. Participant reference in Saranami Hindustanie. Lengues
de Penama.
BURNS, Sir Allen. 1963. Fiji. London: Her Majesty's Stationary Office.
DOMINGUE, Nicole 7. 1971. Bhojpuri and Creole in Maritius: a study of
linguistic interference and its consequences in regard to synchronic
variation and language change. Ph.D dissertation. University of
Texas at Austin.
DURBIN, Mridula A. 1973. Formal changes in Trinidad Hindi as a result of
language adaptation. American Anthropologist 75:5 1290-1304.
. 1977. An identification of a language: a methodology.
In Paul J. Hopper (ed.). Studies in Descriptive and Historical Lin-
guistics. Amsterdam: John Benjumins.
GILLION, K.L. 1977. The Fiji Indians, Challenge to European Dominance
1920-1946. Canberra: Australian National University Press.
GRANT, Kenneth James. 1923. My Missionary Memories. Halifax: The Imperial
Publishing Company.
HAUGEN, Finer. 1978. Bilingualism, language contact, and immigrant lan-
guages in the United States. A research report. In Joshua A. Fishman
(ed). Advances in the Study of Societal Multilingualism. The Hange:
Mouton.
HUISKAMP, A.B. (compiler). 1978. Soeroe Se Soeroe Kar : An audio-visual
course in Sarnami Hindustani for Beginners. Part-one . Paramaribo-
Zuld: Summer Institute of Linguistics.
KINGSLEY, Charles. 1871. At last: A Christmas in the West Indies. New York
and London: Macmillan and Company.
KLASS, Morton, 1961. East Indians in Trinidad. New York: Columbia University
Press.
LA GUERRE, John G. (ed.). 1974. Calcutta to Caroni: The East Indians of
Trinidad. London: Longman Caribbean.
MOAG, Rodney F. 1977. Fiji Hindi. Canberra: Australian National University.
. 1978. The linguistic adaptation of the Fiji Indians. To
appear in Vijay Mishra (ed.) Rama's Banishment: A Centenary Volume on
the Fiji Indians. Heinemann: Australia.
134
MOHAN, Peggy R. 1978. Trinidad Bhojpuri: a morphological study. Unpublished
Ph.D dissertation. University of Michigan, Ann Arbor.
MOISTER, W. 1883. The West Indies; Enslaved and Free. London, Hayman Brothers
and Lilly.
NRALE, Barbara. 1974. Kenya's Asian Languages. In W.H. Whitley (ed.).
Language in Kenya. Nairobi: Oxford University Press.
NEIHOFF, A. and J. Neihoff. 1960. East Indians in the West Indies. Milwaukee:
Public Museum.
SIEGEL, Jeffery. 1975. Fiji Hindustani. Working Papers in Linguistics.
Honolulu: University of Hawaii. 7:3 127-144.
TINKER, Hugh. 1974. A New System of Slavery: the Export of Indian Labor
Overseas 1830-1920. Oxford: Oxford University Press.
TIWARI, B.N. 1979. Fiji bat. (in Hindi). Gagaganacal, Fiji Visesink no. 4.
pp 97-100.
VERRILL, A.H. 1931. West Indies of Today. New York: Dodd , Mead and Company.
Vi/va Hindi Darsan. 1975. Nagpur : Visva Hindi Sammelan.
WOOD, Donald. 1968. Trinidad in transition: The Years After Slavery. London:
Oxford University Press.
Stu d ieti 1 n_t he Linguistic Scl encea
Volume 11, Number 2. Spring l"982
TRINIDAD HINDI: THREE GENERATIONS OF A TRANSPLANTED VARIETY
Tej K. Bhatia
Syracuse University, Syracuse
The aim of thia paper is to study the development of Hindi
in the transplanted environment of Trinidad and Tobago by
examining the formal and functional changes undergone by that
language in three generations of Trinidad Indians. The paper
is primarily divided into two parts. The first part presents
linguistic and soclolinguistic settings of Trinidad Hindi
(henceforth, TH) . The second part is devoted to the salient
linguistic features of TH and generationatlonal similarities
and differences in TH.
In order to achieve the second goal, the speech of three
generations of Hindi speakers is analyzed. The data consists
of a text of about seven to ten hours of taped conversation
among ten Trinidad Indians, representing three generations.
In addition, data on retroflexion and aspiration was also
collected by means of giving a production and perception test,
consisting of thirty occurrences of retroflex consonants and
fifteen occurrences of aspirated consonants to approximately :
thirty-five subjects.
Our investigation has shown that contrary to the diffident
characterization of their competence in Hindi by the speakers of
Trinidad Hindi, their competence goes beyond 'a few words'.
Although the third generation has suffered heavy language
attrition because of adopting a more simplified version of
the older generation's grammar, the loss of some features and
the retention of others is not arbitrary. They tend to retain
only the prestigious or more standard forms. This marks a
radical departure from the first generation which tends to
favor native dialect forms.
0. INTRODUCTION
In recent times, a new dimension which Is frequently termed as
'language death/attrition' has been added to the studies on language contact
and bi-(multl-) llnguallsm. The two main settings in which the phenomenon
of language death/attrition has been freauently studied in recent
sociolinguistic literature are the transplanted /immigrant languages and the
indigenous minority languages. Some noteworthy studies on this topic are by
Dressier (1972), Dorian (1973, 1980) and by several others (See Dressier
and Wodak-Leodolter (1977) ). These studies primarily dwell on the various
sociolinguistic aspects of language death/attrition. The present study is
a step in this direction by way of examining the development-'-/decay of Hindi
136
in the transplanted environment of Trinidad and Tobago. The paper examines
the formal and functional changes undergone by that language in three gen-
erations of Trinidad Indians and is primarily divided into two parts. The
first part pre&ents the linguistic and sociolinguist ic setting of Trinidad
Hindi (henceforth, TH) . The second part is devoted to the salient linguis-
tic features of TH and generational similarities and differences in TH.
In addition to providing a necessary background for TH, the inclusion of
the first part is also crucial to gain understanding of why TH developed the
way it did and why there is a discrepancy between the 'self-reported' com-
petence and the 'actual' competence on the part of TH speakers.
I. LINGUISTIC AND SOCIOLINGUISTIC SETTING OF TH
Trinidad identifies itself as a monolingual nation with English as its
only official language. The failure to recognize other languages has
resulted in serious linguistic neglect. This is seen in the fact that
even basic questions such as how many languages are spoken in Trinidad and
what the number of their speakers is remain either unanswered or inadequate-
ly answered. Pointing out this serious gap, the West Indies Census
(1946) noted that
"In British Guiana and in Trinidad and Tobago the census did not
Investigate language. This would seem to have been an important
omission, since the high illiteracy there among the East Indians
points to an unf amiliarity with the English language and, it may
be concluded that languages other than English are in wide spread
current use both among East Indians and West Indians."
West Indian Census (1946:43-44)
Inspite of this reminder, such omission continues and information on lan-
guages is not collected to this date in Trinidad. In the absence of such
information, investigators have to rely on indirect means to analyze the
linguistic situation of Trinidad. In this paper, I have utilized such in-
direct evidence to abstract demographic information about TH.
In Trinidad, Hindi is predominantly spoken in areas such as Caroni,
Couva, Chaguanas and San Fernando. On the basis of the concentration of
Hindus and Moslems in these areas (see Map 1), it appears that approximately
two thirds of the area of Trinidad can be characterized as a 'Hindi speak-
ing area', shown by heavy lines in Map 2. (Map 2 identifies the areas plot-
ted on Map 1.) This presents the most conservative picture of the Hindi
speaking area, for Hindus and Moslems are not the only ones who employ
Hindi as a significant number of the East Indian Christian Community also
uses Hindi as the first or second language. The validity of these obser-
vations can be supported further by a survey of the linguistic background
of primary school children conducted by the University of the West Indies
in 1969. This survey revealed that about 46% of them were exposed to
Hindi at home (for details see Currington, et. al. 1974). Hence, nearly
half the population of Trinidad usesHindi either as their first or second
language. ^
1
Religious Distribution li
Hindu- Moslems vs. Ottx
I 1
960- Trinidad
?rs
1
if
'^■■^"^'-J' i ijM- '
Jlf|[[[||| ^^^m
^llllll^ ||j|,ij|||ti^
^■■^^^ ?|#
-;;:;;l^
^^^TBli""''
> m
Wi
_I0°30'[IIIIII]0-10%
mm 10-30%
1 .••.••■ ;| 30-50%
1 1 50-70%
l^H OVER 70%
0 14 6 » MiLls
_io°.5'"-^"^^ ^- :.].:<f "-'^N
"W"'"']
'^^^^^MUL^
^MttHi
\:::::|
1
L-- ;■ '. ..' ::::::::::::::::;
6I°45'
1
6I°30' 6I°I5
' 1
MAP I - RELIGIOUS DISTRIBUTION I960
(Compiled from 1960 Census Reports, Government of Trinidad
and Tabogo)
138
MAP 2 -THE HINDI (BHOJPURI) SPEAKING REGION OF TRINIDAD.
139
Hindi was transplanted by East Indian indentureship in Trinidad in
1845. By 1917, some one hundred forty three thousand immigrants, mainly
from U.P. and Bihar, had arrived there. The influx of this new culture
set the stage for ethnic conflict. The sentiments of the ruling white
population towards the new immigrants can be summed up by the following
remarks of Lord Harris:
"They are not, neither Coolies nor Africans, fit to be placed
in a position which the laborers of civilized countries must at
once occupy. "3
The African population also looked upon the Indians as the new slaves.
Since today East Indians and Blacks represent the main ethnic groups of
Trinidad Society, the historical rivalry between the two groups remains
salient and can be seen in developments where Hindi vocabulary items such
as bap 'father' and indrani^' wife of Indra, a female name' have developed
derogatory connotations in black speech.
Historically, Hindi played two important roles in the multi-cultural
and multi-racial setting of Trinidad. One, it served as the ethnic language
of the diverse East Indian Community. Second, it acted as the primary pro-
moter of literacy among East Indians: 80% of the East Indian population came
to Trinidad from the rural areas of North India where English education
was negligible, hence the parents and the children of the immigrants shied
away from formal English education. Also, the government and the planters
took little interest in their education in the earlier periods. However,
the rate of literacy improved significantly when the Canadian Presbyterian
Church employed Hindi as a medium to promote Christianity, literacy and
education among East Indians. Basic textbooks were written and printed
in the Hindi language by the Canadian Mission Press at Tunapuna.^ This
strategy of achieving literacy (and conversion) through the regional lan-
guage, Hindi, proved extremely rewarding and the success of the church set
an example for Surinam and other countries. In 1890, special East Indian
schools were set up in which every teacher was required to know both English
and Hindi. Currently, there are more than forty volunteer organizations
which teach Hindi as a second language in Trinidad. In short, even in a
linguistically unfavorable environment, Trinidad never broke away from the
Hindi language.
It is a well known sociolinguistic phenomenon that in bilingual or
multilingual societies, languages tend to adopt complementary roles.
Since English has taken over the important role in education, Hindi has
given up this role in Trinidad. That explains why the younger generation
is gradually giving up Hindi in favor of English.
In spite of this language shift, Hindi still serves its role as the
ethnic language and is still the main channel for the preservation and
transmission of Indian culture. Although Hindi provides membership in the
East Indian group, it is not to say that Hindi is used by all East Indians
in every setting. Yet for intimate or informal conversation, Hindi is pre-
ferred by educated and non-educated speakers alike, whether rural or urban,
whereas English is used by more educated speakers in formal settings.
140
The linguistic attitude of the speakers of TH is similar to the general
attitude of speakers of transplanted languages towards their language.
The speakers of TH generally feel that they speak corrupt Hindi. Neihoff
and Neihoff (1960: 86) also observed that
"There is a wide spread feeling of inferiority among Indians
towards the dialect of Hindi they speak. The occasional Hindu
or Moslem missionaries or other Indian nationals that the local
people hear, are widely admired for their 'beautiful' Hindi."
In addition, popular Hindi movies seem to serve as a constant reminder of
the distance between their own Hindi and "real" Hindi which in turn is
viewed as the model for their own Hindi. As a result, the language of
Trinidad Indians has registered several changes in favor of standard
Hindi. Such changes are readily apparent in the lexicon and syntax of
TH (See Durbin, 1973) .
So far I have used the term TH for the speech of the East Indians of
Trinidad. The term TH does not imply that there is complete homogeneity in
the use of Hindi in Trinidad. There is wide variation in TH; variation
that can best be characterized as a continuum. On the one end of the
continuum lies Bhojpuri which still has a minority of monolinguals adhering
to it. In the middle is the Creolized Hindi whose speakers are in over-
whelming majority. The other end of the continuum is represented by a near
standard Hindi of the type: ve ne kaha 'he said'.
II. THREE GENERATIONS OF TH
As was pointed out earlier, several factors such as a lack of demo-
graphic information about languages and linguistic variation blurr the
linguistic picture of Trinidad. The case of Hindi is particularly complex
since Indians who admit to knowing "only a few words of Hindi" are sometimes
quite fluent in Hindi. Neihoff and Neihoff attempted to explain the under-
lying reason for this paradox in their verbal behavior. They remark,
"It is our impression that most of these people believe that
their Hindi was so 'bastardized' that they could not honestly
consider most of it as true Hindi at all"
Neihoff and Neihoff (1960: 86)
This paper seeks to study the actual as opposed to the self-reported know-
ledge of Hindi by Trinidad Indians. For this purpose, the speech of three
generations of Hindi speakers is analyzed and their similarities and dif-
ferences are examined.
DATA
The data consist of a text of about ten hours of taped conversations
among ten Trinidad Indians, representing three generations. The data were
collected during my trip to Trinidad in the month of December 1979. In
addition, data on the production and perception of retroflexion and aspir-
ation, the two defining features of Hindi, was also collected by means of
141
giving a production and perception test, consisting of thirty occurrences of
retroflex consonants and fifteen occurrences of aspirated consonants to
approximately thirty-five subjects. These tests aimed at testing the hypo-
thesis that TH is losing its phonemic and phonetic sensitivity to retroflexion
and aspiration because of its contact with English (Creole).
SUBJECTS
The following subjects participated in the recording sessions.
Generation I (Gj^) : G, consists of subjects born in India, ages 59 to 80.
Mr. Chotkan Lalu (age 77) arrived in Trinidad by ship in 1913. Mr. Ramdin
Hetlal, (age 75), is currently a resident of the Claxton Bay area. The only
female member of the first generation represented in our sample was Ms.
Rukamani.
Generation-II (Gy) '• The middle generation consists of Mr. Sunderlal
whose parents came to Trinidad on the last ship in 1917. The other two
members representing this generation were Mr. Basand Shukla (42, resident
of Claxton Bay) and Mr. Dal Chand Musai (35, resident of Aleivao) .
Generation-Ill (G^) : subjects were primarily BA or BSC students at the
University of West Indies at St. Augustine. They were between the ages of
twenty and twenty-five years. Ms. Devi (Siewrettan area), Ms. Shanta
(Arima resident), Mr. Hamid (Diego Martin) and Anjani (21, Cottage Village)
were the representatives of the third generation.
The subjects were selected for their willingness to take part in the
recording. No other special criteria went into the selection of the subjects.
Such an approach did not cause any serious sampling inadequacy for our study
since the population under investigation does not exhibit the high degree of hel
erogeneity, social class complexity and mobility of the urban population.
In addition, the reliance on volunteers was imperative since the study also
aimed at obtaining older dialect forms before they become extinct (See
Trudgill 1974: 20).
ELICITATION TECHNIQUES
The elicitation techniques were essentially the same as described in
Labov (1966). Two complimentary methods were used to abstract data for
this study. The first technique employed was group narrations, discussions,
or free conversation among each of the three groups. Since the subjects
were familiar with each other and were motivated to participate in this
study, no special problem was encountered.
The second method of elicitation was personal interviews employed after
speech samples obtained in the group discussions were tentatively analyzed.
These had two aims: One, to test the hypotheses suggested by the preliminary
analysis. Two, to fill gaps in the data collected.
SALIE^JT FEATURES OF TH
An analysis of our texts reveals the following salient features:
The first important characteristic of TH is the different use of
142
retroflexion and aspiration from Standard Hindi. The uae of aspira-
ation in TH differs in three important respects: One, aspiration is intro-
duced word initially or word finally in those instances which do not have
aspiration as is exemplified in (1) below:
(1) cana 'chick peas' ^ c ana
d31 'lentil' —^ d^al
Several other examples such as pha^r_i 'turban', sabh 'all' further support
the claim that the feature of aspiration is over-generalized in TH. Not
only this, aspiration and retroflexion are considered to be the defining
features of Hindi by Trinidad Indians. That explains why even the basic
vocabulary items are not free from aspiration. Such a practice of different
use of aspiration and retroflexion is more frequent among the third genera-
tion of Hindi speakers of Trinidad. Second, aspiration is dropped in some
instances. For example, words such as jhil_ 'lake' become j_iJ in TH .
Third, it is important to note that the phonetic correlates of retroflexion
in TH are different from those of Standard Hindi. Retroflex sounds are
generally labeled as 'rolled' sounds by Trinidadians. In the Trinidad variety,
although the tongue is rolled, sufficient muscular tension is not created.
Also, the point of articulation is fronted. That explains why the nature
of so called rolled or retroflex sounds are distinctively perceived by native
speakers of Standard Hindi.
Another important characteristic of TH is the loss of gender, number
and case agreement in noun phrases as exemplified in (2) and (3) .
(2) bara xarab ciz
• —
big/very (Masc. Sg) bad thing (Fem. Sg)
A very bad thing.
chota^ larka
little (Masc. Sg) boy (Masc Sg)
A little boy
choja chori
little (Masc. Sg) girl (Fem. S^)
A little girl
(3) rasta ke pas
path of near
Near the path
baniya ke sath
Baniya of with
With the Baniya.
*(3a) raste ke pas
Near the path
baniye ke sath
With the Baniya
Example (2) shows that gender agreement is lost within the noun phrase. This
is related to the loss of grammatical gender in TH. The logical gender is
143
either expressed morphologically or by employing different lexical items.
The grammatical ity of phrases such as (3) and the ill-f ormedness of (3a)
indicate that oblique case is lost in TH . The examples of the loss of
number agreement in noun phrases and the loss of pronominal forms are given
in Durbin (1973) and Mohan (1978), therefore, I will not repeat them here.
What is of interest to note is that such loss of gender, number and case
agreement is characteristic of non-native varieties of Hindi spoken in India
as well.
The phenomenon of language simplification manifests itself in several
forms in the variety of Hindi under discussion. I am using the term
'language simplification' in the sense of 'simplified registers' -re-
ductions of a source language - as employed by Ferguson (1975) and Fer-
guson and DeBose (1977) in the treatment of 'Baby Talk'. One such simplifi-
cation worth noting is the reduction of paradigms in TH. Consider the
future tense paradigm of the verb 'to come' in TH given in (4)
(A) Future paradigm of the verb 'to come'
1st person: sg/pl aib
2nd person: sg/pl aibe
3rd person: sg/pl ai
When we compare this paradigm with the Indian Bhojpuri paradigm which has
eighteen inflexional forms, it becomes clear that there has been a substan-
tial reduction. This reduction involves factors such as elimination of
honorific inflection in the 2nd and 3rd person, loss of singular vs. plural
distinction and loss of feminine inflexional forms in TH. (For details See
Mohan, 1978).
The regularization of irregular verbs has also contributed to simplifica-
tion of TH. Our data show that almost all the suhiects tend to regularize
verbs such as dena 'to give' and lena 'to take' in the imperative and
karna 'to do' in the polite imperative.
Borrowings and innovations in TH mark its independent development. The
primary source of borrowing for TH is English Creole. As a result, it has
developed the following unique lexical and syntactic properties:
Plantation vocabulary and lexical innovations such as those given in
(5) are a natural result of language contact.
(5) bakuli bonded colliee
kulumbar European overseer
barak labourer quarters
sardar/driver Indian headman of labourers
suplai replacing dead plants with new
Needless to say, such coining occured during the indentured period, and some
of these terms are unintelligible to the native speakers of Hindi in India.
Other outstanding characteristics of TH is reverse word order in com-
pound verbs (6 and 7) and the introduction of a new class of compound and
144
complex verbs with English nouns followed by Hindi verbals (8) .
(6) hamar beji ke biha Jala howe.
our daughter of marriage will go happen
Our daughter's wedding will take place.
(7) u dan gail jel
he done went jail
He has already gone to jail.
(8) faitam kar 'fight'
tepam kar ' tape'
laikam kar 'like'
Notice that the word order of compound verbs in Indian Bhojpuri is howe
jala, i.e. happen will go. The reverse order of compound verbs in TH as
exemplified in (6) is due to the influence of English Creole which has
just the opposite word order in compound verbs. Example (7) indicates that
TH has developed a set of new classes of compound verbs. The operator
cukna is often replaced by Creole dan 'done' in Trinidad. As a result,
compound verbs such as dan jana (j_a cukna in Standard Hindi) in (7) are
found invariably in TH. Similarly, lexical borrowing from Creole has result-
ed in the introduction of a new set of complex verbs such as tepam karna
'to tape', phonam karna 'to phone', ^rabal deni 'to give trouble'.
The reverse order and the influence of English can also be witnessed
in the following sentences of TH.
(9) apan na mage jae
I Neg want to go
I do not want to go
(10) polls ke sab na mage dekhe
police to all Neg want to see
They (all) do not want police to see (it).
In the above sentences, the higher verb 'want' and the infinitival form
of the embedded verb follow the word order of English; in addition, verbs
such as want do not follow the like-subject constraint as in Standard Hindi.
It should also be pointed out that TH frequently employs the Creole quota-
tive marker 'se' instead of 'bole'.
Code-Mixing/Switching is wide spread in TH. It is not only employed
in day-to-day conversations but is also used in creative writing to induce
realistic effects. Observe the mixing of English in the following popular
Hindi song from Trinidad.
(11) nana nani ghar se nikle
dhire dhire calti hai
badla se dukan mai dono ja baithe
nana taking white one, nani drinking wine
nana age cale, nani going behind
nana ttaking white one, nani drinking wine
145
Observe an additional example:
(12) a2_ rosy girl
hey.
every time I passing, you are pisai masala
'grinding' 'spices'
In examples (11) and (12), the underlined items are code-mixed from English
and Hindi, respectively.
GENERATIONAL SIMILARITIES AND DIFFERENCES
The linguistic properties shared by all of our subjects are as follows.
With the exception of some basic vocabulary, primarily those characteristics
such as loss of gender, honorifics, basic vocabulary were acquired by the
middle and the third generation which are induced by the simplified model of
!indi adopted from the first generation of TH speakers. The generational
differences are summarized in Table I.
TABLE I CGenerational Differences)
G 1
G 2
G 3
Wide Range of Functions Declining Functions
linguistic security
stylistic choices
maximum fluency
Devanagari script
borrowing
large Hindi vocabulary
specialized vocabulary
rel . pronoun
quotative marker
perfect aspiration/
retroflexion
minimum insecurity
decreased choices
adequate fluency
Roman
code-mixing/ -switching
large Creole vocabulary
question pronoun
Creole 'se' (say. St.
English)
Radically Declining Functions
maximum insecurity
no choice
least fluency
Roman
code- switching
loss of Hindi vocabulary
St. Hindi/English vocabulary
question pronoun
Hindi ki 'that'
different aspiration/
retroflexion
Table I reveals that linguistic security among the first generation is
maximum while it is minimum in the third generation. In our analysis, the
two measures used to determine linguistic insecurity were: (i) hypersensiti-
vity to stigmatized speech which Trinidad Indians themselves use, and (ii)
inaccurate perception of their own speech. These two determinants are in
agreement with the determinants of linguistic insecurity cited by Labov
(1972). In addition, subjects often transmitted their linguistic insecurity
by appending to their response expressions such as 'Do you think it is
right?', 'In India people speak like this', 'I think ', 'Is it right?'
The use of these expressions was maximum by the third generation and minimum by
146
the first. It id interesting to note that it was the lack of stylistic
choices and not the great fluctuation in stylistic variation, as observed
by Labov (1972: 132) in the speech of lower-middle class New Yorkers, which
yields profound linguistic insecurity among the third generation TH speakers.
At the phonetic level, hesitation served as another indicator of linguistic
insecurity whereas fluency was treated as a marker of linguistic security in
our analysis. The Devanagari script is maintained only by some speakers of
the first generation. The following generations have primarily adopted the
Roman script.
As regards the question of vocabulary stock, there are significant
differences between the first and subsequent generations. The first genera-
tion primarily relies on the stock of native Hindi vocabulary whereas the
third generation displays massive loss in this area. Parts of basic vocabu-
lary such as ghoj^S 'horse' ,chora 'boy', kha^ta 'sour', ^h&r 'much' thoya
'less', sat 'seven' were absent in the third generation. The process of
language attrition did not take place randomly. It occurred systematically
by retaining one or two members of a set of near synonyms. Let me give some
examples to clarify the point that I am making. The first and second gener-
ations have two words 4.her and bahut to express the concept 'much' whereas
the third generation has only bahut. Several other such examples can be
cited to support this point. Out of a set of chora and layka 'boy',
lamba 'tall' and bay a 'big', kariya and kala 'black', the lexical items re-
tained by the third generation are lajjlca, baya and kala, respectively.
Also notice that the items preserved by the third generation are from
Standard Hindi instead of Bhojpuri or any other Hindi dialect. The re-
placement of the quotative marker bole by the Hindi compliment izer ki
'that' is yet other evidence of Standard Hindi influence on the third gen-
eration.
The loss of lexical items together with the loss of grammatical items
i.e., substitution of question pronoun for relative pronoun, present
evidence that the process of language simplification is still operative in
TH. Each subsequent generation has evolved a simpler variety of Hindi than
the previous generation.
One of the unique features of the third generation is the loss of
morphological processes. The first and second generation could relate the
adjective barka with bara 'big'. However, the third generation could not
do so.
The first generation restricted itself to lexical borrowing whereas
Code-Mixing/Switching mark the important features of second and third
generation speakers.
The results of our production test revealed that the use of aspiration
and retroflexion is rather different and non-phonemic in the third genera-
tion. The first generation was fully aware of the phonemic distinction pre-
sent in the following two sets:
(A)
roti
'crying'
vs
ro^i
'bread'
(B)
dan
'charity'
vs
dhan
'paddy'
147
The third generation demonstrates merely the subjective preference for retro-
flexion because the lexical items in question are supposed to be Indie in
origin.
The results of our perception test are summarized in Figures 1 and 2.
Tne analyses of results show that although the third generation perceives a
distinction between unaspirates and aspirates and retroflex and non-retro-
flex consonants, their perception is not as acute as registered by the first
or the second generation. Such perceptual differences seem to be the
natural consequence of the unstable and non-phonemic nature of aspiration
and retroflexion in the grammar of the third generation.
CONCLUSION
In conclusion, our investigation has shown that contrary to the
diffident characterization of their competence in Hindi by the speaKers of
Trinidad Hindi, their competence goes beyond 'a few words'. Although the
third generation has suffered heavy language attrition and has undergone
the process of pidginization because of adopting a more simplified version of
tne older generation's grammar, the loss of some features and the retention
of others is not arbitrary. They tend to retain only the prestigious
or more standard forms. This tendency is due to the popularity of Hindi
movies and songs (during my stay approximately sixteen theaters out of
fifty were playing Hindi movies on a regular basis) and formal teaching of
Hindi by some volunteer Hindi organizations. The retention of prestigious
and standard forms marks a radical departure from the first generation
which tends to favor native dialect forms of Bhojpuri.
NOTES
*This work was supported by the Partners of Americas grant. I would
like to express my deep appreciation to Professor Michael Marge and Mr.
Bernard Broadbridge. My special thanks are also due to Professors Claire
Broadbridge, Lawrence D. Carrington, B. Samaroo and Kelvin Singh who
assisted me in every conceivable way during my visit to Trinidad. Without
their generous support, my visit would not have been as productive as it was.
An earlier version of this paper was presented at the third South
Asian Languages Round table at SUNY at Stony Brook, May 1-3, 1981. I
have especially benefited from the comments and suggestions of Professors
Frank Anshen, S.N. Sridhar and Rodney Moag.
1 am employing the term 'language development' for it Is free from
negative connotations which terms such as 'language death' and 'language
decay' denote.
According to Uukhedin-Lalla (l974: 3), "Hindi commands the loyalty of
almost half the Trinidad population." Trinidadians employ various labels
such as Hindustani, Bhojpuri, Urdu, Purbi to characterize their speech.
These labels signify their personal choice of a particular style or dialect
of the Hindi language.
British Sessional Papers 1856 XXIV. Quoted from Patricia DuKhedin-
Lalla a974: 5).
148
100
^
^
50
""'
■^^
■^
0
1
•- III
GENERATIONS
FIG. 1 - GENERATION PERCEPTUAL DIFFERENCES
-RETROFLEXION
8
100
en
UJ
1-
1
^^
NUMBER OF
O
V
\
_i
2 n
FIG
I i nr
GENERATIONS
2- GENERATION PERCEPTUAL DIFFERENCES
-ASPIRATION
149
^Tnis word gained currency because of Its use in a calypso.
^bee Grant (1923: 145). Along with Dr. K.J. Grant, Rev. Babu Lai
Behari who was born in the city of Duraarow, in the district of Arrah,
India did poineering work in the area of education and literacy by means
of Hindi. The oldest known translat ional works in Hindi in Trinidad are in
the forms of Lord's prayers. These translations, popularly known as
'bhajans' were carried out by Dr. John Morton and K. J. Grant.
See McNeill and Chimman Lai (1915: 174). Hindustani was taught to
school children and more than half of the school-going age children attend-
ed school in Surinam.
''Rules under the Education Ordinance of 1840' in Trinidad Royal
Gazette November 19, 1890. Quoted from Dukhedin-Lalla (1974: 10)
°To be specific, it should be added that in the case of unmarked ad-
jectives, gender agreement within noun phrases is sometimes preserved in
some dialects of TH. These dialects present evidence that the tendency
of marking unmarked adjectives for the masculine gender is over-generali-
zed in TH.
"plural nouns in TH are marked for the oblique case. However, singular
nouns remain unchanged in this case. The loss of oblique case in singular
nouns can be attributed to the Indian Bhojpuri.
This is the version used by Trinidad Christians. The version
employed by Trinidad Hindus is slightly different.
REFERENCES
CURRINGTON, L.D. Boreley, C and Knight, H. 1974. Linguistic exposure of
Trinidad children. Caribbean Journal of Education. Vol. 1: 1 p. 12-22.
DORIAN, N. 1973. Grammatical Change in a dying dialect. Language 49:
413-438.
. 1980. Language loss and maintance in language contact situations.
A paper presented at the conference on the "Attrition of Language Skills'
Philadelphia: University of Pennsylvania. May 2-3.
DRESSLER, W. 1972. On the phonology of language death. CLS 8: 448-457.
and Wodak-Leodolter . (eds.) 1977. Language Death. International
Journal of the Sociology of Language. Vol. 12.
DUKHEDIN-LALLA, Patricia. 1974. A Creole trichotomy: the case of Hindi in
Trinidad Creole with reference to Standardization. Unpublished m.s.
DURBIN, Mridula A. 1973. Formal changes in Trinidad Hindi as a result of
language adaptation. American Anthropologist. 75: 5. pp. 1290-1304.
FERGUSON, Charles A. 1975. Baby talk as a simplified register. Papers and
Reports on Child Language Development. 9: 1-27. In C.E. Snow and C.A.
Ferguson, eds.. Talking to Children. London: Cambridge University Press.
and Charles E. DeBose. 1977. Simplified registers,
broken language and pldginization. pp. 99-125. In Valdman, A. (ed.).
Pidgin and Creole Linguistic:3. Bloomington: Indiana University Press.
GRANT, Kenneth J. 1923. My Missionary Memories. Halifax: The Imperial
Publishing Company.
150
LABOV, William. 1966. The Social Stratification of English. Washington, D.C.;
Center for Applied Linguistics.
. 1972. Sociolinguistic Patterns. Philadelphia: University of
Pennsylvania Press.
MOHAN, Peggy B. 1978. Trinidad Bhojpuri: a morphological study. Unpublished
Ph.D dissertation. University of Michigan, Ann Arbor.
MCNEILL, James and Chimman Lai. 1915. Report to the government of India
on the conditions of Indian immigrants in four British Colonies and
Surinam. Pt. II. London: H.M.S.O.
NEIHOFF, Arthur and J. Neihoff. 1960. East Indians in the West Indies.
Milwaukee. Milwaukee Public Museum.
TRUDGILL, Peter. 1974. The Social Differentiation of English in Norwich.
Cambridge: Cambridge University Press.
West Indian Census. 1946. Pt I. Jamica : Central Bureau of Statistics.
Studies in the Linguistic Sciences
Volume 4, Number 2, Fall 1981
INTERNAL CHANGE IN A TRANSPLANTED LANGUAGE
Nicole Domingue
McGill University, Montreal
When speakers of a language emigrate en masse to a new
environment and find themselves in a situation where they commu-
nicate mainly with members of their own group, the language they
speak can undergo some significant internal change which is the
result of the move itself and is independent of the contact
situation.
The hypothesis presented in this paper is that this inter-
nal change occurred along the lines of three general principles:
1) Levelling of most dialectal differences. Bhojpuri as
spoken today in Mauritius is strangely uniform. While most of
the inflectional morphology of Mauritian Bhojpuri can be traced
to the Standard dialect of Continental Bhojpuri, certain forms
of Mauritian Bhojpuri must be related to other dialects. A
great need for unification among speakers of different dialects
in a new environment must have led to this levelling.
2) Reduction of stylistic options. In Mauritian Bhojpuri,
the system of nominal forms is greatly reduced: only 'strong'
and 'long' forms have survived. Similarly, of the three levels
of honorifics found in Continental Bhojpuri, only two have
endured. The lack of diversification of the society may be
responsible for this loss.
3) Use of remaining dialectal differences as stylistic indi-
cators. In Mauritian Bhojpuri, one finds remnants of earlier
regional differences at the phonological level. These alterna-
tions are no longer interpreted as dialectal differences by the
speakers but as stylistic options.
1. It is a well-known fact that languages spoken by groups of immi-
grants often undergo certain specific changes which are due to the contact
with the language (s) of the country of settlement.^ The scope and ampli-
tude of this evolution varies according to the circumstances of the con-
tact situation. Less noticed are the changes which take place in such
transplanted languages not as the result of contact with other languages,
but as the result of the move away from the norms in effect in the
country of origin. The example chosen to illustrate this point is the
specific evolution which has characterized Bhojpuri as it is spoken in
Mauritius.
2. Mauritian Bhojpuri is so different from Continental Bhojpuri that
its speakers are not really sure that they are, indeed, speaking that lan-
guage. In fact, most of them believe that they speak some patois of Hindi,
152
very much like Mauritian Creole speakers believe that they speak a patois
of French. This impression is conveyed to them mainly by the fact that,
recently, Mauritian Bhojpuri has borrowed enormously from Creole, particu-
larly at the lexical level. However, a number of differences between
Continental Bhojpuri and Mauritian Bhojpuri cannot be assigned to contact
with Creole; rather, they appear to be due to the 'move' itself, to the
geographical and historical distances which have severed the links between
the transplanted variety and its origins. The purpose of this paper is to
discuss some of the changes which can be accounted for better by the move
theory than by the contaat theory.
In order to justify the hypothesis that the changes discussed are
indeed innovations in Mauritian Bhojpuri, it is necessary to compare the
linguistic situation at both ends of the migration. It is difficult to
know what Bhojpuri was like, a century and a half ago, when the first in-,
dentured laborers left Bihar for more promising lands. This difficulty is
compounded by the fact that the migration lasted almost a century,^ which
means that different stages of the language spoken in Bihar are represented
today in Mauritian Bhojpuri. The only sources of 19th century Bhojpuri
available to me have been John Beames (1868) which is concerned only with
the Northern dialect of Bhojpuri, and the dialectal statements found in
Kellog (first published in 1875). Since these descriptions do not vary
greatly from the descriptions of contemporary Bhojpuri (mostly Tiwari's
studies), 1 feel justified in using Tiwari's statements when no older in-
formation is available. As for Mauritian Bhojpuri, I assume that the
speech of older speakers represents a state of the language from which ex-
tensive external influence is absent, since the socio-political system of
Mauritius has kept the Bhojpuri speakers quite isolated until about fifty
years ago. The comparison between these two varieties of Bhojpuri shows
that indeed Mauritian Bhojpuri has been innovative.
To understand why such an evolution has occurred in Mauritian Bhojpuri,
one must take two socio-demographic factors into account. First, the in-
dentured laborers who came to Mauritius originated from various regions of
Bihar and West Bengal, which means that they spoke different dialects and
languages. Second, they all belonged to one well-defined economic group:
poor and/or landless farmers. As indentured laborers in the new land, they
had no choice in deciding where to live and found themselves settled on
sugar cane estates, completely isolated from the rest of the population.
The argument presented here is that this socio-demographic situation
produced change in the language along the lines of three general principles :
- levelling of most regional differences;
- use of remaining dialectal differences as stylistic indicators;
- reduction of grammatical alternations and stylistic options.
3. Levelling of regional differences.
A very striking fact about Mauritian Bhojpuri is that it does not
display, according to its speakers, any regional variation. In that. It
differs from Creole which, again according to its speakers (often the same),
is diversified regionally: people can tell what part of the island a per-
son is from just by listening to his/her Creole. The same is allegedly not
153
possible in Bhojpuri. This is not to say that there is no variation in
Mauritian Bhojpuri: but variation is due to religious f actors, ^ stylistic
choices, and interference from Creole.
On the other hand, regional variation in Continental Bhojpuri is quite
extensive. Tiwari, like Grierson, distinguishes between four main regional
dialects in Bhojpuri: the Central (or Standard), Western, Eastern and
Nagpuria dialects. It appears that the Central dialect contributed most to
Mauritian Bhojpuri, but specific morphological and syntactic features of
the other dialects have replaced their counterparts in the central dialect.
The great need for unification felt by the laborers must have led to the
development of a system made up of forms originating from different dialects,
From the Nagpuria dialect, Mauritian Bhojpuri has received the
- first and second person possessive:
(1) hamar, tohar (Nagpuria and Mauritian Bhojpuri) vs. hamar, tohar
(Central dialect)
- the third person in the past tense:
(2) dekhal-ak (Nagpuria and Mauritian Bhojpuri) vs. dekhal-asa
(Central dialect)
The form in -ak also occurs in Maithili, which borders the Nagpuria domain.
- a distinction in the copula, which is governed by the nature of the pred-
icate. The actual forms are different, but the process is the same.
From the Western dialect:
the final short i of many lexical items is lost:
(3) akh, pakh (Western dialect and Mauritian Bhojpuri) vs. akhi, pakhi
(Central dialect)
4. Use of remaining dialectal differences as stylistic indicators.
Here, we deal with phonological alternations, which can be traced to vari-
ous dialects of Continental Bhojpuri and appear in Mauritian Bhojpuri as
indicators of style.
The Eastern dialect trilled r alternates with the Central dialect
retroflex R:
(4) bara/baRa, ghori/ghoRa
The Western dialect Z alternates with the Central dialect r in final
position:
(5) mandil/mandir , dubla/dubar
Similarly, the Western dialect s alternates with the Central dialect h in
medial clusters:
(6) masjid/mahjid, rasta/rahta
154
How do these alternations operate as stylistic indicators? In each pair of
(4), (5) and (6) above, and similar pairs, one of the forms happens to be
also the Standard Hindi form. For instance boRa is the lexical entry for
'big' in both the Central dialect of Bhojpuri and in Standard Hindi. In
Mauritius, as, I believe, in contemporary Bihar, the forms which are iden-
tical with the corresponding Standard Hindi words have acquired prestige
among the Bhojpuri speakers. In view of the statement made earlier to the
effect that Bhojpuri speakers in Mauritius believe that they speak a cor-
rupted form of Hindi, it is easy to see why Hindi-sounding words may give
prestige to the speech of anyone. In addition, as Hindi is taught in many
public schools attended by Bhojpuri-speaking children, Hindi-like Bhojpuri
is a sign that one is educated.
The regional variation which existed in the country of origin has not
survived as such in Mauritius. The specific situation of the Bhojpuri
speakers on the island encouraged them strongly to strengthen their iden-
tity by doing away with distinctions based on place of origin. Only those
regional features which were capable of assuming a stylistic function
remained.
5. Reduction of grammatical alternations and stylistic options.
Here the argument appeals to the second socio-demographic factor men-
tioned above: the suggestion that the migrating Bhojpuri speakers came
from one of the lowest socio-economic classes of the society. Indeed, it
can well be argued that such speakers did not feel much the prestigious in-
fluence of a standard or of a refined language and chose among options
those which were the most normally used in their own groups. As a conse-
quence, some of the alternations found in a more diversified society disap-
peared.
5.1 Gender
Tiwari describes the notion of gender, in Continental Bhojpuri, as
natural and not grammatical: animate nouns must be marked for masculine or
feminine gender :
(7) laika 'boy', laikT 'girl'; ghoR 'horse', ghoRi 'mare'
The adjective optionally agrees with the noun: baRz/boE ghoRl 'the big
mare'. The natural gender is also found in Mauritian Bhojpuri, but the
adjective agreement rule is no longer operative:
(8) baRa laiki 'big girl', baRa laika 'big boy'
In other words, a rule which is optional in Continental Bhojpuri is lost in
Mauritian Bhojpuri.
Gender is also indicated in the verb in Continental Bhojpuri, which
shows different endings for the masculine and the feminine. In Mauritian
Bhojpuri, the feminine forms have disappeared except in the second person
singular of the past tense:
155
(9) tu dekhale 'you saw' (general and masculine)
tu dekhalu 'you saw' (feminine)
Even this feminine marking rule is optional and characteristic of a more
archaic style.
5.2 Case
In Continental Bhojpuri, there exists an instrumental case marker in
-e which alternates with the post-positional phrase construction:
(10) bukhe/bukh se 'out of hunger'
This alternation is not available in Mauritian Bhojpuri, where only the
post-positional construction can be used.
5.3 Nominal system
In the Central and Western dialects of Continental Bhojpuri, the noun,
and its contiguous adjective (s) , can have up to four forms:
(11) , weak ghar 'house' gh5R 'horse' baR 'big'
short t - - ^i_ ■ ti_ 1 I i_ -
strong — ghoRa pothi book baRa
long gharwa ghoRwa pothiya baRkl
redundant gharauwa ghoRauwa pothiyawa baRkawa
In Mauritian Bhojpuri, the redundant form has disappeared, and only one
short form occurs:
(12) ghar/gharwa, ghoRa/ghoRwa, pothl/pothiya, baRa/baRka
The different levels of familiarity and affection which may be expressed by
the original forms of Continental Bhojpuri are consequently erased. The
only function retained by the long form in Mauritian Bhojpuri is that of
reference.** It can be argued that, with respect to the categories of case,
gender and nominal systems, optional rules have become lost in Mauritian
Bhojpuri, resulting in a simplified morphology.
5.4 Honorlflcs system
The most obvious reduction of alternations occurs in the systems of
honorlflcs, both in the pronoun and in the verb.
- In the pronoun:
Continental Bhojpuri has up to three degrees of honorlflcs for the personal
and the possessive pronouns. Of these distinctions, Mauritian Bhojpuri has
kept only those of the second person singular:
(13) personal: tori 'you' [ordinary], tohra 'you' [honorific],
raur 'your honor'
possess.: tor 'your' [ordinary], tohar 'your' [honorific]
156
- In the verb:
The Continental Bhojpuri verbal system shows up to three levels of style
for the second and third persons; Tiwari has called them contemptuous, or-
dinary, and honorific. An additional unmarked form exists for the first
person. In Mauritian Bhojpuri, only two levels are retained: the con-
temptuous level becomes the ordinary, and the unmarked suffix of the first
person is extended to all persons as the honorific . The plural forms of
the past tense of 'see' show the difference between the two types of
Bhojpuri :
(14) Continental Bhojpuri Mauritian Bhojpuri
1st dekhalija (unmarked) dekhalisa [ordinary]
dekhalija [honorific]
2nd dekhalasa [contemptuous] dekhalasa [ordinary]
dekhala [ordinary] dekhalija [honorific]
dekhalT [honorific]
3rd dekhalesa [contemptuous] dekhalansa [ordinary]
dekhal [ordinary] dekhalanja [honorific]
dekhalT [honorific]
In actual fact, the Mauritian Bhojpuri speakers rarely use the honorific,
the form in -co., except in the imperative:
(15) khaja 'please eat'
Regarding the reduction in the honorific systems, and particularly the
loss of the more formal levels, one might hypothesize that the indentured
laborers, belonging to the lower strata of society and being illiterate,
had a poor command of these systems and knew mostly their lowest levels,
which, to them were ordinary. Their need to express politeness led them,
however, to adopt one marker (the wrong one though as far as the verb is
concerned) as the honorific.
6. The situation described above calls for a discussion of the concept
of internal change. One may wonder whether the making of Mauritian
Bhojpuri did in fact involve the changes suggested. It could be hypothe-
sized that much dialect mixture activity, as is common in India, was taking
place in Bihar before the migration and that the laborers all spoke a
common and more or less uniform variety of Bhojpuri even before they set-
tled in Mauritius. It would also be perfectly reasonable to expect that
the same laborers, belonging to a low social class, did not normally use
grammatical and honorific markers which were more likely to be found in the
speech of the educated.^ In other words, it could be assumed that the mi-
grating Bhojpuri speakers took with them their specific variety of the lan-
guage which survived until today in Mauritius. This appears to be in con-
tradiction with the observation that the Bhojpuri spoken by similar labor-
ers in other parts of the world is not only quite different from Mauritian
Bhojpuri but also very varied.^ This, in itself, is not a sufficient argu-
ment since one is not sure that these laborers were indeed similar in
social and geographical background. Our ignorance of the circumstancial
details of the migration might well hide facts which would point to a
157
common and non-diversified origin for a majority of Mauritian Bhojpuri
speakers .
The position taken here is that the argument for internal change is to
be found linked to the concept of 'a language'. If a language, here
Bhojpuri, is defined not in terms of the collection of its varieties but in
terms of an idealization of these varieties, it is quite possible to posit
for an offshoot of one of the varieties (Mauritian Bhojpuri) an idealiza-
tion different from that which would be appropriate for its origins: the
model for Mauritian Bhojpuri could be different from the model for Conti-
nental Bhojpuri. One can then speak of internal change when comparing
Mauritian Bhojpuri with Continental Bhojpuri.
At the beginning of this article, the facts of internal change were
contrasted with those of external change such as one notes in a language
contact situation. The facts of internal change presented here, however,
could also be considered as resulting from contact, contact between
dialects, which led to a merger. The similarity between external and in-
ternal change must stop here as there are obvious differences between the
two phenomena: while, in a language contact situation there is little, if
any^ borrowing of the inflectional material,^ in the dialect contact situa-
tion, this is mainly what happens.
Another difference between the two situations of contact is that con-
tact between dialects decreases diversity while contact between languages
may increase it.^" Indeed contact between two languages can result in the
emergence of a new language which differs from either of its 'parents' and
which may exist alongside of them. The origin of pidgin and creole lan-
guages is often attributed to such events.
As to the reduction in grammatical and stylistic choices, they could
be described as the result of an absence of contact, an impossibility of
interaction between different levels of society since the society was hardly
stratified. This again contributes to the low level of diversity mentioned.
It is interesting to note that this impoverishment of traditional linguistic
options in Mauritian Bhojpuri is not final. New ways of expressing diver-
sity have been found; the phonological variations described above and the
variation due to the contact with Creole,
NOTES
^ A discussion of the contact situation between Mauritian Bhojpuri and
Creole, and its linguistic results, appears in Domingue 1980.
^ 1835 is usually considered as the official year in which the immi-
gration of indentured laborers from India started. It lasted until the
first quarter of the 20th century.
^ Some lexical items serve to distinguish the religion/culture of the
speakers. For instance, Moslems will always use kitab and Hindus pothi to
express the concept 'book'.
158
"* I have argued elsewhere (Domingue 1980:4) that the long form has
become a definite marker.
^ The optional rules discussed in the literature on Bhojpuri may, in
fact, reflect this particular situation.
^ According to Surendra Gambhir (personal communication) , this is at
least valid for the varieties of Bhojpuri spoken in Guyana.
' For the sake of this discussion, I will consider dialects to mean
varieties of a language which are mutually intelligible and which are
spoken inside the same political unit.
^ This is of course a disputable statement which the Boas-Sapir con-
troversy has highlighted and which is still being discussed.
^ Where this material has the same inflectional - and, consequently,
productive - function in the borrowing language as in the donor language.
^^ Contact between languages does not, however, always result in in-
creased diversity. In fact, many studies in areal linguistics have shown
just the opposite: two, or more, languages in contact often become more
alike (see for instance Gumperz and Wilson 1971).
REFERENCES
BEAMES, John. 1868. Notes on the Bhojpuri dialect of Hindi, spoken in
Western Bihar. Journal of the Royal Asiatic Society 3.483-508.
BEATON, P. 1859. Creoles and coolies or five years in Mauritius. London:
J. Nisbet.
BENEDICT, Burton. 1961. Indians in a plural society: A report on
Mauritius. London: HMSO.
CHALLAYE, C.A. de. 1844. Memoire sur 1' emigration des indiens et sur le
travail libre dans les colonies de Maurice et de Bourbon. Paris:
Imprimerie Royale.
DOMINGUE, Nicole. 1980. Syntactic innovations in Mauritian Bhojpuri. Un-
published manuscript.
DOMINGUE, Nicole and Pierre-Marie Moorghen. In press. Multilingualism in
Mauritius. International Journal in the Sociology of Language.
DUTT, Girindranath. 1897. Notes on the vernacular dialects spoken in the
district of Saran. Journal of the Asiatic Society of Bengal 66.194-
212.
GRIERSON, George A. 1929. Linguistic survey of India vol. 1, part 1.
Delhi: Motilal Banarsidas.
GUMPERZ, John and Robert Wilson. 1971. Convergence and creolization: A
case from the Indo-Aryan/Dravidian border. In D. Hymes, ed.: Pidgin-
ization and creolization of languages, pp. 151-168. Cambridge
University Press.
KELLOGG, S.H. 1965. A grammar of the Hindi language. London: Routledge
and Kegan Paul Ltd. First edition: 1875.
159
TIWARI, Udai Narain. 1960. Origin and development of Bhojpuri. Calcutta:
The Asiatic Society, Monograph Series, vol. X.
TIWARI, Udai Narain. 1964. The pronouns of Standard Bhojpuri. Indian
Linguistics 25:227-228, (Baburam Saksena Felicitation Volume)
TRAMMEL, Robert L, 1971. The phonology of the Northern Standard dialect
of Bhojpuri. Anthropological Linguistics 13, no. 4.
Studies in the Linguistic Sciences
Volume 11, Number 2 » Fall 1981
TRANSITIVITY IN HINDI
Rajeshwari Pandharipande
University of Illinois, Urbana-Champaign
The purpose of this paper is (a) to point out the
features of transitivity in Hindi and to argue that the
concomitance of these features is essential for transitivity,
(b) to demonstrate with illustrations that transitivity is
a continuum on which some verbs are more transitive than others,
(c) to provide evidence to show that there is a striking simi-
larity between the features of transitivity and of causativity
in Hindi, and (d) to provide evidence from creative literature
to point out that the notion of transitivity is better under-
stood if the pragmatic function of transitivity is taken
into account
0. Introduction
Transitivity has been discussed both in traditional grammars as
well as in modern linguistic studies for two major reasons: (i) it is
one of the language universals and (ii) it plays an important role in
the grammar of a language.
Traditionally, the following characteristics of the notion of
transitivity have been pointed out: (a) transitivity is an ambiguous
term, i.e., for Jesperson (1964:116), a verb is transitive when used
with an object, and intransitive elsewhere, (b) transitivity is a property
of a sentence and not only of a verb, and (c) transitivity is a property
which signifies an action transferred from an agent to a patient.
Several syntactic as well as semantic aspects of transitivity have been
discussed in the contexts of language typology (Masica 1976) , ergativity
(Anderson 1977, Comrie 1978, Dixon 1977, Hock 1980, Kachru and
Pandharipande 1979), Relational Grammar (Postal and Perlmutter 1974,
Pandharipande 1981), and causativity (Binnick 1976, Kachru 1976,
McCawley 1976, Shabatani 1976) .
The purpose of this paper is (a) to point out the features of
transitivity in Hindi and to argue that the concomitance of these
features is essential for transitivity, (b) to demonstrate with
illustrations that transitivity is a continuum on which some verbs are
more transitive than others, (c) to show that there is a striking similarity
between the features of transitivity and of causativity in Hindi, and
(d) to provide evidence from creative literature to point out that
the function of transitivity is derived from the pragmatic fact that
"... as human beings we are particularly interested in the results
of our purposive actions and the effects that our actions have on
patients" (Lyons 1977:491).
162
2.0 The Markers of Transitivity in Hindi
Traditional grammars (Guru 1952, Vajpeyi 1959) have mentioned the
following markers of a transitive verb in Hindi, i.e., (a) a transitive
verb in Hindi requires its agent to take the postposition -ne in the
perfective aspect (with the exception of lana 'to bring'). For example,
dekhna 'to see' (transitive):
(l)j hamCne ve sundar ghar dekhe
(*haia)
we those beautiful houses saw
We saw those beautiful houses.
jana 'to go' (intransitive)
(2) f bacce 1 ghar gae
/*bacco J ne
children home went
The children went home.
(b) While a transitive verb takes two participants (i.e., subject and
object) an intransitive verb takes only one, i.e., subject. (c) A
perfective participial of a transitive verb in Hindi generally refers
to its object! rather than to its subject, while a perfective parti-
ciple of an intransitive verb agrees with its subject. Compare and
contrast the following perfective participles of transitive (3a) and
(3b) and intransitive (4) verbs, respectively.
Perfective participle referring to the object: likhna 'to write':
(3a) (larki ki) likhi hui kahani
girl poss . wrote aux story
The story written (by the girl)
Participle referring to the subject:
(3b) * (kahani) likhi hui larki
(story) wrote aux girl
The girl who (has /had) written (the story)
girna 'to fall'
(it) gira hui, patta
fell aux leaf
the fallen leaf
Notice that the perfective £articiple of the transitive verb
likhna 'to write', likbi hui 'written', refers to the object (3a)
and not to the subject (3b). In contrast to this, the perfective
participle of the intransitive verb girna 'to fall' refers to its
subject (4).
163
2.1 Problems
The above features of transitivity are not adequate to define
transitivity because (a) there are verbs in Hindi which express acts
which involve two participants but they are not considered to be
transitive, e.g.,
(5) pasand bona 'to like'
liking be
(5a) mujhe vah ghar pasand ht
me to that house liking aux
I like that house.
(6) yad hona 'to remember'
memory to be
(6a) us ko bacpan ki bate yad hf
him to childhood of things memory are
He has the memory of the things about childhood.
(7) kam hona 'for work to happen'
work to happen
(7a) us se kam hua
he by work happened
The work got done (happened) by him.
Notice that in the above examples ((5) and (6)) the verbs take two
participants (i.e., in (5) the one who likes and the one which is liked)
in (6) the one who remembers and the one who/which is remembered) .
However, these verbs are not treated as transitive, i.e., these verbs
do not allow their subject to take the ne-marking in the perfective
aspect, and perfective participles cannot be derived from these verbs.
Similarly, (7) has two participants but the verb is not transitive,
i.e., neither the marking -ne_ is added to its agent nor is a perfective
participle derived from it.
(b) -N£ marking is optional for certain verbs, i.e., nahana 'to
take a bath', hasna 'to laugh', rona 'to cry', bolna 'to speak'.
(c) Compound verbs can be treated as transitive verbs as well
as intransitive verbs depending upon the explicator in the compound
verb, i.e., karna 'to do' is a transitive verb, but when followed by
bfct:hna 'to slt"^ (i.e., kar btthna 'to do by mistake') it loses its
transitivity. Similarly, sona 'to sleep' and hasna 'to laugh' are
intransitive verbs but are optionally treated as transitive verbs when
the explicator lena 'to take' is added to them.
(d) Some transitive verbs allow inanimate subjects and the -ne
marking is freely added to them in the perfective aspect. However,
the perfective participle is not derived from such verbs, e.g., toyna
'to break' .
164
(8) (a) jtez hawane ( ghar ke darvaze jtor diye ?
(b) (corne ) (tore )
(a) (powerful wind? house of doors broke
(b) (thief )
(a) ("The powerful wind) broke the doors of the house.
(b) ( The thief . \
(9) (a)-)|((tez) hawa ke tore hue darwaze
(b) cor ke tore hue darwaze
(a) The doors broken by the power f\il wind.
(b) The doors broken by the thief.
Notice that in (8) both subjects, i.e., animate and inanimate, take
the -ne marking. However, the perfective participle does not refer
to the inanimate subject (9a), but refers to the animate subject (9b).
(e) More often than not, transitive verbs are related (morpho-
logically or semantically) to their intransitive counterparts. The
above facts suggest beyond doubt that in order to understand/interpret
the above facts (i.e., a-e) and to clearly define the link among them,
it is necessary to define transitivity in Hindi.
3.0 Features of Transitivity
If we assume that the definition of transitivity (mentioned in
section 1.0) is correct, i.e., if transitivity is signified by an agent
performing an act on a patient, then, in order to define transitivity
in Hindi, we must ask the following questions:
(a) What kind of agent does the verb take (animate agent, instru-
mental agent, inanimate agent, etc.)?
(b) What kind of act is expressed by the verb (action, state,
etc.)?
(c) How does the agent perform the act (volitionally, non-volition-
ally)?
(d) What kind of patient is involved in the act (overtly expressed,
implied, same as subject/agent other than the subject, etc.)?
Thus, it is clear that transitivity cannot be defined in terms of only
one feature, but rather, it is an amalgam of several features, each
defining one crucial aspect of the whole. A close examination of the
data shows that the following features are essential for transitivity
in Hindi.
(a) Agency--the subject of a transitive verb/clause is an agent
as opposed to an instrument of the act expressed by the verb.
(b) Volitionality--it is essential that the agent's volition
controls the act.
(c) Activity--the act carried out by the agent should express
an action as opposed to a state.
(d) Object--the object should be a patient which may either be
expressed or implied in a sentence.
165
The concomitance of all the above features is essential for transitivity
in Hindi. The absence of any one of these features is enough to mark
a verb/sentence intransitive. Let us consider each feature. Notice
that if the subject of the verb is [-agent] the verb/sentence is intransi-
tive. The morphology of the subject denotes its role (Kachru et . al.
1976, Pandharipande 1979, 1981). Thus a subject with the -se marking
expresses instrumentality of the agent while a subject with^The ko
marking expresses a recipient/experiencer role. Thus the subjec~with
the -se/-ko marking indicates a lack of both agency as well as transitivity,
i.e.,
X ko malum bona 'for X to know'
(10) J" j an ko_ , vah bat malum thi
;^*jan ne {
John-dat. that matter know aux
John knew the matter.
(lOa) wjan ko / malum hui bat
;*jan ki )
[John- to/ knew matter
( ofl
The matter that John knew.
(11) nan sec kam hua
L *ne )
; John by ,' work happened
The work got done by John.
(lla) )*jan se ( hua kam
( ka<,
John by happened work
The work which got done by John.
Notice that the verbs in (10) and (11) do not allow the -ne marking
on their subjects and that participialization is blocked in (10a)
and (lla).
If the verb expresses a non-volitional act, it lacks transitivity.
Examples (10) and (11) lack volitionality as well as transitivity.
It has been argued independently (Pandharipande 1978, 1981) that verbs
expressing a non-volitional act fail to undergo Passive; (10) and (11)
are typical cases of exceptions to passivization in Hindi. Similarly,
verbs such as kar b^thna 'to do by mistake' and kah jana 'to say
(something) unintentionally' fail to undergo passive and fail to qualify
as transitive verbs. Notice that these verbs take two participants
and that their subjects are also agents which are (-volitional) and
therefore, these verbs are treated as (-transitive) . Consider the following
examples (12), (13), and (14), which show that the verb kar bi^hna
fails to undergo passivization (12), -ne marking (13), and participializa-
tion (14), respectively.
166
(12 ) *us se sara kam kar bftha gaya
he by all work do sat went
The whole work was done (by mistake) by him.
(13) )*us nel sara kam kar bLtha
(vah J
he-Ag. all work do sat-past
He did all the work (by mistake).
(lU) *kar bttha gaya kam
do sat went work
The work done (by mistake).
It is argued with evidence from Pandharipande (forthcoming) that these
compound verbs fail to express volitional acts and therefore fail to
undergo passive. (For more discussion on the role of volitionality
in defining transitivity refer to Kachru (1981 in this volume) and Pandhari-
pande (1982 forthcoming) .
Verbs which do not express action (but rather a state, etc.)
fail to qualify as transitive verbs, i.e., verbs such as dikhna 'to
seem' and lagna 'to feel' do not express any action and therefore are
marked as intransitive verbs.
In Hindi, it is essential that two participants, i.e., subject
and object (or agent and patient) are involved in the action expressed
by the verb. However, the subject and the object may be expressed or
implied. As a consequence of this, verbs in Hindi which neither express
nor imply an object are treated as intransitive verbs, «.g., bfct^hna
'to sit', girna 'to fall', etc. A large class of intransitive verbs
in Hindi lack objects. Similarly, verbs such as kam hona 'for work
to happen' neither express nor imply an agent which is evident from
example (15), which lacks an overt agent. Notice that when the adverbial
dhyan se 'carefully' is inserted in (15), the resulting sentence (15a)
is ungrammatical . The adverbial dhyan se 'carefully' only occurs in
sentences with agents (Pandharipande 1981) .
(15) kam hua
work happened
The work got done.
(15a) *dhyan se kam hua
carefully work happened
The work got done carefully.
If an agent was implied in (15), (15a) would be grammatical.
Contrast (15a) with (16) which also lacks an overt subject/agent.
However, the adverbial dhyan se 'carefully' is freely inserted in
the sentence.
167
(l6) dhyan se kam kiya gaya
carefully work did went
The work was done carefully.
While the verb kam hona 'for the work to happen' is intransitive, the
verb kam karna 'to do work/to work' in (16) is transitive.
4.0 The Hierarchy of Transitivity
The discussion so far points out a necessity of the above four
features for transitivity. However, agency, volitionality, and activity
are not discrete and can be quantified only in terms of more or less
Ci.e., more volitional versus less volitional, etc.). Similarly, it is
difficult to imagine that all transitive sentences in the language
have the same proportion of the above features. The data points out
that there is a hierarchy of transitivity on which some verbs are more
transitive than others. The following evidence supports this hypothesis.
Ca) Notice that verbs such as karna 'to do', parhna 'to read,
study', dena 'to give', lena 'to take' and torna 'to break' contain
all four features. However, if the subject of the above type of verbs
is inanimate, (recall examples (8) and (9)) the volitionality is reduced
and as a result the agency is reduced, too. Notice that such verbs
show a lower level of transitivity as compared to transitive verbs
with animate agents. Recall that although the ne- marking is added on
to the inanimate subject (8), the participle cannot be derived (9).
(b) Now let us consider verbs such as hasna 'to laugh', ronS
'to cry', and nahana 'to bathe, to take a bath' which can take/ imply
only cognate objects such as hgsi 'laugh', rona 'cry', and snan 'bath'.
Native speakers feel these verbs are not clearly transitive since they
do not freely allow objects other than cognate objects. This lowers
their transitivity and therefore the n£ -marking optionally shows up
on their subjects only, i.e.,
(it) (ml \ hasa
( ml. ne )
I laughed.
(c) Native speakers feel that verbs such as janna 'to know',
pahcanna 'to recognize', etc., are lower in activity and volitionality
as compared to verbs such as karna 'to do', (although they do express
more volitionality as compared to verbs such as pata hona 'to know
which is evident from the fact that while janna 'to know' and pahcanna
'to recognize' undergo passive, pata hona 'to know' does not) .^
The above verbs are felt to be less transitive than karna 'to do'.
This intuition of the speakers is supported by the fact that the
perfective participles of janna 'to know' and pahcanna 'to recognize'
are not as acceptable as those of verbs such as payhna 'to read',
karna 'to do', etc.
168
(l6) ?yah meri jani hui |la:rki ( ht
^kahani \
this my known igirl ? is
(story j
Thisf girlcis known to me.
CstoryN
(l8a) ?yah meri pahcani hui la:{-ki h^
this my recognized aux girl is
This is the girl (who is) recognized by me.
However, the -ne marking is freely added on to their subject, i.e., C18b)
(l8b) mt ne )jana I ht ki...
( pahcana \
I [have known /> that .
(recognized
Cd) In Hindi the process of compounding verbs produces transitive
or intransitive verbs (Hook 1974, Kachru 1980). Consider the following
examples.
C19) karna 'to do' (transitive) y kar b£.thna/jana (intransitive)
'to do by mistake/ unintentionally '
(20) bolna 'to speak' (transitive) ) bol a) b^t^hna (intr.)
b) uthna (intr. )
c ) jana (intr. )
to say/ speak a) by mistake
b ) suddenly
c) unintentionally
Notice that when the explicators bfcthna 'to sit', uthna 'to get up',
and jana 'to go' are added to the transitive verbs karna 'to do' and
bolna 'to speak', the resulting verbs are intransitive, i.e., (19) and
(20). The following evidence supports this hypothesis; karna 'to do'
takes n£ (21), kar bithna 'to do by mistake' and kar jana 'to do
unintentionally' do not (22, 23). Similarly, the perfective participle
derived from karna 'to do' (24) refers to its object while this process
is blocked for kar btthna 'to do by mistake' and kar jana C24a) 'to do
unintentionally'. Consider the following examples.
(21) fus ne? kam kiya
(*vah )
he work did
He did the work.
(22) 1 vah ? kam kar b^tha
Q*usne )
he work do sit
He did the work by mistake.
169
(23)/vah ? kam kar gaya
(*usne)
he work do went
He did the work iinintentionally.
(2U) us ka kiya hua kam
his did aujc work
The work done hy him.
(2^a) us ka kar (bi_tha/ hua kam
/gaya )
his do ssat • aux work
/went \
The work he did iby mistake (^
/unintentionally \
In contrast to the above, the process of verb compounding produces
transitive verbs from their intransitive counterparts in some varieties
of Hindi (e.g., in the variety of Hindi spoken in Madhya Pradesh (central India)
Consider the following examples.
(25) sona (intr.) 'to sleep' ) so lena (tr. ) 'to sleep for oneself
(26) hona (intr. ) 'to be' )ho lena 'to become'
(27) bittina (intr.) 'to sit' )blth lena 'to sit for oneself
While the intransitive verbs in (25) -(27) do not take the -ne marking,
their derived counterparts optionally do, as is seen in the following
examples.
(28) ve soe
*unhone soya
They slept.
(28) a) ve so liye
b ) unhone so liya
They slept for themselves.
(29) ham un ke sath the
*ham ne
we them with were
We were with them.
(30) a) ham un ke sath a) ho liye
b)?ham ne b) ho liya
we them with became took
We joined them.
(31) ham btthe
jtham ne bftha
We sat.
(31) a) ham ne bfth a) liya
b) haim b) liye
we sit took
We sat for ourselves .
170
Notice that native speakers feel that the derived verbs are not as
transitive as karna 'to be' which is evident from the fact that ne
is optionally added to the subject. Native speakers do not have~very
clear judgments about the placement of the -ne marking in (28), (30),
and (31a). Also, perfective participles are not derived from these
derived transitive verbs. Consider the following example.
(31t) *btth liye log
sit took people
People who Chad ? slept for themselves,
(have )
Native speakers intuitively feel that while the explicator lena 'to
take' increases the volitionality of the above intransitive verbs
it does not provide an object (i.e., the act bt.thna 'to sit' itself is
vaguely felt to be the object). Therefore, transitivity is low in
the verbs in (28), (30), and (31a). The addition of an explicator
(transitive) does not necessarily increase transitivity, i.e., both
calna 'to leave' (intr.) and its counterpart cal dena 'to leave'
leave give
(emphatic), to go away' are felt to be intransitive verbs. This intuition
is supported by the fact that ne is blocked for both calna 'to leave'
and cal dena 'to leave, to go away'. Similarly, perfective participles
are not derived from them.
The above evidence shows that we have to assume a cline of transi-
tivity in Hindi. Bhatia's (1981, in this volume) discussion on features
of transitivity supports the above hypothesis.
5.0 Transitivity and Causativity in Hindi
A close examination of the Hindi data shows that the above-mentioned
features of transitivity strikingly resemble the features of causativity
in Hindi. This hypothesis explains why the above four features cluster
around transitivity in Hindi. In what follows I will provide evidence
to support this hypothesis. The relationship between transitivity and
causativity has been discussed in various studies. It is claimed within
the generative semanticists ' framework that a transitive verb has an
embedded intransitive sentence predicate with a cause node,
(Dowty 1972, McCawley 1968, 1971). At this point, I am not making
any claims about the exact derivational history of transitive verbs in
Hindi, but rather, my purpose here is to demonstrate the similarity
between the notions of transitivity and causativity in Hindi.
(a) The major process of deriving transitive verbs in Hindi is
to add the causative morpheme /a/ to intransitive verbs, (Guru 1952,
Kachru 1980, Vajpeyi 1959).
(32) calna 'to walk' > cal-a-na 'to drive'
(33) u^hna 'to get up' ---^ uth-a-na 'to wake up (someone)'
b)
kam karna
(35) a)
band bona
b)
band karna
171
Cb) There are pairs of conjunct verbs in Hindi which seem to hold
a causal relationship. (For further discussion on conjunct verbs refer
to Kachru 1980) .
'for work to happen'
'to do the work'
'to be closed'
'to close'
Pairs such as (34a and b) and (35a and b) form a large class in
Hindi. (34a) and (35a) have the intransitive verb bona 'to be' and
(34b) and (35b) have the transitive verb karna 'to do'. Notice that
the causal relationship between (34a) and (34b) as well as between (35a)
and (35b) is very clear from the gloss. Notice that (34a) and (35a) do
not have other causative counterparts besides (34b) and (35b), respectively.
Therefore, it is reasonable to assume that (34b) and (35b) hold a causal
relationship with (34a) and (35a), respectively.
(c) The processes of causativization and transitivization as
discussed in (b) above, are also functionally similar, i.e., they both
create objects from basic subjects. Recall that two participants (subject/
agent and object) are required for a sentence to qualify as a transitive/
causative sentence (section 3) .
(d) There is a very large class of transitive verbs in Hindi which native
speakers feel are semantical ly related to their respective intransitive
counterparts.
(36) tutna 'to be broken' torna 'to break'
(37) bikna 'to be sold' becna 'to sell'
(38) \ dikhai )dena ? /'to be seen'
) (^ parna\ y
dekhna 'to see'
dikhna
(39) chutna 'to leave' choyna 'to leave'
The causal relationship between the above intransitive verbs and their
transitive counterparts is pretty clear. It is observed that 'morpho-
logically unrelated verbs in certain languages stand in semantic relation
to one another as do pairs of verbs that are related by means of a
productive morphological construction in other languages' (Lyons:
1977:489).
(e) An interesting piece of evidence to support our hypothesis
comes from the phenomenon of language change in at least one (North
Eastern) variety of Hindi. Verbs such as khana 'to eat' and pina 'to
drink' do not have intransitive counterparts in Hindi. In the above
variety of Hindi the intransitive counterparts are constructed by
172
native speakers. Such forms are freely used in the spoken/ informal style
but as yet are not used in the formal style of the language.
Transitive Intransitive
(40) khana 'to eat' khilana 'to be eaten'
*khilna
(41) pina 'to drink' pilana 'to be drunk'
*pilna
The above examples indicate that native speakers feel that transitive
verbs hold a causal relationship with their intransitive counterparts.
(f) Another piece of evidence to support our hypothesis comes
from the fact that verbs which are lower on the transitivity hierarchy
do not have causative counterparts, i.e., janna 'to know' and pahcanna
'to recognize' do not have causative counterparts such as * j anwana 'to
make/cause to know' and *pahcan(w)ana 'to make to recognize', respectively.
If we assume that the above verbs are lower on the transitivity
hierarchy, (recall the discussion in section 4(c)) and thereby express
a lower level of causation, we can explain why these verbs do not have
intransitive (syntactic/semantic) counterparts in Hindi. In such cases
the gap is filled by using conjunct verbs of which the first member
is a passive participle from Sanskrit and the second part is the
auxiliary verb hona 'to be', i.e,
{k2) gyat hona 'to be know' (intr. of janna 'to know')
known to te
(ii3) paricit hona 'to be familiar' (intr. of pahcanna 'to recognize)
familiar to be
know
3
(g) Causativity and transitivity in Hindi share the above four
features in common, e.g., both need a volitional agent to be the subject.
Notice that if the subject is instrumental, neither a transitive (44)
nor a causative (45) verb is used. Consider the following sentences.
(hk) *(meri) jkhusis ne yah kam kiya
(kosis ])
(my) Chappiness ? Ag. this work did
[effort ]
(My) [happiness/ did the work.
( effort 3
(ii5) *(meri) ikhusi) ne mujh se yah kam karaya
( kosis(
(my) ChappinessP Ag. me by this work make-do-past
(effort ) ^
(My) jhappinesE^ made me do this work,
(effort
173
{k6) (meri ) jkhusi ( ke karan yah kam hua
(kosis 1
(my) i happiness/ because of this work happened
/ effort j
The work got done because of my (happiness.
' effort.
(h) If an intransitive verb in Hindi has a causative (morphological)
counterpart then it does not have another transitive (morphological/
semantic) counterpart in the language. This is predictable within
our hypothesis that causative and transitive verbs express a similar
causal relationship. Therefore, the existence of another transitive
verb (besides causative) is only redundant.
(i) Intransitive verbs such as jana 'to go' which have transitive
(semantically causative) counterparts, i.e., bhejna 'to send' fail to
undergo the process of causativization. Thus causative verbs such as
janana/janwana 'to make go, send' are absent in the language.
6.0 Conclusion
The major points in the preceding discussion can be summarized
as follows: (a) agency, volitionality, activity and the involvement
of two participants are four essential features of transitivity in Hindi,
(b) transitivity is a continuum on which some verbs are more transitive
than others, and (c) the features of transitivity in Hindi are similar
to those of causativity in Hindi.
There are some questions which need to be answered in this context.
(a) Since Hindi has a morphological process of deriving transitive
verbs from their intransitive counterparts by using the causative suffix
/a/ and since transitive verbs express a causal relationship with
their intranstive counterparts, is it plausible to propose that all
transitive verbs in Hindi are derived from their intransitive counterparts?
If so, what is the process of the derivation? Should we assume (47)
to be the basic structure (pre-lexical) for the transitive verb karna
'to do'?
(47)
Cause
hua
happened
If the above structure is assumed, we will have to define transformations
which would transform the subject kam into the object by substituting
the verb karna 'to do' for hona 'happen'.
(b) Should we treat the processes of ko/se markings (on the subject)
as filters which block the derivation of transitive verbs? If so,
at what stage in the derivation should they apply?
174
The above questions have definite theoretical/empirical implications
and a great deal of research needs to be done before making any conclusive
statements.
(c) How do we define the relative status of a transitive verb?
Should inchoative type verbs, i.e., janna 'to know', etc. be placed
lower on the hierarchy than transitive verbs with inanimate agents?
While inchoative verbs would be placed very low on the activity scale,
verbs with inanimate agents would be placed low on the volitionality
scale. The question then is: Are there any language particular/ language
independent parameters to judge the relative status of the transitive
verb?
(d) Pairs of verbs such as kam hona 'for work to happen' versus
kam karna 'to do the work' exist also in a number of other Indo-Aryan
languages, such as Marathi, Gujarati, etc. It would be interesting
to find out the historical development of such pairs of verbs in these
languages .
(e) A close examination of the data shows that the link between
transitivity and causativity in Hindi is better understood if we examine
their function in the language use. Lyons 1977:511 points out "...
so far as transitivity and causativity are associated with motion from
a source to a goal, there may well be grounds for believing, as many
scholars have done, that in referring first to the agent one is adopting
as the communicative point of departure what is also the more natural
cognitive point of departure."
In the context of the four features of transitivity I examined
some Hindi short stories in order to identify the contexts where transitive/
causative versus intransitive verbs are used. There is clear evidence
to assume that transitive/causative verbs are used where the theme
moves ahead, i.e., where the author describes the events in the story,
while intransitive verbs are used where the background information
necessary to develop the theme is presented. Thus descriptions of the
psychological states, nature, or the effect of the situations on the
characters is presented; more often than not, intransitive verbs are
used. In contrast to this, where deliberate actions are described,
transitive and causative verbs are used by the authors.
Consider the following paragraph (48a) which is from a Hindi short
story entitled 'ek balig aurat ka faisla' 'the decision of a mature
woman. ' The story focuses on a woman named Rita whose life has been
controlled by other people, in particular her brother-in-law, Harish.
The story portrays two sides of Rita's personality, i.e., Rita as a
weak woman and Rita as a mature woman who makes a decision about her
own marriage. The major event of the story is Rita's conversation with
her brother-in-law where Rita announces her decision. She despises her
brother-in-law for being selfish and hypocritical and who has been
pretending to be her guardian for his own selfish interests.
175
(48a) Transitive Verbs:
umra bhar ki cuppi ke "bad aj us ne pahli bar zaban kholl thi aur
vah bahut-kuch kahna cahti thi magar Haris ne kahne ka awSar nahi
diya. vah sayad tevar bhap gaya aur cupcap kamre se bahar cala gaya.
Jane ke bad hi rita ne use 'dhogi' pakhandi kahkar dilki bhanyas mikali .
vah iske alava bhi bahut kuch kahna cahtl thi aur Haris ko
sunakar kahna cahti thi. Isliye tanik rukkar phir budbudati hi
"me janti hu ki turn ne saj janta ka yah mukhauta laga rakha ht , use ye
ujle kapfe nahi chipa sakte. '
After long years (life time) of silence, she had _o££ned.her mouth
today for the first time and she wanted to say a lot, but Harish did
not give her any opportunity to do so. Perhaps he anticipated her mood
and quietly slipped out of the room. After, he had left, Rita called
him 'hypocrit', 'fraud' and released the tension of her heart. However,
she wanted to say a lot more and she wanted to say it in Harish 's presence
so that he would listen to it. Therefore, after a little while she
says angrily "1 know you have put on this mask of a gentlemen, your
smile is fake, too. Your soul is full of deceit and even your clean
clothes cannot hide it."
(48b) Intransitive Verbs:
Rita ke cehre ki tani hui rage kuch dhili par jati h^. Use apne
bhitar ek tarah ki rahat mahsus hoti ht. 'Adhik nahi kah sakti to na
sahi; kahne-simne ka uddesya pura ho gaya. Haris sirf ciyha
hi nahi .jal-bhun gaya, bilbila utha. 'Acchal' kewal sabda nahi,
uski ghayal atma se nikli hui cikh thi .
The strained nerves on Rita's face became loose and relaxed. She
had a feeling of satisfaction. She could not say more - it was all
right. The whole purpose of her talk was to get Harish angry and to
sooth her wounded heart. That goal was achieved. Harish not only
got angry but he was almost burnt with it. "Good."'- was not only a word -
a sound, it was a scream of her wounded soul.
Ek balig aurat ka faisla 'The decision of a mature woman': Rahbar: 1965
In 'Hindi ke lokpriya kathakar; paccis sre^th kahaniya. Kewal S. ed.,
Delhi, N. D. Sahgal and Sons. Pp. 99-110.
Notice the use of transitive as well as causative verbs underlined
in C48a) . Contrast (48a) with (48b) where the effect of the above
event on Rita is described. Notice the use of intransitive verbs in
(48b) .
Now consider the paragraph (49a) which is from nafrat ki maut
'the death of hatred'. One of the major events in the story is described
in (49a). Chandraprakash sends his own story to a publisher who eventually
rejects it; (49a) described Chandraprakash 's effort in putting the
176
story together and in sending it to the publisher while (49a) described
the reaction of Chandraprakash 's wife to the letter of rejection by the
publisher. Notice that more causative/transitive verbs are used in
(49a) while more intransitive verbs are used in (49b) .
(49a) Transitive verbs:
is kahani par Candraprakas ne mahino parisram kiya tha bar bar
kata tha, joya tha, ghataya tha^ bayhaya tha. hhejne se pahle usne
apne mitr Ramcandra se use mote hare kagaz par t^ip karaya tha. ba^-i
asa ke sath Candraprakas ne use ek si patrika ke sampadak ke pas bheja
tha jo use bahut pasand thi . savitri se usne pucha bhi tha, "kahani
thik ht • chap jaegi- Aur Savitri ne apne us kamzor k§ap me Candraprakas
ki drsti me jam! asa ko dekhkar kaha tha "ha zarur chap jaegi".
Candraprakash had worked hard on this story - he had cut it apart -
and joined it back several times. He had shortened it and added to
itt Before he sent it out, he had gotten it typed on thick green paper
by'his friend Ramcandra. With great hope, he had sent it to the editor
of a magazine which he liked a lot. He had asked Savitri also "Is
the story all right? Will it get published?" She saw the hope in
Candraprakash ' s eyes and in that weak moment of hers she said, "Yes,
it will be certainly published."
(49b) Intransitive verbs:
Pandrah din aur bit gae aur patrika se koi jawab nahi aya to
Candraprakas ne ek din kaha tha "kahani abhi tak vapas nahi ai .
Sayad kahani chap j aegi . "
Fifteen more days passed. When he did not get any reply from
the magazine, then Candraprakash had said one day, "The story has not
come back yet. Perhaps it will be published."
Intransitive verbs:
Aur ab "kahani sadhanyavad lautai ja rahi h£ , kyo ki ham iska
upyog karne me asamarth ht."
Yah kahani ka aswikrtipatr nahi tha, savitri ki sari abhilasaoka
aswikftipatr tha. . .
Savitri ke hoth simat gae, uska dil sima^ gaya, uska dimag simat
gaya aur uske antar me sab kuch nafrat ke ek bindu me kendrit hokar
rah gaya. Aur jyo jho samay bitta gaya sekand» mina-f,, aur ghan^e
bitte gae, yah bindu uske man me dhasta gaya.
And now "the story is being returned to you, we are unable to
make use of it."
177
This was. not the letter or rejection of the story it was the
letter of rejection of all of her aspirations...
Savitri's lips puckered, her heart became motionless, her mind
stopped functioning. Everything got concentrated on one point in
her heart. As the time passed - second, minutes, hours, this spot
kept s^tling_ deeper and deeper into her heart.
Nafrat kl maut 'The death of hatred', kulbhusan. 1965. In
Hindi ke lokpriy kathakar: Paccis sresth kahaniya.' Shaktipal kewal
ed., Delhi N.D. Sahgal and Sons.
Examples C48a)-C49b) show that the causative/ transitive verbs are
used to describe volitional acts which are central to the theme of the
story while intransitive verbs are used to describe states (or the effect
of the major events on the characters) . This observation is in
consonance with the hypothesis in the preceding discussion according to
which transitive/ causative verbs express deliberate acts.
NOTES
An earlier version of this paper was presented at the 10th Wisconsin
Conference on South Asia held in November 1981.
There are a few verbs in Hindi (e.g., pahanna 'to wear') whose
perfective participles may agree with either of the two participants in
the act (i.e., agent or patient). However, the class of the above verbs
is very small. The reason for this phenomenon is discussed in detail
in Pandharipande and Kachru 1976.
2 - - - -
Both janna 'to know', and pahcanna 'to recognize' undergo Passive
as is evident from the following examples,
(a) yah jani, gaya h£ ki amrika ab sastra nahi bhejegi
this knew went aux that America now weapons neg. will send
It has been found out that America will not send the weapons, now.
(b) Jvah I (mujhse) pahcana nahi gaya
( usko )
he (by me ) knew not went
a) He was not recognized by me.
b) I could not recognize him.
3
Causativity is used as a technical term in linguistic studies
(Lyons 1977, etc) to denote "the notion of causality,
according to which agents are seen as causes of the situations which
they bring into existence" (Lyons 1977:490). While causality refers
to the cause-effect relationship in general, causitivity refers to a
specific type of causality as mentioned above.
178
4
In the metaphorical usage of the language (44) and (45) are
acceptable. In this case khusi 'happiness '/kosis 'effort' are viewed
as agents.
REFERENCES
ANDERSON, S. 1977. On the mechanisms by which languages become
ergative. In Charles Li, ed.: Mechanisms of syntactic change.
Austin, TX: University of Texas Press.
BHATIA, T. K. 1981. Treatment of transitivity in the Hindi grammatical
tradition. (in this volume)
BINNICK, R. 1976. The iffiness of transitive verbs. In M. Shibatani, ed.:
Syntax and semantics volume 6. The grammar of causative constructions,
NY: Academic Press.
COMRIE, B. 1978. Ergativity. In Winfred P. Lehman, ed: Syntactic
typology: studies in the phenomenology of language., pp. 329-394.
Austin, TX: University of Texas Press.
DIXON, R, M. W. 1977. The Dyirbal languages of North Queensland.
Cambridge: Cambridge University Press.
. 1979. Ergativity. Language 55.1.59-138.
DOWTY , D. R. 1972. On the syntax and semantics of the atomic predicate
cause. In Papers from the Eighth Regional Meeting of the Chicago
Linguistic Circle. Chicago: Chicago Linguistic Circle.
GURU, K. P. 1952. Hindi vyakrafl. Banaras : Kashi Nagri Pracharini
Sabha .
HOCK, H. H. 1980. The modern Indo-Aryan ergative and passive constructions:
a historical perspective. Paper presented at the South Asian
Roundtable held at University of Illinois, Urbana.
HOOK, P. E. 1974. The compound verbs in Hindi. Ann Arbor, Michigan
Center for South and Southeast Indian Studies, University of
Michigan.
JESPERSON, 0. 1964. Essentials of English grammar. Alabama: University
of Alabama Press.
KACHRU, Y. 1976. On the semantics of the causative construction in
Hindi-Urdu. In M. Shibatani, ed.: Syntax and semantics volume 6.
The grammar of causative constructions. NY: Academic Press.
. 1980. Aspects of Hindi grammar. New Delhi: Manohar Publica-
tions.
. 1981. Transitivity and volitionality in Hindi. (in this
volume)
, B. Kachru and T. K. Bhatia. 1976. The notion 'subject':
a note on Hindi-Urdu, Kashmiri, and Punjabi. In M. K, Verma, ed.:
The notion of subject in South Asian languages. Department of
South Asian Studies, University of Wisconsin: Madison Publication
Series .
and R. Pandharipande. 1979. On ergativity in selected
South Asian languages. In B. B. Kachru, ed.: South Asian Languages
Analysis, volume I. Urbana-Champaign: University of Illinois.
KULBHOSAN. 1965. Nafrat ki maut. In Shaktipal Kewal, ed.: Hindi ke
lokpriya kathakar: paccis srejfh kahaniya, Delhi: Sahgal and Sons.
179
LYONS, J. 1977. Semantics volume II. Cambridge: Cambridge
University Press.
MASICA, C. P. 1976. Defining a linguistic area: South Asia. Chicago:
University of Chicago Press.
MCCAWLEY, J. 1968. Lexical insertion in a transformational grammar
without deep structure. In B. Darden, C. Bailey, and A. Davison,
eds.: Papers from the fourth regional meeting. Chicago: Chicago
Linguistic Society.
. 1971. Prelexical syntax. In O'Brien, ed. : Report of the
22nd annual roundtable meeting on linguistics and languages studies,
Washington D.C.: Georgetown University Press.
1976. Remarks on what can cause what. In M. Shibatani, ed.:
Syntax and semantics volume 6. The grammar of causative. NY:
Academic Press.
PANDHARIPANDE, R. 1978. Exceptions and rule government: the case of
the passive rule in Hindi. Studies in the Linguistic Sciences 8.1.
. 1979. Postpositions in passive sentences in Hindi. Studies
in the Linguistic Sciences 9.2.
. 1981. Syntax and semantics of the passive construction in
selected South Asian languages. Ph.D. dissertation. University
of Illinois at Urbana-Champaign.
. 1982. A constraint on passive: more evidence from Hindi.
(to be presented) Third International Conference on South Asian
Languages and Linguistic to be held in Mysore, India. January 1982.
POSTAL, P. and D. Perlmutter. 1974. Lectures on relational grammar.
Delivered at the Amherst Linguistic Institute (unpublished) .
RAHBAR, H. 1965. ek balig aurat ka faisla. In Shaktipal Kewal, ed.:
Hindi ke lokpriya kathakar: paccis sresth kahaniya. Delhi, N.D.:
Sahgal and Sons .
SHIBATANI, M. 1976. The grammar of causative constructions: a conspectus,
In M. Shibatani, ed.: Syntax and semantics volume 6. The grammar
of causative constructions. NY: Academic Press.
VAJPEYI. 1959. Hindi shabdanushisan. Banaras : Kashi Nagri Pracharini
Sabha .
Studies in the Linguistic Sciences
Volume 11, Number 2 , Fall 1981
TRANSITIVITY AND VOLITIONALITY IN HINDI -URDU
Yamuna Kachru
University of Illinois, Urbana-Champaign
This paper discusses the syntactic and semantic categories
of verbs such as intransitive-transitive, stative-inchoative-
active, volitional-non-volitional, and attempts to answer
the question: Are all active verbs volitional in Hindi-
Urdu? In order to answer this question, the paper discusses
the grammatical devices that signal the volitional-non-
volitional distinction in Hindi-Urdu. The productive
processes that change volitional verbs into non-volitional,
or vice-versa, are also discussed. Finally, crucial evidence
is provided to show that transitivity, volitionality and
action are three distinct parameters that vary independently
of each other in the language.
INTRODUCTION^
1.0 On the basis of syntactic facts, Hindi verbs are classified
into categories such as the following:
l.a. intransitive-intransitive-complex transitive-ditransitive-
■z
causative-^
4
b. unmarked subject-oblique subject
In addition, on the basis of semantic facts with important syntactic
consequences, verbs are categorized as follows:
c. stative-inchoative-active
d. volitional-non-volitional^
The categories of stative-inchoative, dative subject, and non-volitional
verbs coincide and are not necessarily mutually exclusive. Verbs that
are stative or inchoative and take a dative subject are invariably
non-volitional; however, not all non-volitional verbs take dative
subjects. Some intransitive and transitive verbs take unmarked
subjects, or subjects marked with postpositions other than n£ which are
non-volitional. All active intransitive and transitive verbs (including
complex transitive, di transitive and causative verbs) are volitional.
The facts presented so far seem fairly straightforward. These
may, then, suggest the following question: If all active verbs are
volitional, what justification, if any, is there for positing voli-
tionality as an independent characteristic of Hindi verbs? Would
not 'active' redundantly specify 'volitionality' as well? In this
paper, I will attempt to justify the notion 'volitionality' independent
of the notion 'action'.
182
GRAMMATICAL PROCESSES THAT SIGNAL "VOLITIONALITY"
2.0 There are several grammatical devices by which the volitionality
feature associated with a verb may be signalled. In the first part
of this section, 1 will discuss the pairs or triples of verbs which
express the distinction 'volitional - non-volitional', then I will
discuss the productive grammatical processes that are used to signal
this distinction.
The first relevant grammatical fact is the pairing of intransitive -
transitive verbs through the morphological process of causativization.
Consider, for example, the following:
2. vah phuulo kii maalaa le kar jaa rahaa thaa. kisii ke dhakke
se maalaa kiicaR me gir gaii.
He was going with a garland in his hand. He dropped the
garland (accidentally) when someone dashed into him ( acci dentally )
3. vah phuulo kii maalaa le kar jaa rahaa thaa. kisii ke dhakke
se *usne maalaa kiicaR me giraa dii.
garland into the mud when someone dashed into him (accidentally) .
h. vah phuulo kii maalaa le kar jaa rahaa thaa. kisii ne dhakka
de kar maalaa kiicaR me giraa dii .
He was going with a garland in his hand. Someone dashed
into him and caused the garland to be dropped into the
mud deliberately.
In 2, an accidental happening is described; in 3, the adverbial
kisii ke dhakke se suggests an accident, but the use of the transitive
verb giraa dii suggests a deliberate act, hence, the sentence is not
well-formed. In 4, both the adverbial and the verb express deliberate
action and the sentence is well-formed.
The second grammatical process is that of using a select number
of verbalizers with adjectives or nouns to create new verbs, which
is productive. This process has resulted in a large number of verbs
described as nominal compound or conjunct verbs (Kachru 1980:61-62).
The verbalizers used are mainly honaa for stative, honaa, bannaa , or
aanaa for inchoative, and karnaa for active (see also Bhatia (in this volume))
The following sets illustrate this process. Note that the stative
inchoative verbs made up of noun + verbalizer take dative subjects.
5. rajnii ko tumhaarii baate yaad hai. (stative)
Rajnii remembers all that you said.
6. ye patra dekhte hii rajnii ko tumhaarii baate yaad aaengii.
(inchoative)
Rajni will remember all that you said as soon as she sees
these letters.
183
7. usne baarah gune baarah tak kaa pahaaRaa yaad kar liyaa hai .
(active)
He has memorized multiplications tables up to twelve times twelve.
8. sushma xush hai . (stative)
Sushma is happy.
9. apnii saheliyo ko dekh kar sushmaa xush ho gaii. (inchoative)
Sushma was (became) happy to see her friends.
10. kaii kiimtii upahaar de kar raanii ne sushmaa ko xush kar liyaa.
(active)
Rani made Sushma happy by giving her many expensive gifts.
Adverbials such as jaan-Duujh kar 'knowingly', man lagaa kar
'attentively', soc-samajh kar 'after careful consideration', koshish
karke 'with effort' and others that express deliberate action occur
in sentences such as 7 and 10; their use in sentences such as in
5-6 and 9-10 yield ill-formed sentences, as is evident from 11-13.
11. *rajnii ko jaan-buujh kar tumhaarii baate yaad hai .
Rajni deliberately remembers all that you said.
12. *sushmaa jaan-buujh kar xush ho gaii .
Sushma deliberately became happy.
13. *vah jaan-buujh kar diikhii hai ki aapse nahii mil sakaa.
He is deliberately sorry that he could not meet you.
Evidence for the claim that this process of selecting appropriate
verbalizer and using dative subject construction is productive comes
from ongoing borrowing of English nouns, adjectives, and verbs to
create new conjunct verbs such as the following.
ik . kamiTii ko aapke bajaT par aabjekshan thaa, isliye aapkaa
aplikeshan rijekT ho gayaa.
The committee had objections on your budget, therefore,
your application was rejected.
15. kamiTii ke kaii membaro ne aapke bajaT par aabjekshan kiyaa/
uThaayaa , isliye aapkaa aplikeshan rijekT kar diyaa gayaa.
Several members of the committee raised objections on your
budget, therefore, your application was rejected.
Again, 14 expresses a state of affairs with no implication of deliberate
action on anyone's part, whereas 15 expresses a volitional act on
the part of an agent. Therefore, an adverbial such as aapko pareshaan
karne ke liye 'in order to harass you', if inserted in 14 would result
in an ill -formed sentence, but would yield a grammatical sentence
in 15.
A third process, rather restricted, is that of intransitivizing
184
transitive verbs by using morphological processes similar to the ones
used for causativization ('anticausative' in Masica 1976). The
following illustrate this process.
16. mai kapRe dho rahii huu. (active, transitive, volitional)
I am washing clothes .
17. kapRe dhul rahe hai. (inchoative, intransitive, non-volitional)
The clothes are getting washed.
18. usne dhyaan se tasviire dekhiJ . (active, transitive, volitional)
He looked at the pictures attentively.
19. bacca rone hii vaalaa thaa ki duur se use apnii maa aatii
huii dikhii aur vah uskii or da-uRaa. (inchoative, intransitive,
non- volitional )
The child was about to cry when he saw his mother coming
from a distance and ran toward her.
20. j&h merii koii naii kitaab chap kar aatii thii to mere ghar
me ek utsav man jaataa thaa. ' Baccan ' : niiR kaa nirmaaN phir
Whenever one of my new books was printed and delivered, a
festival got celebrated in my house. (intransitive mannaa
derived from transitive manaanaa 'celebrate')-
Verbs such as dhulnaa 'get washed', dikhnaa 'become visible', mannaa
'be celebrated', kahlaanaa 'be named' are used as intransitive or
dative subject verbs to express non-volitionality. In Eastern Hindi,
the first causal of affective verbs (Kachru 1980:96) are used to signal
the non- volitional meaning as in the following.
21. zor se bolo, kuch naliii sunaataa.
Speak loudly, nothing is audible (i.e., I can't hear you).
22. hamse ThaNDaa khaanaa nahii khilaataa.
I can't eat cold food.
A fourth process is that of using abstract verbal nouns derived
from active transitive verbs or non-stative adjectives with appropriate
verbalizers to express the non-volitional meaning. Compare the
following.
23. kal mai ek mahfil me gayaa thaa jahaa *mujhe baRii acchii
Gazle sunaaii paRJi.
Yesterday I went to a concert where *I accidentally heard
very good Ghazals .
2it. kal mai ek mahfil me gayaa thaa jahSa mai ne baRii acchii
Gazale sunii.
Yesterday I went to a concert where I listened to many good
Ghazals .
185
25. vahaa uskii baRii buraaii ho rahii thii,
There he was being maligned.
26. suresh ramesh kii buraaii kar rahaa thaa.
Suresh was maligning Ramesh.
27. *mujhe dhyaan se acchii Gazle svmaaii paRJi.
I (accidentally) heard very good Ghazals attentively.
28. *vahaa jalan ke maare uskii baRii buraaii ho rahii thii.
He was being maligned there because of jealousy.
29. suresh jalan ke maare ramesh kii buraaii kar rahaa thaa.
Suresh was maligning Ramesh because of jealousy.
Sentence 23 is ill-formed because after specifying a context for
deliberate action, the second conjunct indicates an accidental event.
Sentence 27 is ill -formed because the adverbial suggests a deliberate
action whereas the verb expresses a non-volitional, accidental event.
The same is true of 28. The adverb suggests a volitional agent capable
of feeling jealous, the verb expresses a non-volitional process.
Sentences 24, 25, 26, and 28 are well-formed since there are no contra-
dictions in them.
A fifth, productive process for signalling non-volitionality is
that of verbal compounding. Consider the following.
30. raakesh ne xuub soc-samajh kar sab baate kahJi.
Rakesh said things after careful consideration.
31. raakesh binaa soce-vicaare sab baate kah gayaa.
Rakesh blurted out things without thinking.
32. *raadhaa soc-vicaar kar sab baate kah jaatii hai .
Radha blurts out everything after careful consideration.
33. *soc-samajh kar pati-patnii aapas me laR baiThe.
After careful consideration, the couple had a fight.
3^. vah jaan-buujh kar apne bhaaii se laR rahaa hai .
He is deliberately fighting with his brother.
Verbs such as kahnaa 'say' and laRnaa 'fight' are active, volitional
verbs, hence, sentences 30 and 34 are grammatical. Compounding with
j aanaa 'go' or baiThnaa 'sit' negates the volitionality of the main
verb, hence, 32 and 33 with the deliberate action adverbials are ill-
formed. Sentence 31, on the other hand, contains an adverbial that
expresses a lack of deliberate action which is in consonance with the
compound verb kah j aanaa 'blurt out', hence, it is well -formed.
The explicators that signal the lack of volitionality are j aanaa 'go',
uThnaa 'rise', baiThnaa 'sit', and paRnaa 'fall' - all intransitive.
Note that compounds made up of transitive verbs and these explicators
are not used in imperative sentences either, which confirms the hypo-
thesis that these compound verbs are non-volitional .6 Sentences 35-37
are ill-formed.
186
35. *sa'b baate kah jaaol Bl\irt out everything!
36. *is peR ko kaaT baJTho.' Chop down this tree.'
37- *baRo ke biic me bol uTho! Speak up among the elders!
A sixth device utilized for signalling diminished volitionality
is the use of adverbials such as beman se 'unwillingly', na caahte
hue bhii 'without wanting to', anicchaa se 'without wanting', etc.
That these adverbials reduce volitionality is clear from the fact
that these sound odd in direct imperatives.
38. ??beman se sab baate bataao !
Tell everything unwillingly!
39. ??na caahte hue bhii gaanaa sunaaol
Sing a song without wanting to !
They, however, sound perfectly natural in constructions that express
some external compulsion.
ko . beman se hii sahii , tumhe sab baate bataanii paRengii.
You will have to tell everything even though you are unwilling.
kl. na caahte hue bhii use baas ke yahaa paarTii me jaanaa paRaa.
Although he didn't want to, he had to go to the boss's party.
The main verb in the infinitive followed by the item paR- in 40 and
41 with the dative subject expresses external compulsion (Kachru
1980:50).
So far, I have discussed devices used to signal volitionality,
and processes used to negate or reduce the volitionality of volitional
verbs. I shall now discuss the devices that make non-volitional verbs
volitional .
The first productive device is compounding with lenaa 'take',
denaa 'give', Daalnaa 'pour or add", and maarnaa 'hit'.? Verbs that
express reflex action can be made volitional by compounding with
explicators from this set. Consider the following.
k2. mere calte vaqt use chJik aa gaii to vah baRaa sharmindaa huaa.
He sneezed when I was about to leave and felt ashamed of
himself.
U3. mere calte vaqt usne ek chiik maarii aur yah kah kar mujhe
rok liyaa ki is apshakun ke baad ravaanaa honaa Thiik nahii.
He sneezed just as I was about to leave and stopped me by
saying that it would not be proper to set out after this
bad omen.
Sentence 42 expresses a reflex action, sentence 43 a deliberate
one. Adverbs such as jaan-buujh kar 'deliberately' are not appropriate
in 42, they would be appropriate in 43.
187
Similarly, transitive verbs such as j aanaa 'know', and samajhnaa
'understand', which are neither clearly volitional, nor non-volitional,
acquire clear volitional or non-volitional force by compounding. 8
Consider the following.
hk. *kal tak is savaal kaa javaab j aano !
Know the answer to this question by tomorrow.'
i+5- yaii jaan lo ki vah tumhaarii madad nahii karegaal
You must realize that he will not help you in any way.
k6. *ramesh koshish karke jaan gayaa ki paise kahaa rakhe hai .
Ramesh knew with effort where the money was kept .
U7. unke binaa kahe hii vahOaan 7 gayaa ki unhe uskii baate
lii vahriaan 7 j
/ sama.ih\
acchii nahii lag rahii hai.
Without their saying anything, he knew that they did not like
what he was saying.
k8. mai ne sam.jhaayaa to usne savaal acchii tarah se samajh liyaa.
I explained and he understood the question well.
Since j aanaa 'know' is not volitional, 44 is ill-formed. Sentence
45, on the other hand, is grammatical because compounding with lenaa
'take' makes it volitional. Sentence 46, with jaan j aanaa, is
ill-formed because the adverbial koshish karke 'with effort' suggests
a deliberate action, but the compound verb expresses a non-volitional
process ('event' in the sense of Lyons 1977). Both 47 and 48 are
well -formed because there are no contradictions in them, they express
a non-volitional process and a volitional act respectively.
It is interesting to contrast the explicators lenaa and denaa
with the explicators paRnaa and j aanaa with a variety of verbs.
The verb hasnaa 'laugh', for instance, may express a reflex action,
or a deliberate action. Note the following where the compound has
paRnaa denotes the reflex action whereas the compound has denaa expresses
deliberate action.
k9, is tarah udaas kyo baiThe ho, *zaraa has to paRo!
Why are you sitting there looking so sad, burst out laughing.
50. is tarah udaas kyo baiThe ho, zaraa has to do!
Why are you sitting there looking so sad, laugh a little.
The compound has paRnaa expresses spontaneous reaction, hence
is inappropriate in an imperative sentence such as in 49. The compound
has denaa is entirely appropriate in 50. The same is true of kho j aanaa
'lose (accidentally)' and kho denaa 'lose (deliberately)' as in
51 and 52.
51. baKTii kii cabhii kho gaii thii , isliye use der tak baahar
hii maa kaa intazaar karnaa paRaa.
Banti's key got lost, so he had to wait for his mother for
a long time outside.
52. BaNTii nahii caahtaa thaa ki uskii maa naiikrii kare, isliye
baar baar apnii caabhii kho detaa thaa. 9
Banti did not want his mother to work, so he would lose
his key repeatedly.
Higher verbs such as koshish karnaa 'try', caahnaa 'want', tay
karnaa 'decide' that clearly express volitionality impose a volitional
interpretation on the lower verbs even if they are basically non-
volitional .
53. vah jaan-buujh kar bhiiR me kho jaane kii koshish kar rahaa
thaa par mohan ne use dekh hii liyaa.
He was deliberately trying to lose himself in the crowd,
but Mohan spotted him.
5^+. vah caahtaa thaa ki dal ke har vyakti ke baare me sab kuch
jaan jaae.
He wanted to know everything about everyone in the party.
Both 53 and 54 are well-formed since the higher verbs determine
the nature of the subordinate verbs.
The process of using adverbials such as jaan-buujh kar 'deliberately'
and svecchaa se 'of one's own volition' to mark non-volitional verbs
as volitional is also utilized.
55. use dekhte hii jaan-buujh kar gir paRo taaki use vishvaas
ho jaae ki tumhaarii tabiiyat Thiik nahii.
As soon as you see him, deliberately fall down so that he
concludes that you are not well .
56. hasii na bhii aae to koshish karke uske har mazaak par
hasnaa.
Even if you can't lavigh (naturally), try and laugh at all
his jokes.
In 55, gir paRnaa is volitional and so is hasnaa in 56.
CONCLUSION
3.0 The picture that emerges from the above discussion is rather
complex and worth exploring further. It is clear that transitivity,
action, and volitionality are independently variable parameters and
various devices are utilized to vary them in Hindi. The following
sentences show that in fact, transitivity, action, and volitionality
are independently variable by utilizing the devices discussed so
far. The verb kho jaanaa is intransitive and non-active, non-volitional
in 57, hence 58 is ungrammatical .
189
57. merii cathii kho gaii.
My key got lost.
58. *merii cabhii kho jaa rahii hai.
My key is getting lost.
The verb kho denaa, on the other hand, is transitive, active, and
volitional, and yet does not denote action, hence 60 is not well-formed.
59. usne apnii kalam kho dii .
He lost his pen (intentionally).
60. *vah apnii kalam kho de rahaa hai ,
He is losing his pen.
Note that the ungrammaticality of 60 is not due to the fact that
kho denaa is momentary, even inserting an adverb such as baar baar
'repeatedly' does not improve it. Other verbs such as kuudnaa ' j ump ' ,
which also express momentary dynamic situations, occur with progressive
and have the interpretation 'repeated' action, since they are verbs
of action. It may be argued that compound verbs signal the
perfective meaning (Porizka 1967-69, Hook 1974; but see Kachru 1979),
therefore, it is natural that they would not occur in the progressive.
Note, however, that compound verbs such as calaa jaanaa 'depart',
de denaa 'give away' and bataa denaa 'divulge secrets', which are
action verbs, do occur in the progressive.
It has already been established that volitionality is crucial
for passivization in Hindi (Pandharipande 1981) . One way of establish-
ing volitionality as an independent characteristic of the verb in
Hindi would be to test if the processes discussed above for increasing
or reducing volitionality have any consequences for passivization.
Although verbs such as hasnaa 'laugh', ronaa 'cry', maarnaa 'hit',
kahnaa 'say' are passivizable in Hindi, the compounds hgs paRnaa,
ro uThnaa, maar baiThnaa and kah jaanaa are not. However, compounds
such as kho denaa, jaan lenaa are readily passivized as is clear from
the following examples.
61. thakaavaT ke maare usse hasaa bhii nahJi jaa rahaa thaa.
He could not laugh because of being tired. (i.e.. He was
so tired he could not laugh . )
62. *uskii bet\ikii baate sun kar bhii usse has paRaa nahJi gayaa.
Even though he heard his nonsense, he covild not burst out
laughing.
63. pahle sab baate Jaan lii jaae, tab koii faislaa ho.
First every information should be gathered, then decisions
should be made.
190
There are many points that can be discussed with regard to the
notions volitionality, action, and transitivity. One such point is
that the interaction of distinctions in terms of transitivity, stativity, and
vjiitionalitv offers immense possibilities for stylistic exploitation.
(See Appendix for some examples.) A study of this phenomenon, both
in speech events and in literary texts of various genres should prove
to be both fascinating and rewarding.
NOTES
An earlier version of this paper was presented at the Tenth
Wisconsin Conference on South Asia held at Madison, Wisconsin, November
8, 1981.
2
See the Appendix for examples of all the classes of verbs mentioned
in l(a)-Cd).
3
This classification is strictly on the basis of valency, i.e.,
the number of arguments a verb takes (see Lyons 1977, 481-488 for an
extended discussion of valency) . The notions relevant to this paper
are the following: 'stative' refers to a 'static situation or state';
'inchoative' refers to 'dynamic situations (either momentary or
extended through time) which are not under an agent's control'; 'active'
refers to 'dynamic situations under the control of an agent'. (See
Pandharipande (in this volume) for the syntax of transitivity.)
Oblique subjects are marked with postpositions such as ne_ and
se (active and passive agentives) and ko^ ('dative' subject construction).
Notional subjects may be marked with other postpositions also, e.g.,
instrumental se_, genitive kaa, (see Bhatia (in this volume)).
See Kachru 1980 for a discussion of these distinctions.
See Kachru and Pandharipande 1980 for a detailed discussion of
the relevance of the notion 'volitionality' for verb compounding in
Hindi-Urdu and other selected South Asian languages.
See footnote above. See also Kachru 1979 for a discussion of
the semantics of the explicator verbs.
a
Note that jaanaa 'know' and samajhnaa 'understand' do not behave
like regular transitive verbs in all contexts. The verb j aanaa
'know', when stative, does not occur in the progressive, neither does
the compound verb jaan lenaa as it denotes an 'act' rather than an
•action' (Lyons 1977, loc.cit.). The verb sama j hnaa takes both an
unmarked subject and a marked (with ne) subject, and, when active,
occurs in progressive. A detailed investigation of the properties of
these verbs is yet to be done. Pandharipande (in this volume) discusses
the syntax of these verbs.
191
9
Verbs such as kho jaanaa are not stative. They are non-volitional
and express 'events' (Lyons 1977, loc.cit.). Verbs such as kho denaa,
on the other hand, are volitional 'acts' or 'achievements' (Vendler 1967).
As such, they do not express 'actions' and do not occur in the progressive.
APPENDIX
Sub-classes of Verbs
Ci) laRkaa ro rahaa hai.
The boy is crying,
[inchoative, unmarked subject, active, non-volitional/volitional]
Cii) bacce titlii pakaR rahe hai.
The children are catching butterflies,
[transitive, unmarked subject, active, volitional]
(iii) makaan baRaa hai.
The house is big.
[intransitive, unmarked subject, stative, non-volitional]
(iv) patte jhaR rahe hai.
The leaves are falling.
[intransitive, unmarked subject, inchoative, non-volitional]
(v) aaj shaamvaale naaTak me siimaa shankuntalaa ban rahii hai.
Sima is becoming (i.e., acting as) Shakuntala in this evening's
play.
[intransitive, unmarked subject, inchoative, volitional]
(vi) bacce ko pyaas lag rahii hai.
The child is feeling thirsty,
[intransitive, dative subject, inchoative, non-volitional]
(vii) mujhse gilaas TuuT gayaa.
1 broke the glass (unintentionally) .
[intransitive, inchoative, non-volitional; unmarked subject:
gilaas ; instrumental adverb: mujhse]
(viii) raaj ne mohan ko bevakuuf banaayaa.
Raj made a fool of Mohan,
[complex transitive, marked subject, active, volitional]
(ix) riitaa ne mitraa ko ghaRii dii.
Rita gave a watch to Mitraa.
[ditransitive, marked subject, active, volitional]
(x) mSi usse sab kaam karvaaungaa.
I will make him do all the work,
[causative, unmarked subject, active, volitional]
192
II. Note the use of bhulaanaa 'forget deliberately', bhuulnaa 'forget
accidentally', dhikhaaii paRnaa 'see', dekhnaa 'look', and
other underlined items in the following passages.
a. dhuup kii garmii se sukhii ho kar vah cintaa bhulaane kaa
prayatna kar rahaa thaa ki kisii ne pukaaraa...
Feeling content with the heat from the sun, he was trying
to forget his worries when someone called.
b. 'bhale aadmii, rahe kahSa.' saalo par dikhaai paRe. . . '
Where have you been, my good man.' I haven't seen you for
ages. . . '
c. sharaabii ne cituk kar dekhaa. koii jaan-pahcaan kaa to maaluum
hotaa thaa; par kaun hai, yah Thiik-Thiik nahii jaan sakaa.
The drunk was startled and looked at (him) . He seemed to be
familiar; but exactly who he was, the drunk could not
think of.
d. 'oho, raamjii, turn ho, bhaaii, mai bhuul gayaa thaa...'
Oh, it's you, Ramji, I had forgotten. . .
'Prasad': madhuaa
e. sacetan mashtishka ko j in baato kii yaad karne se bhii rok
diyaa gayaa thaa, unhllko acetan mashtishka jiitaa,
bhogtaa, jheltaa, sahtaa.
The unconscious mind lived, experienced, endured, and suffered
the very things that the conscious mind was forbidden even
to recall.
'Baccan': niiR kaa nirmaaN phir
f . barsaatii baahar rakh do, saaraa kamraa bhigo rahe ho.
Go put the raincoat out. You are making the whole room wet.
is ghar mS aNDe kaa naam le rahii hoo.. muh bhraST ho gayaa
hogaa.
(You) are mentioning egg in this house... (Your) mouth
must have become polluted.
caar din jo aNDe khaa liye hat, ve chilkO samet vasuul
ho jaaenge.
The eggs that (you) have eaten for a few days, (price) will
have to be paid for not only the eggs but also the shells.
. . . saare ghar kaa gangaa ishnaan ho jaaegaa.
There will be bathing in Ganges for the whole house.
193
...aur jahaa tak ammaa kaa savaal hai, arninaa inhe naalii mg
paRe hue bhii nahli dekhengii .
As far as mother is concerned, she will not look at them
even (if they are) lying in the drain (in plain sight) .
Raakesh: aNDe ke chilke
REFERENCES
BHATIA, Tej K. 1981. Treatment of Transitivity in Hindi grammatical
tradition. (In this volume) .
HOOK, Peter. 1974. The compound verb in Hindi. Ann Arbor, MI.
The Michigan Series in South and Southeast Asian Languages
and Linguistics. The University of Michigan.
KACHRU, Yamuna. 1979. Pragmatics and verb serialization in Hindi-
Urdu. SLS 9.2.157-169.
. 1980. Aspects of Hindi grammar. New Delhi, India:
Monohar Publications.
and Rajeshwari Pandharipande. 1980. Toward a typology of
the compound verb in South Asian languages. SLS 10.1.113-124.
LYONS, John. 1977. Semantics II. Cambridge, MA: Cambridge University
Press .
MASICA, Colin P. 1976. Defining a linguistic area: South Asia.
Chicago, IL: University of Chicago Press.
PANDHARIPANDE, Rajeshwari. 1981. Passive in selected South Asian
languages. Urbana-Champaign, IL: University of Illinois
Ph.D. dissertation.
. 1981. Transitivity in Hindi. (In this volume).
PORIZKA, Vincence. 1967-69. On the perfective verbal aspect in
Hindi. Archiv Orientalini 35.64-88, 208-231; 36.233-251;
and 37.19-47, 345-364.
VENDLER, Zeno. 1967. Verbs and times. In Linguistics in philosophy.
NY: Cornell University Press,
studies in the Linguistic Sclencea
Volume 11, Number 2, Spring 1982
THE TREATMENT OF TRANSITIVITY IN THE HINDI GRAMMATICAL TRADITION*
Tej K. Bhatia
Syracuse University, Syracuse
The aim of this paper is to discuss the treatment of transi-
tivity in the grammatical tradition of Hindi, beginning with the
grammars of the seventeenth century. For this purpose, at least
fifty traditional grammars, covering the approximately three
century old tradition, were analyzed. The treatment of transiti-
vity in the Hindi grammatical tradition reveals the following:
One, transitivity in Hindi involves a number of components, only
one of which is the presence of an object of the verb. Two,
other components require and interact with notions such as case,
agency, verb agreement, postpositions, precedence of animate ne
subjects over inanimate ne subjects and syntactic deletions, and
treat them as the language-specific properties of transitivity.
Three, the dynamic and complex nature of transitivity can be
witnessed in the number of transitivizing and intransitivizing
processes. Four, Hindi verbs cannot be classed into a water-
tight traditional classification of transitive and intransitive
verbs; there is a significant degree of overlapping between verbs
with respect to transitivity and intransitivity . The paper reveals
that to the list of the properties of transitivity presented by
traditional grammarians, notions such as volitionality and prag-
matics must be added in order to arrive at an adequate continuum
or treatment of transitivity in Hindi.
The picture that emerges from this historical discussion is
not only important from the view point of the syntax and semantics of Hindi
but also very crucial for linguistic theory as well as language
pedagogy, description and historical development of categories
such as compound and causative verbs, dative subjects, ergativity
and passivization.
0. INTRODUCTION
This paper presents a brief account of the treatment of transitivity
in the grammatical tradition of Hindi, beginning with the grammars of the
seventeenth century. For this purpose, at least fifty traditional grammars,
covering the approximately three century old tradition, were analyzed.
What emerges from this analysis is the following: One: transitivity in
Hindi involves a number of components, only one of which is the presence of
an object of the verb. Two: other components require and interact with
notions such as case, agency, verb agreement, postpositions, particle,
etc. Three; the dynamic and complex nature of transitivity can be seen in
the number of transitivizing processes and changeable behavior of verbs in
the Hindi language. And, Four_: the discussion of transitivity in tradition-
al grammars is not only important from the view point of Hindi syntax and
semantics but also very crucial for linguistic theory as well as from the
point of view of language pedagogy.
196
1.0 THE SYNTAX OF TRANSITIVITY
The primary goal of traditional Hindi grammarians was not to engage in
theoretical debates involving transitivity. Placing high priority on lan-
guage pedagogy, their main aim was to present the paradigms of intransitive
and transitive verbs in different tenses. During the initial stages of
their task, they were delighted to observe that the conjugation of intran-
sitive and transitive verbs was identical (noting few irregular forms).
They loved the Hindi language for its simplicity and identical treatment
of the two classes of verbs. Consequently, Ketelaar (1698), Schultz (1745),
Hadley (1772) and Ferguson (1773) overgeneralized the identical syntactic
behavior of transitive and intransitive verbs and treated sentences such as
(1) and (2) as well-formed.
(1) ham piya
we drank
(2) tum piya
you drank
However, towards the end of the 18th century, their love for the Hindi
language began to go sour and the Hindi language began to receive labels
such as 'peculiar' and 'strange'. At the root of this estranged relation-
ship was the 'peculiar' syntax of the transitive verbs, formed with past
participle. Although some traces of their distinct behavior could be wit-
nessed in the following past tense data presented by Ketelaar (1698) and
Schultz (1745),
(3) inne piye the (Ketelaar, 1698)
this (oblique) had drunk
'He had drink'
(4) un karaa (Schultz 1745, from
he (oblique) did Vecchoor, 1976:93)
'He did'
they did not pay any serious attention to the occurrence of oblique forms
of third person pronouns such as ^il ('this') and un ('he-') with the past
tense, nor did they pause to analyze the ne in (3).
The first Hindi grammar to note the distinct syntactic behavior of
transitive and intransitive verbs in the past tense was written by Gilchrist
in 1796. The ne^ construction mystified Gilchrist and his contemporary
grammarians; and they felt embarrassed for lack of a grammatical termin-
ology to describe the behavior of this construction. Forbes (1855) best
summarizes their sentiments regarding their treatment of this construction:
"Our great grammarians have succeeded wonderfully well
in mystifying the very simple (though singular) use and
application of this particle ne. Dr. Gilchrist, in the first
edition of his grammar, seems to have felt greatly embarrassed
by it, without exactly knowing what to make of it."
Forbes (1855: 105)
197
In spite of his doubt, Gilchrist succeeded well in unravelling
the mysteries of the ne construction. In his grammar (1796) and several
other works, he observed that the use of ne is governed by the 'power of
transitive verb' as exemplified in (4).
(4) mt ne khana khaya (17 96)
I agent food eat
I ate food .
Gilchrist noted that in the tenses formed with past participle, a subject
takes n£ with it. As a result, the verb does not agree with a subject.
Instead it agrees with an object.
1.1 TRANSITIVITY AND CASE
The integral relationship between ne and object-verb agreement also
revealed that transitivity interacts with case in a meaningful way in Hindi.
The lack of verb agreement with n£ subjects called for a reconsideration of
the nominative case in Hindi. Prior to Gilchrist, Hindi grammarians in-
variably assigned subjects a nominative case for reasons of subject-verb
agreement. The object-verb agreement in the ne_ construction now led them
to conclude that in such instances, object takes the nominative case in
Hindi. Although Hindi grammars noted sentences such as (5),
(5) larke ne la^ki ko dekha^
boy agent girl obj saw (M.S^ past form)
A boy saw the girl
no attempt was made to point out the difficulty of assigning a nominative
case to an object NP which does not involve verb agreement and which is
marked with the postposition ko. Monier Williams was the first to object
to the practice of assigning nominative case to such object NPs. In order
to find a solution to this problem, or avoid the whole issue, grammarians
such as Yates (1827) and Forbes (1855) turned their attention to the ne
subjects again. On the basis of diachronic and quasi-semantic (transla-
tional) evidence, they postulated cases such as ablative or instrumental
for the ne^ subjects. Drawing evidence primarily from Sanskrit and second-
arily from Latin, Forbes (1885) concluded that in sentences such as (4) and
(5),
"Nominative case changed into that form of the
ablative expressive of agent."
(Forbes, 1855: 53)
The basis of this conclusion was his observation that the ne construction
originated from the Sanskrit apadana karak (marker ejja) and in Sanskrit
ablative case-*- is found wherever n£ is used in Hindi. Such a conclusion
was also arrived at, with some minor changes though, by Yates. Rather
than utilizing historical linguistic evidence, he employed quasi-semantic
or translational evidence and concluded that
198
"The instrumental case is always used instead of the
nominative, before transitive verbs in the perfect
tense or its formatives, (in such instances the tran-
sitive verb), though active in its termination, is
passive in its signification."
(Yates, 1827: 20)
This led Yates to interpret sentences such as (6) as "by the woman,
an answer was given".
(6) aurat ne zavab diya (Yates, 1827: 20)
Woman by answer gave
By the woman, an answer was given.
Finally, during the twentieth century, grammarians closed the whole issue
of case marking by postulating a new and more satisfactory case_, "Agentive
case-ne" (See Scholberg, 19A0) . This case was not only semantically motiva-
ted but also explained the syntactic behavior of nouns which is not shared
by subject NPs in the nominative case. At the same time, it also dismissed
the objections raised against assigning instrumental or ablative case to
the ne subject.
1 . 2 AGENCY
In addition to the interaction of transitivity with subjects, verb
agreement and cases, Hindi grammarians, in the middle of the nineteenth
century, also noticed the ability of transitivity to interact with a
specific class of agents. During the discussion of instrumental or
passive case, Monier Williams (1876) made an insightful observation
that the ne construction is not only sensitive to a transitive verb but
also to the animacy of a subject. He observed that although ne can
occur with inanimate subjects as is evident by sentences such as (7) and
(8).
(7) badshah ke lahu ne jos mara
king of blood agent excitement took
The king's blood boiled.
(Williams, 1876: 127)
(8) is bat ne ' mujhe xarab kiyS
this matter agent me bad did
This thing ruined me.
(Williams, 1876: 127)
it is not a common practice to use ne_ with an inanimate subject. Instead,
it is preferred to use an intransitive verb and se^ postposition with the
inanimate subject as in (8a)
(8a) is bat se m« xarab hua
this thing by I bad happened
This thing ruined me.
This led Hindi grammarians to postulate the following two types of
agents in the Hindi language:
199
Agent
The Agent with ne
(Generally Animate)
The Agent without ne
(Animate and Inanimate)
Forbes and others stressed that "the agent with ne should be termed
'special agents', because they are restricted in use." They are generally
animate subjects which occur with a class of transitive verbs (formed with
past participle) . The -ne agents are employed by intransitive or neuter
verbs and transitive verbs such as bolna 'speak.', lana 'bring', larna 'to
fight' > and they can be either animate or inanimate. '
1.3 TRANSITIVITY AND VERB CLASSIFICATION
On the basis of their analysis of Hindi, it became quite apparent to
Hindi grammarians that the traditional notion of transitive verbs (object
N/NP) was not sufficient to account for transitivity and intransitivity in
Hindi. Therefore, they also used additional criteria to classify Hindi
verbs. However, to do so, they did not abandon the traditional classifi-
cation of intransitive and transitive verbs. They attempted to capture the
idiosyncratic syntactic behavior of Hindi verbs without sacrificing the
traditional notion of transitivity. In addition to the traditional criterion^
they listed some new features of transitivity in Hindi and finally, arrived
at the following classification of the Hindi verb:
Verb
Action
Verb
neuter
(non-active
state)
L-nel
transitive verbs
such as bolana
'to speak'
Non-Action
Verbs
(janna 'to know'
p abac anna ' to
recognize')
Action
Verbs
The inclusion of ne to define transitivity in Hindi did not win approval
without debate; it registered several protests in the Hindi grammatical
tradition. Notably Platts (1873: 171) opposed the tendency of treating
-ne verbs such as lana 'to bring' and compound verbs such as llkh b€thna
'to write' as intransitive on the basis of the fact that they are consid-
ered transitive verbs by the native speakers and grammarians.
1.4 POSTPOSITION/PARTICLE/EXPLETIVE
Having observed the interaction of ne with transitive verbs, Hindi
grammarians also addressed themselves to the question of treating ne as
a postposition. Gilchrist assigned a totally new label, "expletive",
for ne^. Although his followers employed this label for about half a
century, it was frequently felt that the label was inadequate and lacked
any serious motivation. Therefore, they began to search for a new label
and, consequently, switched in favor of a better term - particle. It was
200
further noticed that the term 'particle' fails to account for the fact
that ne acta like a postposition in the sense that it requires an oblique
case with an NP . This led the 19th century grammarians and modern grammar-
ians to treat ne as a postposition, especially sensitive to transitivity.
They also observed that postpositions such as se_, ko and ^ are sensitive
to transitivity in Hindi. The intransitive subjects take either s£, i_ or
ko postposition whereas ko is generally used with animate objects and
sometimes with inanimate objects.
1.5 TRANSITIVITY AND DELETION
With regards to deletion and transitivity, the traditional grammarians
stressed the following two facts: One : if the two subjects of a conjoined
sentence are corref erential but one is the non-ne^ subject and the other is
a +ne subject, the second subject can be deleted. Consider the following
two sentences:
(9) faqir aya ar 6 sawal kiya
saint came and ^ question did
A saint came and i asked a question.
(Williams 1876: 127)
(10) me ne ye bat suni Or i bola
I agent this talk heard and i spoke
I heard this speech and i said
(Williams 1876: 127)
Williams and others pointed out that in (9), the L+rie] subject (i.e., faqlr
ne) in the second conjunct does not need to be spelled out and similarly,
the t-ne3 subject (i.e., mT 'l') can undergo deletion in (10). However, in
spite of the deletion, the deleted subject continues to remain sensitive to
the transitivity or intransitivity of a verb, for native speakers will
supply only the C+nel subject and t^e"]] in (9) and (10), respectively and
not vice versa. Two: a surface deletion of an object NP in some verbs,
converts a transitive verb into an intransitive verb. Consider the follow-
ing sentences (ll-12a).
(11) m£ ne gana gaya
I agent song sang
I sang a song.
*(lla) mt gaya
I sang
I sang.
(lib) mt- ne gaya
I Agent sang
I sang.
(12) bacce ne khel khela
child Agent game played
A child played a game.
(12a) bacca khela
child played
A child played.
201
Notice that in (11) if an object is deleted, such deletion does not trigger
the deletion of ne. This explains the ill-formedness of (11a). However,
if an object of the verb khelna 'to play' is deleted, the verb does not
obligatorily require the ne^ with the subject. This explains why (12a) is
acceptable to Hindi speakers. Therefore, the traditional grammarians
assigned a small class of verbs such as khelna 'to play' dual membership
and explained that they are both transitive and intransitive.
2.0 DYNAMIC/NON-STATIC NATURE OF HINDI TRANSITIVITY
What captivated the traditional grammarians the most was the
changeable behavior of verbs in Hindi; this led them to a very detailed
discussion of the transitivizing and intransitivizing processes of Hindi
verbs. Their analysis of Hindi and its comparison with their native lan-
guage (i.e., English) soon revealed that several syntactic and morphologi-
cal processes can convert a transitive verb into an intransitive verb or
vice versa. Before the emergence of the twentieth century, they had
ennumerated the following main transitivizing and intransitivizing
processes.
2.1 TRANSITIVIZING PROCESSES
The earliesC^^indi grammarians (17th and 18th century) not only
observed some important transitivizing processes such as causativization,
but went beyond purely linguistic observations. They recorded their inter-
esting and subjective reactions (value judgements) together with a list of
such processes. Consider for example, John Ferguson's (1773) reaction to
causative verbs:
"There is a particular species of verbs in the
Hindustan (Hindi) for which it is difficult to find a
proper name, but whose character is to express not any
action but causing of an action. It is no small advan-
tage in the Hindustan (Hindi) to have a particular species
of verbs to express this particular meaning."
Ferguson (1773: 57)
2.1.1 CAUSATIVIZATION/TRANSITIVIZATION PROCESSES
From Gilchrist (1796) onwards, every grammarian discussed the process
of causativization. They unanimously agreed on the derivational nature of
causative verbs; and hence, derived them primarily from intransitive verbs.
A general rule presented by them to derive causative verbs is given below:
(13) By inserting long vowel -a^ and wa to the root
of an intransitive verb, the verb becomes
transitive and causal, respectively. Example:
Intransitive Transitive Causal
pakna 'to be cooked' •=^ pakana "^^^ pakwana
The derivation and basic patterns of causative verbs from intransitive
verbs and transitive verbs is summarized in Table 1 and Table 2, respec-
tively.
202
TABLE 1
Pattern
1.
Intran
sltive
Transit
ive
Non-active
Active
Transitive
C-causalJ
First Causal
Second Causal
sona
'to
sleep'
sulana
'cause X to
sleep'
sulwana
'to cause X
to cause Y to
sleep'
2.
banna
' to be
made'
- -
banana
'to make'
banawana
'to cause to
be made'
""
3.
pijna
'to be
beaten'
pi^na
'to beat'
pifwana
'to cause to
beat'
— —
4.
bigaj-na
'to be
spoiled '
bigarna
'to spoil'
bigarwana
' to cause to
be spoiled'
— —
5.
ghirna
'to be
surround-
ed'
gherna
' to surround '
ghirana
'to cause to
be surround-
ed'
6.
calna
•to
move'
calana
' to cause
X to move'
calwana
'to cause X
to cause Y
to move'
7,
- -
btthna
'to
sit'
- -
btthana
'to seat'
bithawana
'to cause X to
cause Y to sit'
8.
bhagna
'to
run'
bhagana
' to cause
X to run'
bhagwana
' to cause X
to cause Y to
run'
Traditional grammarians noticed that non-active/statlve verbs yield only
transitive and first causaitive. They fail to yield 2nd causal verbs
(except sona) . Active intransitives, on the other hand, generate both
the causals but do not produce transitives ^-causalj • In this sense
they exhibit a complementary behavior.
TABLE 2
Transitive
Pattern
Transitive
[^-causalj
First Causal
Second Causal
1
2
karna
'to do'
khana
'to eat'
karana/karwana
'to cause X to do'
khilana
' to cause X to eat'
khilwana
'to cause Y to cause X to eat'
203
2.1.2 CONJUNCT VERBS
Transitivity can be induced in nouns, pronouns, adjectives and adverbs;
and they are converted into transitive verbs by the addition of transitive
verbs such as karna 'to do', lena 'to take', dena 'to give', lagana, and
marna as shown in the following examples:
(14)
bat karna
apna karna
ace ha karna
dhire karna
nishana marna
'to talk'
'to make one's own'
'to cure'
' to slow down '
'to aim'
2.1.3 NOMINAL VERBS
Some few transitive verbs are formed by the addition of the infinitive
marker ni or by inserting -a between a noun or pronoun and infinitive marker,
Consider the following examples:
(15) dukh (noun) 'pain'r^ dukhana 'to cause pain'
apna (pro.) 'reflexive ;=^ apnana 'to make one's own'
pronoun
2.2.2 INTRANSITIVIZING PROCESSES
2.2.2.1 Complex Verbs
Intransitive verbs are derived by adding intransitive verbs such as
hona 'to be', ani 'to come', lagna 'to a noun, pronoun, adjective and adverb.
(16)
pa sand
hona
to
like'
gussa
ana/ hona
to
become/be angry
accha
lagna
to
like'
accha
hona
to
recover'
dhire
hona
to
slow down'
apna
hona
to
be one's own'
2.2.2.2 Compound Verbs
Transitive verbs behave like an intransitive verb when followed by an
intransitive verb. For example, the traditional_grammarians pointed out
again and again that although verbs such as khana 'to eat' are transitive
verbs in the sense that they take ne with a subject in the past tense. How-
ever, if a compound verb kha bg.thna or kha cukna is employed instead of
its simple verb counterpart, the compound verb ceases to be a transitive
verb, i.e., J^+ne^ verb, as shown in the following examples of Gilchrist.
(17) mfc ne khana khaya
I Agent food ate
I have eaten my dinner.
(18) mf khana kha cuka/bfctha
I food eat done/sat
I have done eating.
(Gilchrist 1808: 31)
(Gilchrist 1808: 31)
204
The above obdervation enabled them to explain the intransitive behavior of
the verb lana 'to bring'. According to them, it is derived from lena 'to
take' and ana 'to come'. For the same reason, the verb lena which is a
transitive verb (+ne verb) acts like intransitive verbs in the following
set of compound verbs.
(19) le ana 'take + come' = 'to bring'
le jana 'take + go ' = ' to carry'
le calna 'take + walk' = 'to carry/load'
2.2.2.3 "dikhai dena" - type
It was also noticed by Gilchrist and others that transitive verbs such
as lena and dena always behave like intransitive verbs when formed as a part
of a verb with sunai and dhikhai.
2.2.2.4 Onomatopoeic
Traditional Hindi grammars also reveal that onomatopoeic verbs are
intransitive in Hindi. For example, verbs such as chahcahana 'to twitter'
are intransitive.
3.0 THE SEMNTICS OF TRANSIimTY
Although traditional grammarians made use of grammatical as well as
semantic notions such as case, animacy, direct and indirect agents, action
and non-action verbs in their treatment of transitivity and causation, semantics
represented one of their weakest areas. Traditional grammarians sometimes
gave this impression that they were reluctant to enter into this area of Hindi
grammar. They engaged themselves in a long discussion of distinguishing
between passive voice and passive case in sentences such as (20) and (21) .
(20) mujh se galtl hui
me by mistake happened
I committed a mistake.
(21) mujh se kitab parhi gayi
me by book read went
A book was read by me.
According to them, in (20), the subject NP is in the passive case because of
the presence of an active intransitive verb, whereas (21) is an example of
passive voice. They rightly observed that the ne marks an agent of a transi-
tive verb and preferred to treat it as an agentive case marker rather than a
passive, ablative or an instrumental case marker. They also dwelled on the
dative marker ko. However, they seriously fell short of capturing the semantic
core of transitivity. They failed to observe the relationship between transiti-
vity and volitionality as explained in Verma (1971) and Kachru (1982).
Also, no attempt was made to discuss dative subjects nor was any relationship
between psychological predicates/intransitive verbs (such as used in (20)),
and non-volitionality established. Occasionally they attempted to
account for the semantics of the ko subjects, however their
205
analysis was inadequate because they treated the ko with a subject as a
marker of possession of material, mental and moral qualities. ((See Greaves
(1919: 106), Kellogg 1875: 403) ). In this sense, traditional grammars
did not have much success in accounting for the semantics of transitivity
or intransitivity in Hindi.
4.0 CONCLUSIONS
From the above discussion, the following conclusions can be drawn about
transitivity in Hindi. Traditional grammars employed and listed a number of
defining features of transitivity. In addition to the presence of an object,
they listed ne subjects, object-verb agreement, postpositions ko, se, pre-
cedence of animate ne_ subjects over inanimate ne subjects, syntactic dele-
tion as other properties of transitivity in Hindi. While listing the set
of additional features, they often reminded their readers that these features
distinguish between transitive and intransitive verbs in a 'peculiar' way,
implying that these additional features represent idiosyncratic or language
specific features of Hindi transitivity and by no means are treated as univer-
sal. Also, they observed that a static or universal notion (presence of an
object) was not sufficient to account for transitivity in Hindi. They en-
umerated several processes which convert a noun, adjective, pronoun, adverb
and an intransitive verb into a transitive verb; and a transitive verb, noun,
adjective, adverb, pronoun and an onomatopoeic item into an intransitive
verb. In addition, they noticed that Hindi verbs cannot be classified into
a water-tight traditional classification of transitive and intransitive
verbs; there is a significant degree of overlapping between verbs with
respect to transitivity and intransitivity (See Comrie 1979). This over-
lapping can best be captured by way of postulating a continuum of transi-
tivity as proposed by Hopper and Thompson (1980) . Hindi verbs do not
simply represent two end points in a continuum of transitivity but also
occupy several intermediate points on the continuum as shown in Table 3.
The above observation is in agreement with Pandharipande' s (1982)
findings. Furthermore, it should be added that some verbs exhibit an in-
herent potential to occupy the two poles on the continuum at the same time
because of their dual membership in the two classes of verbs, i.e., transitive
and intransitive verbs. What is even more important is the fact that if all
intransitivizing and transitivizing processes are to be accounted for a
dynamic conception of the continuum, where the points will not occupy a
fixed position, is necessary. To the list of the properties of transiti-
vity presented by traditional grammarians, another feature, volitionality
proposed by Verma (1971) and Kachru (1982), must be added in order to account
for the fact that verbs such as nlshana cukna 'to miss a target' are
inherently intransitive in Hindi and they never permit the ne construction
and passivization. Consider the following sentences (22-24).
(22) mujh se nishana cuka
me by target missed
I missed the target.
*(23) mt ne nishana cuka/cukaya
I agent target missed
I missed the target.
206
A
\
1 \
r
1 \
rH
s» > \
Id
=■ cu \
a
•o *-> > \
o
UJ C O 1-1 0) \
',
•■
•r-l
4-1 4J 3 C 4J > \
o
4-1
cd o o 3 cd T^ \
; 1 to 13
•H
B <U a. -r-l '-0 'fl \
1 c
iH
•H -r-l B d 3 fl \
1 u o
O
a;
C XI 0 O Cd Cd \
' CB 4-1
>
C
cd o u o o a.
. ^ -
1 +
_+
+ + + + + +,
(
1-1
1
> >
»
cd
J^
c
•V 4-1
It)
o
oj COO)
0)
•H
4J 4-1 3 C >
D.
4J
cd O O 3 -H
ltd CO
•H
B <u (X ---i »
1 "=
r-l
•'^ •'-1 B C to
T -H o
0
01
C XJ O 0 td
O 4-1
>
c
to o o CJ o.
jQ »
1 + +1+ +1+ 1 +,
"rH
> > f
cd
••
iJ C
TJ 4-1
4-1
x: o
a) c CJ 01
<u
60 -H
4J 4J 3 C >
i)
-H 4J
Cd O O 3 -H
lit) B
led >4-l -H
B <u ex -r-l to
c
C r-l
•H ---i B C to
■» ^ o
V*iO o
0)
C XI O O cd
•H 4-1
cd 4J >
c
cd 0 o o (x
E -
^' ( +
1
+ + + 1 + 1
1-1
1
> >
•■
Id
a
C
4J 13
OJ
o
<U u C <U
<u
•H
■u C 3 >
■-•
4-1
cd 3 O -H
to
•H
B T-i a «
•H C B 03
1 n)
1-1
, C o
o
<u
o a* n o o td
0 4->
>
C JSii <a cd. a o D-
a) -
+
1
1 1 +1 1 1 +,
r-l
1
> >
td
•
c
4-t -a
0)
o
0) O C (U
B
•H
4J C 3 >
o
4-1
cd 3 O -H
CJ
•H
a •r^ D. to
•a C B M
r-l
.1 tfl o
o
0)
(U O C O O cd
' a 4J
>
c
to ^ cd, o o a.
1 + +1+ + +,
1 CO -
+
1
1 —
iH
> >
A)
c
4-1 T3
o
0) o c <u
—
•H
4-1 C 3 >
0)
4J
cd 3 O -H
J3
•H
a 1-) D. to
•H C B «
1 m
rH
► c o
o
(U
(U o c 0 o cd
O u
>
c
to ^ td, a o a
x: -
lL
1
+ + +1+ + 1 ,
rH
1
>
Cd
>
•
c
4J T)
M
o
(U O C 4)
CO
•H
W C D >
•H
4J
td 3 O -H
led S
•H
B i-i o. to
^ o
rH
•H C S 3)
o
v
o « c o o td
1 3 *J
>
a
1
^ CO cd u u o<
1 + + + 1 1
r_
'
207
*(24) mujh se Tdbhana cuka/cukaya gaya
me by target miased went
The target was misded by me.
The verb nishana cukna 'to miss a target' is inherently intransitive in
Hindi because one does not miss a target intentionally under normal circum-
stances. Finally, it should be added that the tests/features outlined above
do not carry equal weight. In terms of their weight, the following ten-
tative rank ordering of features is imperative in order to arrive at an
adequate continuum or treatment of transitivity in Hindi.
Pragmatics^ Volitionality > Passivization > ne_ ;^ object-verb agreement/object
NP _> animacy
Pragmatics or volitionality is likely to yield a higher degree of
trauaitivity in a verb than does the ne construction or the presence of an
(object) NP. Similarly verbs which permit passivization are likely to
be more transitive in Hindi than verbs which allow the ne^ construction.
NOTES
*This paper was presented at the tenth annual conference on South Asia,
University of Wisconsin, Madison. My thanks are due to Professor Y. Kachru
and Professor Hans H. Hock for their comments on this paper. The paper is
especially benefited by the comments of Professor M.K. Verma who served as
a discussant during the conference.
According to Professor Hans H. Hock, European grammarians employed
the term 'Instrumental Case' to refer to the Sanskrit 3rd case whereas
the term 'Ablative' was used in the context of the Latin language.
Although the modal cukna is not a verb, the traditional grammarians
treated it as a verb because of its identical syntactic behavior with
operators.
3
For the motivations for including pragmatics on the continuum,
see Pandharipande (1982) .
REFERENCES
COMRIE, B. 1979. Some remarks on Ergativity in South Asian Languages. SALA
1: 1979. 211-219.
FERGUSON, John. 1773. A Dictionary of The Hindustani Language To Which Is
Prefixed A Grammar Of The Hindustani Language.
FORBES, Duncan. 1855. A Grammar of the Hindustani Language. London: W.H.
Allen and Company.
GILCHRIST, John B. 1796. A Grammar of the Hindustani Language. Menston:
The Scolar Press Ltd. (1970 Reprint).
. 1808. The Stranger's East India Guide to the Hindoostanee.
London: Black, Party and Kingsbury.
GREVES, Edwin. 1919. Hindi Grammar. London: Oxford Press.
208
HADLEY, George. 1772. Grammatical Remarks on the Practical and Vulgar
Dialect of the Indostan Language. Menston: The Scolar Press Ltd.
(1967 Reprint).
HOPPER, P.J. and S.A. Thompson. 1980. Transitivity in grammar and dis-
course. Language. 56: 2 pp. 251-299.
KACHRU, Yamuna. 1982. Transitivity and volitionality in Hindi (in this
volume) .
KELLOGG, S.H. 1875. A Grammar of the Hindi Language. London: Routledge
and Kegan Paul Ltd. (1955 Reprint).
KETELAAR, J. 1698. Instructie off onderwijsinge der Hindostane en Persianse
talen? nevens hare declinatie en conjugative etc. M.S.
PANDHARIPANDE, R. 1982. Transitivizing and intransitivizing processess
in Hindi, (in this volume).
PLATTS, John J. 1873. A Grammar of the Hindustani or Urdu language.
Delhi: Munshiram Manoharlal . (1967 Reprint).
SCHOLBERG, H.C. 1940. Concise Grammar of the Hindi Language. London:
Oxford University Press.
SCHULTZ, B. 1745. Hindustani Grammar. Translated by Vetchoor , M. 1976. Hindi
ke tin prarambhik vyakarana. Allahabad: St. Paulus Prakashan.
VERMA, M. 1971. Review of Y. Kachru's Introduction to Hindi Syntax. Indian
Linguistics. 32: 2 pp. 156-164.
WILLIAMS, Monier. 187 6. A practical Hindustani Grammar. London: Longmans,
Green and Company.
YATES, W. 1827. Introduction to the Hindustani Language In Three Parts,
Calcutta: The Baptist Mission Press.
studies in the Linguistic Sciences
Volume 11, Number 2, Fall 1981
REFERENCE MATERIAL IK HINDI: STATE OF THE ART
Narindar K. Aggarwal
University of Illinois Library, Urbana-Champaign
In the post-independence period of India we have seen considerable
activity in major Indian languages for making available reference
materials, generally not available before 19^7. There is no doubt
that in initiating and coordinating such activities a significant
role is being played by the regional language academies and other
learned societies.
I shall briefly discuss below the current state of such reference
materials in Hindi, the national language of India. In its various
varieties Hindi is spoken by about U6 percent of India's population
and its use is fast spreading in the Sub-continent. It is therefore
essential to evaluate various types of resources in this important
link language.
In this study I have taken a sample of about 350 sources in
Hindi published since 19^7. The sample has been divided into the
following categories:
Table Showing the Types of Reference Materials
Dictionaries
= 50
= 181*
= 26
Hindi-Hindi
= 2k
Hindi-English
= 2li
English-Hindi
Hindi-Other languages
(except English)
Terminology
= 70
Usage, idioms
= 16
Encyclopaedias
General
= 6
Subject
= 20
Yearbooks
= 3
Directories
= 3
Bibl iographies
(all types, including = UO
subject )
Statistical Sources = 3k
210
Literature
(Dictionaries, concordances, etc.) = 28
Religion
(Concordances, dictionaries, etc.) = l6
Biographical Dictionaries = T
Geographical Sources = h
Miscellaneous = h
Total 3^9
See Appendix for a list of the above Reference Materials,
Dictionaries
Dictionaries (sabdakosa) constitute the biggest portion (ap-
proximately 55 percent) of all the Hindi reference sources; it is
indeed a large percentage and this type of imbalance does not exist
in languages like English. However, the earlier tradition of
lexicography in Hindi essentially follows the tradition of Sanskrit
lexicography, though later there has been great influence of English
lexicographical tradition, too. After Indian independence several
good Hindi dictionaries have appeared: Nalanda visala sabdasagar,
edited by Navalaji (Delhi: Adis b\ik depo, 1950); Manaka Hindi kosa,
edited by Ramcandra Varma (Prayag: Hindi sahitya sammelana, 1962-66.
5 vols.); Brhat Hindi kosa, (sabda samkhya 1,38,000) edited by
Kalika Prasad and others (3.edo Varanasi: Jnanamandala , I965);
Hindi sabdasagara, edited by Syamasundara Dasa (rev.enl. & new ed,
Varanasi: Kasi Nagari Pracharini Sabha, 1965-75. llvols.); A
practical Hindi-English dictionary (Vyavaharika Hindi -Angreji kosa)
edited by Mahendra Chaturvedi & B.N.Tiwari (Delhi: National Publishing
House, 1970) are useful dictionaries. English to Hindi dictionaries
are also n\amerous : Byhat Angreji-Hindi kosa (Comprehensive English-
Hindi dictionary) by Hardev Bahri (Varanasi: Jnanamandala, 1960);
A comprehensive English-Hindi dictionary of governmental &^ educational
words & phrases, by Raghu Vira (i+th ed. New Delhi: International
Academy of Indian Ciilture, 1963); Afigreji-Hindi kosa (An English-
Hindi dictionary) by Camille Bulcke ( Ranchi : Kathalik Pres, 1968);
Manaka AAgrezi-Hindi kosa (Standard English-Hindi dictionary) edited
by Satyaprakasa and Balabhadra Prasada Misra ( Prayag: Hindi Sahitya
Sammelana, 1971) ; The Oxford progressive English-Hindi dictionary =
Oksafard progreslv Angrezi-Hindi kosa, by S.K.Verma and R.N.Sahai
(Delhi: Oxford University Press, 1977) are all helpful to learners
and users of Hindi. However, there still is need for a authentic
Hindi-English and English-Hindi bilingual dictionary.
There are several other bilingual dictionaries of Hindi_and
other Indian languages, e.g. Samana srota aura bhinna vartani ki
sabdavali , Asamiya-Hindi aura Hindi-Asamiya, by Vijayaraghava Reddi
( Agra : Kendriya Hindi Samsthana, 1976); HTndi Bangala abhidhana
211
(Calcutta: Bengal Mass Education Society, 1958); Hindi-Marathi
sabdakosa, edited by G.P„Nene and Sripada Josi (P\me: Maharashtra
Rashtrabhasha sabha, 196?); Amara kosa; Hindi-Marathi-Hindi , edited
by Siunera and Lilavati (Solapur: Surasa granthamala, 1963); IChindi-
russkii slovar (Moscow, 1972 )._ There are a large number of
dictionaries of terminology: Akasavanl sabdakosa, A. I.E. lexicon
(New Delhi: Akasavani ke samacara vibhaga dvara prakasita, 1970);
Angreji-Hindi Sri Aravinda sabdakosa, English-Hindi dictionary
of Sri Aurobindo 's terms , by Kesavadeva Acarya (Pondicherry : Diva
jivana sahitya prakasana, I969); Padanama sabdavali , Angrezi-Hindi
(Glossary of designational tenns, English-Hindi) put out by Standing
Commission for Scientific and Technical Terminology (New Delhi,
1965). These are mainly from English to Hindi. The existence of
a large number of such publications is understandable. When Hindi
became the national language of India it was realized that standard
terminology did not exist in that language o A special body (Standing
Commission for Scientific and Technical Terminology) was set up
in October I961 for coining terminologies. Central Hindi Directorate,
New Delhi, under the Ministry of Education and some other bodies
also engaged themselves with terminology building. Some state
governments also set up bodies for building terminologies mainly
for official purpose, e.g. the states of Bihar, Madhya Pradesh,
Uttar Pradesh. All these efforts resulted in several volumes of
Hindi terminology. These were overlapping and caused even some
confusion. The most authoritative work is Byhat paribhashika sab da
sangraha ( Comprehensive glossary of technical terms : Science)
in 2 volumes published in 1973 by the Standing Commission for
Scientific and Technical Terminology. This glossary consists of
technical terms in science which has so far been finalized by the
Commission for Scientific and Technical Terminology. It comprises
equivalents of about 1,30,000 terms in botany, chemistry, geography,
geology, home science, mathematics, physics and zoology up to the
postgraduate level. The work on this glossary started in 1952 and
was the result of concerted efforts of the Ministry of Education
and its different organizations through a careful process of evolution,
standardization and coordination. It is now xondergoing a revision.
Encyclopaedias
The first Hindi encyclopaedia Hindi visva kosa (Encyclopaedia
Indica) was published in 1916-32 (25 voliimes) which was a Hindi
version of Bengali encyclopaedia of Nagendranath Vasu with additions
and alterations. It was followed by Hindi visva bharati (1939)
from Lucknow visvabharati karyalaya, in 10 volumes (second edition
in 1958-65). Both of these are out of date. Sachitra visva kosa
(1967) in 10 volumes published by Rajpal which is based on American
golden book encyclopaedia fills the gap to some extent.
212
That leaves us with Hindi visva kosa (196O-TO) in 12 volvimes
puhlished by Nagari Pracharini Sabha, Varanasi. This was financed
by the government of India, and received cooperation of several
scholars. But most of the technical articles are translations
from the Encyclopaedia Britannica and other sources. However,
some articles on Indian religion, history, culture, etc. are
original. It is being revised now, however, only the 1st volume
has come out so far (1973), which shows that the rate of revision
is very slow.
Other encyclopaedias like Visvajnana s agar a (A Hindi encyclo-
paedia, 1962) and Visvajnana kosa (196U) are one volume encyclopaedias
and are not very comprehensive » Recently K.C.Suman has launched
a reference book project called Diwangat Hindi -Sewi Cpast scholars
of Hindi]; this project provides for publication of an encyclopaedia
in 10 volijmes, the first of which has already been released.
Compiling an encyclopaedia is an expensive proposition. It
requires a great deal of planning, organizing acumen and selfless
devoted work. Its success also depends on the availability of
contributors who know the subject well and can express complicated
ideas in a non-technical style. Hindi and other Indian languages
do not always fulfil these prerequisites. A good Hindi encyclopaedia
under the present circumstances is a difficult proposition unless
a permanent body is created for the purpose on the lines of the
one on Hindi terminology. It may not be a viable commercial propo-
sition, otherwise, a commercial publisher could be involved in
this work.
Subject encyclopaedias are rather late comers. They are small,
mainly confined to social sciences and humanities and among the
sciences, only ayurveda . Invariably they have very short articles,
unsigned and without any references or bibliographies.
One of these is Manaviki paribhashika kosa (Encyclopaedia of
Hiimanities) edited by Nagendra (Rajkamal, Delhi, I965-68). So far
only three volumes have been published: these volumes cover sahitya
(literature), darsana (philosophy), and manovijfiana (psychology) .
The editors have planned to finish up the whole project in five
vol\ames and the remaining two volvimes will cover samaja sastra
(sociology) and lalita kala (fine arts). This is the first venture
of its kind and promises to be a very comprehensive encyclopaedia.
Some of the others like siksha kosa, arthasastra kosa are arranged
by English terms with Hindi equivalents and meanings, followed by
brief articles. Atmanand Misra, editor of siksha kosa, gives one
reason for this practice. He claims that Hindi terminology has
not been standardized as yet. For one English term there are several
Hindi equivalents, none of them as yet accepted as the equivalent,
and all of them used simultaneously by various authors. This appears
to be a genuine problem as has been discussed in several studies,
e.g. see Kachru 1976.
213
Yearbooks
The only yeartook published is Bharata , a translation of A
reference annual , issued by the Ministry of Information and Broad-
casting, Government of India. Some state governments like Bihar
and Uttar Pradesh also publish yearbooks but they are very irregular.
Even Bharata cannot be compared to yearbooks such as Statesman's
Yearbook, or even Manor ama Yearbook. Because it covers only one
country and gives only the official point of view, it is more of
an official handbook. There is a great need for a Hindi yearbook
which should be a commercially viable proposition.
Bibliographies
There have been several attempts at Hindi bibliography including
Indian National Bibliography (INB). Such attempts however are
rather amateiorish and irregular. INB is based on the books deposited
at the National Library, Calcutta, under the provisions of the
Delivery of Books (Public Libraries) Act of 195^, as amended by
Act no. 99 of 1956, but as we know, it has not been a great success.
Besides the time lag and irregxilarity, INB is not by any means a
current comprehensive bibliography of Hindi books. The only regular
and most current bibliography available these days is the Accessions
List published by the U.S. Library of Congress Office, New Delhi.
However, it is not meant to be a Hindi bibliography.
There does not exist any retrospective comprehensive bibliography.
Hindi sahitya sarini (1971-T^; in 2 volumes) covers upto I96U. It
is a detailed bibliography of Hindi books published through 196^.
Hindi sahitya: aloe ana grantha suci, published in 1971 by Bharatiya
grantha niketana, Delhi, covers publications from 19^7-1971. Hindi
alocana kosa published by Bharatiya grantha niketana, Delhi, came
out in 1978. Both of these are detailed bibliographies on Hindi
literature and list publications pertaining to criticism and lin-
guistics. Hindi grantha kosa, 1976, 1977, 1978, published in 1979
by Bharatiya grantha niketana, Delhi, covers publications in Hindi
published during 1976, 1977 and 1978. It is an uphill task for
any reference librarian to respond to bibliographic questions about
Hindi publications.
Even subject bibliographies are rare. My own bibliography
A Bibliography of Studies on Hindi Language and Linguistics published
in 1978 by Indian Documentation Service, Gurgaon, covers the
publications in Hindi language and linguistics. It has filled up
some of the gaps in this particular area and its revised edition
is actively in progress. Regarding this bibliography Yamuna Kachru
of the University of Illinois has said that "it is the single source
of information on what has been/is being done on Hindi in India
and centers of Hindi studies in the Western Hemisphere". There
have also been sporadic attempts at bibliographies of Hindi literature.
It seems to me that it is in this field where librarians can
contribute the maxirnxom.
214
As far as bibliographic control of periodical publications
goes, there continues to be a serious gap. Only one volume of
Hindi sandarbha (1969) was published by the Rajasthan University
Library, Jaipur. We badly need a publication on the lines of Index
India, or Guide to Indian Periodical Literature. Again, librarians
are the only organized group of professionals who can undertake
such major projects with competence.
Biographical Sources
There is hardly any biographical source of Who's who type
or even a comprehensive biographical dictionary. Hindi sahitya
kosa was published in 1958-63 in 2 volumes by Jnanamandala, Varanasi.
Volume one of this kosa gives definitions on almost all aspects
of literature by authorities on the subject; volume two contains
biographical account of authors born until 1915 and their works
published until 1950. This work is arranged alphabetically emd
gives descriptive account of the historical characters frequently
used in Hindi literature. Bholanath Tiwari's Hindi sahitya ki
antarkathaem, published in I962 by Kitab mahal, Allahabad is one
of the latest biographical dictionaries of historical names occuring
in Hindi literature. Sahityika kosa published in 1973 by Sahitya
samaroha, New Delhi, describes 2500 literary authors and journalists
and includes a "Who's who" of 200 literary persons in Hindi. There
seems to be a great need for a good retrospective biographical
dictionary as well as a current biographical Who's who in Hindi.
Geographical Sources
A dictionary of historical and religious places, a geographical
dictionary, a contrived tourist guide and a national atlas (bilingual)
and a couple of school atlases - these provide the total of
geographical sources in Hindi .
Directories
There is hardly any directory available in Hindi worth the
name. This is an area which deserves special attention of publishers
and Hindi organizations.
Statistical Sotirces
The number of statistical sources in Hindi is increasing day
by day. Ministries of the Central government. Central Statistical
Organization, State governments of Hindi speaking states, especially
Madhya Pradesh, Rajasthan and Uttar Pradesh, and the Reserve Bank
of India are bringing out several statistical sources in Hindi.
Some of them are bilingual (English-Hindi); others are exclusively
in Hindi. In the sample under discussion, 3*+ sources are listed
and this list certainly does not exhaust the statistical sources
available in Hindi.
215
It can be hoped that many more statistical sources will be
available in Hindi with the passage of time.
Literary Dictionaries, Concordances , etc.
The predominance of reference books on literature as compared
to other subjects becomes immediately apparent when one takes a
census of Hindi reference sources. That only shows that Hindi is
essentially a literary language and is still struggling to be a
language for scientific study and expression.
In the sample, 28 literary sources are listed. There coiild
be a few more. These are in the nature of dictionaries, concord-
ances and a few literary encyclopaedias. These cover either the
entire literary forms, or one form like drama, or an individual
author. The latter type of works are new comers but their number
is fast increasing. Inspite of numbers, a good literary encyclo-
paedia on the lines of Oxford Companions has yet to come. The
most comprehensive one of these is Hindi sahitya kosa, edited by
Dhirendra Varma and others (2.ed. Jnanamandala , Varanasi, 1963-6*+.
2 volumes) but it does not include authors or their works. It
is mainly concerned with literary forms.
Religion: Concordances, Dictionaries, etc.
There is a strong hold of religion on the Indian mind which
is evident in the number of reference works. In the sample, l6
such works are included. These cover entire Hinduism or Jainism,
or are confined to sacred books like Mahabharata, Puranas, Ramayana,
Vedas, etc. Apparently these works are inspired by their Sanskrit
counterparts and we should be ready to receive more of such works
in our libraries.
This brief survey of the current Indian reference sources in
Hindi show that the picture is not at all encouraging. There are
however, some bright spots such as dictionaries of religion, language
and literature and to some extent statistical sources. The darkest
spots are bibliographies, directories, encyclopaedias, yearbooks,
biographical sources. One might ask: Why are there such substantial
gaps, and why the quality of some of the reference materials is
poor. Perhaps some of the reasons are: Inspite of all-out efforts
by the government of India and some dedicated individuals, Hindi
terminology has yet to be standardized. Terminology is the pre-
requisite of several types of reference books. In the beginning
there must be the word; but there is also another force which
influences the development of language; it is individuals and
groups who can generate a momentum around a linguistic cause among
the users of the language. This momentum takes the shape of a
movement with varying degrees of cohesion and sustenance. The
216
movements bring about change by forcing the government to pursue
a particular policy or by vinning the people to a particular point
of view. Exclusively government directed efforts in this direction
are doomed to fail. Artificial coining of terminology can only
succeed to a very limited extent. There will always be new subjects
and new terms without equivalents in Hindi. The reason being that
our scientist neither read nor write in Hindi. This language is
not in the forefront of knowledge. There is a tremendous gap between
knowledge existing in international languages like English and
that available in Hindi. And this gap is widening everyday.
As already stated, coining of terminology by committees is
a very cumbersome and time-consuming process. It is artificial,
too. Terms coined by this process do not become very popular.
The policy of coining terms from Sanskrit roots has also something
to do with the slow progress of terminology. This policy has been
partially modified lately and now terms from other languages are
accepted as such in the Hindi language. But it seems that this
policy has to be used more generously.
The rate of translation from English and other foreign lan-
guages is very slow. Translation work has been confined mainly
to the textbooks. Original monographs and periodical articles
are hardly being translated. To increase knowledge in Hindi,
translation work has to be organized on the lines it is done in
the Soviet Union, China and Israel. It is very essential that a
body of knowledge exists in a language before it could be sifted,
condensed and conveniently arranged in the form of reference books.
It is a moot point whether the compiling of reference sources
is dependent on the type of reference service offered in the
Indian libraries. Necessity is the mother of invention, as the
saying goes and there is some truth in it. Librarians collectively
can point out the type of reference books in Hindi which are urgently
required. And they can get some government body or commercial
publishers interested in the compilation and publication of some
types of reference books.
During my research and acquisition trips to India in the past
several years I have felt that librarians have not played their
part very effectively in indicating the need for reference books
in Hindi ; nor have they done much to compile them and to get them
published. I know that there are many young and bright librarians
in India, a small committee of whom could go into the details of
unrepresented areas for Hindi reference sources and give suggestions
for their compilation and publication. Some of the reference books,
like a yearbook, should be a viable proposition commercially. But
the greatest contribution librarians could make is in the field of
compiling bibliographies which are the core of any reference materials.
217
REFERENCES
MNAMALAI , E. , ed. 1979. Language movements in India. Mysore:
Central Institute of Indian Languages.
BAHRI, Hardev. 1965. Hindi: udbhava, vikasa aura rupa.
Allahabad: Fitab Mahal.
CAUDHARI, Ananta. 1972. Hindi vyakarana ka itihasa. Patna:
Bihara Hindi grantha akadami .
CHATTERJI, Suniti Kumar. I960. Indo-Aryan & Hindi. Rev. & enl.,
2d ed. Calcutta: K.L.Mukhopadhyay.
CHATURSEN. 19^6. Hindi bhasha aura sahitya ka itihasa. Lahore:
Meharchand Lakshmandas.
INDIA; a reference annual. 1980. New Delhi: Ministry of Information
and Broadcasting, Government of India.
NARULA, Shamsher Singh. 1976. Hindi language: a scientific history.
2d ed. Delhi: Oriental Publishers & Distributors.
SHARMA, Ram Gopal. 1963. Hindi bhasha aura usaka itihasa.
Allahabad: Kitab Mahal.
The TIMES of India directory & yearbook including Who's Who. 196U-65.
London: Bennett, Coleman.
The TIMES of India directory & yearbook including Who's Who, 1979, I980.
Bombay: The Times of India Press.
TIWARI, Bholanath. I966. Hindi bhasha. Allahabad: Kitab Mahal.
TIWARI,Udai Narain. 1961. Hindi bhasha ka udgama aura vikasa.
2d ed. Prayag: Bharati Bhandar.
VARMA,Dhirendra. 1962. Hindi bhasha ka itihasa. 7th ed. Prayag:
Hindustani Akademi .
APPENDIX
A Partial List of Resource Materials Available in Hindi
AGGARWAL, Narindar K. 1978. A bibliography of studies on Hindi
language and linguistics. Gurgaon: Indian Documentation Service.
BAHRI, Hardev. I96O. Brhat Angreji-Hindi kosa. Comprehensive
English-Hindi dictionary. Varanasi : Jfianamandala.
BAJORIA, Ratanlal Kaluramji. 1972. Byhad rashtrabhasha kosa. Vardha:
Rashtrabhasha prachar samiti.
BARANWAL, Sukhdeo Prasad, ed. I962. Visvajflana-sagara. A Hindi
encyclopaedia. New Delhi: Vidya prakashan bhavan.
BHARATA. 195^- New Delhi: Ministry of Information & Broadcasting,
Government of India.
BHATIYA, Kailasacandrao I968. Hindi kl besika sabdavali. Aligarh:
Aligarh Muslim visvavidyalayao
BULCKE, Camille. I968. Aftgreji-Hindi kosa. An English-Hindi
dictionary „ Ranchi : Kathalik Pres.
CHATURVEDI, Mahendra, and B.N. Tiwari , eds. 1970. A practical
Hindi-English dictionary. Vyavaharika Hindi -Angreji kosa.
Delhi: National Publishing House.
DAS, Syamasundar, ed. 1965-75. Hindi sabdasagara. Rev. enl. new ed.
Varanasi: Kasi Nagari Pracharini Sabha„ llv.
HINDI visva-bharatl. 1958-65. Rev. ed. Lucknow: Hindi visva-bharati,
lOv.
218
HINDI visvakosa. 1962-73. Rev. ed. Varanasi : Nagari Pracharini
Sabha„ 12v.
INDIA (Republic) Ministry of Education. 1958. Jflana sarovara.
New Delhi: Siksha mantralaya, Bharata sarakara. v. 2
INDIA (Republic) Standing Commission for Scientific and Technical
Terminology, 1973. Brhat paribhashika sabda sangraha. Compre-
hensive glossary of technical terms: science. New Delhi. 2v.
The INDIAN national bibliography. 1957- Calcutta: Central Reference
Library.
JCACHRU, Braj B. 1976. General linguistic studies in Hindi: a review
of resources. Lingua, 38:335-55.
KENDRIYA Hindi Samsthana, Agra. 1967. Hindi ki adharabhuta sabdavali.
LIBRARY of Congress. Library of Congress Office, New Delhi. 1962-
Accessions list.
MAHAJAN, Yash Pal, 1978. Hindi alocana kosa. Delhi: Bharatiya
grant ha niketan.
MAHAJAN, Yash Pal. 1979- Hindi grantha kosa: 1976-1978 mem
prakasita sabhi vishayom ki vistrta namavali. Delhi: Bharatiya
grantha niketan. 3v.
MAHAJAN, Yash Pal. 1971. Hindi sahitya: alocana grantha suci. Delhi:
Bharatiya grantha niketan.
NAGENDRA, ed. 1965-68. Manaviki paribhashika kosa. Encyclopaedia
of humanities. Delhi: Rajkamal. 3v,
NARAIN, Pitamber. 1971-7^. Hindi sahitya sahitya sarin! . Hoshiarpur:
Visvesvarananda-samsthana. 2v.
NAVALAJI, ed, 1950. Nalanda visala sabdasagara. Delhi: Adis biik depo.
PRAKASH, Satya, and B. Misra, eds. 1971. Manaka Angrezi-Hindi kosa.
Standard English-Hindi dictionary. Prayag: Hindi sahitya sammelan.
PRASAD, Kalika, and R. Sahaya, eds. 1965. Brhat Hindi kosa;
sabdasarakhya 1,38,000, 3.ed. Varanasi :Jf5anamandal a.
RAGHU Vira. 1963. A comprehensive English-Hindi dictionary of
governmental & educational words & phrases. . .Uth ed. New Delhi:
International Academy of Indian Culture.
RAUCH, Irmengard, and Gerald F. Carr, eds. 1979. Linguistic method,
essays in honor of Herbert Penzl. The Hague: Mouton.
SACHITRA visvakosa. I967. Delhi: Rajpal. lOv,
SHARMA, Om Parkash. 1973. Sahityika kosa. New Delhi: Sahitya
samaroh.
TIWARI, Bholanath. 196*+. Hindi muhavara kosa. Rev. 2d ed.
Allahabad: Kitab Mahal,
. 1962. Hindi sahitya ki antarkathaen. Allahabad:
Kitab Mahal,
. 1962. Vrhat paryayavaci kosa, 2d. rev, ed.
Allahabad: Kitab Mahal.
VARMA, Dhirendra, ed. 1963-6i+. Hindi sahitya kosa. 2d. ed, Varanasi:
Jflanamandala .
VARMA, Ram Chandra, ed, 1962-66. Manaka Hindi kosa. Prayag: Hindi
sahitya sammelan. 5v.
. 1971. Sankshipta Hindi sabdasagara. 7.ed. Kashi :
Nagari Pracharini Sabha,
VASU, Nagendranath, and Visvanath Vasu, comps. 1916-32. Hindi
visvakosa. Encyclopaedia Indica. Calcutta: Visvakosa Press. 25v.
219
VERMA, Shivendra Kishore, and R.N. Sahai „ 1977. The Oxford progressive
English-Hindi dictionary=Oksafard progresiv Angrezi-Hindi kosa.
Delhi: Oxford University Press.
VIDYALANKAR, Avanindra Kumar, ed. 196^*. Visvajfiana kosa. Delhi:
Hindi Pocket Books.
YADAV, Shree Ran, and Satyapal Goyal, eds. 1973. Hindi sandarbha-
grantha nirdesani. Hindi reference books in print, l800-1971.
Sangrur: Sarcamets.
The following special issues are in preparation:
Papers on Diachronic Syntax
Editor: Hans Henrich Hock
Studies in Language Variation:
Nonwestern Case Studies
Editor: Braj B. Kachru
STUDIES IN THE LINGUISTIC SCIENCES
Spring 1974
Vol.4,
No
Fall 1974
Vol.4,
No
Spring 1975
Vol.5,
No
Fall 1976
Vol. 6,
No
Spring 1977
Vol. 7,
No
Fall 1977
Vol. 7,
No
Spring 1978
Vol. 8,
No
Fall 1978
Vol. 8,
No
Spring 1979
Vol.9,
No
Fall 1979
Vol. 9,
No
Spring 1980
Vol. 10,
No
Fall 1980
Vol. 10,
No
Spring 1981
Vol. 11,
No
Fall 1981
Vol. 11,
No
following issues are available:
1, Papers in General Linguistics
2, Papers on Phonetics and Phonology
Editors: Charles W. Kisseberth and
Chin-W. Kim
1 , Papers in General Linguistics
2, Papers on African Linguistics
Editors: Eyamba G. Bokamba and
Charles W. Kisseberth
1 , Papers in General Linguistics
2, Studies in East Asian Linguistics
Editors: Chin-chuan Cheng and
Chin-W. Kim
1 , Papers in General Linguistics
2, Linguistics in the Seventies:
Directions and Prospects
Editor: Braj B. Kachru
1 , Papers in General Linguistics
2, Relational Grammar and Semantics
Editor: Jerry L. Morgan
1 , Papers in General Linguistics
2, Studies in Arabic Linguistics
Editor: Michael J. Kenstowicz
1 , Papers in General Linguistics
2, Dimensions of South Asian
Linguistics
Editor: Yamuna Kachru
Orders should be sent to:
SLS Subscriptions, Department of Linguistics
4088 Foreign Languages Building
University of Illinois
IrhariQ TllinnJc ^1801
Live Music Archive
Librivox Free Audio
Metropolitan Museum
Cleveland Museum of Art
Internet Arcade
Console Living Room
Books to Borrow
Open Library
TV News
Understanding 9/11