UNv RSITY OF

AT URBANACHAMPWQN
NATURAL HIST. SURVEY

)•-)

t FIELDIANA ...

Zoology """"""mm

Published by Field Museum of Natural History

AUG 6 1985

20

LIBRARY

ON THE PHYLETIC WEIGHT OF MENSURAL
CRANIAL CHARACTERS IN CHIPMUNKS
AND THEIR ALLIES (RODENTIA: SCIURIDAE)

BRUCE D. PATTERSON

July 29, 1983
Publication 1348

ON THE PHYLETIC WEIGHT OF MENSURAL

CRANIAL CHARACTERS IN CHIPMUNKS
AND THEIR ALLIES (RODENTIA: SCIURIDAE)

FIELDIANA
Zoology

Published by Field Museum of Natural History

New Series, No. 20

ON THE PHYLETIC WEIGHT OF MENSURAL
CRANIAL CHARACTERS IN CHIPMUNKS
AND THEIR ALLIES (RODENTIA: SCIURIDAE)

BRUCE D. PATTERSON

Assistant Curator

Division of Mammals

Department of Zoology

Field Museum of Natural History

Accepted for publication July 15, 1982
July 29, 1983
Publication 1348

Library of Congress Catalog Card Number: 82-84350

ISSN 0015-0754

PRINTED IN THE UNITED STATES OF AMERICA

CONTENTS

Abstract 1

Introduction' 1

Methods and Materials 3

Results 6

Univariate Analyses 6

Bivariate Analyses 8

Multivariate Analyses 12

Discussion 15

Mandibular Morphology 17

Cranial Morphology in Chipmunks and Related Forms 18

Acknowledgments 21

Literature Cited 22

Appendix: Specimens Examined 24

LIST OF ILLUSTRATIONS

1. Geographic locations of the five sampled populations 3

2. Skull of Eutamias minimus from the LaSal Mountains, Utah 5

3. Histogram of correlation coefficients among the 28 mensural characters in
chipmunks 11

4. Plot of canonical variate loadings of 86 chipmunks on the first canonical variates
for "taxonomic" and "mandibular" characters 13

LIST OF TABLES

1. Means, coefficients of variation, and sample size of 28 cranial and mandibular
characters of least chipmunks 7

2. Product-moment correlation matrix for "taxonomic" characters 9

3. Product-moment correlation matrix for "mandibular" characters 10

4. Product-moment correlation matrix for "taxonomic" and "mandibular"
characters 10

5. Results of canonical correlation analysis, showing Bartlett's test for the remaining
eigenvalues 12

6. Canonical variate loadings, variance extracted, and redundancy for the first two
canonical variates 14

7. Simple linear regression analyses of loadings on the first canonical variates vs.
body size 15

ABSTRACT

This study outlines a method for investigating the phyletic information or
"weight" in mensural cranial characters in the Sciuridae. Five isolated populations
of least chipmunks (Eutamias minimus) form the frame of reference for this analysis.
The study examines the hypothesis that divergence in a set of mainly cranial
characters used in taxonomy ("taxonomic" characters) differs from that evident in
a set of "mandibular" characters. Mandibular characters in many other animals are
known to be genetically and ontogenetically responsive to ecological conditions.
Therefore, if taxonomic characters are to provide unique information on the
phyletic relationships of chipmunks, patterns in their variability should contrast
with mandibular patterns.

Taxonomic and mandibular characters differ in mean values among the five
populations to comparable extents. Although mandibular characters tend to be
more variable, they tend also to be more highly correlated with one another than
are taxonomic characters. Correlations of characters within the taxonomic and
mandibular sets are apparently no greater than between-set correlations; both are
low. The relative interdependence of taxonomic and mandibular characters is
emphasized by a canonical correlation analysis which suggests that individuals
tend to occupy the same relative positions in each of the two morphological
spaces. Results of the canonical correlation analysis must be considered tentative
in view of the facts that (1) relatively little of the variation in original characters is
represented in the canonical variates, and (2) no comparable study exists for
contrast.

Nevertheless, this study helps to explain some confusing patterns of cranial
morphology in chipmunks and, by extension, their allies. Cranial morphology in
these animals is apparently tied to environmental conditions, to an extent at least
comparable with mandibular morphology. In this respect, chipmunks resemble
tree squirrels more closely than ground squirrels. Because of this responsiveness,
phenetic studies of cranial morphology in chipmunks may lead to better
classifications of ecological conditions than of phyletic relationships (cf . Patterson,
1981).

INTRODUCTION

Characters of the skin and skull have served for many years as the foundation
of mammalian classification. Higher taxa are commonly distinguished by one or
more qualitatively different characters, whose variation is thereby amenable to
"phylogenetic" analysis (e.g., Marshall, 1977). In contrast, mammalian species
and subspecies commonly lack such distinctive attributes, and the study of their
interrelationships is thus typically limited to "phenetic" approaches (e.g.,

2 FIELDIANA: ZOOLOGY

Genoways, 1973). With the advent of computers and the development of
multivariate statistical packages, mammalian systematists have increasingly
employed mensural cranial characters in taxonomic decision-making.

The use of mensural cranial characters in mammalian classification is clouded by
our ignorance of components of cranial variability (cf. Leamy, 1977; Straney &
Patton, 1980). While distinct patterns of cranial morphology may reflect degrees of
phyletic relatedness, it is equally possible that they may result from the distinctive
responses of a given underlying genome to different environmental conditions.
Problems raised by environmental determination of cranial morphology are
exacerbated by the fact that phyletic lineages and environments are both based in
geography. Thus, simple geographic patterns of cranial morphology can,
themselves, neither support nor refute the importance of environmental
determination. Are complex, laboratory analyses of morphological genetics (e.g.,
Atchley et al., 1981) the only reasonable avenue of inquiry?

Another means of approaching this question is afforded by the intercorrelations
of characters. Specifically, if one can establish a priori a character set that is
genetically or ontogenetically responsive to ecological conditions, then one can
assess the concordance of these characters with those in question. Close
correspondence of characters in the two sets would argue for the importance of
ecological determination. Conversely, independent responses would permit
cranial characters to have "phyletic weight" [sensu Mayr, 1969) and hence
taxonomic utility.

Available data indicate that characters of the mandible are often
environmentally plastic. Functional morphologists generally recognize that the
"size and shape [of the mandible] is maintained in being as a response to the
primary morphogenetic demands of functionally related tissues" (Moss &
Meehan, 1970, p. 11). Characters of the mandible can thereby reflect the
developmental importance of environmental conditions, such as nutrition and diet
(McCance, 1962; Moore, 1965; Tonge & McCance, 1965). Characters of the
mandible may also reflect divergent selection pressures resulting from different
resource-use patterns (the "Darwin principle"; e.g., Johnston, 1969; Behle, 1973;
Bamett, 1977). In either case, such characters should be accorded "particularly low
taxonomic value" (Mayr, 1969, p. 125). If the assumption is valid that mandibular
morphology closely corresponds to environmental conditions, then it is possible
to assess the taxonomic worth of traditionally employed characters of the general
conformation of the cranium.

This paper compares patterns of variability in taxonomic and mandibular char-
acters among isolated, montane populations of chipmunks, genus Eutamias (Ro-
dentia: Sciuridae). A recent study of geographic variation in the Colorado chip-
munks (Eutamias quadrivittatus group) suggested that cranial and mandibular
characters being used in chipmunk taxonomy may be responsive to environmental
conditions associated with the habitats that different montane populations occupy
(Patterson, 1981). Another study suggested that the cranial morphology of Doug-
las squirrels (Tamiasciurus douglasii) reflects the conifer species they feed upon
(Lindsay, 1982). If such conclusions are warranted, then patterns of morphology
in taxonomic and mandibular character sets should largely coincide, both being
principally determined by ambient environmental conditions. If, in contrast, tax-
onomic (chiefly cranial) characters more closely reflect phyletic affinities, discor-
dance in the two character sets should be evident, because mandibles can be

PATTERSON: CRANIAL CHARACTERS IN CHIPMUNKS 3

expected to reflect ecological conditions. The test of this hypothesis examined
cranial and mandibular characters in a group of chipmunks, all currently classified
as Eutamias minimus operarius, which occupy a spectrum of environmental condi-
tions.

METHODS AND MATERIALS

Five montane populations of least chipmunks were selected for study, repre-
senting a considerable latitudinal extent (fig. 1). Populations are as follows: (1)
Sierra Blanca, New Mexico (lat. 33°23'N, long. 105°48'W), (2) northeast of Santa
Fe, New Mexico (lat. 35°47'N, long. 105o49'W), (3) northeast of Canjilon, New
Mexico (lat. 36°28'N, long. 106°26'W), (4) west, northwest of Tres Piedras, New
Mexico (lat. 36°43'N, long. 106°14'W), and (5) LaSal Mountains, Utah (lat.
38°23'N, long. 109°11'W). Precise locality information for specimens comprising
these populations is given in the Appendix. On Sierra Blanca in southern New

200 km

Fig. 1. Geographic locations of the five sampled populations.

4 FIELDIANA: ZOOLOGY

Mexico, least chipmunks are confined to habitats within the glacial cirque on that
peak, which are characterized by extensive boulder fields (Conley, 1970). In north-
ern New Mexico, £. minimus occurs in mesic coniferous forests, sage brushlands,
and alpine meadows (Bailey, 1931). In contrast, I coUected the sample from the
LaSal Mountains in 1980 from an open, xeric pine forest along the southeastern
foothills of the mountains proper. The five studied populations therefore occur in
a variety of ecological conditions.

Information on head-and-body length and weight was taken from the collector's
tag; other data were gathered by me from skulls and jaws using dial calipers and
were recorded to the nearest 0.1 mm. Cranial and mandibular characters used in
the present analysis are defined by numbers in the various aspects (denoted by
letters) of Figure 2. The characters may be classified into three groups, the first two
being combined as "taxonomic" characters, consisting of cranial and mandibular
characters used in making taxonomic inferences in various systematic studies of
chipmunks, and the third consisting of exclusively "mandibular" characters se-
lected in this study for their probable functional significance. The first group is
comprised of characters I employed in systematic studies of the Eutamias quadri-
vittatus species complex (Patterson, 1980, 1981; see also Johnson, 1943; Conley,
1970). These include: GLS, greatest length of skull — 1, 2 (A); ZB, zygomatic
breadth— 3, 4 (A); CB, cranial breadth— 5, 6, (A); ML, mandibular length— 33, 34
(E; see Levenson et al., in review, for another study drawing taxonomic conclu-
sions from mandibular morphology in chipmunks); NL, nasal length — 1, 9 (A);
MTR, length of maxillary molar toothrow — 16, 21 (C); IOB, least interorbital
breadth— 11, 12 (A); NW, nasal width— 13, 14 (A); DOL, diagonal length of
orbit— 22, 23 (C); PL, premaxillary length— 1, 10 (A, B); CD, cranial depth— 17, 18
(B). The second group consists of other characters of general cranial conformation,
some of which have been used in taxonomic investigations by Howell (1929) and
Hoffmeister & Ellis (1979). These are: PTL, palatillar length— 15, 24 (C); MD,
length of maxillary diastema — 15, 16 (B, C); LAB, length of auditory bulla — 25, 26
(C); OIL, distance between orbit and incisive foramen — 22, 27 (C); PW, palatal
width at third premolar— 28, 29 (C); BOS, basioccipital length— 30, 31 (C); IL,
length of incisive foramen — 27, 32 (C); TZ, length of zygomatic arch at posterior
union with skull— 19, 20 (B); and FW, frontal width— 7, 8 (A).

Patterns of variation in these "taxonomic" (mainly cranial) characters were
compared with those shown by mandibular characters. Mandibular characters
examined (in addition to ML, above) are: MC, distance between mental foramen
and articular condyle — 33, 35 (D); MAD, distance between anterior alveolus of M3
and articular condyle — 33, 36 (E); CVH, distance between tips of coronoid and
ventral angular processes — 37, 38 (D); CDH, distance between tips of coronoid
and dorsal angular processes — 37, 39 (D); AG, greatest width of angular
processes — 38, 40 (E); AL, least width of angular processes — 41, 42 (D); AF,
distance between mandibular foramen and articular condyle — 33, 43 (E); WMM,
width of mandibular molars (M2 and M3) — 36, 44 (E).

Descriptive statistics and univariate analyses of variance were obtained for adult
specimens (i.e., permanent P4 showing wear) via BMDP7D (Dixon, 1977). The
equality of group means was assessed via analysis of variance, assuming equality
of group variances. Inspection of Welch and Brown-Forsythe test statistics (Dixon,
1977), which do not assume equal variances, provided highly similar results.

Relative variability of characters was determined by the coefficient of variation,
calculated on a desk calculator as the standard deviation divided by the mean

|1 3

s I c

O 0)
S'<3 3

a-3 1

<U c
m- ■£ 0(
o _ c

£<?
00 1

3

u. ^ «

6D

^ — ■ -• ~ -— .

_ ^- m <u ^

sii-si

.2 "S o «* "^

ri

"S o in c c

V,- « c w

•c . 2 hjj

.£ "m J2 o -a

6 FIELDIANA: ZOOLOGY

(BMDP7D) times 100. Coefficients of variation were then pooled across characters
and localities within character sets to determine differences in the relative vari-
ability of characters of taxonomic and mandibular sets. The equality of relative
variability in these sets was determined by the Mann-Whitney U test (Siegel,
1956), a distribution-free nonparametric test. This test was considered one-tailed
because characters responsive to environmental influences might be expected a
priori to be more variable than those marked by great phyletic weight.

Bivariate correlations between characters were examined using Pearson
product-moment correlation coefficients (BMDP6M; Dixon, 1977). Examinations of
within- and between-set correlations were made by simple proportions and by
histograms.

Correspondence between the sets of taxonomic and mandibular characters was
assessed via canonical correlation analysis (BMDP6M; Dixon, 1977). Canonical
correlation provides a convenient summary of multiple bivariate correlations be-
tween previously defined sets of normally distributed variables. The procedure
relies on eigenvector-eigenvalue solutions to a series of linear functions; it is
canonical in the sense that the number of variates, and hence their non-zero
correlations, is reduced to a minimum number equal to the number of variables in
the smaller original set. Excellent discussions of this methodology may be found
in Cooley & Lohnes (1971) and Pimentel (1979).

Geometrically, canonical correlation analysis indicates the extent to which indi-
viduals occupy the same relative positions in terms of the separate sets of mea-
surements (Cooley & Lohnes, 1971). It is similar to principal components analysis
in that original variable axes are rotated to a new set of axes that is mutually
orthogonal and does not warp the original euclidean spaces. It differs from prin-
cipal components analysis in that axes are selected so that between-set pairs have
the maximum possible correlation; principal components are selected to maximize
extracted variation (Pimentel, 1979).

Bartlett's chi-square test (Dixon, 1977; Pimentel, 1979) was used to determine
the significance of remaining eigenvalues and hence the number of canonical
variables necessary to express the dependency between the two sets of characters.
In order to assess the influence of body size variation on the canonical variates, I
regressed canonical variates on head-and-body length and the cube root of
weight, both statistically independent variables, using regression routines pro-
grammed in a TI-59 desk calculator. The cube root of weight, rather than weight
perse, was used to satisfy statistical assumptions of normality. The transformation
may be visualized as one reducing a three-dimensional property, weight, into a
one-dimensional one compatible with linear measurements or transformations of
them. The significance of these regressions was judged via tables given in Rohlf
& Sokal (1969).

RESULTS

Univariate Analyses

Means and coefficients of variation of the "taxonomic" and "mandibular" char-
acters are given for five populations in Table 1. Significant mean differences exist
among these populations in 19 of 28 (68%) characters. Within the taxonomic
character set, 13 of 20 characters (65%) differ significantly among the five popu-
lations (P < .05), and five characters (25%) differ at highly significant levels

• » • • » • •

Tt<infnTrrsior^rsitN^drvisdodrsiod<-"<Nrsi<N<N

i— i OS O i— i
Ov tJ> r-i <N

oooot^^mrsivO'T

N'tOOMc'lONO^

_ oq in fCT Sv" cT oq £*" $? fT Co rn rn 5j" oo r*- Co rn £*" Co Co ^~oq Co ST 6^" ^ Co oo
Jfi i^rsirsii^uS(NTto6rvifnrvirsirorsir«Sror<SrvuSTt rsirvirninininTtTr

o1*-

w — ' <Nir>>Trooooinso«Nooo^r*icooot^\omm<N<NON

O i— i i— i r— r-c r-i

Toot^mmmi-i

Jz

Ni-<oq\onNoo't(N<Nt»iovOoinf^^(No
i^rM<Nr^\DrncnKrvicorvirvirj(rirsir^rnK^df<S

0\rOTti<r-tin(N\0

i-irjci'i'LnTj'^'Tj'

m r-i r-> r-c i-c i-H iH

I JS T3

R^noovooofvnnr-oowfsN riooinooR

fN|Tj<r-ivO'!rt^h>'tl

rHr-MrHTrM'tvOrJiniNrilNlNINmiNN'trt

(NtNromvOsomTt1

'tooMnnin

« >^3
c jr.

"3

8

ns m 3

W 01 <*.

«£ °

< oO «;

H oi "5!

k« Ot

re is

C_ »Co q y in PPCo in ^n CoCoRS'iNin'oo oo 5" int^cn r-^ro nTr? Co
O J r£^^r-<r^rNiTto6rornrsir-<r\ir^njroro'nvoin nirnrnininTfinrn

eii ■^•pm^'rfosTtTtfnxoinrHO
«Z ^*no4r«N'iNn«t\ciN

U~

« _

£.3

W o

— 1/1

c oj £ 2 £.

.2§ fc

«|

> o -2

1= '■= 3
o A a.

sis.

Oi o o

•0 a->

> ^S

S2

rt(Nrtoo

ooa; n n rs N
■to>dt\ifinifiri

'^r^^rJf^^o^^o'^^fn'oo'i^rH'fN'o'rH'o'
rvirvirviinrsiinr^icNrorNirNir'ScNir^fiSrSininin

inn innoMnt n
(Ntom^n't'tin

^NlflNnO^OMNNniN

Ho6^ododih\drvioda^<^roo6odsdinin<N(sioN iriddsin'riirirH

s ______________ _____ ________

C _ so m ». ** cr ; so 5f© co <n vq fsTro CoPTo'infn 5" c?" Co rCfoo o<n a^inin

"™ ii ._ ._._._ — _. __ ._. -. _ _.. - _._ -.

qa;(l6rHNriooNNNiNiniNaiO>OiNiN'l,(> 0\OCT»i— 'i—r^O^t

niK<ncdcdin3r^cco^rnroo6KvO\bin<N<No6 -*ocTvtNiinrrjinr-I

CO 1— it— ■ i— i i— « i— i i— < r-i

-J ca co rd -J fc
UN USZS

Si^eSlSoiBdB

ul5gg2sl

8 FIELDIANA: ZOOLOGY

(P < .001). For mandibular characters, six of eight (75%) differ significantly
(P < .05), two of these (25%) highly so (P < .001). The pattern of mean differences
for the two character sets appears grossly comparable.

However, there are apparent differences in the variability of taxonomic and
mandibular characters. The most frequently occurring coefficient of variation in
the taxonomic data set is 2.7 (10 occurrences); that for mandibular characters is 4.5
(four occurrences). One-half the coefficients of variation for taxonomic characters
are less than or equal to 3.0; the median for mandibular characters is 3.7. The
tendency of mandibular characters to exhibit greater variation relative to mean
values can be demonstrated via the Mann-Whitney test of 100 taxonomic vs. 40
mandibular coefficients of variation (U ■ 1487.5; t = 2.36; P < .01, one-tailed
test). It should be noted, however, that the greater relative variability of man-
dibular characters does not obviate mean differences in mandibular characters
among these populations (Table 1).

The univariate analyses show that: (1) significant differences in mean values
exist among sampled populations in the character sets analyzed; (2) significant
differences among populations are not confined to one character set or the other,
but are typical of both, to roughly comparable extents; and (3) the mandibular
characters analyzed tend to show greater variability relative to mean values than
do the taxonomic ones. However, the univariate analyses do not address them-
selves to the intercorrelations of characters: taxonomic and mandibular characters
may be either fully integrated, so that changes in one set demand changes in the
other, or largely independent but parallel, producing the same univariate effects.
Bivariate and multivariate analyses are necessary to distinguish between these
possibilities.

Bivariate Analyses

If "taxonomic" and "mandibular" characters comprise two distinctly different
morphological domains which have undergone comparable degrees of change
during the divergence of these populations, closer correlations can be expected
between characters of the same set than between characters of different sets.
Conversely, if taxonomic and mandibular characters are fully integrated, either
genetically or ontogenetically, so that changes in one set imply corresponding
changes in the other, then such differences in the degree of intercorrelation should
not exist.

It is possible to distinguish three sets of correlation coefficients from the set of
378 unique bivariate correlations of the 28 characters analyzed: correlations within
the taxonomic character set, correlations within the mandibular character set, and
correlations between taxonomic and mandibular characters. Coefficients for these
three sets of correlations are presented in Tables 2, 3, and 4, respectively. Coeffi-
cients of correlation in all three sets are low. The number of correlations with
negative coefficients is at first surprising, given that all correlations are of mensural
characters confounding size variation. However, only one negative correlation
differs significantly from zero (TZ and NW, table 2), and this correlation holds at
the 5% level but not at the 1% level. Generally, those characters showing the
greatest number of negative or nonsignificant correlations with others (e.g., NW,
IL, and WMM) are similar in having small mean values (table 1). The lack of
correlation between these and others may be attributed in part to the relative

■5

<NOr-"r-«T-c<S<NOrvlrr)r-irHOCiOOi-<^0

ooooooooooooooooooo

I III I

NrtO(SririONnOOnrtOtOr<Nn
OOOOOOOOOOOOOOOOOC)

'-' ooooooooooooooooo

II III

jk *4qnnSNririri5nniNrtri
5 dddddddddddddddd

i
I

5

0

ooooooooooooooo

I I

, mcnr^rr>t^i-ii-io»-it^'^<<NO(^

►J ininq<tnNr;riqHOvo>ON

Q oooooooooooooo

j ooooooooooooo

Q vOinr<^NrirHp<riOr<N

2 cJoocdoooooooo

g

E

v00*v0 5>J>On(Nh0MH

cJoooooooooo

S2

« o

oooooooooo

ddddddddd

go

1"

1°

En

C rM

O .-:

<g n

? i

<U O

ll

$*

TJ o>

p Oi

J" •■"

— 00

. o>

CM "O

< Oi

H >

q odocJoodo

^ NNrlOrlH Ji
^ 1— l i— c r-l O O O i— I

T d d d d d d d
"ill l

q NNnnNH

» d d d d d d

a

o o o o d

-J

oe ins

vO v© CM

odd

EC

u

fx ri

CN fM

d d

N

d

J BD D d |J

UnuSz

S 2 z S E u C s 5 o E 2 rf P

10 FIELDIANA: ZOOLOGY

Table 3. Product-moment correlation matrix for "mandibular" characters (within-set
correlations). Correlation coefficients are based on 86 cases and have 84 degrees of free-
dom. P(r = 0.212) a .05; P (r - 0.277) a .01.

MAD

CVH

CDH

AGW

ALW

ALF

WMM

MC

0.77

0.54

0.47

0.50

S 0.31

0.49

0.05

MAD

0.52

0.41

0.44

0.28

0.64

0.01

CVH

0.71

0.73

0.40

0.44

0.06

CDH

0.34

0.16

0.45

-0.06

AGW

0.55

0.38

0.18

ALW

0.13

0.29

ALF

-0.01

Table 4. Product-moment correlation matrix for "taxonomic" (rows) and "mandibular"
(columns) characters (between-set correlations). Correlation coefficients are based on 86
cases and have 84 degrees of freedom. P (r = 0.212) a .05; P {r = 0.277) ■ .01.

MC MAD CVH CDH AGW ALW ALF WMM

GLS

0.64

0.63

0.56

0.46

0.51

0.20

0.42

-0.00

ZB

0.63

0.62

0.72

0.60

0.62

0.31

0.46

0.07

CB

0.18

0.22

0.22

0.12

0.15

0.22

0.25

0.14

ML

0.75

0.73

0.57

0.50

0.48

0.18

0.55

-0.01

NL

0.27

0.28

0.12

0.07

0.04

-0.08

0.02

-0.15

MTR

0.38

0.32

0.29

0.27

0.36

0.25

0.17

0.35

IOB

0.29

0.28

0.23

0.05

0.14

0.22

0.06

0.22

NW

-0.09

-0.05

-0.13

-0.17

-0.12

0.03

-0.16

0.02

DOL

0.47

0.52

0.40

0.36

0.29

0.06

0.36

-0.02

PL

0.24

0.27

0.13

-0.04

0.14

-0.00

0.08

-0.04

CD

0.22

0.18

0.25

0.22

0.10

0.13

0.29

0.12

PTL

0.59

0.51

0.54

0.50

0.51

0.25

0.33

-0.10

MD

0.43

0.40

0.48

0.46

0.45

0.12

0.30

-0.15

LAB

0.34

0.31

0.29

0.26

0.13

0.04

0.09

-0.02

OIL

0.48

0.39

0.42

0.35

0.45

0.23

0.22

-0.12

PW

0.14

0.01

0.11

0.15

-0.03

0.18

0.06

0.10

BOS

0.48

0.36

0.37

0.29

0.19

0.17

0.10

0.03

IL

-0.05

-0.01

-0.03

-0.04

0.03

0.22

-0.04

0.28

TZ

0.31

0.32

0.24

0.18

0.30

0.08

0.29

-0.03

FW

0.02

0.12

0.08

-0.04

0.14

0.06

0.04

0.15

imprecision with which these characters were measured with dial calipers; each
0.1-mm increment corresponds to 6.9% of the mean value of WMM vs. only 0.3%
of GLS (table 1).

The extent of correlations of characters within taxonomic and mandibular sets
differs substantially. Seventy-nine of 190 (36.3%) intercorrelations of taxonomic
characters (table 2) differ significantly from zero (P < .05), whereas 20 of 28
(71.4%) correlations of mandibular characters (table 3) are significant. The greater
proportion of significant correlations among characters of the mandibular set can
likely be attributed to the relative structural and functional simplicity of the man-
dible in comparison with that of the cranium.

The extent of correlation between taxonomic and mandibular data sets (table 4)
appears intermediate to correlations within these sets. Eighty-four of 160 between-
set correlations (52.5%) are significant (P < .05) vs. 99 of 218 (45.4%) total within-
set correlations. The relation of correlations within and between character sets is
made evident in the histogram of all 378 correlation coefficients (fig. 3). The
median correlation coefficient for within-set correlations is 0.19 (P > .05), whereas

S LU

o

n

0) J

■*-• ^^

</>

<1> rf

*-»

</)

■r-;

d)

o

(o

c Jl

1

0)

a>

a>

n

a

CsJ

O

I

5

—

o

<2 c

S .9

</) <£

O &

<N v

•c 2

Cl n

° 2

1j 60
c '«

U o

5 rS

C n
01 01

c 1o

01 (ft

■= c

60.2

c .2

0 >*-
C <u

« 8

1 c

£.2
0 _m

i?

o> -r

si

m o<
60 t

o o

— u
<n ^

I- Ot

CO

I-
z

UJ

—

Li.
Li.
LU
O
O

1

11

12

FIELDIANA: ZOOLOGY

that for between-set correlations is 0.31 (P < .01). There is no evidence from the
comparison of correlation coefficients to indicate that taxonomic and mandibular
characters are less closely related to one another than characters of either set are
to each other.

Multivariate Analyses

The complexity of relationships shown by the bivariate analysis can be readily
summarized by canonical correlation analysis. Table 5 contains the results of the
canonical correlation analysis. Of the eight pairs of canonical variates calculated
from the original matrix, only the first is judged significant (P < .001) by Bartlett's
test for the remaining eigenvalues (Dixon, 1977). The correlation of scores on the
first canonical variate is high (R. = 0.91310), indicating that 83.4% of the variance
on the first canonical variates is shared. A close correspondence exists between the
relative positions that individuals occupy on the first canonical variates.

Results of the canonical correlation analysis are presented graphically in Figure
4. Only the first canonical variates are shown, as only these are statistically signifi-
cant. As indicated by the strong correlation between canonical variates, individu-
als tend to occupy the same relative positions in the two morphological spaces.
The possession of a given mandibular morphology strongly implies a correspond-
ing cranial morphology.

Because canonical variates represent linear transformations of original variables,
the importance of any canonical correlation is critically dependent upon the suc-
cess of the initial transformation. Table 6 contains canonical variate loadings (cor-
relations between original variables and canonical variates), variance extracted,
and redundancy for the first two canonical variates. Most variables show positive
correlations with elements of the first canonical vectors and negative correlations
with elements of the second.

'Taxonomic" and "mandibular" character sets are expressed with unequal pre-
cision by their respective canonical variates. The proportion of variance extracted
by each canonical variate is analogous to the variance accounted for by each
principal component (Pimentel, 1979). The first canonical variate extracts 43.2% of
the variance exhibited by mandibular characters, whereas the corresponding
canonical variate for taxonomic characters accounts for only 23.6%. This difference
in the extent to which canonical variates account for variance in the two character

Table 5. Results of canonical analysis, showing Bartlett's test for the remaining eigen-
values. Bartlett's test indicates the number of canonical variables necessary to express the
dependency between the two sets of characters; in the present analysis, only one is
necessary. See text for discussion.

Canonical

No. of

eigenvalue

correlation

eigenvalues

x2

d.f.

Probability

0

271.79

160

0.00000

0.83376

0.91310

1

145.29

133

0.21995

0.46498

0.68190

2

101.20

108

0.66556

0.42337

0.65067

3

62.38

85

0.968%

0.27130

0.52086

4

40.07

64

0.99172

0.18206

0.42669

5

25.90

45

0.99000

0.15116

0.38880

6

14.35

28

0.98454

0.13347

0.36533

7

4.25

13

.0.98827

0.05846

0.24179

PATTERSON: CRANIAL CHARACTERS IN CHIPMUNKS 13

3

0

-1

2

. **/

-2-10 1 2 3

CV 1 -Mandibular Set

Fig. 4. Plot of canonical variate loadings of 86 chipmunks on the first canonical variates
for "taxonomic" and "mandibular" characters. Of the variation along these axes, 83.3% is
shared, indicating a close relationship between the relative positions individuals occupy in
the two morphological spaces. Only one pair of canonical variates is necessary to express
the dependency between "taxonomic" and "mandibular" data sets.

sets can be attributed to the differential intercorrelations of taxonomic and man-
dibular characters (tables 2 and 3), as well as to differences in dimensionality. The
space defined by measurements of mandibular characters has eight dimensions,
whereas that defined by taxonomic characters has 20 dimensions. The greater
dimensionality of taxonomic characters is also reflected in cumulative variance
extracted: together, the eight canonical variates completely account for variance
among mandibular characters, but account for only 56% of the variance exhibited
by taxonomic characters. Such differences are inevitable in comparisons of spaces
differing initially in intercorrelation and dimensionality.

Redundancy provides a measure of the actual overlap of sets in canonical space.
Redundancy is calculated as the proportion of total variance extracted by a canon-
ical variate times the canonical correlation between the variate and the correspond-
ing canonical variate of the other set. Measures of redundancy, therefore, vary in
parallel with extracted variance: for the first pair of canonical variates, redundancy

14

FIELD1ANA: ZOOLOGY

Table 6. Canonical variate loadings (correlations between original variables and canon-
ical variates), variance extracted, and redundancy for the first two canonical variates. Only
the first canonical correlation is significant (P < .001).

Canonical variates

Canonical variates

'Taxonomic"
character

"Mandibular"
character

1

2

1

2

GLS

0.794

-0.079

MC

0.845

-0.394

ZB

0.858

-0.075

MAD

0.822

-0.185

CB

0.193

-0.264

CVH

0.786

-0.088

ML

0.864

-0.167

CDH

0.692

0.032

NL

0.254

-0.143

AGW

0.725

0.199

MTR

0.448

-0.191

ALW

0.269

-0.366

IOB

0.229

-0.544

ALF

0.611

0.123

NW

-0.164

-0.164

WMM

0.015

-0.394

DOL

0.584

-0.090

Variance

PL

0.233

-0.130

extracted

0.4322

0.0678

CD
PTL

0.212
0.736

-0.295
-0.003

Redundancy

0.3604

0.0315

MD

0.633

0.237

LAB

0.369

-0.260

OIL

0.582

0.029

PW

0.038

-0.422

BOS

0.437

-0.501

IL

-0.069

-0.204

TZ

0.399

0.097

FW

0.090

0.022

Variance

extracted

0.2358

0.0603

Redundancy

0.1966

0.0281

is 36% for mandibular characters and 20% for taxonomic characters. Because only
the first canonical correlation is significant, no justification exists for interpreting
subsequent canonical variates. Loadings, extracted variance, and redundancy for
the second canonical variates are included in Table 6 for illustrative purposes only.

Is the correspondence between the taxonomic and mandibular canonical vari-
ates based simply on body size variation, or does it represent a more complex
interdependency? The relationship between loadings on the first canonical vari-
ates and body size was assessed by means of regression analysis. Two indepen-
dent variables related to body size were considered: head-and-body length and
the cube root of weight. The results of these regressions are presented in Table 7.
Both measures of body size are positively and significantly correlated with both
canonical scores. A stronger relationship is evident between the cube root of
weight and canonical variate scores than between the latter and head-and-body
length. This is noteworthy in that (1) weight is known to fluctuate daily and
seasonally, and (2) the specimens examined herein were collected by several rather
than dozens of collectors. Both conditions would suggest that head-and-body
length would be a better criterion of size than weight; instead, the reverse seems
true.

In spite of the significant correlation between canonical variate loadings and
body size, size explains no more than one-fourth of the variation in canonical
variate scores. The remaining variation is typically considered under the general
rubric of "shape." The implication of the finding that canonical variates are not
comprised of simply size-related variation seems clear: the canonical variates sum-

PATTERSON: CRANIAL CHARACTERS IN CHIPMUNKS 15

Table 7. Simple linear regression analyses of loadings on the first canonical variates
(CV1) vs. body size. Two independent measures of size have been used: HBL — head-and-
body length, in millimeters, and CUBEWT — cube root of weight, in grams. Each entry
includes a least-squares equation, a correlation coefficient (r), a coefficient of deter-
mination (R2), and a statement of the likelihood that the regression slope is not different
from zero.

'Taxonomic" set

CV1 = -5.6056 + 0.0496 HBL

r = 0.2954 R2 = 0.0873 P < .01

CV1 = -15.0598 + 4.2660 CUBEWT

r = 0.5380 R2 = 0.2894 P < .001

"Mandibular" set

CV1 = -4.7336 + 0.0419 HBL

r = 0.2494 R2 = 0.0622 P < .05

CV1 = -14.2736 + 4.0433 CUBEWT

r = 0.5099 R2 = 0.2600 P < .001

marize a complex morphological hyperspace with no unitary explanation. The
relatively small coefficients of determination in these regressions imply that cranial
and mandibular morphology may vary in an integrated manner, irrespective of
changes in body size.

DISCUSSION

The present analysis has uncovered some interesting results which bear on the
taxonomic value of mensural, cranial characters in chipmunks. The analysis of
mean values (table 1) demonstrates that significant differences exist among the
studied populations in both "taxonomic" and "mandibular" characters. Compara-
ble proportions of characters of each set show significant differences. A check on
the ecological and evolutionary significance of such differences is provided by the
geographical proximity of least chipmunk populations from Canjilon and Tres
Piedras (see fig. 1). Reference to Table 1 indicates that these two populations agree
closely in both cranial and mandibular morphology, implying that significant
differences in characters among the remaining populations probably have a strong
geographical component.

Analyses of the relative variability of these characters suggests that mandibular
characters tend to be relatively more variable than taxonomic ones. Unfortunately,
the strength of this conclusion is limited by the analytical resolution afforded by
coefficients of variation; no more rigorous parametric test appears as appropriate.
The greater variability of mandibular characters may be attributed to either the
variety of developmental factors known to affect mandibular morphology or the
lesser number of constraints operating on the generally simpler structure of the
mandible. It could be argued that these data support the operation of diversifying
selection on a character set related to resource-use (e.g., Van Valen, 1965), but this
would appear unlikely for reasons given in Patterson (1981).

The correlations of characters within character sets were surprisingly low, es-
pecially those for taxonomic characters. While low correlations of small characters
may be attributed to experimental error in measurement, loose correlations of
larger characters defy a simple explanation. Perhaps the behavioral plasticity of
chipmunks lessens their reliance on any single constellation of cranial characters
(cf. Patterson, 1981).

16 FIELDIANA: ZOOLOGY

Certainly, the morphological differences between even different species complexes
of chipmunks do not prevent them from functioning in an ecologically similar
manner. Where members of the minimus complex of chipmunks are absent from
mountain ranges in the American Southwest, members of the quadrivittatus group
occupy habitats characteristic of E. minimus populations where these are found
(Findley, 1969; Patterson, 1981). Similarly, £. dorsalis (a member of the toumsendii
group) expands the range of habitats it occupies to include those of both £.
minimus and chipmunks of the quadrivittatus complex where these are absent from
a montane biota (Findley, 1969). Loose correlations of cranial morphology may
facilitate such niche shifts by presenting a wider spectrum of cranial configurations
to natural selection. It seems more probable, however, that such loose correlations
exist in the absence of selection for the close integration of characters, and that
behavioral plasticity inhibits selection among particular suites of characters.

The correlation of characters within taxonomic and mandibular sets apparently
differs. The mandibular character set would appear to be more highly integrated,
based on the proportion of significant intercorrelations (tables 2 and 3). This
difference in degree of intercorrelation of characters is difficult to test, however.
Tests of the equality of correlation coefficients between sets suffer from the inter-
dependence of correlations: if characters A and B and characters B and C are
positively correlated, then a positive relation between A and C is implied. In
addition, the two sets under consideration differ in dimensionality, obviating the
possibility of a principal components approach to determining the proportion of
shared variation among characters. The elementary course followed here of
presenting the proportions of significant correlations within sets seems sufficient
to establish the difference (but not the significance thereof) in degree of character
correlations within sets.

A difference in the degree of intercorrelation of characters comprising the two
sets may not be surprising. The morphology of bones is currently recognized as
constituting a response to the demands of functionally related tissues (or periosteal
matrices), especially muscles (e.g., Moss & Meehan, 1970). The periosteal set
affecting mandibular growth and maintenance is likely to be far simpler than that
affecting the cranium because of the radical difference in the complexity of these
two structures. By this interpretation, one could expect that the greater number of
constraints on cranial morphology would lead to reduced variability of cranial
characters, but a greater degree of independence among them, because each
subset of characters would be responding to its own periosteal matrix. Each of
these patterns of morphology has been suggested by the preceding analyses.
Although tentative, this explanation accounts for two seemingly contradictory
morphological patterns, namely, the greater variability, yet greater intercorrela-
tion, of mandibular characters.

The canonical correlation analysis indicated that the dependency between the
two character sets could be described by a single pair of canonical variables (table
5). The strength of this correlation is such that more than 80% of the variance
expressed by the first canonical variates is shared. These data demonstrate that
individuals tend to occupy the same relative positions in the two character sets and
that changes in one character set imply corresponding changes in the other.

However, these conclusions can be no stronger than the canonical correlation
analysis itself. The transformation of original data to canonical axes was accom-
plished with only moderate success: 24% of the variation in taxonomic characters

PATTERSON: CRANIAL CHARACTERS IN CHIPMUNKS 17

and 43% of the variation in mandibular characters is represented in the first
canonical variates. Loose correlations among characters is the probable cause of
such low proportions of extracted variance. Thus, the actual overlap between
characters in canonical space ("redundancy," table 6) is too low to warrant sweep-
ing generalizations regarding th£ functional or developmental interdependency of
taxonomic and mandibular characters.

It is noteworthy, however, that such loose correlations between characters have
also plagued other studies of chipmunk craniometries. In principal components
analyses, axes are derived which extract a maximum amount of variation in the
original data matrix. In a principal components analysis of Eutamias quadrivittatus
chipmunks involving three external and 12 cranial characters, Hoffmeister & Ellis
(1979) found only 48% of the variation among characters was extracted by the first
principal component; the second component expressed only 20%. These results
are compatible with those of the present analysis.

Mandibular Morphology

The present analysis is critically dependent on the assumption that mandibular
characters are responsive to ecological conditions. Two different avenues for the
ecological responsiveness of the mandible are possible: mandibular morphology
may be subject to selection for a specific adaptive function, or mandibular mor-
phology may reflect developmental history or patterns of individual use within a
specific set of ecological conditions.

Clear-cut demonstrations of ad hoc adaptations of the mandible among popu-
lations of mammalian species are few, but such specializations are often assumed
to underlie the radiation of trophic morphologies between species. Much study
has been directed toward the bills of birds and their relations to food resources
(e.g., Kear, 1962; Grant, 1967; Willson, 1971). Bill morphology has been found to
parallel, and may in fact determine, the range of food resources used by popu-
lations. Characters associated with particular adaptations are unreliable bases for
taxonomy (under the "Darwin principle"), and those "associated with shifts in the
food niche are particularly susceptible to a rapid attainment of conspicuous differ-
ences or, conversely, of convergent similarities" (Mayr, 1969, p. 223). Mayr illus-
trated this point by comparing the piscivorous bills of fish-eating mergansers
{Mergus) with more typical ducklike bills of related goldeneyes (Bucephala). Evo-
lutionary adaptation to particular "stations in life" may obscure underlying phyle-
tic relationships by convergence or spurious divergence of characters involved in
that adaptation.

Ontogenetic modifications of mandibular morphology in response to prevailing
ecological conditions are also known. A variety of studies has elucidated com-
ponents of mandibular variability in laboratory rats and mice. For example, experi-
mental studies indicate that the final dimensions of bones, including the mandible,
may depend on levels of nutrition during growth (McCance, 1962; Park, 1968; Park
& Nowosielski-Slepowron, 1971; McAnulty, 1977). The type of nutrition may also
influence mandibular development. Moore (1965) fed laboratory rats an abnor-
mally soft diet, thereby lessening the utility of masticatory muscles. Effects of this
regimen included a 12% reduction in the mass of the cranium and mandible, a 1%
to 2% decrease in the dimensions of the cranial and facial skeleton, and a 4%
decrease in the length of the angular process of the mandible (Moore, 1965).

18 FIELDIANA: ZOOLOGY

Subsequent studies of the effects of surgical removal of masticatory muscles
corroborate these findings, suggesting a direct role of muscle development and
use in mandibular growth and maintenance. Moss & Meehan (1970, p. 12) showed
that "temporalis muscle removal in the juvenile rat is followed rapidly by a marked
change in the form of the coronoid process," the related microskeletal unit of the
temporalis. Moore (1973), studying the effects of bilateral masseterectomy in rats,
found a 25% reduction in the dimensions of the angular process; lesser effects (all
involving reductions) were also discerned in the length (11 %) and width (17%) of
the condyloid process, in the height of the condyloid (10%), and in the length of
the base (5%). Again, Mayr (1969) has argued that characters marked by high
variability are not useful in taxonomic decision-making.

Thus, both genetic and ontogenetic mechanisms exist for the correspondence of
mandibular morphology with ecological conditions. Holbrook (1982) has recently
documented a case in which free-living Peromyscus mamculatus in different, nearby
habitats exhibit different mandibular morphologies. Dietary differences between
habitats may account for the mandibular differences observed in these deer-
mice, but genetic studies of morphological traits are necessary to fully distin-
guish between genetic and ontogenetic effects (e.g., Leamy, 1974, 1977; Atchley
et al., 1981).

Mammalogists have long recognized the problems inherent in a taxonomy
based on mandibular characters; this is reflected in the common choice of cranial
over mandibular characters. However, the present study indicates that cranial and
mandibular morphologies in these chipmunks are not as insulated from one an-
other as the traditional taxonomic use of cranial characters would suggest.
Changes in one character set apparently imply changes in the other. I believe the
parallel responses of cranial and mandibular characters warrant a reassessment of
those cranial characters used to classify chipmunks and their allies.

Cranial Morphology in Chipmunks and Related Forms

Complex patterns of variation in cranial and external morphology have long
puzzled students of chipmunk taxonomy. Allen (1877, p. 797), in the North Amer-
ican Rodeutia, stated: "While the five varieties of T. asiaticus [that were then recog-
nized and] above characterized so thoroughly intergrade that they are not to be
trenchantly defined, the extreme phases of differentiation are often quite widely
diverse, and would require recognition as distinct species were they not found to
be so inseparably connected." Several years later, in the description of a new
species, Merriam (1886, p. 28) observed: "That this genus is peculiarly susceptible
to environmental influences is amply attested by the number and perplexing
characteristics of the incipient species already known from the United States."
Allen (1889, p. 178) believed that, with additional collecting, "Many well-marked
and easily definable forms will be found confined to very limited regions, and to
peculiar environmental conditions." These early assessments of morphological
variation in chipmunks are best summarized by Allen (1890, p. 53) in his generic
revision:

From the extreme susceptibility of this plastic group to the influences of environment,
it is one of the most instructive and fascinating groups among North American
mammals. No one can doubt its comparatively recent differentiation from a common
stock, and its dispersion from some common centre. Whether the type originated at
some point in North America, or in the northern part of Eurasia, it is perhaps idle to

PATTERSON: CRANIAL CHARACTERS IN CHIPMUNKS 19

speculate, but that it has increased, multiplied, spread, and become differentiated to
a wonderful degree in North America is beyond question; as it is found from the Arctic
regions to the high mountain ranges of Central Mexico, and has developed some
twenty to thirty very palpable local phases. Some of them easily take rank as species,
others as subspecies. Probably a more striking illustration of evolution by environ-
ment cannot be cited. •

Nearly 40 years later, Howell (1929) undertook a generic revision of chipmunks,
involving more than 13,000 specimens and 60 taxa in Eutamias alone. He empha-
sized external and pelage characters, as well as geographic distribution, to the
exclusion of cranial characters. Howell's reduced reliance on cranial characters is
reflected in his generic key (Howell, 1929, pp. 30-33); in no case does a cranial
character other than size (greatest length of skull) define a bifurcation in the key.
Howell's classification is still generally accepted (e.g., Hall, 1981), albeit with
minor alterations of taxonomic status and descriptions of new forms (e.g., White,
1953; Callahan & Davis, 1977).

A more recent evaluation of cranial (and other) characters in chipmunks was
offered by White (1953, p. 565):

The differences in anatomy and color between many species of chipmunks are subtle,
and refined techniques are required to discover them. When "measuring" chipmunks
taxonomically, it is necessary to use a "chipmunk scale" and not, for example, a
"pocket-gopher scale." In explanation, some species of pocket gophers closely allied
to each other, and even some subspecies of the same species, differ markedly in color
and in size and shape of parts of the skeleton; comparable differences are not so
pronounced among many species of chipmunks.

Other characters, especially those of the baculum, have since been shown to be
superior to cranial characters in making inferences about reproductive limits in
chipmunks (e.g., White, 1953; Sutton & Nadler, 1974; Callahan, 1980).

Nor are such confusing patterns of cranial morphology confined only to chip-
munks; they appear to be characteristic of many members of the Sciuridae. In fact,
the baculum was first named as such in a taxonomic work on Sciurus, in which
Thomas (1915, p. 383) stated:

There has always appeared to be something wrong with the inclusion of the Oriental
squirrels in the same genus as Sciurus vulgaris, although when classifying the group
some years ago I was unable to find any material differences in their skulls and teeth.

Now, however, I have found a character by which such squirrels as are still put in
Sciurus may be sorted into several groups, each sharply defined from the others.

This is the structure of the os penis, which shows very striking differences between
the various groups of species, and may evidently be of great service in classifying the
members of this difficult family.

Shortly after, Pocock (1923, p. 211) surveyed the baculum in many sciurids, alter-
ing previous classifications based on skulls and teeth: ". . . the conclusion very
forcibly suggested by the literature of the subject is the untrustworthiness of such
characters." An excellent example of this is provided by Sciurus and Tamiasciurus,
forms considered congeneric on the basis of cranial and dental characters until the
highly significant differences in the male reproductive tracts of these forms were
recognized (Mossman et al., 1932).

Bryant (1945), in his review of all Nearctic squirrels, acknowledged three major
subdivisions of the family: flying squirrels, tree squirrels, and chipmunks and
ground squirrels; however, "The significant phylogenetic differences apparent in
the skulls of sciurids are principally associated with the masticatory mechanism
and with the differential development of the parts of the brain" (p. 288). These
characters would therefore appear to document the major branches of the adaptive

20 FIELDIANA: ZOOLOGY

radiation of the Sciuridae, but would appear impotent, under the "Darwin prin-
ciple" (Mayr, 1969), for assessing the phyletic relations of taxa within these adap-
tive zones.

Moore (1959, p. 159) judged the value of cranial characters to sciurid taxonomy
as follows:

For the tree squirrels in particular, no one has been able to report any satisfactory skull
characters that distinguish, even at the generic level, the tree squirrels of the Palearctic
and Nearctic regions (Sciurus) from Callosciurus and others of the Indo-Malayan re-
gion, or Heliosciurus and others of the Ethiopian region, or Guerlinguetus and others
from the Neotropical region.

Moore's "assiduous" search for skull characters supporting these taxonomic dis-
tinctions did reveal some meristic and qualitative ones (e.g., the number of trans-
verse septa in the auditory bulla, the inclination and shape of foramina, suture
ankylosis), but, in sum, "the various genera that may be ecologically classified as
tree squirrels possess few distinguishing skull characters" (p. 192). Specific differ-
ences in skull characters are known for semiterrestrial and ground squirrels, how-
ever (e.g., Moore & Tate, 1965).

Others studies of ground squirrels also suggest that they may show patterns of
cranial morphology contrasting with those exhibited by chipmunks and tree squir-
rels. In his study of North American ground squirrels, Howell (1938) listed a
number of cranial and dental characters by which the various genera and sub-
genera could be recognized. The relative stability of ground squirrel classification
in comparison with those of chipmunks and tree squirrels is also suggestive of the
greater phyletic information embodied in the cranial characters of ground squir-
rels. It appears likely, therefore, that results of this study are more pertinent for
character divergence among tree squirrels than among ground squirrels.

The present study thus provides quantitative support for conclusions reached
earlier by Allen, Merriam, and others: mensural cranial characters of chipmunks
appear responsive to environmental conditions, and hence are unreliable grounds
for phyletic inference. Purely pheneric studies of cranial morphology in these
animals could describe patterns of variability more closely related to ecological
conditions than to phyletic history (cf . Patterson, 1981), especially in areas of great
topographic and ecological heterogeneity. In such cases, taxonomic distinctions
based on cranial morphology might reflect nothing more than contrasting environ-
mental conditions in different parts of a taxon's geographical range.

Problems raised by this source of cranial variability can be illustrated by the
systematics of Southwestern chipmunks. For nearly half a century (cf. Bailey,
1931), chipmunks from the Organ Mountains, New Mexico, were assigned to
Eutamias cinereicollis cinereus. As such, they were thought to be a transition series
joining widely separated populations on the Sacramento Mountains to the east
with populations inhabiting the Mogollon Rim to the west under E. cinereicollis.
Fleharty (1960) subsequently noted the distinctive baculum of the eastern mem-
bers of this group, elevating these populations to specific status as £. canipes.
Patterson (1980) then found that both karyotypic and bacular evidence indicated
that the Organ Mountains population in fact represents a southern derivative of
£. quadrwittatus. Rather than representing a transition series between E. canipes
and E. cinereicollis (cf. Findley et al., 1975), the Organ Mountains population of E.
quadrwittatus documents a more complex biogeographic history detailed elsewhere
(Patterson, 1982). Previous considerations of E. canipes, E. cinereicollis, and E.
quadrwittatus australis as conspecific populations reflect the convergent responses

PATTERSON: CRANIAL CHARACTERS IN CHIPMUNKS 21

of populations inhabiting a region whose climatic regimen is dominated by the
northern Chihuahuan Desert (Patterson, 1980).

An additional example of the taxonomic effects of this evolutionary potential is
afforded by red squirrels. The Baja California red squirrel, Tamiasciurus mearnsi,
was accorded specific status because cranial characters showed it to be as distinct
from T. douglasii and T. hudsonicus as these two are from one another (Lindsay,
1981). For the last two species, the "test of sympatry" is present in the Pacific
Northwest and affords a "phenetic yardstick" by which to assess allopatric cases
(e.g., Mayr, 1969). However, the cranial distinctiveness of Baja red squirrels may
not signify specific divergence, but instead reflect the divergent environmental
conditions pertaining there. Lindsay's (1982) more recent work on northern red
squirrels indicates that patterns of cranial morphology can be traced to the differ-
ent types of conifers in which various populations live. Is the Baja population of
Tamiasciurus distinctive because of its phyletic history or the forests it occupies? At
present, no concrete answer seems possible.

The present study of chipmunk morphology would be strengthened by an
explicit demonstration that mandibular characters in chipmunks are closely speci-
fied by ecological conditions. I have not attempted such an analysis because the
environmental determination of mandibular morphology is expected on both gen-
etic and ontogenetic grounds. In addition, other studies demonstrate a re-
lationship between mandibular morphology and environmental conditions, both
in other rodents and in more distantly related taxa. Thus, assuming a relationship
between mandibular morphology and environmental conditions in chipmunks
seems warranted, if not substantiated.

It is unfortunate that the most pertinent kind of data for conclusions of this
nature seems unavailable. Because chipmunks and tree squirrels do not acclimate
readily to laboratory conditions, controlled genetic studies of morphological vari-
ability cannot be conducted (cf. Atchley et al., 1981). Barring such direct ap-
proaches, this alternative, employing correlations among characters, appeared
most fruitful.

In spite of these limitations, it is recommended that strictly phenetic studies of
cranial morphology not be incorporated into the taxonomy of chipmunks and tree
squirrels. Many other characters exist in these taxa, and they may serve in studies
of phyletic relationships (e.g., karyotypes, pelage, and postcranial morphology,
especially of the reproductive tract, baculum, and feet). On the other hand, the
noteworthy plasticity of cranial characters in these groups may provide much
insight into patterns of ecological and evolutionary divergence and help to explain
the notable success of the family Sciuridae.

ACKNOWLEDGMENTS

This project was conceived while I was a graduate student at New Mexico State
University; my thanks to fellow graduate students and Profs. C. S. Thaeler and
J. R. Zimmerman for enlightening discussions on systematics. Field expenses were
partially defrayed by the Department of Biology, N.M.S.U., and multivariate
analyses were conducted at the Computer Center, N.M.S.U. Rosanne Miezio
drew Figure 1 and has otherwise given unstintingly of her time and attention.
Sarah Derr Bruner typed the manuscript. I wish to thank L. R. Heaney, S. L.
Lindsay, and R. M. Timm for their comments throughout this work.

22 FIELDIANA: ZOOLOGY

LITERATURE CITED

ALLEN J A 1877. Monographs of North American Rodentia. XL Family Sciuridae, pp.

637-939. In Coues, E., and J. A. Allen (eds.). Monographs of North American Rodentia.

U.S. Geological Survey, Washington, D.C.
. 1889. Notes on a collection of mammals from southern Mexico, with descriptions

of new species of the genera Sciurus, Tamias, and Sigmodon. Bulletin of the American

Museum of Natural History, 2: 165-181.

1890. A review of some of the North American ground squirrels of the genus

Tamias. Bulletin of the American Museum of Natural History, 3: 45-116.
An m i "i W R . J ). Rutledce, and D. E. Cowley. 1981. Genetic components of size and

shape. II. Multivariate covariance patterns in the rat and mouse skull. Evolution, 35:

1037-1055.
BaILEY, V 1931. Mammals of New Mexico. North American Fauna, 53: 1-412.
Bakm n. R J 1977. Bergmann's rule and variation in structures related to feeding in the

gray squirrel. Evolution, 31: 538-545.
Behle. W H 1973. Clinal variation in white-throated swifts from Utah and the Rocky

Mountain region. Auk, 90: 299-306.
Bryant. M D 1945. Phvlogenv of Nearctic Sciuridae. American Midland Naturalist, 33:

257-390.
Callahan. J R 1980. Taxonomic status of Eutamias bulleri. Southwestern Naturalist, 25:

1-8.
Callahan, J. R.. and R Davis. 1977. A new subspecies of the cliff chipmunk from coastal

Sonora, Mexico. Southwestern Naturalist, 22: 67-75.
CONLEY . W H 1970. Geographic variation in the least chipmunk, Eutamias minimus, in New

Mexico and eastern Arizona. Journal of Mammalogy, 51: 695-702.
Cooley. W W , and P. R. Lohnes. 1971. Multivariate Data Analysis. John Wiley & Sons,

New York, 364 pp.
Dixon. W J 1977. BMDP Biomedical Computer Programs. University of California Press,

Berkeley, 880 pp.
FlNDLEY, J S 1969. Biogeography of southwestern boreal and desert mammals, pp.

113-128. In Jones, J. K., Jr. (ed.), Contributions in Mammalogy. Miscellaneous

Publication, Museum of Natural History, University of Kansas, 51: 1-428.

FlNDLEY, J S.. A H Harris. D. E. Wilson, and C. Jones. 1975. Mammals of New Mexico.

University of New Mexico Press, Albuquerque, 360 pp.
Fleharty. E D. 1960. The status of the grey-necked chipmunk in New Mexico. Journal of

Mammalogy, 41: 235-242.

G I noways. H. H 1973. Systematics and evolutionary relationships of spiny pocket mice,

genus Liomys. Special Publications of the Museum, Texas Tech University, 5: 1-368.
Grant. P R 1967. Bill length variability in birds of the Tres Marias Islands, Mexico.

Canadian Journal of Zoology, 45: 805-815.
Hall. E R 1981. Mammals of North America, vol. 1. John Wilev & Sons, New York,

600 + 90 pp.
Hoffmeister. D. F.. and L. S. Ellis. 1979. Geographic variation in Eutamias quadrivittatus

with comments on the taxonomy of other Arizonan chipmunks. Southwestern

Naturalist, 24: 655-665.
Holbrook. S J 1982. Ecological inferences from mandibular morphology of Peromyscus

maniculatus. Journal of Mammalogy, 63: 399-408.
Howell. A H 1929. Revision of the American chipmunks (genera Tamias and Eutamias).

North American Fauna, 52: 1-157.

. 1938. Revision of the North American ground squirrels, with a classification of

the North American Sciuridae. North American Fauna, 56: 1-256.

Johnson. D H 1943. Systematic review of the chipmunks (genus Eutamias) of California.
University of California Publications in Zoology, 48: 63-148.

PATTERSON: CRANIAL CHARACTERS IN CHIPMUNKS 23

Johnston, R F. 1969. Character variation and adaptation in European sparrows. Systematic
Zoology, 18: 206-231.

Kear. J. 1962. Food selection in finches, with special reference to interspecific differences.
Proceedings of the Zoological Society of London, 138: 163-204.

Leamy. L. J 1974. Heritability of osteometric traits in a randombred population of house
mice. Journal of Heredity, 65: 109-120.

. 1977. Genetic and environmental correlations of morphometric traits in random-
bred house mice. Evolution, 31: 357-369.

Levenson, H., C. F. Nadler, and R S Hoffman In review. Systematics of the Holarctic
chipmunks (Tamias). I. Morphological relationships and subgeneric affinities. Journal of
Mammalogy.

Lindsay. S L. 1981. Taxonomic and biogeographic relationships of Baja California chick-
arees (Tamiasciurus). Journal of Mammalogy, 62: 673-682.

. 1982. Morphologic variation and its systematic significance in tree squirrel spe-
cies (Tarn iasciurus) in western North America. Unpublished Ph.D. dissertation, Memphis
State University, Memphis, 133 pp.

Marshall. L. G 1977. Evolution of metatherian and eutherian (mammalian) characters: A
review based on cladistic methodology. Zoological Journal of the Linnean Society, 66:
369-110.

Mayr. E 1969. Principles of Systematic Zoology. McGraw-Hill, New York, 428 pp.

McAnulty. P. A 1977. Growth and skeletal development of the rat during prolonged

undernutrition and subsequent rehabilitation. Acta Anatomica, 98: 39-51.
McCance. R. A. 1962. Food, growth, and time. Lancet, 2: 621-626.
Merriam, C H 1886. Description of a new species of chipmunk from California {Tamias

macrorhabdotes sp. nov.). Proceedings of the Biological Society of Washington, 3:

25-28.
Moore. J C. 1959. Relationships among living squirrels of the Sciurinae. Bulletin of the

American Museum of Natural History, 118: 157-206.
Moore. J C . and G. H. H. Tate 1965. A study of the diurnal squirrels, Sciurinae, of the

Indian and Indochinese subregions. Fieldiana: Zoology, 48: 1-351.
Moore. W J 1965. Masticatory function and skull growth. Journal of Zoology, 146:

123-131.
. 1973. An experimental study of the functional components of growth in the rat

mandible. Acta Anatomica, 85: 378-385.
Moss. M L., and M. Meehan. 1970. Functional cranial analysis of the coronoid process in

the rat. Acta Anatomica, 77: 11-24.
Mossman. H W , J. W. Lawlach, and J. A. Bradley. 1932. The male reproductive tract of

the Sciuridae. American Journal of Anatomy, 51: 89-155.
Park. A W 1968. Aspects of oral development in relation to litter size in rats. Acta

Anatomica, 70: 403.
Park. A W . and B. A. J. Nowosielski-Slepowron 1971. The effects of litter size on rat body

growth. Acta Anatomica, 79: 15-26.
Patterson. B D 1980. A new subspecies of Eutamias quadrivittatus (Rodentia: Sciuridae)

from the Organ Mountains, New Mexico. Journal of Mammalogy, 61: 455-464.
. 1981. Morphological shifts of some isolated populations of Eutamias (Rodentia:

Sciuridae) in different congeneric assemblages. Evolution, 35: 53-66.

-. 1982. Pleistocene vicariance, montane islands, and the evolutionary divergence

of some chipmunks (genus Eutamias). Journal of Mammalogy, 63: 387-398.
Pimentel, R A 1979. Morphometries. Kendall/Hunt, Dubuque, Iowa, 276 pp.
Pocock. R I 1923. The classification of the Sciuridae. Proceedings of the Zoological Society

of London, 15: 209-246.
Rohlf. F. J., and R R Sokal 1969. Statistical Tables. W. H. Freeman, San Francisco, 253 pp.
Siegel. S. 1956. Nonparametric Statistics for the Behavioral Sciences. McGraw-Hill, New

York, 312 pp.

24 FIELDIANA: ZOOLOGY

Straney. D O. and J. L. Patton. 1980. Phylogenetic and environmental determinants of
geographic variation of the pocket mouse Perognathus goldmani Osgood. Evolution, 34:

Sutton. D. A., and C F. Nadler. 1974. Systematic revision of three Townsend chipmunks

(Eutamias toumsendii). Southwestern Naturalist, 19: 199-212.
Thomas. O. 1915. The penis bone, or "baculum," as a guide to the classification of certain

squirrels. Annals & Magazine of Natural History, 15: 383-387.
Tonce, C. H.. and R. A. McCance. 1965. Severe undernutrition in growing and adult

animals. 15. The mouth, jaws and teeth of pigs. British Journal of Nutrition, 19: 361-372.
Van Valen. L. 1965. Morphological variation and width of ecological niche. American

Naturalist, 99: 377-390.
White. J. A. 1953. Taxonomy of the chipmunks, Eutamias quadrivittatus and Eutamias umbri-

nus. University of Kansas Publications, Museum of Natural History, 5: 563-582.
Willson. M F. 1971. Seed selection in some North American finches. Condor, 73: 413-429.

APPENDIX

Specimens Examined

All specimens in the following list are currently classified as Eutamias minimus
operarius and are deposited in the Vertebrate Museum, New Mexico State
University.

NEW MEXICO: Lincoln Co.: N. Ridge, Sierra Blanca Pk., 11,500 ft (3). Otero Co.: East
face. Sierra Blanca Pk., 11,500 ft (8); V* mi N top, Sierra Blanca Pk., 11,500 ft (2). Rio Arriba
Co.: 1.9 mi N, 1.7 mi E Canjilon, 8,200 ft, T26NR5Esec9 (2); 5.0 mi E, 4.7 mi N Canjilon,
9,800 ft, T27NR6Esec30 (12); 15.0 mi W, 3.1 mi N Tres Piedras, 9,750 ft, T29NR7Esec30 (3);
15.3 mi W, 3.1 mi N Tres Piedras, 9,800 ft, T29NR7Esec31 (10); 15.4 mi W, 3.0 mi N Tres
Piedras, 9,750 ft, T29NR7Esec31 (1); 15.4 mi W, 3.0 mi N Tres Piedras, 9,800 ft,
T29NR7Esec31 (1); 15.5 mi W, 5.8 mi N Tres Piedras, 9,800 ft, T29NR7Esecl9 (3); 24.1 mi
W, 3.8 mi N Tres Piedras, 9,800 ft (10). Santa Fe Co.: Little Tesuque Creek, 8,200 ft (1); Santa
Fe Ski Basin, Little Tesuque, 8,200 ft (1); Santa Fe Ski Basin (3); Santa Fe Ski Basin, 10,000
ft, T18NRllEsec7 (2); Santa Fe Ski Basin, 10,500 ft (1); Santa Fe Ski Basin, upper (1); Upper
Santa Fe Ski Basin, 10,500 ft, T18NR11E (4).

UTAH: San Juan Co.: 6.8 mi S, 1.3 mi E Mount Peale, LaSal Mts., 7,650 ft,
T28SR25Esec30 (14); 8.2 mi S, 0.3 mi E Mount Peale, LaSal Mts., 7,650 ft, T28SR24E-
sec36 (4).

Field Museum of Natural History
Roosevelt Road at Lake Shore Drive
Chicago, Illinois 60605-2496
Telephone: (312) 922-9410

t\v

UNIVERSITY OF ILLINOIS URBANA

3 0112 018406972