i

o

II B HAHY

OF THE
UNIVERSITY
Of ILLINOIS

550.5

Fl
y-IO--ll

GEDLQdl

UNIVERSITY OF

ILLINOIS LIBRARY
AT URBANA-CHAMPA1GN

GEOLOGY

Return this book on or before the
Latest Date stamped below.

GEOLOGY LIBRARY

University of Illinois Library

dec 30 m :mi

198

4r

A,

IfC t«» MAR* 5-^

FEB 3 1967

JM

MAY

it up

JUL j ;

WAfK

NOV 2

deU>

s,

JS68

1972

WW

172004

L161— H41

EVOLUTION AND CLASSIFICATION
OF THE OSTEOSTRACI

ROBERT H. DENISON

Curator of Fossil Fishes

FIELDIANA: GEOLOGY

VOLUME 11, NUMBER 3

Published by

CHICAGO NATURAL HISTORY MUSEUM

JANUARY 19, 1951

IHtlWRARVOFlHE

FES* ,j51

PRINTED IN THE UNITED STATES OF AMERICA
BY CHICAGO NATURAL HISTORY MUSEUM PRESS

INTRODUCTION

The past twenty-five years have seen a tremendous increase in
our knowledge of the Osteostraci. In two monumental works on
the cephalaspids of Spitsbergen and Great Britain, Stensio (1927,
1932) has made a notable contribution to our understanding of the
detailed anatomy of the group. The ateleaspids (hemicyclaspids),
particularly those from Norway, have been exhaustively treated by
Heintz (1939). Robertson has produced a number of papers dealing
with the fauna from the Island of Oesel in the Baltic. New forms
from Spitsbergen, Great Britain, Poland, Norway and North
America have been described by other authors. In spite of all this
work, however, there is still a lack of agreement concerning the
general evolutionary trends within the order, and a poor under-
standing of the relationships of the various genera.

The present paper is the result of a study of the Late Silurian
Osteostraci from Oesel, collected by William Patten and preserved
in the Dartmouth College Museum. This fauna is of particular
interest not only because it includes the earliest known members of
the order, but also because of the considerable variety of forms by
which it is represented. The Oesel genera are of the utmost im-
portance in the study of the evolution and classification of the
Osteostraci, with which this paper is primarily concerned.

TAXONOMIC REVISIONS

There is no attempt in this paper to present a complete taxo-
nomic revision of the Osteostraci. In the course of this study, how-
ever, the need for certain changes has become apparent, and they
are discussed at this time in order to avoid confusion and misunder-
standings in later sections of this work.

Robertson (1939a) established the genus Witaaspis with Cephal-
aspis schrenkii Pander1 as genotype. In 1940, he described a
second species of the genus, W. patteni. Both are from the Wita
Quarry on the Island of Oesel, and both are of similar dimensions.
W. patteni was distinguished in part by the ornamentation of its

1 Not C. schrenckii as used by Robertson, nor C. schrencki as used by Stensio,
927.

157

158 FIELDIANA: GEOLOGY, VOLUME 11

cephalic shield, but such differences as appear are due to the manner
of preservation; they depend on whether the matrix has broken
away fully from the surface of the shell, or has removed with it
part or all of the exoskeleton. The apparent large size of the dorsal
field and the low median crista of W. schrenkii are again due to
faulty preservation. In the absence of any other characteristics
to distinguish it, Witaaspis patteni Robertson must be referred to
W. schrenkii (Pander).

In another study of the Oesel Osteostraci, Robertson (1938a,
p. 288) established a new cephalaspid genus, Saaremaaspis, for the
species Tremataspis mickwitzi Rohon. In a later paper (1938b,
p. 489) he described Rotsikilllaspis obrutchevi, a new genus and
species, and referred it to the Dartmuthiidae. According to Robert-
son, there are only two known specimens of S. mickwitzi, the holotype
in the Academy of Sciences in Moscow (No. 256:536) and one
specimen in the Dartmouth College Museum (No. 38-71-9526). It
appears that Robertson's differentiation of Rotsikilllaspis from Saare-
maaspis was the result of a misinterpretation of the inadequate
material of the latter genus. The two species agree quite closely
in size. Robertson was in error in giving the length of S. mickwitzi
as 18 mm. and the maximum width as 19 mm. A comparison of
the measurements is given below:

Total length Maximum width
mm. mm.

S. mickwitzi 27 28

(type from Robertson, 1938b, pi. 60, fig. 8)

S. mickwitzi 29 26

(measured from D.C.M. 38-71-9526) (estimated)

R. obrutchevi 29 24

(type, D.C.M. No. 38-71-9404)

The most striking differences that are apparent from Robertson's
restorations (1938b, text figs. 1, 2) are the lengths of the shields of
the two, and the presence of short, broad cornua and deep pectoral
sinuses in Saaremaaspis. His restoration of Saaremaaspis is based
largely on the photograph of the type (1938b, pi. 60, fig. 8) in which
the posterior boundary appears to be a break rather than an actual
edge of the shield. Cornua in Osteostraci are developed at the
posterior corners of the cephalic shield, not far back in the trunk
region as Robertson has shown them. Thus there is good reason
to believe that Saaremaaspis lacked cornua, that the type was
incomplete, and that the total length of the shield was greater than
27 mm. None of the other differences of shape and proportion that
Robertson mentions is particularly significant when it is considered

DENISON: OSTEOSTRACI 159

that all of the known specimens are crushed or distorted to some
extent. Another character used by Robertson to distinguish these
genera is the length of the lateral fields. The boundaries of these
fields are indistinct in the type of Saaremaaspis mickwitzi, but,
according to my interpretations, the fields are similar in extent to
those of Rotsikullaspis obrutchevi; this is certainly true of D.C.M.
No. 38-71-9526, referred by Robertson to Saaremaaspis. Robertson
mentions differences in the ornamentation of the exoskeleton, but
does not state what they are; none is apparent.

No valid grounds for the separation of these two species are
left, so it becomes necessary to refer Rotsikullaspis obrutchevi to
Saaremaaspis mickwitzi as a synonym. The relationships of this
species to other Osteostraci are discussed below.

For reasons to be presented in the discussion of the classification
of the Osteostraci that follows, the following taxonomic changes are
proposed :

Cephalaspis oeselensis Robertson is referred to Procephalaspis,
gen. nov.

Cephalaspis staxrudi Stensio is referred to the genus Securiaspis
Stensio.

Cephalaspis woodwardi Stensio is referred to Stensiopelta, gen. nov.

In order to avoid confusion in the discussion of the evolution of
the Osteostraci, a classification is presented below. This differs
from earlier classifications in many respects, but a discussion of it
will be deferred until the evolution has been considered.

Order Osteostraci

Family Tremataspidae

Subfamily Tremataspinae Tremataspis

Subfamily Dartmuthiinae { £££*,,

Subfamily Oeselaspinae Oeselaspis

Subfamily Didymaspinae Didymaspis

Family Sclerodontidae Sclerodus

Hemicyclaspis

Hemiteleaspis

Family Ateleaspidae | 22S#

Aleleaspis
{ Wilaaspis
Family Cephalaspidae

f Thyestes

Subfamily Cephalaspinae I Procephalaspis

{ Cephalaspis

160 FIELDIANA: GEOLOGY, VOLUME 11

f Securiaspis
I Benneviaspis

Subfamily Benneviaspinae \ Hoelaspis

?Boreaspis
Stensiopelta
Family Kiaeraspidae Kiaeraspis

THE EVOLUTION OF THE OSTEOSTRACI

The Osteostraci are a relatively compact group, first appearing
in the Late Silurian, flourishing in the Early Devonian, but repre-
sented by only a few survivors in the Middle and Late Devonian.
Throughout this time they were quite conservative in many respects,
although there are obvious differences in certain characteristics.
The evolutionary significance of these differences has been variously
interpreted. Stensio (1927, 1932) came to the following conclusions:

(1) The trunk carapace was lengthened by the incorporation
of trunk scales.

(2) The pectoral fins, pectoral sinuses, and cornua have been
reduced in such forms as Kiaeraspis and Didymaspis, and lost in
Tremataspis.

(3) Both the exoskeleton and endoskeleton have shown a de-
generation of ossification.

(4) The paired lateral fields of Tremataspis are the result of a
subdivision of primitive single fields.

(5) The originally metameric nerves supplying the lateral fields
have been modified and reduced in number by fusion.

These views have been supported recently by Wangsjo (1946).
On the other hand, Westoll (1945) has come to diametrically opposed
conclusions. In his opinion, the Osteostraci with long carapaces
and small or absent pectoral sinuses and cornua are primitive, this
view being based largely on their supposed dominance in the earliest
deposits. The solution to the problem is not obvious, however,
when we consider the variety of forms described from the Late
Silurian of Oesel. Among them are genera with long carapaces and
those with short, genera with well-developed pectoral sinuses and
those without, genera with subdivided lateral fields and those with
only one pair, and, finally, genera with strong endo- and exo-skeletons
as well as those with poorly developed skeletons.

In order to clarify the situation, an attempt has been made in
the present paper to study the controversial characters as objectively
as possible, in order to determine whether any evolutionary trends
are demonstrable. Where possible, the characters are expressed as

DENISON: OSTEOSTRACI

161

*T~*~

^_^LJ J

Fig. 20. Dorsal shield of Procephalaspis, illustrating the measurements used
in this paper. A. Distance from pineal opening to line connecting centers of
external openings of endolymphatic ducts. B. Prepineal length (excluding rostral
process, if present). C. Postpineal length, median. D. Postpineal length,
measured to level of postero-lateral corners of cephalic or trunk shield. E. Dis-
tance from pineal opening to line connecting the anterior ends of the pectoral
sinuses. F. Depth of pectoral sinus. G. Length of lateral field. H. Maximum
distance between external borders of lateral fields.

measurements, or more often as ratios of measurements. Since most
actual specimens are crushed, distorted, or incomplete, the data have
been derived mainly from restorations, some original but many from
the published figures of Stensio (1927, 1932), Wangsjo (1937) and
Zych (1937). The species are grouped by age in three categories,
Ludlow, Downtonian and Early Devonian, and the numerical
expression of the characters is represented graphically.

It has been found necessary at the outset to eliminate a con-
siderable number of the described Osteostraci from this evolutionary
study. Many are too poorly preserved to furnish the required
measurements. This, unfortunately, is true of all the Middle and
Late Devonian species. Others have been found in strata whose
geological age is uncertain, or whose age is determined in large part
by the supposed evolutionary stage of their contained Osteostraci.
This applies at present to the forms from Norway and Scotland.
Thus, this study is limited to the species found in the West Midlands
)f England, Spitsbergen, and Oesel, and the one species described as
/et from Poland. The measurements used are indicated in figure 20.

162 FIELDIANA: GEOLOGY, VOLUME 11

The approximate stratigraphic occurrence of the genera is indicated
in figure 31, page 194.

THE SIZE OF THE LATERAL FIELDS

Tremataspis and Oeselaspis are unusual among the Osteostraci
in having two pairs of lateral fields; all the other described genera
have but a single pair. This characteristic has been used in classifica-
tion as a primary basis for distinguishing these genera as separate
families (Stensio, 1927; Robertson, 1938a), suborders (Heintz, 1939),
or even orders (Berg, 1940). Two pairs of lateral fields might be
considered as a primitive character since they occur in the earliest
fauna, or they might be regarded as the result of subdivision of ori-
ginally single lateral fields. The latter view was apparently taken
by Stensio (1927, p. 308; 1932, p. 180). Wangsjo (1946, p. 360)
mentions an undescribed form from Spitsbergen which "shows a
still more advanced subdivision of the lateral electric fields into
separate portions than we find in Tremataspis and Oeselaspis. ..."

It is proposed here to determine whether there is any evidence
supporting a trend towards an increase or decrease in the size (that
is, the length) of the lateral fields in the Osteostraci. The length is
denoted by G (fig. 20) ; in Tremataspis and Oeselaspis the sum of the
lengths of the two fields is used. Since we are concerned in this case
not with the absolute length, but with the relative length, a suitable
base measurement has been sought with which G can be compared
in the form of a ratio. The distance from the pineal opening to the
level of the apertures of the endolymphatic ducts (A, fig. 20) is the
only measurement of this sort that has been discovered. Determined,
as it is, by brain structures, it is presumably stable, but unfortunately
is not available in a large number of species.

The ratio, G/A, is plotted against an arbitrary time scale in
figure 21. This demonstrates the relatively small size of the lateral
fields in all of the Ludlow species (G/A less than 2.5). On the other
hand, the Early Devonian forms have relatively large lateral fields
(G/A greater than 2.5). The Downtonian fauna, though inade-
quately represented, is of an intermediate nature. Clearly the
geological evidence favors the view that small lateral fields were
primitive and that there was a trend towards an increase in size
through the Downtonian and Early Devonian. On purely theoretical
grounds this might be expected. The lateral (and dorsal) fields are
peculiar to the Osteostraci, and thus must have developed within
the group as a specialization, perhaps as electric organs for defense.
Such specialized structures would first appear in the course of evolu-

DENISON: OSTEOSTRACI

163

©

*®

©

©

©

©

©

®

9

©Xa) ®@

1.0 1.5 2.0 2.5 3 0 3.5

G/a = Relative length of lateral fields

Fig. 21. Relative length of lateral fields (G/A) plotted against an arbitrary
time scale. The encircled figures refer to different species plotted here and in
figures 24-26 as follows: 1, Tremataspis mammillata; 2, T. schmidti; 3, Oeselaspis
pustulata; 4, Dartmuthia gemmifera; 5, Saaremaaspis mickwitzi; 6, Didymaspis
grindrodi; 7, Witaaspis schrenkii; 8, Hemicyclaspis murchisoni; 9, Thyesies verru-
cosus; 10, T. egertoni; 11, T. salteri; 12, Procephalaspis oeselensis; 13, Cephalaspis
kozlowskii; 14, C. lankesteri; 15, C. salweyi; 16, C. heintzi; 17, C. whitbachensis;
18, C. ftoefa'; 19, C. arcticus; 20, C whitei; 21, C. ton^i; 22, Securiaspis kitchini;
23, Benneviaspis lankesteri; 24, J5. holtedahli; 25, B. anglica; 26, Hoelaspis angulata;
27, Boreaspis rostrata; 28, Kiaeraspis auchenaspidoides.

tion as rudiments, and these are most closely approximated among
known forms by the two pairs of lateral fields of Tremataspis and
Oeselaspis. The single lateral fields of other Osteostraci are pre-
sumably the result of enlargement and fusion of such rudiments.
The maximum development is shown by Cephalaspis; the observed
range of G/A in this genus, including several species not plotted
because of uncertainty regarding their age, is 3.8 to 5.3. The
determination of the status of the species with subdivided fields from
Spitsbergen mentioned by Wangsjo (1946, p. 360) must await its
description, but secondary subdivision may be involved.

No separate study of the dorsal field is presented, since it shows
;he same tendencies as the lateral fields. It is always single, but is
;mallest in the Late Silurian Tremataspis and Oeselaspis and largest
n the Early Devonian Cephalaspis.

Fig. 22. Cranial nerves and blood vessels of Ludlow Osteostraci. A, Tre-
mataspis mammillata, dorsal (Xl.5); B, Oeselaspis pustulata, dorsal on right,
ventral on left (X2.0); C, Dartmuthia gemmifera, dorsal (Xl.l); D, Thyestes
verrucosus, dorsal (X2.2). alf, anterior lateral field; asc, anterior semicircular
canal; b, boundary of endoskeletal component; dfn, dorsal field nerve; If, lateral
field; lfnl-6, nerves of lateral fields; med, medulla oblongata; plf, posterior lateral
field; psc, posterior semicircular canal; vcl, vena capitis lateralis; veil, preorbital
part of vena capitis lateralis; vlsl-6, dorso-lateral superficial veins; VI, N. pro-
fundus; V2, N. trigeminus; VII, N. facialis; IX+X, root of N. glossopharyngeus
and N. vagus.

164

DENISON: OSTEOSTRACI

165

THE NERVES OF THE LATERAL FIELDS

The arrangement of the nerves supplying the lateral fields and
their relationship to other cranial nerves have been used by Stensio
(1932, p. 75) and accepted by Westoll (1945, p. 351) as primary
bases for distinguishing the families and subfamilies of Osteostraci.
Thus the Cephalaspinae were defined in large part by the fact that
the first two nerves of the lateral fields divide close to the orbit,
and N. trigeminus mandibularis lies between them. On the other
hand, the Kiaeraspinae were distinguished by having the first two
nerves of the lateral fields united most or all of the way, and lying
behind N. trigeminus mandibularis. Stensio (1927, p. 241) considered
that the nerves of the lateral fields were primarily metameric in
disposition, as was suggested to him by their alternation with the

Fig. 23. Cranial nerves and blood vessels of Ludlow Osteostraci, dorsal view
on left, ventral on right. A, Procephalaspis oeselensis (X2); B, Witaaspis schrenkii
(X3). For abbreviations, see figure 22.

166 FIELDIANA: GEOLOGY, VOLUME 11

branchial nerves in Cephalaspis; other arrangements, such as that
found in Tremataspis, he considered as secondary, due to fusion of
nerves.

Caution in the use of nerves of the lateral fields in classification
is demanded by the fact that their disposition may vary within a
species, or even on the two sides of one individual. Thus Heintz
(1939, fig. 27) demonstrated considerable differences within individ-
uals of Micraspis gracilis. The same variation is shown in Securiaspis
kitchini (Stensio, 1932, p. 39, footnote). Among the species from
Oesel, variations have been noted in different individuals of Oeselaspis
pustulata, Witaaspis schrenkii, and Procephalaspis oeselensis. Figures
22 and 23 show the disposition of the cranial nerves of the Oesel
Osteostraci as far as has been determined by dissections. It should
be noted that Tremataspis has only three lateral field nerves, as
recognized by Wangsjo (1946, p. 360), not "at least four" as stated
by Stensio (1927, p. 305), and Robertson (1938a, pp. 201-202).

As an approach to the problem of the evolution of the nerves of
the lateral fields, the number of nerves has been counted in each
species where they are known, and the number plotted against an
arbitrary time scale, just as in the analysis of the lateral fields. A
nerve that divides half way between the brain and the lateral field
is counted as one and a half. By this system, Cephalaspis has 5.4
to 5.7 nerves. Where the number varies, the median value is taken.

The evidence from the temporal distribution of the Osteostraci,
as represented graphically in figure 24, supports the view that there
was a trend towards an increase in the number of nerves supplying
the lateral fields. Only one of the Ludlow species has more than
five nerves (that is, shows any subdivision of the first two nerves),
while this is true of all the Early Devonian representatives, with the
exception, perhaps, of certain individuals of Kiaeraspis. The Down-
tonian fauna again appears to be intermediate. Stensio's theory
that the number of nerves was reduced by fusion is plainly contrary
to the geological evidence.

There is certainly a high degree of correlation between the size
of the fields and the number of nerves supplying them. Thus
Tremataspis, which was shown above to have the smallest lateral
fields, has only three pairs of nerves; the evidence indicates that this
is the most primitive state among the known Osteostraci. On the
other hand, Cephalaspis, with the largest fields, has 5.4 to 5.7 nerves,
exhibiting as advanced a subdivision as is known in the group.
Genera intermediate in this respect — including the Benneviaspine

@(J8

©

®

©

©

©

O

®©@

®

©

3.0 3.4 3.8 4.2 4.6 5.0

Number of nerves supplying each lateral field

Fig. 24. Number of nerves supplying each lateral field plotted against an
arbitrary time scale. For a key to the species represented by the encircled numbers,
see figure 21.

© /-* ©

©

© ©

©

©

©

©

-i 1 1 1-

1.0 1.2 1.4 1.6 1.8 20

B/A = Relative length of prepineal shield

Fig. 25. Relative length of the prepineal shield (B/A) plotted against an
irbitrary time scale. For a key to the species represented by the encircled numbers,
ee figure 21.

167

— . *- ■ ■

168 FIELDIANA: GEOLOGY, VOLUME 11

forms, also Kiaeraspis, Thyestes, Didymaspis and Sclerodus — were
grouped together by Stensio in the Kiaeraspinae. As is shown
below, this is an artificial assemblage, since it includes otherwise
divergent forms that have arrived at a similar stage in the enlarge-
ment of the lateral fields and in the subdivision of their nerves.

If the condition found in Tremataspis is primitive, the basic
metamerism of the nerves, which Stensio believed was exemplified
by Cephalaspis, is questionable. To be sure, it is probable that the
nerves of the lateral fields were derived from branches of metameric
cranial nerves. They are closely associated with the roots of the
facial and glossopharyngeal nerves, and may also be related to the
trigeminal and vagus. Their only claim to metamerism is in probable
derivation from some of these nerves. As the lateral fields enlarged
in the evolution of the Osteostraci, the primitive three nerves of
Tremataspis subdivided and spread out, culminating in the Ceph-
alaspis condition, where they came to alternate more or less with
other cranial nerves. But this is a secondary condition, not a primary
metameric arrangement.

THE PREPINEAL SHIELD

As is illustrated in figure 25, there is clear evidence that the
prepineal shield was lengthened in the course of evolution of the
Osteostraci. This is measured by the ratio of the prepineal length,
B, to the base measurement, A (see fig. 20). B/A is less than 1.5
in all but one of the Lower Ludlow Osteostraci, and greater than 1.5
in all of the Early Devonian species; the Downtonian forms are
intermediate. As is the case with the other characters discussed
above, Tremataspis is to be considered most primitive in having the
shortest prepineal shield, while Cephalaspis is most advanced in
possessing the longest. Oeselaspis and Didymaspis, which in other
respects are very primitive, show a precocious lengthening of the
prepineal shield.

The meaning of the enlargement of the anterior part of the
shield is not at once obvious. It would, of course, give added space
for the expansion of the lateral fields, but it is improbable that there
is any relation, since the available space is never completely occupied.
It is probably related to the enlargement of the oralo-branchial
chamber. As Romer (1946, p. 43) has stressed, the Osteostraci were
"food strainers" in which the oralo-branchial chamber has more to
do with food intake than with respiration. The living "food
strainers," the whalebone whales and the whale shark, have a much

DENISON: OSTEOSTRACI 169

enlarged mouth and pharynx that allow a greater intake of water
with its contained food particles. Thus it is logical to assume that
the enlargement of the prepineal shield was connected with a per-
fection of this type of food-getting mechanism in the Osteostraci.

EXTENT OF THE TRUNK CARAPACE

Of paramount importance in the study of osteostracan evolution
is the problem of whether the portion of the carapace covering the
trunk was reduced, as thought by Westoll (1945, pp. 349-350), or
lengthened by the incorporation of trunk scales, as Stensio believed
(1927, p. 30). For analysis, it has not been possible to measure the
length of the trunk carapace directly, since there is no way by which
the carapace can be separated into its cephalic and thoracic portions
externally. Used as a substitute measurement is the postpineal
length (C, fig. 20), which includes a portion of the cephalic shield,
but whose magnitude is dependent to a large degree on the amount
of thoracic shield included. The relative length of the postpineal
shield is given by the ratio, C/A.

Plotting this character against the time scale (fig. 26) does not
give as clear a picture of an evolutionary trend as is the case with
the characters discussed above. Analysis of the graph, however,
does bring out the following points: (1) With one exception, the
Early Devonian species have short postpineal shields, with C/A less
than 3.2 (the exception is Kiaeraspis, in which the trunk shield is
very long, with C/A estimated to be 4.5) ; (2) the earliest Osteostraci
from Oesel show a wide range, but Tremataspis, which is surely
primitive in other characters, has the longest shield (C/A greater
than 4.1). Thus the evidence favors the theory that there was a
trend towards the reduction of the trunk shield. This view is
supported by Westoll's point (1945, p. 349) that the short-shielded
forms were much more efficient swimmers. In the absence of any
convincing evidence supporting Stensio's contrary theory, it will be
assumed that a long trunk shield is a primitive character in the
Osteostraci, and that this has been reduced in most later genera by
subdivision into body scales.

In this connection an interesting point is to be observed in
Thyestes verrucosus. In the restoration made by Patten (1912,
fig. 235, A), transverse grooves indicate the presence of two thoracic
segments included in the shield posterior to the pectoral sinus. This
may be the most common condition in this species, but certain
specimens have a longer thoracic shield, including three, or even

170 FIELDIANA: GEOLOGY, VOLUME 11

four segments. Thus, within the species, C/A may range between
2.7 and 3.3. The same variability may occur in Thyestes egertoni.
A specimen figured by Stensio (1932, pi. LII, fig. 3) clearly shows
three segments and may contain a fourth in its thoracic shield. In
another specimen (ibid., pi. XL, fig. 2) the segments are not clear,
but there is not room for more than two or three. Stensio interprets
these as transverse rows of scales that have fused to form the thoracic
shield. In view of the evidence that the shield was reduced, the
apparent segmentation should be considered as a preliminary stage
of the subdivision into scales of the solid thoracic shield.

DEVELOPMENT OF EXOSKELETON AND ENDOSKELETON

As a result of his study of the Spitsbergen forms, Stensio (1927,
pp. 31, 34) concluded that the Osteostraci represent a degenerating
series with regard to the degree of ossification of both the exoskeleton
and endoskeleton. A trend toward reduction of ossification has
recently been demonstrated in other groups of fishes, but apparently
it is not clearly shown by the Osteostraci. In the first place, Stensio's
conclusions were based originally on a comparison of species from
the Red Bay and Wood Bay Series, whose ages are Late Downtonian
and Early Devonian, respectively. In other words, they are all
relatively late forms in the evolutionary history of the group.
Secondly, the Wood Bay Series includes forms with well-developed
exoskeletons and endoskeletons, Cephalaspis lata, C. isachseni, and
probably Boreaspis and Cephalaspis brevicornis. Third and finally,
a good proportion of the earliest species from the Ludlow of Oesel
have a much reduced skeleton.

Considering first the exoskeleton, the species may be classified
in four categories. (The information is derived from Stensio [1927,
1932] and from study of the Oesel Osteostraci.)

Group 1. Exoskeleton unreduced; superficial layer complete;
sensory canals opening to surface by pores.

Group 2. Superficial layer slightly reduced; sensory canals con-
tinuously open to surface, resulting in external division of shield
into polygonal areas.

Group 3. Superficial layer considerably reduced or absent, ex-
cept perhaps on tubercles; middle layer may be slightly reduced
between tubercles.

Group 4. Superficial layer absent, except perhaps on tubercles;
middle layer greatly reduced or absent, so that the sensory canals
are exterior to the exoskeleton.

DENISON: OSTEOSTRACI

171

c
o
'c
o

>
0)

O

>»

o
111

@ ®

©

c
o
°c
O
c
i
o

a

® ®

®

*

o

3
-1

@© © ©

©

® ®

©

1.0

C/A*

1.5 2.0 2.5 3.0

Relative length of postpineal shie

3.5

d

4.0

4.5

Fig. 26. Relative length of the postpineal shield (C/A) plotted against an
arbitrary time scale. For a key to the species represented by the encircled numbers,
see figure 21.

Group 2

Group 3

Group U

5

0

1

2

4

0

0

2

4

1

5

1

The species whose stratigraphic position and exoskeletal develop-
ment are known, are grouped in the above categories and by age
as follows:

Group 1

Wood Bay Series 2

Dittonian and Upper Red Bay Series. .15
Downtonian and Lower Red Bay Series . 5
Ludlow 4

This grouping gives little support to Stensio's theory; in fact, it
would be possible to make a case for strengthening of the exoskeleton.
For example, in the family Cephalaspidae the majority of the Early
Devonian species have a well-developed exoskeleton with unreduced
superficial layer, yet Thyestes and Procephalaspis, which are the
earliest and in other respects the most primitive members of the
family, have a greatly reduced exoskeleton. The same apparently
applies to the Ateleaspidae. The exoskeleton is most strongly de-
veloped in Hemicyclaspis murchisoni, and perhaps in Hemiteleaspis
heintzi, both relatively late members of the family. The earliest
ateleaspid, Witaaspis, and the contemporary and probably related
genus, Saaremaaspis, have much reduced exoskeletons. Genera
that are presumed to occupy an intermediate stratigraphic position—

172 FIELDIANA: GEOLOGY, VOLUME 11

Ateleaspis, Micraspis, and Aceraspis — appear to be intermediate in
exoskeletal development, judging from Heintz's descriptions (1939) ;
in all of them the network of the sensory canals is open to the surface
by grooves, at least on part of the shield, forming polygonal areas
or "tesserae." But this is not conclusive, since Hemicyclaspis
lightbodii, contemporaneous with H. murchisoni, has the exoskeleton
reduced, retaining the superficial layer only on tubercles.

On the other hand, Tremataspis might be used to support the
view that a well-developed exoskeleton is primitive within the
Osteostraci. This genus, which has been shown in other characters
to occupy a central and basal position as regards osteostracan evolu-
tion, has a continuous superficial layer pierced by pores of the
sensory canal network (Denison, 1947). These sensory canals have
a relatively simple structure when compared to the apparently
secondarily subdivided canal systems described by Stensio in many
Downtonian and Early Devonian species. The vascular canals of
the middle layer form an irregular network, differing from the more
highly organized "radiating canals" of Cephalaspis. All in all, the
exoskeletal structure is simple and probably primitive.

Whether a highly ossified or a reduced exoskeleton be considered
primitive, it becomes obvious that there is no clear and general
trend within the Osteostraci towards reduced or increased ossifica-
tion. The variable development of the exoskeleton within two
species, Cephalaspis pagei and C. powriei brevicornis (Stensio, 1932,
pp. 99, 111), suggests that it may have been subject to genetic
fluctuations throughout much of the evolutionary history of the
order. Study of the histology of various exoskeletons does not
support Westoll's suggestion (1945, p. 351) that the loss of the
superficial layer may be due to resorption or wear from burrowing
in the sand or to post-mortem abrasion.

It is not possible at present to make a satisfactory analysis of
the degree of endoskeletal ossification. Since we are not dealing
with surface structure, the determination of its presence and develop-
ment depends upon the availability of enough specimens so that
suitable preparations can be made. Usually this is not possible in
species known from only a very few specimens. Stensio's experience
with the Spitsbergen Osteostraci (1927) led him to believe that there
was a trend towards reduction of ossification. It is worth noting,
however, that one of the latest members of the order, Boreaspis,
from the Wood Bay series, has the most completely ossified endo-
skeleton of any of the Osteostraci ; in fact, it is the only known form
in which the actual body of the endoskeleton, rather than just its

DENISON: OSTEOSTRACI 173

surfaces, is ossified. At the other extreme, the Oesel fauna includes
three species in which the perichondria! ossification is thin: Saare-
maaspis mickwitzi, Witaaspis schrenkii, and Thyestes verrucosus. The
other Oesel Osteostraci have a strongly ossified endoskeleton; this
includes Tremataspis, Dartmuthia, Oeselaspis, and Procephalaspis.

In the present state of our knowledge, it seems hardly possible
to demonstrate any well-defined evolutionary trend with regard to
the ossification of either the exoskeleton or the endoskeleton.

CORNUA AND PECTORAL FINS

In the consideration of the evolution of the Osteostraci, one of
the most interesting and at the same time most difficult questions
is whether or not the ancestral members of the group possessed paired
fins. Stensio (1927, p. 302) believed that they did. According to
his theory, they have been reduced in such forms as Didymaspis
and lost in Tremataspis. This belief is hard to reconcile with the
relative stratigraphic occurrence of genera with and without pectoral
fins, and requires the postulation of a hypothetical ancestral form
that on the one hand could give rise to groups retaining the fins,
and on the other hand to those losing them. It is reasonably certain
that all the Early Devonian genera possessed pectoral fins. They
are known to be present in Cephalaspis, and are clearly implied by
the proven presence of canals for nerves and blood vessels in Kiaer-
aspis and Hoelaspis. Although they have not been demonstrated
in Benneviaspis, Boreaspis, and Securiaspis, the possession of pectoral
sinuses, and the close relationship of these genera to Cephalaspis
and Hoelaspis, makes it safe to infer their presence. Turning to
the earliest osteostracan fauna, it is important to note that paired
fins were absent in the majority of genera and species. Witaaspis,
Thyestes, and Procephalaspis may well have possessed them, although
their presence has not been demonstrated as yet. But the strongest
argument against Stensio's theory is their undoubted absence in
Tremataspis, Dartmuthia, and Saaremaaspis, and almost certain
absence in Oeselaspis, for these are the genera that have been shown
to be most primitive in other respects.

Stensio argues that Tremataspis possessed an endoskeletal
shoulder girdle component, but this has not been demonstrated.
There are, to be sure, in the postero-lateral cephalic region endo-
skeletal structures that are concerned with the enclosure of nerves
of the lateral fields, marginal blood vessels, etc., and the post-
branchial wall is closely associated with them. It is more probable

174 FIELDIANA: GEOLOGY, VOLUME 11

that these structures later came to support the pectoral fins in such
forms as Cephalaspis, than that they are primarily shoulder girdle
elements.

The evidence points to the independent acquisition of paired
fins within the order Osteostraci. From a general evolutionary
point of view this is an important point, since it means that they were
acquired more than once by the vertebrates. (As a matter of fact,
it is probable that paired fins were evolved independently at least
three or four times among the vertebrates — within the Anaspida,
within the Osteostraci, within the Placodermi, and perhaps inde-
pendently by the ancestors of modern fish groups.) To make the
evolutionary origin of fins more complicated, it appears that within
the Osteostraci, fins were evolved concurrently yet independently
in two different lines; the manner in which they evolved was in
general similar, but differed in details, and the overall phylogenetic
picture, to be discussed later, indicates that the phyla separated
prior to the acquisition of paired fins.

Admitting for the Osteostraci the primitive nature of Tremataspis
and the advanced position of Cephalaspis, the stages in the evolution
of pectoral fins are moderately well documented. The first step is
the development of a lateral fold that may properly be considered,
in part at least, as a "fin fold." No such fold is present in Tremataspis
mammillata. A transverse section through the exoskeleton of the
body in the thoracic region suggests that this part of the body
completely filled the carapace. Laterally the exoskeleton is un-
thickened and unconstricted, and there is no indication of any
structure that might be morphologically ancestral to a fin (fig. 27, A) .
In other species of Tremataspis a thickening of the lateral curve of
the carapace may be seen in transverse sections; it is pronounced
in both T. schmidti (fig. 27, C) and T. milleri (fig. 27, B), where it
may be discerned on superficial inspection of the shield. This
thickening is interpreted as the first indication of the development
of a lateral fin fold, and is correlated with a lateral constriction of
the thoracic and abdominal parts of the body. The posterior part
of the shield of Didymaspis shows a similar development of the lateral
rim, judging by the section figured by Stensio (1932, fig. 61, C).
Oeselaspis exhibits an interesting stage in this evolution. In the
posterior cephalic and anterior thoracic regions there is pronounced
lateral thickening of the exoskeleton, especially of its basal layer,
and it is probable that there was dorso-ventral compression to form
a real rim (fig. 27, E). More posteriorly, in the narrower part of
the exoskeletal shield (that is, in the region of the so-called "pectoral

Fig. 27. Transverse sections of the lateral margin of the shield in various
Ludlow Osteostraci. A, Tremataspis mammillaia (X20); B, T. milleri (X25);
C, T. schmidti, juvenile (X25); D, Dartmuthia gemmifera (X25); E, Oeselaspis
pustulata (X25); F, Thyestes verrucosus (X15). 6c, basal vascular cavity; bl,
basal layer; ds, dorsal shield; e, enamel; gr, groove by which sensory canal opens
to surface; I jo, lateral fold; Ir, lateral rim of lateral fold; m, margin of oralo-branchial
fenestra; ml, middle layer; p, pore by which sensory canal opens to surface; pf,
pectoral fin; sc, sensory canal; si, superficial layer; t, tubercle; vs, ventral shield.

175

176 FIELDIANA: GEOLOGY, VOLUME 11

sinus"), this rim is not developed, and it is presumed to have been
lost. In Dartmuthia the lateral fold is very strong and distinct
(fig. 27, D). In both Oeselaspis and Dartmuthia it may be considered
to consist of two elements. First is the more lateral part, or lateral
rim proper, which consists of the three exoskeletal layers separated
only by channels for nutrient vessels and nerves. Second is the more
medial part, which is distinguished from the rest of the body only
by the fact that it is somewhat compressed dorso-ventrally, yet is
still appreciably open to the mesodermal and endodermal tissues of
the body. It is suggested that the lateral rim was morphologically
ancestral to the cornua of the Cephalaspidae and to the thickened
lateral edge of the ateleaspid fin, while the more medial part was to
give rise to the fin proper in both groups. For the most part the
genera illustrating the stages described above form a morphological
but not a phylogenetic series.

For reasons to be entered into below, it is thought that Saarema-
aspis is close to the ancestry of the Ateleaspidae. Unfortunately,
this genus is rare and incompletely known, but, since it is not far
removed from Dartmuthia, it is probable that the two genera were
similar in the development of the lateral fold. In any case it is
likely that such a stage was passed through by the ancestors of those
Osteostraci that evolved paired pectoral fins. The evolution of the
paired fins within the Ateleaspidae has been discussed by Heintz
(1939, p. 101). Since the morphological stages in the evolution of
their fins do not correlate with the presumed evolution of other
characters, it is probable that the known genera do not represent
a phylogenetic series; nevertheless they do illustrate more clearly
than any other group of Osteostraci some of the details of fin evolu-
tion. The processes involved are:

(1) Further emphasis of the lateral fold, particularly of its more
medial part, which is to give rise to the fleshy fin. As the lateral
fold is compressed, the thoracic part of the body is laterally con-
stricted more and more. An advanced stage is well illustrated by
Heintz's sections of Aceraspis (fig. 28, C, D).

(2) Reduction and sometimes complete loss of the posterior part
of the lateral fold. An early stage in this change is shown by Oesel-
aspis (which is not related to the Ateleaspidae), and more advanced
reduction is shown by Ateleaspis, Aceraspis, and Micraspis. It is
probable that the posterior reduction has a functional importance,
since a pectoral fin of the length of the post-cephalic fold of Dart-
muthia or Saaremaaspis would be an awkward appendage. It is

DENISON: OSTEOSTRACI

B /

177

Fig. 28. Transverse sections of the lateral margin of the shield in some
early Osteostraci. A, Procephalaspis oeselensis (X30); B, Cornua of same (X20);
C and D, Aceraspis robusta (from Heintz, 1939). For explanation of abbrevia-
tions, see figure 27.

correlated in this family, but not in all Osteostraci, with the assump-
tion of mobility in the abdominal and posterior thoracic regions.

(3) Subdivision of the exoskeletal shield of the lateral fold into
scales to give mobility to the incipient fins. This process is correlated
with the segmental breakup of the carapace of the body proper into
scales, which gives mobility to the whole post-cephalic body.

(4) Separation of the "pectoral flaps" from the thorax so that
they become freely movable fins. Heintz has shown that Ateleaspis,
Aceraspis, Micraspis, and Hemicyclaspis, in that order, exemplify
stages in this separation (Heintz, 1939, pp. 101-102).

Nothing is known of the stages of fin evolution within the Cephal-
aspidae. Heintz (1939, p. 101) considered that the cephalaspid
cornua was derived by fusion and separation of the thickened lateral
rim of the ateleaspid fin, leaving a free and separate fin in the result-
ing pectoral sinus. This is unlikely since, as will be shown below,
the different phyletic lines represented by the two families may have
been distinct prior to the evolution of fins. The manner of fin
evolution within the Cephalaspinae was probably similar to that
described for the Ateleaspidae, with one exception: the lateral rim,
retaining a solid, undivided exoskeleton attached to the cephalic
shield, became the cornua (fig. 28, B), while the more medial part
of the lateral fold, attaining mobility and freedom from the thorax
and cornua, became the fin.

178 FIELDIANA: GEOLOGY, VOLUME 11

In the Cephalaspinae, the development of pectoral fins probably
went hand in hand with the reduction and subdivision of the thoracic
and abdominal shield. This is not so in Kiaeraspis. This genus
retains a long trunk shield. In section (Stensio, 1927, fig. 79, D)
this shield has broadly rounded lateral contours showing no trace
of a lateral fold. However, the thoracic and abdominal regions have
been constricted laterally, indicating that a lateral fold was developed
and then lost by the ancestors of Kiaeraspis. The anterior part of
it was modified as a pectoral fin, whose presence is indicated by
canals and foramina for the subclavian and brachial arteries. Kiae-
raspis lacks prominently projecting cornua, and it is not known
whether the lateral rim was lost, or whether it was retained as a
thickened margin on the fins.

It is unlikely that Oeselaspis possessed paired fins. Only the
posterior part of the lateral fold has been lost, and the position,
contour and serrate edge of the posterior boundary of the remaining
lateral fold all argue against the possession of pectoral fins.

Didymaspis, judging from Stensio's description (1932, p. 172,
fig. 61, C), possessed a small lateral fold. This is present along the
lateral edge of the shield except in the posterior part of the cephalic
region, where there is a constriction forming the so-called "pectoral
sinus." But the weak development of the lateral fold and the small
size of the sinus indicate that fins, if present, must have been small
and ineffectual organs.

Sclerodus, which occupies an isolated position among the Osteo-
straci, probably did not possess any paired fins. But the lateral
fold is exceedingly strongly developed, even in the head region,
where it is fenestrated. The exoskeleton of the trunk proper has been
reduced (or more probably subdivided into scales), but the lateral
fold retains its solid exoskeletal sheath in this region, remaining as
greatly elongate "cornua." It is probable that the "cornua" include
the whole lateral fold, not just its lateral rim, so for this reason they
are not exactly comparable with the cornua of cephalaspids.

CLASSIFICATION AND PHYLOGENY

It was not until the appearance of Stensio's important descrip-
tions of the Spitsbergen and British Osteostraci that the group was
well enough understood to allow any satisfactory analysis of the
interrelationships of the various genera. Thus the first important
classification appeared in those works (1927, 1932). Since that
time, Heintz (1939) has described certain Ateleaspidae (Hemi-

DENISON: OSTEOSTRACI 179

cyclaspinae) in detail, and discussed their classification in relation-
ship to the rest of the order. Robertson (1945) has studied the
Oesel representatives and has suggested changes and additions
based on his knowledge of those forms. Westoll (1945), in describing
a Scottish ateleaspid, has considered the relationships of the various
genera of the order. All of these later classifications have been
founded largely on that of Stensio, who used as primary characters
the following:

(1) The length of the trunk shield.

(2) The presence or absence of pectoral sinuses and fins.

(3) The subdivision and development of the lateral fields.

(4) The subdivision of the lateral field nerves, and their relation-
ship to other cranial nerves.

It has been shown above that all of these characters are involved
in general evolutionary trends that pervade the order. A classifica-
tion based almost entirely on such characters is apt to be a purely
horizontal one in the sense that it may well include forms belonging
to quite unrelated phyletic lines that have happened to arrive at
the same stage of evolutionary development in these respects. Such
is certainly the case with most of Stensio's (1932) groups. To take
a specific example, a subgroup of the family Kiaeraspidae, including
Thyestes, Didymaspis, and Sclerodus, was characterized by the rela-
tively weak development of the lateral and dorsal fields, the coales-
cence of the first two nerves of the lateral fields, and the position
of the trigeminus and facial nerves in front of the nerves of the
lateral fields. It is contended here that these three genera are
almost as distantly related as any Osteostraci can be, that the three
characters used to unite them are closely correlated, and that they
agree only in exemplifying a rather primitive stage in the evolution
of these traits.

An alternative classification was presented in outline form
above (p. 159). It is based primarily on such characters as the
manner of development of the pectoral fins, pectoral sinuses and
cornua, and on the proportions of the shield, as well as on other
traits that sometimes appear to characterize phyletic lines. It also
considers those features that show progressive evolution throughout
the Osteostraci, but with due regard for the fact that their rates
and modes of evolution may be characteristic of certain phyla. Such
characters include:

(1) The relative length of the lateral (and dorsal) fields.

180 FIELDIANA: GEOLOGY, VOLUME 11

(2) The number and arrangement of nerves to the lateral fields.

(3) The amount of reduction of the thoracic shield.

(4) The lengthening of the prepineal part of the shield.

(5) The development of pectoral sinuses and fins.

The grouping is largely phylogenetic or vertical. Since, however,
the Osteostraci converge as they are traced backwards into the Late
Silurian, it has been found impractical to hold entirely to a phylo-
genetic subdivision of the more primitive forms; they are therefore
grouped together in one horizontal or primitive family: the Tre-
mataspidae. The proposed classification is defined and discussed
below.

Order OSTEOSTRACI Lankester 1868

Family TREMATASPIDAE1 Woodward 1891

Originally named for Tremataspis Schmidt alone, this family is
here extended to include a number of primitive genera from the
Ludlow, as well as one from the Downtonian. It includes the basal
stock from which other Osteostraci were derived, yet within the
family may be recognized the beginnings of phyletic divergence in
a number of directions, some of which trend towards later families.
The classification of such an assemblage is a matter of convenience,
and depends on the bias of the individual investigator. Here it is
considered more practical to group them together in one family,
emphasizing the convergence towards a central ancestral type, and
at the same time to subdivide them into subfamilies, recognizing
their incipient phyletic divergence. The Tremataspidae, as used in
this way, may be characterized by the following primitive traits:

(1) Long, relatively unreduced trunk shield (C/A greater than
2.8).

(2) Relatively short lateral fields (G/A less than 2.2).

(3) Nerves of the lateral fields few (3 to 5 in number).

(4) Pectoral fins and sinuses absent, or at the most in an incipient
stage of development.

1 Fowler (1947, p. 4) referred Tremataspis to Stigmolepis, and the family
Tremataspidae to Stigmolepidae. He was not aware, apparently, that a recent
student of this group considered Tremataspis and Stigmolepis doubtfully cogeneric
(Robertson, 1947). And he could not have known that the difficult taxonomic
problem involving the genotype of Tremataspis had been referred to the Inter-
national Commission on Zoological Nomenclature. It was agreed at the 1948
meeting of this body in Paris to designate T. schmidti Rohon 1892 as the type
species of Tremataspis Schmidt 1866, making Fowler's use of Stigmolepis and
Stigmolepidae entirely unnecessary.

DENISON: OSTEOSTRACI 181

Subfamily TREMATASPINAE new rank
(=TREMATASPIDAE Woodward, 1891)

Represented only by Tremataspis Schmidt (fig. 29, A; Lower
Ludlow, Oesel), this subfamily includes forms considered to be the
most primitive known Osteostraci, which from a morphological point
of view could well be ancestral to all the other genera. It is charac-
terized by its primitive features, as follows:

(1) Trunk shield long, covering both the thoracic and abdominal
regions (C/A= 4.2-4.8).

(2) Two pairs of small lateral fields (G/A=1.2).1

(3) Only three lateral field nerves, two supplying the anterior
field and one the posterior field.

(4) No pectoral sinuses, pectoral fins, or cornua, and the lateral
fold absent or only moderately developed.

(5) Prepineal part of shield very short (B/A=l.l).

(6) Exoskeleton well developed with unreduced superficial layer,
simple sensory canal network, and unspecialized vascular canal
system.

Other characteristics, some or all of which may be primitive,
include a well-ossified endoskeleton, strong sclerotic ossifications,
large oralo-branchial plates, absence of a definite dorsal fin, and
relatively few and large caudal scutes.

Subfamily DARTMUTHIINAE new rank
(=DARTMUTHIIDAE Robertson 1935)

Dartmuthia Patten and Saaremaaspis Robertson (including
Rotsikilllaspis Robertson) form a group that is very primitive in
most respects. These two genera show tendencies suggesting that
they may be close to the basal stock from which both the Cephalas-
pidae and Ateleaspidae arose. The subfamily is characterized as
follows:

(1) Trunk shield only slightly reduced (C/A= 2.9-3.6).

(2) Lateral fields single, but still relatively short (G/A= 2.0-2.2).

(3) Nerves of lateral fields five in number (in Dartmuthia); the
first two nerves united all the way to the lateral fields. (Saaremaaspis
is not well known in this respect.)

1 G is the sum of the lengths of the two fields.

182 FIELDIANA: GEOLOGY, VOLUME 11

(4) No pectoral sinuses, pectoral fins, or cornua, but the lateral
fold strongly developed.

(5) Prepineal part of the shield relatively unexpanded (B/A= 1.2) .

(6) Superficial layer of exoskeleton may be reduced or absent.

In most of these respects the Dartmuthiinae are still very primi-
tive. However, they do show an advance over the Tremataspinae
in the following: the slight reduction of the trunk shield; the moderate
elongation of the single pair of lateral fields; the greater subdivision
of the nerves of the lateral fields; the strengthening of the lateral
fold; and exoskeletal modifications.

Saaremaaspis (Lower Ludlow, Oesel; fig. 29, C) exhibits tend-
encies suggesting that it may have been close to the ancestry of
the Ateleaspidae. Many of these are so slight that they are hardly
definable, but it does possess a moderately narrow shield (H/A=3.1),
the lateral fields are quite long for the Ludlow (G/A=2.2), and the
nerves of the lateral fields may be subdivided more than in Dart-
muthia. In all of these respects it approaches the early Ateleaspidae.

Dartmuthia (Lower Ludlow, Oesel; fig. 29, D) is considered to
be closer to the base of the cephalaspid stem. Its head shield is
slightly broader (H/A=3.2), its lateral fields are shorter (G/A=2.0)
and comparable in relative length to those of Thyestes, and large
tubercles are developed in longitudinal rows in the exoskeleton, a
feature of certain Cephalaspidae.

Subfamily OESELASPINAE new rank

(=OESELASPIDAE Robertson 1935)

Only Oeselaspis Robertson (Lower Ludlow, Oesel; fig. 29, B) is
included in this subfamily. In the following respects it is very
primitive :

(1) The trunk shield is long (C/A=3.9).

(2) There are two pairs of small lateral fields (G/A^l.3).1

(3) The nerves of the lateral fields are relatively few (4.0-4.6
in number).

On the other hand, Oeselaspis is definitely advanced beyond
Tremataspis in the following characters:

(4) A lateral fold is strongly developed anteriorly, and pre-
sumably has been lost posteriorly in the trunk region; this is probably
to be considered as an intermediate stage in the evolution of pectoral
fins.

1 G is the sum of the lengths of the two fields.

Fig. 29. New restorations of dorsal shields of Ludlow Tremataspidae. A,
Tremataspis mammillata (X2); B, Oeselaspis pustulata (X2); C, Saaremaaspis
mickwitzi (Xl.9); D, Dartmuthia gemmifera (Xl.2). alf, anterior lateral field;
co, cornua; de, external opening of endolymphatic duct; df, dorsal field; ifc, in-
fraorbital lateral line canal; 11, main lateral line canal; If, lateral field; nh, naso-
hypophysial opening; or, orbit; ps, pectoral sinus; pi, pineal opening; plf, posterior
lateral field.

183

184 FIELDIANA: GEOLOGY, VOLUME 11

(5) The prepineal shield is expanded (B/A=1.6).

(6) The superficial layer of the exoskeleton is lost except on the
large tubercles, and the middle layer is reduced.

(7) The cephalic shield is moderately broad (H/A=3.9).

It is possible that Oeselaspis was descended from a Tremataspis-
like form. Any relationship to any of the other genera of Osteostraci
is unlikely. An affinity with Didymaspis has been suggested, but
the very long trunk shield and the unique manner in which the lateral
fold is reduced in the latter makes this highly improbable. Because
of the wide gap between Oeselaspis and other Osteostraci, it is safer
at this time to consider that it diverged from the primitive stock
in early times and left no known descendants in later periods.

Subfamily DIDYMASPINAE new rank

(=DIDYMASPIDAE Berg 1940)

This is another monotypic subfamily, including only the genus
Didymaspis Lankester (Downtonian, England) . It retains the follow-
ing primitive characters:

(1) Very long trunk shield (C/A=5.0).

(2) Moderately short lateral fields (G/A=2.2).

(3) Relatively few nerves of lateral fields (4.5 in number).
It is definitely specialized as compared to Tremataspis in:

(1) The presence of small "pectoral sinuses" that may or may
not have contained rudimentary pectoral fins.

(2) Long prepineal shield (B/A=1.8).

(3) Loss of the superficial layer and reduction of the middle
layer of the exoskeleton.

(4) Moderately broad cephalic shield (H/A=4.5).

As far as its ancestry is concerned, Didymaspis might have been
derived from Tremataspis, but that is not certain, since the trunk
shield is even longer than in the latter genus. Oeselaspis has been
considered as a possible relative, but, as was stated above, the
manner in which the lateral fold is reduced is entirely different in
the two genera. Among other long-shielded forms, Dartmuthia and
Saaremaaspis are excluded from an ancestral position by being
more advanced in several respects.

It is probable that Didymaspis represents a sterile side branch
of the primitive tremataspid stock, not ancestral to any of the known
later genera. There is, however, a remote possibility that Boreaspis

DENISON: OSTEOSTRACI 185

was descended from Didymaspis. This is suggested by the inturning
of the posterior ends of the lateral fields and by the anterior position
of the pectoral sinuses, and is supported by a similarity in general
proportions. As will be shown in the discussion of the Bennevia-
spinae, Boreaspis differs in certain respects from other members
of that subfamily and might possibly represent a parallel develop-
ment from a non-cephalaspid stock. But because of the many
resemblances of Boreaspis to other Benneviaspinae, and because of
the large structural and temporal gap between Boreaspis and
Didymaspis, the former is retained provisionally in the Bennevia-
spinae, and the relationship of the two genera is considered as
insufficiently supported by the available evidence.

Family SCLERODONTIDAE Fowler 1947
(=SCLERODIDAE Berg 1940)

This is an aberrant family of Osteostraci, including only the
genus Sclerodus Agassiz (Downtonian, England).

This genus is characterized particularly by the very long, so-
called "cornua," which have been shown above to consist of the
entire lateral fold encased in exoskeleton. Between the "cornua"
the trunk shield has been reduced, and presumably subdivided into
scales. Pectoral fins were surely not developed. The wide lateral
fold is continued onto the head where it is fenestrated. The peculiar
nature of the exoskeleton has been emphasized by Stensio and
Wangsjo. It is considerably reduced superficially but not necessarily
much specialized in other respects and is probably derivable from
the primitive Tremataspis type.

In spite of its peculiar specializations, Sclerodus retains some
very primitive features, as follows:

(1) The lateral fields are very short.

(2) The nerves of the lateral fields are few; the first two nerves
are not subdivided, and it is possible that the posterior two may
be united for part of their course.

(3) The prepineal part of the shield is moderately short.
Stensio (1932, p. 176) and with less certainty Westoll (1945,

p. 352) considered that Sclerodus was related to Thyestes and Didym-
aspis. The only similarity between these genera is that they have
all progressed to about the same degree in the lengthening of the
lateral and dorsal fields, and in the subdivision of the nerves of the
lateral fields. Sclerodus is more probably an early and divergent

186 FIELDIANA: GEOLOGY, VOLUME 11

branch from the central Tremataspis stock, highly specialized in
the development of the lateral fold and in the reduction of the trunk
shield, yet retaining some very primitive characteristics.

Family ATELEASPIDAE Traquair 1899
(= HEMIC YCLASPINAE Heintz 1939)

A clearly defined and natural group recently discussed by Heintz
(1939) and Westoll (1945). It includes the following genera: Wita-
aspis Robertson (Lower Ludlow, Oesel); Ateleaspis Traquair (Lud-
low or Downtonian, Scotland); Aceraspis Kiaer (Ludlow or Down-
tonian, Norway); Micraspis Kiaer (Ludlow or Downtonian, Nor-
way) ; Hemiteleaspis Westoll (?Downtonian, Scotland) ; Hemicyclaspis
Lankester (Downtonian, England; ?Downtonian, Norway).

The Ateleaspidae are distinguished by the following characters:

(1) The cephalic shield is relatively narrow (H/A= 2.8-3.7).
Narrowness of the shield is a trait that clearly distinguishes the
Ateleaspidae from the Cephalaspidae.

(2) Pectoral fins are developed, but there are no cornua or
distinct pectoral sinuses; the cornua of the Cephalaspidae find their
homologues in the thickened marginal rim of the ateleaspid pectoral
fin.

(3) The thoracic shield is greatly shortened (C/A= 1.4-2.0).
Ateleaspids are more progressive even than later cephalaspids in
this respect.

(4) The lateral fields are only moderately long (G/A= 2.0-3.0).
They may extend far anteriorly, but because of the absence of
cornua and the shortness of the shield, they cannot extend far
posteriorly.

(5) The number and disposition of the nerves of the lateral fields
vary because of evolution within the group; the first two nerves are
always subdivided to some extent, except in Witaaspis.

(6) The prepineal shield is not much expanded (B/A= 1.3-1.6);
this is less than in most Cephalaspidae.

(7) The posterior contour of the cephalic shield is emarginate
medially (except in Witaaspis) instead of having a posteriorly pro-
jecting spine as in the Cephalaspidae.

(8) There are usually strong sclerotic ossifications.

Witaaspis was referred by Robertson (1939a, p. 652) to the
Cephalaspidae, but as restored here (fig. 30, B) it agrees closely

DENISON: OSTEOSTRACI
„ifc

187

Fig. 30. New restorations of dorsal shields of Ludlow Cephalaspidae and
Ateleaspidae. A, Thyestes verrucosus (X2.3); B, Witaaspis schrenkii (X2.3); C,
Procephalaspis oeselensis (Xl.5). For abbreviations, see figure 29.

with the Ateleaspidae. Pectoral fins have not been preserved, but
the configuration of the postero-lateral corner of the cephalic shield
agrees with that of other members of this family. As would be
expected, since this is the earliest known genus assignable to the
family, it is more primitive in some respects than the Scottish and
Norwegian genera. The lateral fields are relatively short (G/A= 2.0),
the first two nerves of the lateral fields are completely united, the
third and fourth nerves are partly fused, and the thoracic shield is
longer (C/A=2.0).

The evolution of the Ateleaspidae has been discussed at some
length by Heintz (1939), who demonstrates a convincing series of
stages in the development of the pectoral fins in various genera.
Unfortunately, the evolution of other characters does not coincide
with that exhibited by the fins. The obvious conclusion is that the
Ateleaspidae do not represent a monophyletic group, but, instead,
several branches in which various characteristics were differently
developed. At the present time any study of the evolution of this
family is impractical, since the relative stratigraphic position of
most of the known forms is highly controversial. It is quite certain

188 FIELDIANA: GEOLOGY, VOLUME 11

that Witaaspis is the oldest known form. It is beyond doubt that
Hemicyclaspis kiaeri is younger than Aceraspis and Micraspis. But
correlation of deposits containing these genera with the geologic
sections in which Ateleaspis, Hemiteleaspis, and Hemicyclaspis
murchisoni occur is not yet possible.

Family CEPHALASPIDAE1 Huxley 1861

(=CEPHALASPIDES Agassiz 1843)

This family forms a well-defined and apparently natural group
when restricted to the two subfamilies, Cephalaspinae and Benne-
viaspinae. Among the Osteostraci it is clearly the most successful
family, judging by the number of genera, species, and individuals
that have been discovered; moreover, it is the only family known
to have survived the Early Devonian, for a few species referred to
Cephalaspis have been recorded from the Middle and Late Devonian.
It is characterized primarily by:

(1) The presence of cornua and distinct pectoral sinuses.

(2) Pectoral fins developed, but lacking the thickened lateral
rim of the ateleaspid fin, this being represented by the cornua.

(3) Shield broader than in the Ateleaspidae (H/A= 3.7-7.3).

(4) Postero-median spine on the dorsal shield (with one or two
exceptions) .

The Cephalaspidae may be considered as progressive in the
following tendencies:

(1) Elongation of the prepineal shield.

(2) Reduction of the thoracic shield.

(3) Elongation of the lateral and dorsal fields.

Subfamily CEPHALASPINAE Stensio 1932

The following genera are included: Thyestes Eichwald (Lower
Ludlow, Oesel; Downtonian, England); Procephalaspis, gen. nov.
(Lower Ludlow, Oesel); Cephalaspis Agassiz (Dittonian, England;
Lower and Middle Old Red Sandstone, Scotland; Red Bay Series

1 Fowler (1947, p. 4) has felt called upon to emend many of the names of
ostracoderms. Thus the name Pteraspidae is altered to Pteraspididae, apparently
with some orthographical justification. Why he did not feel that this correction
should apply to Cephalaspidae, Atelaspidae (sic), and Kiaeraspidae is not ex-
plained when he does apply it to the subfamily name Cephalaspidinae. In view
of the inconsistency with which the emender uses this correction, and the awk-
wardness which would result from its use, it has not been followed in the present
work.

DENISON: OSTEOSTRACI 189

and Wood Bay Series, Spitsbergen; Czortkower Stage, Podolia;
Upper Siegen and Lower Koblenz Beds, Germany; Early Devonian,
New Brunswick and Wyoming; Middle and Late Devonian, Quebec).
Cephalaspis, which does not appear until the late Downtonian,
is typical of the subfamily, while Procephalaspis and Thyestes are
early and primitive representatives. The subfamily may be denned
as follows:

(1) Thoracic shield typically less shortened than in the Benne-
viaspinae (C/A= 2.4-3.4; 1.9 in Procephalaspis).

(2) Pectoral sinuses shallower than in the Benneviaspinae (F/D=
0.1-0.5); the shallowness is due largely to their more posterior
position (E/A= 1.2-1.9) .

(3) Shield generally narrower than in the Benneviaspinae
(Cephalaspis hoegi, C. lata, C. brevicornis and C. laticornis have very
broad shields, but they are incompletely known and of uncertain
systematic position).

(4) Lateral fields with simple posterior termination.

The later Cephalaspinae are more progressive than the con-
temporaneous Benneviaspinae in the following respects:

(1) The lateral fields are greatly elongate (in Cephalaspis G/A=
3.8-5.3).

(2) The nerves of the lateral fields are highly subdivided (5.4-
5.7 in number).

(3) The prepineal shield is much expanded (B/A= 2.2-2.5).

"Cephalaspis" oeselensis Robertson (1939b), from the Ludlow,
resembles later species of Cephalaspis in general form and charac-
teristics, but differs in a number of traits which are to be considered
as primitive. These include the short dorsal and lateral fields (G/A=
2.4), undivided or only slightly divided anterior nerves of the lateral
fields (number of nerves= 5.0-5.2), and relatively short prepineal
shield (B/A=1.5). Other features, which may debar it from actual
ancestry of Cephalaspis, are the specialized exoskeletal structure,
the greatly reduced thoracic shield (C/A= 1.9), and the more anterior
position of the pectoral sinus (E/A=1.2; in other cephalaspids,
E/A= 1.3-1.9). It is necessary to distinguish this form as a distinct
genus for which the name Procephalaspis is proposed (fig. 30, C).

Thyestes (fig. 30, A) is the most primitive of the Cephalaspidae
and may represent the stock from which both the later Cephalaspinae
and Benneviaspinae arose. The thoracic shield has been reduced
somewhat, but it is still longer than in most of the later members of

190 FIELDIANA: GEOLOGY, VOLUME 11

the family, and contains a variable number of segments (C/A=
2.5-3.4). The lateral fields are short (G/A= 1.9-2.0) and the
nerves of the lateral fields are relatively undivided. The prepineal
part of the shield is short (B/A=1.3). It is also probably primitive
in having undivided lateral body scales, and in lacking any indication
of a dorsal fin, even a median series of dorsal ridge scales. The
greatly reduced exoskeleton, specialized in its large tubercles, may
indicate that the known species of Thyestes are not ancestral to most
later cephalaspids.

Subfamily BENNEVIASPINAE, subfam. nov.

With the exception of Stensiopelta, all of the genera referred to
this subfamily, as well as Kiaeraspis, Thyestes, Didymaspis, and
Sclerodus, were included by Stensio (1932, p. 151) in his subfamily
Kiaeraspinae. In this paper, Kiaeraspis has been referred to a family
of its own, the Kiaeraspidae, while Thyestes, Didymaspis, and
Sclerodus have been referred to the Cephalaspinae, Tremataspidae
and Sclerodontidae, respectively. The following genera are included
in the Benneviaspinae : Securiaspis Stensio (Dittonian, England;
Upper Red Bay Series, Spitsbergen); Benneviaspis Stensio (Ditton-
ian, England; Middle and Upper Red Bay Series, Spitsbergen);
Hoelaspis Stensio (Upper Red Bay Series, Spitsbergen); Boreaspis
Stensio (Lower Wood Bay Series, Spitsbergen); Stensiopelta, gen.
nov. (?Dittonian, England).

They are definitely Cephalaspidae in possessing cornua and pre-
sumably the cephalaspid type of fin. Their close relationship to the
Cephalaspinae is shown by intermediate forms, namely, Securiaspis
and Stensiopelta, and it is probable that they were derived from the
Cephalaspinae at a relatively late date. The Benneviaspinae are
characterized by the following features, which distinguish them from
the Cephalaspinae:

(1) Thoracic shield greatly reduced, more so than in the Cepha-
laspinae (C/A= 1.7-2.3; 2.8 in Stensiopelta).

(2) Pectoral sinus deeper than in the Cephalaspinae (F/D=
0.6-0.8) ; this is due in large part to the greater anterior invagination
of the sinus (E/A= 0.4-0.8).

(3) Shield generally broad, with a decided tendency towards a
lateral flare in the postero-lateral region.

(4) The lateral fields usually show a two-pointed posterior
termination, one point extending onto the cornua, and one postero-
mesially. Exceptions to this are Stensiopelta and Securiaspis, al-

DENISON: OSTEOSTRACI 191

though the latter may show an incipient development of this
character.

The Benneviaspinae were less progressive than contemporary
Cephalaspinae in the following respects:

(1) The lateral fields only moderately lengthened (G/A=2.6-
4.0).

(2) Nerves of the lateral fields less subdivided than in the
Cephalaspinae (5.1-5.3 in number).

(3) The prepineal part of the shield, excluding the rostrum,
generally less expanded (B/A= 1.5-2.3).

Securiaspis is clearly intermediate morphologically between the
Benneviaspinae and Cephalaspis, which it resembles in the general
shape and proportions of the shield, and in the absence of any
clearly indicated branching of the lateral fields posteriorly. It
differs from Cephalaspis, however, and is shown to be related to
the Benneviaspinae, by the greatly reduced thoracic shield (C/A=
1.8), relatively unexpanded prepineal shield (B/A=1.8), only moder-
ately lengthened lateral fields (G/A=2.6), and deep pectoral sinus
(F/D=0.6) extending far anteriorly (E/A=0.6). On the basis of
these characters, "Ceplialaspis" staxrudi Stensio (1927, p. 272, text
fig. 68) certainly belongs to Securiaspis.

"Cephalaspis" woodwardi Stensio (1932, p. 140, text fig. 50)
differs markedly from other members of that genus. Its lateral
fields extend posteriorly far onto the cornua without any indication
of a medially directed lobe, as is typical in Cephalaspis, and it agrees
with the latter in the moderately long thoracic shield (C/A=2.8).
But in other characters it clearly resembles the Benneviaspinae:
The prepineal shield is relatively short (B/A=1.8) and the lateral
fields are not greatly lengthened (G/A=3.4). Very characteristic
of the Benneviaspinae are the deep pectoral sinuses (F/D=0.6) ex-
tending far anteriorly (E/A=0.6), and the broad shield with a
pronounced postero-lateral flare. These features indicate that
"Cephalaspis" woodwardi should be separated from other species of
Cephalaspis; it is here referred to Stensiopelta, gen. nov. and is
referred to the Benneviaspinae. It is considered to be a relatively
unspecialized member of the subfamily which has paralleled the
Cephalaspinae in some respects.

Benneviaspis and Hoelaspis are highly specialized and divergent
members of the subfamily, characterized particularly by the greatly
broadened shield.

192 FIELDIANA: GEOLOGY, VOLUME 11

Boreaspis differs in certain respects from other Benneviaspinae.
In spite of its late age, the thoracic shield is less reduced than in
other members of the subfamily (C/A=2.3) and the shield is rela-
tively narrow. The highly ossified endoskeleton is a distinctive
feature that may be a specialization. The remote possibility has
been suggested above that Boreaspis was descended from Didymaspis,
but this is not demonstrable at present.

It is unlikely that the ancestry of the Benneviaspinae is to be
sought within the genus Cephalaspis in spite of the annectent forms,
Securiaspis and Stensiopelta. The few known Downtonian species
of Cephalaspis were already more advanced in the lengthening of
the lateral fields, subdivision of their nerves, and in the expansion
of the prepineal part of the shield. On the other hand, there is
nothing (excepting perhaps exoskeletal structure) to debar Thyestes
from the ancestral position, and it is possible that this genus gave
rise to both Cephalaspis and the Benneviaspinae.

Family KIAERASPIDAE Heintz 1939

As used here, this group is restricted to the genus Kiaeraspis,
and is raised to family rank. Kiaeraspinae was originally used by
Stensio (1932) to include not only those genera here referred to the
Benneviaspinae, but also Thyestes, Didymaspis, and Sclerodus. The
classification of Kiaeraspis with these forms was based largely on a
similarity in the degree of enlargement of the lateral and dorsal
fields and on the position of the first two nerves of the lateral fields,
as well as on their lack of subdivision. In an earlier part of this
study it has been shown that these are characters subject to general
evolutionary trends within the Osteostraci, so by themselves they
are of little value in demonstrating relationships. In other respects
Kiaeraspis occupies a rather isolated position.

Among the Osteostraci, only the Cephalaspidae can be considered
as possible relatives, but the resemblances to any known members
of that family are not particularly close. The form of the trunk
segments and the posterior outline of the trunk shield, as well as
the shape of the pectoral sinus and the postero-lateral corner of the
cephalic shield, are not far from the condition of Thyestes verrucosus.
On the other hand, Kiaeraspis differs widely from all Cephalaspidae
in the following important respects:

(1) The trunk shield is long and apparently unreduced in length.
This is to be considered as a primitive character that has been
retained into the Devonian by this genus alone. All of the Cephal-

DENISON: OSTEOSTRACI 193

aspidae, even the earliest representatives from Oesel, have greatly-
shortened shields.

(2) No distinct cornua are developed. It is perfectly possible
that they were present in the ancestors of Kiaeraspis and have been
reduced, but it is also possible that they were never developed and
that the lateral rim of the lateral fold was incorporated in the
pectoral fin as in the Ateleaspidae. Only the discovery of the fins
themselves can solve this problem.

(3) The peculiar inturning of the posterior ends of the lateral
fields is unusual, and can be compared only with the situation in
Didymaspis and the Benneviaspinae. In the former it is certainly
related to the anterior position of the "pectoral sinuses," which
have, in a sense, forced the posterior part of these fields to turn
inward. In the Benneviaspinae, it is correlated with the angulation
and restriction of the size of the cornua, which permitted further
posterior expansion of the lateral fields only in a medial direction.
In Kiaeraspis, however, the pectoral sinuses are posterior to the
lateral fields, and the space on the postero-lateral corners of the
cephalic shield is unoccupied; so the inturning of the fields is ap-
parently related to other factors.

These differences make any close relationship to the Cephal-
aspidae improbable. But the few resemblances, slight though they
are, suggest that Kiaeraspis may have been derived from the same
stock as that family. Phyletic separation must have taken place
earlier than the Thyestes evolutionary stage, where reduction of the
trunk shield was already well under way. Since the manner in
which the pectoral fins evolved in the Cephalaspidae is unknown,
two possibilities must be considered: (a) The common ancestor had
begun to develop pectoral fins from its lateral fold, but retained a
long trunk shield, in which case both Thyestes and Kiaeraspis could
be derived from this hypothetical form. (6) Pectoral fins evolved
concurrently with the reduction of the trunk shield in the Cephal-
aspidae, in which case the common ancestor of Thyestes and Kiaeraspis
must be sought in some such form as Dartmuthia, and Kiaeraspis
must have developed its fins independently.

The Kiaeraspidae, as the name is here used, may be briefly
defined by the following characters:

(1) Trunk shield unreduced in length.

(2) Pectoral fins present, but cornua absent or perhaps reduced.

(3) Lateral fields of moderate length, curving inwards posteriorly.

194

DENISON: OSTEOSTRACI 195

(4) First two nerves of lateral fields undivided, or perhaps very
slightly separated.

(5) Prepineal shield of moderate length.

SUMMARY

The conclusions reached in this paper concerning the evolution
and relationships of the Osteostraci are most concisely summarized
by the accompanying phylogenetic chart (fig. 31). This suggests —
especially during the early history of the group — many contempo-
raneous phyletic lines, each with its own particular specializations,
yet all sharing the general evolutionary trends of the order. The
first part of this work has been concerned with the demonstration of
these general trends in a few of the more obvious characteristics of
the Osteostraci, and, through this approach, the determination of
what characters may be considered as primitive within the order.
In the second part of this paper an attempt has been made to dis-
tinguish different phyletic lines and to base a classification upon
them; the phyla have been distinguished not only by qualitative
characteristics, but also by different rates of evolution in certain
features.

The phylogeny represented in figure 31 is admittedly tentative.
The Osteostraci from Oesel indicate that a wide evolutionary radia-
tion had already taken place by Lower Ludlow times. Many of
these early genera are clearly not ancestral to known later Osteo-
straci, while the suggested phyletic connections of some of the other
genera to subsequent groups is unfortunately based upon less com-
plete evidence than could be desired. It is hoped that future re-
search and discoveries will help to clarify the many uncertainties
that still remain in the evolution, classification and phylogeny of
the Osteostraci.

REFERENCES

Berg, L. S.

1940. Classification of fishes both recent and fossil. Trav. Inst. Zool., Acad.
Sci. URRS., 5, livr. 2, pp. 346-517 (English text), figs. 1-190.

Denison, R. H.

1947. The exoskeleton of Tremataspis. Amer. Jour. Sci., 245, pp. 337-365,
figs. 1-13, pis. I-III.

Fowler, H. W.

1947. New taxonomic names of fish-like vertebrates. Not. Nat., 187, pp. 1-16.

196 FIELDIANA: GEOLOGY, VOLUME 11

Heintz, A.

1939. Cephalaspida from Downtonian of Norway. Skr. Norske Videnskaps-
Akad., I, Mat. Nat. Kl., 1939, no. 5, pp. 1-119, text figs. 1-35, pis. I-XXX.

Patten, W.

1912. The evolution of the vertebrates and their kin. Blakiston's, Philadelphia.
xxi+486 pp., 309 figs.

Robertson, G. M.

1935. The ostracoderm Order Osteostraci. Science, 82, pp. 282-283.
1938a. The Tremataspidae. Amer. Jour. Sci., (5), 35, pp. 172-206; 273-296,

figs. 1-5, pis. I-III.
1938b. New genera of ostracoderms from the Upper Silurian of Oesel. Jour.

Pal., 12, pp. 486-493, figs. 1-3, pi. LX.
1939a. The status of Cephalaspis schrenckii Pander from the Upper Silurian

of Oesel. Jour. Geol., 47, pp. 649-657, figs. 1-6.
1939b. An Upper Silurian vertebrate horizon, with description of a new species,

Cephalaspis oeselensis. Trans. Kansas Acad. Sci., 42, pp. 357-363, 1 pi.

1940. Witaaspis patteni, a new ostracoderm from the Upper Silurian of Oesel.
Ibid., 43, pp. 297-298, 1 fig.

1945. Cephalaspids from the Upper Silurian of Oesel, with a discussion of
cephalaspid genera. Amer. Jour. Sci., 243, pp. 169-191.

1947. Proposed suspension of the Regies for Tremataspis Schmidt, 1866 (Class
Cephalaspidomorphi, Order Osteostraci). Bull. Zool. Nomencl., 1, pt. 10,
pp. 237-238.

Romer, A. S.

1946. The early evolution of fishes. Quart. Rev. Biol., 2t, pp. 33-69, figs. 1-31.

Stensio, E. A.

1927. The Downtonian and Devonian vertebrates of Spitsbergen. Part I.

Family Cephalaspidae. Skr. Svalbard Nordishavet, 12, pp. i-xii, 1-391,

figs. 1-102, pis. I-CII.
1932. The cephalaspids of Great Britain. British Museum (Natural History).

pp. i-xiv, 1-220, figs. 1-70, pis. I-LXVI.

Traquair, R. H.

1899. Report on fossil fishes collected by the Geological Survey of Scotland in
the Silurian rocks of the south of Scotland. Trans. Roy. Soc. Edinburgh,
39, pp. 827-864, figs. 1-6, pis. I-V.

Wangsjo, G.

1937. On a new species of Benneviaspis from the Red Bay Series in Spitzbergen.

Bull. Geol. Inst., Univ. Upsala, 27, pp. 209-211, figs. 1-2.
1946. On the Genus Dartmuthia Patten, with special reference to the minute

structure of the exoskeleton. Ibid., 31, pp. 349-362, figs. 1-5, pis. V-VII.

Westoll, T. S.

1945. A new cephalaspid fish from the Downtonian of Scotland, with notes on
the structure and classification of ostracoderms. Trans. Roy. Soc. Edinburgh,
61, pt. II, pp. 341-357, figs. 1-7, 1 pi.

Woodward, A. S.

1891. Catalogue of the fossil fishes in the British Museum (Natural History).
Part II. British Museum (Natural History), London, pp. i-xli, 1-567,
figs. 1-68, pis. I-XVI.

Zych, W.

1937. Cephalaspis kozlowskii, n. sp., from the Downtonian of Podole (Poland).
Arch. Towar. Nauk., Lwowie, sec. Ill, 9, no. 1, pp. 1-100, 4 pis.

THE EXOSKELETON OF EARLY
OSTEOSTRACI

ROBERT H. DENISON

Curator of Fossil Fishes

FIELDIANA: GEOLOGY

VOLUME 11, NUMBER 4

Published by

CHICAGO NATURAL HISTORY MUSEUM

JANUARY 19, 1951

PRINTED IN THE UNITED STATES OF AMERICA
BY CHICAGO NATURAL HISTORY MUSEUM PRESS

INTRODUCTION

This paper is primarily a discussion of the structure of some of
the earliest known vertebrates, the Osteostraci from the Lower
Ludlow of the Island of Oesel in the Baltic. Recent descriptions
of most of these forms have disregarded the microscopic structure
of the exoskeleton. That of Tremataspis is well known from the
researches of several authors, and the Dartmuthia exoskeleton is the
subject of a recent paper; but as far as the other Oesel Osteostraci
are concerned, only brief, inadequate, and often incorrect notes
have appeared.

The exoskeletal structure is often a diagnostic character of at
least the higher categories, and for this reason alone it is desirable
that descriptions be available. Added significance has been given
to the structure of the exoskeleton by the recent suggestion that
the heavy armor of the earliest vertebrates is to be considered a
primitive trait, and that subdivision, reduction, and even loss of
armor came later. With this in mind, I have made some comparisons
of the Ludlow Osteostraci with later genera, and have given general
consideration to the problem of the evolution of the exoskeleton with-
in the order.

Genus Tremataspis

Since the exoskeleton of Tremataspis has been described in detail
in earlier publications (Stensio, 1927; Gross, 1935; Denison, 1947),
I have given here only a brief account of its structure. This has
been included so that there will be a basis for comparison in the
discussion of other genera, and also to define the terminology here
used.

The surface of the exoskeleton of Tremataspis mammillata (fig.

52, A) is smooth except for occasional small dorsal tubercles, and

he pores and grooves of the sensory canal system and of the related

ateral lines. Externally there is a thin layer of enamel, underlain

by a much thicker layer of dentine-like tissue, the two forming the

. uperficial layer. The dentine is pervaded by tubules that arise

: rom a network of small vascular canals at the very top of the middle

lyer. This network, called the subepidermal vascular plexus by

199

200 FIELDIANA: GEOLOGY, VOLUME 11

Stensio, is divided in this species into polygons that are delimited
by the sensory canals deeper in the middle layer; within the polygons
the arrangement of the vascular canals is irregular. The subepi-
dermal vascular plexus is connected by vertical canals (external
branches of the ascending vascular canals) with a deeper vascular
network near the bottom of the middle layer. The canals of this
network have been called radiating canals in Cephalaspis by Stensio,
but since in Tremataspis they have no regular arrangement, the
name is not appropriate; Wangsjo (1946) calls them horizontal
submucous vascular canals. In Tremataspis this network of vascular
canals lacks a well-defined polygonal arrangement and has numerous
connections below the sensory canals. The thick and strongly
laminated basal layer contains large cavities (basal cavities of
Wangsjo, 1946) thought to have housed vascular sinuses and con-
nected by narrow passages (ascending vascular canals) with the
submucous vascular network of the middle layer. Small canals
also pass internally from the basal cavities, either emerging from
the inner surface of the exoskeleton, or passing into a subaponeurotic
vascular plexus in the endoskeleton, where this is present. Finally,
and definitely characteristic of the Osteostraci, is the very regular
polygonal network of canals in the middle layer, called "mucous
canals" by Stensio (1927), but more recently shown to be sensory
canals (Denison, 1947). In T. mammillata, these form relatively
large and simple polygons, opening to the exterior by pores, or by
open grooves where this system is specialized to form lateral line
canals.

Other species of Tremataspis show some modifications of this
simple structure. In both T. schmidti and T. milleri (Denison,
1947, text figs. 4-5) there is clear evidence of subdivision of the
polygons of the sensory canals by the development of smaller and
more superficial connecting canals across the primary polygons.
These secondary connecting canals are homologous with the intra-
areal canals of cephalaspids, while the larger primary canals are
equivalent to the circumareal canals (of Gross, 1935; equals interareal
canals of Stensio, 1927). There are differences in the vascular net-
works of the middle layer, but the canals lack any regular arrange-
ment; in fact, even their polygonal subdivision is not apparent in
T. schmidti.

Genus Dartmuthia

The exoskeleton of Dartmuthia was the subject of a recent paper
by Wangsjo (1946). In spite of the fact that he had very limited

DENISON: THE EXOSKELETON OF EARLY OSTEOSTRACI 201
A P J .1 7p d y

^^m^^^^^^^^^il&^^^^^m.

smp

be

svp ml

Fig. 32. Transverse sections of exoskeletons of Oesel Osteostraci. A, Trem-
itaspis mammillata (X35); B, Dartmuthia gemmifera, ventral shield (X45); C,
D. gemmifera, dorsal shield (X40). avc, ascending vascular canal; be, basal vascular
<:avity; bl, basal layer; d, dentine; dvc, descending vascular canal; e, enamel;
• avc, external branch of ascending vascular canal; gr, groove by which sensory
•anal opens to surface; ip, intertubercular process; ml, middle layer; p, pore by
vhich sensory canal opens to surface; re, radiating canal; sc, sensory canal; smp,
i ubmucous or lower vascular plexus of middle layer; sre, sinus in radiating canal
system; svp, subepidermal vascular plexus; t, tubercle; x, horizontal septum divid-
ing sensory canal.

material at his disposal, his description is for the most part correct,
although it is incomplete since it does not include the ventral shield.
However, his conclusion (1946, p. 359) that "Dartmuthia, as far as
is exoskeleton is concerned, on the whole agrees more closely with
tie Cephalaspids proper than with other Osteostraci" is not sup-
ported by the facts that he himself marshalls; it is hardly defensible
when the ventral shield is considered, since the latter is very similar
to the Tremataspis exoskeleton.

Excepting the marginal region, the ventral shield of Dartmuthia
( igs. 32, B; 33, A) is smooth, without any of the elevated tubercles

202 FIELDIANA: GEOLOGY, VOLUME 11

that characterize the dorsal surface. It is subdivided into polygonal
areas by the network of sensory canals which open to the surface
by grooves (fig. 33, A, gr), instead of opening by pores as in Trem-
ataspis. The grooves are extremely sinuous superficially, as if
they had been formed by the union of numerous adjacent pores
(fig. 34, B). The superficial layer is well developed, and consists
of a shiny enamel film on the surface, underlain by a layer of dentine-
like substance; it is similar to that of Tremataspis, although thinner.
In the middle layer, the system of sensory canals is, for the most
part, comparable to that of Tremataspis. Its polygons are relatively
large, and there is no indication of any secondary canals except near
the marginal area. Along the ventral surface of the marginal area,
there are elongated areas with scalloped edges, capped with enamel,
which correspond to the tubercles of the dorsal shield, except that
they are flatter (fig. 33, B, st) ; these are separated by areas in which
the polygons are small and lack enamel. The small size of the poly-
gons here suggests that the canals separating them are comparable
to intra-areal canals, while the canals separating the polygons of the
rest of the ventral shield are similar to those of Tremataspis mammil-
lata and to the circumareal canals of cephalaspids. It is possible
to recognize in some of the sections the thin horizontal septum, first
discovered in Tremataspis (Denison, 1947, p. 341), which divides the
sensory canals into an outer part in direct communication with the
exterior, and an inner part that presumably housed the sensory
endings, and into which the vascular canals open (fig. 32, B, x).
The vascular canals of the middle layer are much as in Tremataspis.
The subepidermal vascular plexus is divided into polygonal areas
by the grooves of the sensory canals, but the submucous plexus
extends freely below the sensory canals and shows no indication of
any regular arrangement as radiating canals. The ascending vascular
canals from the basal layer are concentrated near the center of the
polygons, but their external branches are numerous throughout the
whole polygonal area. The basal layer resembles that of Tremataspis;
the basal cavities are larger in the sections studied, but this may not
be characteristic of fully adult exoskeletons.

The dorsal shield (figs. 32, C; 34, A) is much modified, its most
obvious specialization being the presence of tubercles, some of which
are large and arranged in rows. The tops of the tubercles show thin
enamel and thick dentine layers, representing the superficial layer
of the exoskeleton. Just below are canals of the subepidermal
vascular plexus, connected by external branches of the ascending

/

\

/

\

/

\

svp

St

1
\

gr

mt

I
I
I
I
_ 1

Fig. 33. Dartmuthia gemmifera. A, Surface of ventral shield seen as a semi-
t-ansparency under liquid (X35); B, Ventral surface of lateral marginal area
( <30). gr, groove by which sensory canal opens to surface; ip, intertubercular
1 rocess; mt, marginal tubercles; st, small tubercle-like areas; svp, subepidermal
^ ascular plexus.

203

204 FIELDIANA: GEOLOGY, VOLUME 11

vascular canals with the submucous vascular plexus at the base of
the tubercles. The arrangement of the latter plexus is modified
from the primitive condition found in the ventral shield and in
Tremataspis, and approaches the cephalaspid pattern. In the center
of the tubercles an irregular plexus is formed by canals uniting the
ascending vascular canals from the basal cavities; but radially dis-
posed branches extend from this central plexus in all directions,
passing under the sensory canals to unite with radiating canals
from other tubercle areas (fig. 34, A, smp). Although it is obvious
there must have been a blood supply both to and from the exo-
skeleton, it does not seem possible to distinguish, as in cephalaspids,
any clearly differentiated canals representing the beginnings of the
venous system; the canal labeled "descending vascular canal" by
Wangsjo (1946, pi. VI, fig. 3) is not strictly comparable to the
canals so named in Cephalaspis.

Between the tubercles, and nearly enclosing the sensory canals
that surround the tubercles, arise processes with vertical columns
basally, and horizontal laminae superficially (fig. 32, C, ip). The
latter were named "intertubercular plates" by Wangsjo, although
they are not separate from the rest of the exoskeleton. The vascular
supply of the intertubercular processes is derived from branches of
the underlying radiating canals; there are ascending canals in the
vertical columns, from which branches spread out irregularly in
the more superficial horizontal lamellae. The latter are certainly
part of the subepidermal vascular plexus, which is significant, since
it shows that the processes consist largely of the middle exoskeletal
layer. Wangsjo (1946, p. 355) considered that the "intertubercular
plates" belonged to the superficial layer, but this is disproved by
the presence in them of the subepidermal plexus. It is probable
that a very thin layer coating the intertubercular processes represents
the superficial layer here, but I have not been able to identify any
structures that would characterize this coating as dentine; it is
certainly not enamel. In any case, it is clear that the superficial
layer has been largely lost between the tubercules.

The sensory canals ("mucous canals") have been described in
some detail by Wangsjo, who gave different names to canals sur-
rounding the tubercles and to those lying between tubercles. It
should be noted that the canals are in the central or lower part of
the middle layer, and that they are always open to the surface by
grooves (fig. 32, C, gr), in contradiction to statements by Wangsjo.
The lateral line canals are interstitial canals (as distinguished from

svp

smp

sc

I
1

•1

\

ml
I
1

\

\
I

\
avc

svp
i

Fig. 34. Dartmuthia gemmifera (X30). A, Obliquely tangential section of
dorsal shield; B, Obliquely tangential section of ventral shield, avc, ascending
vascular canals; be, basal vascular cavity; bl, basal layer; d, dentine; e, enamel;
gr, groove by which sensory canal opens to surface; ml, middle layer; sc, sensory
canal; smp, submucous or lower vascular plexus of middle layer; svp, subepidermal
vascular plexus and external branches of ascending vascular canals; t, tubercle.

205

206 FIELDIANA: GEOLOGY, VOLUME 11

canals called circumtubercular and cross commissural by Wangsjo)
that are specialized only in their linear arrangement. Thus in es-
sentials, the relationship of the lateral line system to the sensory
canal network is the same as in Tremataspis, where it has been
described more fully in an earlier work (Denison, 1947, pp. 350-
353).

The basal layer does not differ from that of the ventral shield
or from the basal layer of the Tremataspis exoskeleton. The rela-
tively large basal cavities in the figured sections (figs. 32, C; 34, A,
be), as in the case of the ventral shield, probably indicate that the
exoskeleton was not completely grown in these individuals, since in
other sections, not figured, the basal cavities are much smaller. That
there were horizontal connections between the basal cavities as
indicated by Wangsjo (1946, p. 351) is doubtful; in the section he
figures to demonstrate this (1946, pi. VI, fig. 2), these connections
are clearly due to breaks and displacement, not to natural perfora-
tions. Such breaks are invariably present in this relatively weak
part of the exoskeleton in the sections examined, but in no case was
there any clear evidence of a natural communication.

The structure of the ventral shield of Dartmuthia strongly
supports the view that this genus is closely related to Tremataspis.
The dorsal shield, although considerably specialized superficially, is
still unspecialized basally. It could have been derived from the
simple Tremataspis type by the emphasis of tubercles, by their
arrangement in rows, by the reduction of the superficial layer
between the tubercles, and by the partial acquisition of a radial
pattern in the submucous vascular plexus.

There is good evidence that the growth of the exoskeleton was
similar to that described in Tremataspis (Denison, 1947). One
individual that has been sectioned lacks the basal layer and has
the middle layer only partly developed, incompletely enclosing the
vascular canals. As I mentioned above, the large cavities in the
basal layer in the figured specimens would probably have filled in
partly, upon maturity.

Genus Oeselaspis

The structure of the exoskeleton of Oeselaspis has not been de-
scribed before, except for brief notes derived from observation of
the surface by Robertson (1935). The dorsal and ventral shields
are similar and exhibit a characteristic and somewhat specialized
condition (fig. 35, A).

DENISON: THE EXOSKELETON OF EARLY OSTEOSTRACI 207

Fig. 35. Transverse sections of dorsal exoskeletons of Oesel Osteostraci.
A, Oeselaspis pustulata (X40); B, Thyestes verrucosus (X35); C, Procephalaspis
oeselensis (X40). For abbreviations, see figure 32.

Large, widely scattered tubercles with broadly rounded tops are
the outstanding superficial feature. The outer part of the tubercles
is covered with a rather thick layer of enamel, penetrated by numer-
ous fine tubules perpendicular to the surface. Below is a layer of
dentine exhibiting tubules continuous with those of the enamel
layer. Sometimes these tubules look like ordinary cell lacunae of
the middle layer, except that they are elongated perpendicular to
:he surface. This, and the fact that the superficial and middle layers
ire not clearly differentiated, suggest that this tissue is intermediate
n character between true dentine and bone. Only in the tubercles
s the superficial layer preserved.

Vascular canals at the top of the middle layer in the tubercles
; re shown to belong to the subepidermal vascular plexus by the fact
i hat the dentinal tubules arise from them. Lateral branches descend-

208 FIELDIANA: GEOLOGY, VOLUME 11

ing more or less parallel with the surface of the tubercles connect
them with a network of canals just below the grooves of the sensory
canals, while external branches of the ascending vascular canals
unite the subepidermal vascular plexus with the vascular plexus near
the base of the middle layer. The latter, which is equivalent to the
submucous vascular plexus in Tremataspis, is specialized in Oesel-
aspis by the development of an irregular radiating pattern centered
under the tubercles.

In the intertubercular spaces, the superficial layer and the upper-
most part of the middle layer are absent. The surface consists
of open grooves that house the sensory canals (fig. 35, A, sc) and
that are separated by projections of the middle layer, giving the
"minutely spiculate" effect described by Robertson (1935, p. 457).
The middle layer is thinner here. Near its base are the radiating
canals, from which vascular canals extend toward the surface; many
of them pass the short distance to the base of the sensory canal
grooves, where they expand and are separated from the grooves
only by a thin bony partition (fig. 35, A, x). It is probable that
this partition is homologous with the thin, horizontal septum
dividing the sensory canals of Tremataspis and Dartmuthia, and
that the "vascular" expansion below the partition is to be identified
with the lower part of the sensory canals of those genera. There is
good indication in certain sections of Oeselaspis that this thin
septum is pierced by a few minute pores, perhaps for sensory endings.
A few of the vascular branches from the radiating canals appear to
extend into, or along the edge of, the bony projections between the
sensory canals; they probably lead to a part of the subepidermal
vascular plexus external to the exoskeleton.

The sensory canals form a fine-meshed network over the entire
shield, except on the top of the tubercles (fig. 37, C, sc), and are
housed in open grooves, since in these regions the superficial layer
and the top of the middle layer are absent. No large polygonal
sensory canal areas are recognizable, and the small size of the polygons
suggests that most of the canals are intra-areal or secondary, rather
than circumareal. The lateral lines are clearly a part of this sensory
canal system (fig. 37, C, ifc).

The basal layer is thick and strongly cross-laminated. It possesses
numerous basal cavities, widely distributed both below and between
the tubercles. Ascending vascular canals feed the vascular networks
of the middle layer and extend superficially, while basal canals lead
to the internal surface of the exoskeleton. The only indication of a

DENISON: THE EXOSKELETON OF EARLY OSTEOSTRACI 209

large polygonal pattern in the shield is given by the partitions be-
tween the basal cavities. Large polygons, corresponding to the
tubercle areas, are subdivided by the partitions between the basal
cavities within those areas. In certain cases the partitions appear
as ridges on the inner surface of the skeleton because the intervening
weaker parts of the basal layer have been crushed. Narrow vertical
canals in the basal layer of the intertubercular region are occasionally

Fig. 36. Superficial view of the
exoskeleton of Thyestes verrucosus
(X40), showing the tubercles (t), and
the submucous or radiating vascular
canals (smp), which are drawn as
transparencies.

-smp

— t

seen in sections. They extend from the radiating canals to the
internal surface of the exoskeleton, and may represent venous canals,
comparable to the descending vascular canals of Cephalaspis (fig.
35, A, dvc).

Genus Thyestes

A few notes on the minute structure of the shield of Thyestes
egertoni and T. salteri have been given by Stensio (1932). In those
species, as also in T. verrucosus to be described here, the exoskeleton
is specialized and greatly reduced.

The tubercles are the most striking feature of the dorsal shield
of Thyestes verrucosus (fig. 35, B; 36, t). Some of them are arranged
in longitudinal rows and are very large and wide-based, tapering to
i rather sharp tip. Between the rows of large tubercles, and some-
times on their lower slopes, are numerous smaller tubercles, often
rregularly arranged but similar in shape. Along the margin of the
;hield are two or three rows of moderate-sized tubercles, with
straight crests and very much flattened dorso-ventrally, so that they
ippear very slender and sharp in transverse sections. The only
namel that has been observed in this species is on the tips of these
narginal tubercles. The other tubercles are capped with dentine

210 FIELDIANA: GEOLOGY, VOLUME 11

of the usual type found in Osteostraci. Elsewhere on the shield
the superficial layer is completely lacking.

The middle layer is also greatly reduced. In the tubercles it is
represented by a thickness of non-laminated bone, pervaded by a
complicated network of vascular canals representing the subepidermal
vascular plexus, ascending canals, and submucous plexus or radiating
canals. The latter may be expanded into a sinus from which the
lower network of vascular canals extends in all directions toward
the periphery of the tubercle as an irregular network, with only a
slight suggestion of a radiating pattern (fig. 36, smp). Between the
tubercles, the middle layer may be completely missing on the lateral
part of the shield. More medially it is just thick enough to enclose
a stratum of vascular canals that open frequently to the surface.
These represent the submucous vascular plexus, a continuation of
that of the tubercles. The polygonal areas are marked only by
the pattern of these canals, each polygon containing a central
tubercle. The subepidermal vascular plexus (except in tubercles)
and the network of sensory canals is completely outside the exo-
skeleton. Lateral lines may be recognized superficially on the shield;
they are marked only by the linear arrangement of the tubercles on
either side, and by the faintly indicated edges of the polygons that
underlie the lateral lines.

The basal layer is well developed and very strongly laminated,
making up the bulk of the exoskeleton. It contains relatively few
small basal cavities, communicating with the interior of the shell
and with the submucous vascular plexus by means of ascending
vascular canals. Under the large tubercles, the basal cavities are
tremendously enlarged (fig. 35, B, be), so that the upper part of the
basal layer bends sharply upward and extends high into the tubercle.

Sections of the exoskeleton of Thyestes egertoni and T. salteri
have not been figured, but judging from Stensio's descriptions (1932,
pp. 167, 169) the structure is similar to that of T. verrucosus. The
greatly reduced exoskeleton, apparently characteristic of the genus,
is presumably a specialization and is surprising in so early a form,
which in other respects could very well have been ancestral to later
Cephalaspidae.

Genus Procephalaspis

Procephalaspis oeselensis was originally described as a species of
Cephalaspis (Robertson, 1939), but it differs from any described
member of the latter genus in its exoskeletal structure as well as in

DENISON: THE EXOSKELETON OF EARLY OSTEOSTRACI 211

other characters. Superficial reduction, affecting the superficial
layer and to some extent the middle layer, is considerable, though
less than in Thyestes. The most striking specialization is the peculiar
development of tubercles (fig. 35, C, t). Over the dorsal shield these
tubercles are generally tall, and often slender and columnar, with a
rounded top; on the lateral rim they are smaller, close set, elongated,
and usually expanded distally. No evidence of any enamel has been
found in any sections. The cap of the tubercles is nearly structure-
less in most of the sections examined, only occasionally showing
tubules. But, since this tissue is superficial to the most external
vascular canals, it is certainly part of the superficial layer and is
considered to be a modified osteo-dentine.

Judging by the extent of the vascular canals, the middle layer
extends well up into the tubercles, although bone cell lacunae have
not been found as far superficially as the canals. The system of
vascular canals is simple, but not unmodified from the Tremataspis
type. Ascending vascular canals from the basal layer are centered
below the tubercles and lead into large sinuses in the middle layer
at the base of the tubercles (fig. 35, C, src). From these sinuses
the submucous vascular plexus extends horizontally below and to
the bases of the grooves of the sensory canals; it has a typically
cephalaspid radiating pattern. External branches of the ascending
vascular canals continue in a superficial direction from the sinuses
into the subepidermal vascular plexus in the tubercles.

Between the tubercles the sensory canals are represented by
relatively large grooves, left open superficially by the absence of the
outer part of the middle layer. Separating the sensory canals are
processes of the middle layer, among which the tubercles may be
included (fig. 37, A, pml). Superficial inspection indicates that both
circumareal and intra-areal canals are present, the former repre-
sented by the larger polygonal grooves surrounding the tubercle
areas, the latter by the complex network within the polygons and
around the tubercles (fig. 37, A, cac, iac). In the limited number of
thin sections available, it has not been possible to distinguish these
two types of sensory canals. The canals of the lateral line are
circumareal canals differing only in that they are arranged in a
linear fashion (fig. 37, A, ifc).

The basal layer shows no obvious modifications. It is clearly
laminated, contains moderate-sized basal cavities below the tubercles,
and occasionally exhibits between the tubercles narrower canals
that may represent descending vascular canals.

212 FIELDIANA: GEOLOGY, VOLUME 11

Genus Saaremaaspis

Since few specimens belonging to Saaremaaspis are known, no
thin sections have been made, and the following rather incomplete
description has been derived from a study of the inner and outer
surfaces and broken edges of the dorsal and ventral shields.

The exoskeleton is very thin. Along the lateral margin there are
small, flattened tubercles that appear to be coated with enamel,
but only on these tubercles has the superficial layer been recognized.
Over the rest of the shield the middle layer forms the surface and
seems to be reduced to some extent. The fine, "granular orna-
mentation" that covers the whole surface and that Robertson identi-
fied as tubercles of the superficial layer (1938, p. 493) are not com-
parable to the tubercles of other Osteostraci but are small projections
of the middle layer, similar to those of Oeselaspis (fig. 37, D, pml).
Between these projections are open canals forming a polygonal net-
work of very fine mesh. Their identification as canals of the sensory
canal network is confirmed by the fact that they communicate freely
with the canals of the lateral line, which have the appearance of
linear, larger, and somewhat deeper members of the same canal
system (fig. 37, D, ifc); circumareal and intra-areal canals cannot
be differentiated. Below the sensory canals in the middle layer,
the submucous vascular plexus is occasionally discernible, and prob-
ably lacks any regular arrangement. The basal layer is sometimes
present, and is typically cross-laminated; it has numerous small
basal cavities, each with a pore beneath it on the smooth inner surface
of the exoskeleton.

As has been noted elsewhere (Denison, 1951), there are no recog-
nizable differences in the structure of the exoskeleton of Rotsikul-
laspis, which, for this and other reasons, has been referred to Saarema-
aspis.

Genus Witaaspis

Witaaspis is another genus so rare that thin sections of the
exoskeleton have not been made. A few slides, made by Patten
from his Oesel collections, and labeled "Thyestes reticulata," almost
certainly belong to Witaaspis, but are unsatisfactory and show
little that cannot be observed by study of the surface of the exo-
skeleton.

The exoskeleton is very poorly developed except marginally and
around the median dorsal structures, where it may be considerably
thicker. As in Saaremaaspis, the superficial layer is present only

coc-

pml —

B

2

Mfc

--pml —

Fig. 37. Preparations of the exoskeleton in which some of the shell material
i ias been removed from below so as to show the relationships of the sensory canal
i etwork to the lateral lines; all are viewed from the inside. A, Procephalaspis
(eselensis (X25); B, Witaaspis schrenkii (X30); C, Oeselaspis pustulata (X60);
]), Saaremaaspis mickwitzi (X75). cac, circumareal sensory canals; cot, circum-
< rbital tubercles; iac, intra-areal sensory canals; ifc, infraorbital lateral line canal;
5 ml, processes of middle layer between intra-areal sensory canals; sc, sensory
( anals (probably intra-areal) ; t, tubercle.

213

214 FIELDIANA: GEOLOGY, VOLUME 11

on the tips of the small tubercles of the lateral margin. The surface
of the rest of the shield is formed by the middle layer and is marked
by grooves forming polygons of moderate size, which undoubtedly
housed the circumareal sensory canals (fig. 37, B, cac). The polygons
are subdivided by shallower and narrower grooves for the intra-
areal sensory canals, which form a fine network (fig. 37, B, iac).
Lateral line canals may be seen to be part of the sensory canal system,
closely resembling the circumareal canals in size and depth (fig. 37,
B, ifc). Between the sensory grooves are processes of the middle
layer, often pointed in section, and pierced by canals of the vascular
system.

The basal layer is very poorly developed and cannot be observed
in many specimens. In one of Patten's sections of "Thyestes reticu-
lata," possibly a juvenile individual, only a thin layer representing
the most superficial part of the basal layer or basal part of the middle
layer is present. Another section shows a very thin basal lamina,
connected infrequently with the more superficial exoskeleton by
vertical columns of laminated bone. Whether the basal layer was
really reduced in this genus, or whether the specimens examined
are juvenile individuals with incomplete development of the basal
layer, it is not possible to say with certainty.

Genus Sclerodus

The peculiar nature of the exoskeleton of the Downtonian genus
Sclerodus has been emphasized by Stensio (1932, pp. 176, 179) and
Wangsjo (1946, pp. 356-357), who considered it to be distinct from
that of all other Osteostraci. Unfortunately it has not been figured,
except for a section of one of the "cornua" (Stensio, 1932, pi. LVI,
fig. 1), which cannot be considered as typical. The superficial layer
is reported to be reduced or entirely absent, and the middle layer
to be well developed. The latter contains a complicated system of
vascular canals, but no trace of the sensory canals has been recog-
nized. Stensio argues that the irregular arrangement of the sub-
mucous vascular plexus ("radiating canals") indicates that the
sensory canal system had been lost, since the polygonal arrangement
of the latter system determines the arrangement of the vascular
canal areas. It is more likely, however, that the superficial part
of the middle layer was lost, that the sensory canal system was
exterior to the exoskeleton, and that the irregular arrangement of
the vascular network was a primitive feature, comparable to that
in Tremataspis and in the ventral shield of Dartmuihia.

DENISON: THE EXOSKELETON OF EARLY OSTEOSTRACI 215
EVOLUTIONARY MODIFICATIONS OF THE EXOSKELETON

In a recent general analysis of the exoskeleton of the Osteostraci
(Denison, 1951), it was concluded that there was no clearly demon-
strated trend towards the reduction of ossification in the evolution
of this order. In the same paper Tremataspis was shown to be
primitive among the Osteostraci in so many respects that the
assumption was made that its well-developed exoskeleton with
unreduced superficial layer was also primitive. Although it is not
always true that evolution proceeds in the direction of increasing
complexity, this is usually the case, so the relatively simple structure
of the Tremataspis exoskeleton is an added argument in favor of
its primitive condition. The surface shows only an insignificant
development of tubercles, and these are merely swellings of the
superficial layer. The system of sensory canals, at least in T. mam-
millata, consists of a simple network of large polygons, uncomplicated
by secondary subdivisions. The vascular networks of the middle
layer are irregular, lacking any orderly pattern or well-defined sub-
division into areas. The basal layer does not reflect the polygonal
arrangement of the middle layer to any degree, and lacks any
recognizable distinct canal systems for the return of venous blood.

Turning to other Osteostraci, the following modifications, all
presumably specializations, appear in various genera and species:

(1) Development of tubercles, involving a protuberance of the
middle as well as the superficial layer. Since tubercles are superficial
structures, they often retain an enamel crown, as in the dorsal shield
of Dartmuthia, and also in Oeselaspis, Aceraspis, Hemicyclaspis
lightbodii, and many species of Cephalaspis. In other genera,
Procephalaspis and Thyestes, the enamel is lost and the tubercles
are capped with dentine. In Witaaspis, Saaremaaspis, and perhaps
Sclerodus, tubercles are present only on the margins of the shield;
the projections that cover the rest of the shield consist only of the
middle layer and are not to be considered as tubercles in the strict
iense.

(2) Superficial reduction of the exoskeleton. Except in Trem-
Uaspis, the superficial layer is always reduced between the tubercles
;tnd may be completely lost. Reduction and loss of the superficial

ayer is also common in non-tuberculated forms. The outer part

< »f the middle layer may also fail to ossify, leaving the sensory canals
videly open. In a few species only the base of the middle layer is

ossified, so that the entire system of sensory canals is outside the

< xoskeleton; this is the case in Thyestes, Didymaspis, and probably

216 FIELDIANA: GEOLOGY, VOLUME 11

in Sclerodus and some species of Cephalaspis. In the most extreme
cases of reduction the middle layer is completely absent, as in
Cephalaspis borealis, C. oblongus, and parts of the shield of Thyestes
and Didymaspis. There is no well-established case where the basal
layer is also reduced; this may be so in Witaaspis, but the fact that
this layer ossifies last in ontogeny suggests the possibility that those
exoskeletons lacking a well-developed basal layer may be those of
juvenile individuals.

(3) Subdivision of the sensory canal network. The sensory canals
of Tremataspis mammillata and of the ventral shield of Dartmuthia
form a polygonal network of relatively large mesh; these are the
primitive circumareal canals. Subdivision of the polygons by the
growth of smaller, and usually more superficial, intra-areal canals
has taken place in other forms. The beginnings of this subdivision
may be seen in Tremataspis milleri and T. schmidti (Denison, 1947,
fig. 4), and both circumareal and intra-areal canals can be distin-
guished in most of the other genera whose exoskeletal structure is
sufficiently well known. The network of sensory canals may become
very complex in Downtonian and Early Devonian forms, especially
in some species of Cephalaspis, such as C. campbelltonensis and
C. powriei, where the circumareal canals themselves are multiple
(Stensio, 1932, fig. 5).

(4) Organization of the vascular system of the middle layer. In
Tremataspis and in the ventral shield of Dartmuthia both the sub-
epidermal vascular plexus and the submucous plexus are irregular.
In most other Oesel genera, and in the majority of later Osteostraci,
the polygonal pattern defined by the circumareal sensory canals
conditions a subdivision of the vascular network into polygonal
areas. Typically, the blood vessels are supplied from within at the
center of a polygon, or under the center of a tubercle, and blood is
transported to the periphery of the polygon or tubercle area by
canals with a clearly radiating pattern at the base of the middle
layer. While originally the radiating canals occupied a single level,
they may subdivide so as to occupy a considerable thickness of bone
in Cephalaspis.

(5) Modifications of the basal layer. The basal layer is remark-
ably conservative in its structure throughout the Osteostraci. The
basal cavities are consistently placed below the centers of the
polygons, the partitions between them reflecting to varying degrees
the polygonal pattern of more superficial layers. In some species
of Cephalaspis, such as C. salweyi and C. powriei, the polygonal

DENISON: THE EXOSKELETON OF EARLY OSTEOSTRACI 217

pattern of the basal layer is emphasized by the development of
descending vascular canals and ring sinuses beneath the periphery
of the polygons. An unusual modification of the basal layer is found
in Thyestes verrucosus, where the basal cavities are tremendously
expanded under the larger tubercles.

This review of exoskeletal modifications supports the view that
there was a trend towards increasing complexity of structure in the
history of the Osteostraci. The various specializations do not as
yet appear to fit into any phyletic pattern. The exoskeletal structure
may be characteristic of a genus, but families (excepting monogeneric
ones) cannot now be recognized on this basis. Superficial reduction
is common, but the absence of any well-defined trends in this respect,
and the presence of considerable variability not only between species
but also within species (Cephalaspis pagei and C. powriei) make
it unsafe to assume that the loss of the superficial part of the exo-
skeleton must necessarily debar a species from the ancestry of forms
in which the exoskeleton is well developed. Thus the possibility
cannot be excluded that such genera as Saaremaaspis and Witaaspis,
which have greatly reduced but relatively simple exoskeletons, may
still be ancestral to later Ateleaspidae in which the superficial layer
is usually present. On the other hand, the enormous tubercles of
Thyestes, underlain by greatly enlarged basal cavities, are unique,
a specialization that would seem to exclude the known species of
this genus from the ancestry of most later Cephalaspidae. Our
present knowledge of the exoskeleton of most Downtonian and
Devonian Osteostraci is insufficient to permit the distinguishing of
phyletic groups on the basis of their exoskeletal structure alone,
although eventually this feature may help to indicate their relation-
ships to Ludlow genera.

REFERENCES

Denison, R. H.

1947. The exoskeleton of Tremataspis. Amer. Jour. Sci., 245, pp. 337-365,

figs. 1-13, pis. I-III.
1951. Evolution and classification of the Osteostraci. Fieldiana, Geol., 11,

no. 3, pp. 155-196, figs. 20-31.

Gross, W.

1935. Histologische Studien am Aussenskelett fossiler Agnathen und Fische.
Palaeontogr., 83, Abt. A, pp. 1-60, 7 figs., 7 pis.

Robertson, G. M.
1935. Oeselaspis, a new genus of ostracoderm. Amer. Jour. Sci., (5), 29,
pp. 453-461, figs. 1-4.

218 FIELDIANA: GEOLOGY, VOLUME 11

1938. New genera of ostracoderms from the Upper Silurian of Oesel. Jour.
Pal., 12, pp. 486-493, figs. 1-3, pi. LX.

1939. An Upper Silurian vertebrate horizon, with description of a new species,
Cephalaspis oeselensis. Trans. Kansas Acad. Sci., 42, pp. 357-363, 1 pi.

Stensio, E. A.

1927. The Downtonian and Devonian vertebrates of Spitsbergen. Part I.

Family Cephalaspidae. Skr. Svalbard Nordishavet, nr. 12, pp. xii+391,

figs. 1-102, pis. I-CII.
1932. The cephalaspids of Great Britain. British Museum (Natural History),

pp. xiv+220, figs. 1-70, pis. I-LXVI.

Wangsjo, G.

1946. On the genus Dartmuthia Patten, with special reference to the minute
structure of the exoskeleton. Bull. Geol. Inst., Univ. Upsala, 31, pp. 349-
362, figs. 1-5, pis. V-VII.