Skip to main content

Full text of "Bulletin of the Museum of Comparative Zoology at Harvard College"

See other formats



Library of the 

Museum of 

Comparative Zoology 

SuL Latin OF TH 

Museum of 



Studies in Organismic and Evolutionary 

in honor of A. W. Crompton 



FEB 2 2 2002 

Parish A. Jenkins, Jr., 

l\/lichael D. Shapiro, 




Tomasz Owerkowicz 




10 OCTOBER 2001 





Breviora 1952- 

bulletin 1863- 

Memoirs 1865-1938 

JOHNSONiA, Department of Mollusks, 1941-1974 

Occasional Papers on Mollusks, 1945- 


1. Whittington, H. B., and W. D. I. Rolfe (eds.), 1963 Phylogeny and 
Evolution of Crustacea. 192 pp. 

2. Turner, R. D., 1966. A Survey and illustrated Catalogue of the Tere- 
dinidea (Mollusca: Bivalvia). 265 pp. 

3. Sprinkle, J., 1973. Morphology and Evolution of Blastozoan Echino- 
derms. 284 pp. 

4. Eaton, R. J., 1974. A Flora of Concord from Thoreaus Time to the 
Present Day. 236 pp. 

5. Rhodin, A. G. J., and K. Miyata (eds.), 1983. Advances in Herpetology 
and Evolutionary Biology: Essays in Honor of Ernest E. Williams. 

725 pp. 

6. Angelo, R., 1990. Concord Area Trees and Shrubs. 118 pp. 

Other Publications. 

Bigelow, H. B., and W. C. Schroeder, 1953. Fishes of the Gulf of Maine. 
Reprinted 1964. 

Brues, C.T., A. L. Melander, and F. M. Carpenter, 1954. Classification of 
Insects. (Bulletin of the M. C. Z, Vol. 108.) Reprinted 1971. 

Creighton, W. S., 1950. The Ants of North America. Reprinted 1966. 

Lyman, C. P., and A. R. Dawe (eds.), 1960. Proceedings of the First In- 
ternational Symposium on Natural Mammalian Hibernation. {Bulletin 
of the M. C. Z, VoL 124.) 

Ornithological Gazetteers of the Neotropics (1975-). 

Peters Check-hst of Birds of the World, vols. 1-16. 

Proceedings of the New England Zoological Club 1899-1947. (Complete 
sets only.) 

Proceedings of the Boston Society of Natural History. 

Price list and catalog of MCZ publications may be obtained from Publica- 
tions Office, Museum of Comparative Zoology, Harvard University, Cambridge, 
Massachusetts 02138, U.S.A. 

This publication has been printed on acid-free permanent paper stock. 

©The President and Fellows of Harvard College 2001. 



Introduction 1 

A Probainognathian Cynodont from South 
Africa and the Phylogeny of 
NonmammaUan Cynodonts. By James A. 
Hopson and James W. Kitching 5 

On Microconodon, a Late Triassic Cynodont 
from the Newark Supergroup of Eastern 
North America. By Hans-Dieter Sues 37 

A Cynodont from the Upper Triassic of East 
Greenland: Tooth Replacement and 
Double-Rootedness. By Michael D. 
Shapiro and Parish A. Jenkins, Jr. 49 

On Two Advanced Carnivorous Cynodonts 
from the Late Triassic of Southern Brazil. 
By Jose F. Bonaparte and Mario Costa 
Barberena 59 

The Inner Ear and Its Bony Housing in 
Tritylodontids and Implications for 
Evolution of the Mammalian Ear. By 
Zhexi Luo 81 

A New Specimen and a Functional 

Reassociation of the Molar Dentition of 

Batodon tenuis (Placentalia, Incertae 

Sedis), Latest Cretaceous (Lancian), North 

America. By Craig B. Wood and William 

A. Clemens 99 

The Evolution of Mammalian Development. 

By Kathleen K. Smith 119 

Sldn Impressions of Triassic Theropods as 
Records of Foot Movement. By Stephen 
M. Gatesy 137 

A Diminutive Pterosaur (Pterosauria: 
Eudimorphodontidae) from the 
Greenlandic Triassic. By Farish A. Jenkins, 
Jr., Neil H. Shubin, Stephen M. Gatesy, 
and Kevin Padian 151 

Immature Rhizondontids from the Devonian of 
Nortli America. By Marcus C. Davis, NeU 
H. Shubin, and E. B. Daeschler 171 

How Do Mysticetes Remove Prey Trapped in 

Baleen? By Alexander J. Werth 189 

Tongue-Jaw Linkages: The Mechanisms of 

Feeding Revisited. By Karen M. Hiiemae 

and Jeffrey B. Palmer 205 

Extrinsic Versus Intrinsic Lingual Muscles: A 

False Dichotomy? By Kurt Schwenk 219 

Electromyographic Pattern of the Gular Pump 
in Monitor Lizards. By Tomasz 
Owerkowicz, Elizabeth L. Brainerd, and 
David R. Carrier 237 

Synchronization of Electromyographic Activity 
in Oral Musculature During Suckling and 
Drinking. By A. J. Thexton and Rebecca 
Z. German 249 

Sonomicrometry and Kinematic Estimates of 
the Mechanical Power of Bird Flight. By 
Douglas R. Warrick, Bret W. Tobalske, 
Andrew A. Biewener, and Kenneth P. 
Dial 257 

Trade-off Between Modeling and Remodeling 
Responses to Loading in the Mammalian 
Limb. By Daniel E. Lieberman and 
Osbjorn M. Pearson 269 

Muscle Force and Stress During Running in 
Dogs and Wild Turkeys. By Thomas J. 
Roberts 283 

Regulation of Skeletal Muscle Regeneration 
and Bone Repair in Vertebrates. By Uri 
Oron .-'..- 297 

Bull. Mus. Comp. ZooL, 156(1): i, October, 2001 i 


Fuzz Crompton 

INTRODUCTION ly eclectic array of papers spans a range of 
In grateful tribute to Fuzz Crompton, subjects and approaches that constitute a 
this volume presents papers delivered at a challenge to any conventionally inclusive 
symposium held in Fuzz's honor on 15 title. And yet the diversity herein is rep- 
May 1999 at the Museum of Comparative resentative of the breadth of Fuzz's inter- 
Zoology, Harvard University. The seeming- ests, the scope of his inspiration, and the 

Bull. Mus. Comp. ZooL, 156(1): 1-3, October, 2001 1 

2 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

extent of his influence in both teaching electromyographic data from living sys- 

and research. Among this volumes authors tems provided the essential tools wdth 

are present and former students (and stu- which to interpret the evolutionary history 

dents of former students), postdoctoral fel- and transformations embedded in fossils, 

lows, colleagues, and collaborators, all of From his work our understanding of the 

whom gladly joined to celebrate his man- tribosphenic dentition — that once fear- 

ifold contributions to organismic and evo- some array of cusps, crests, cristi, and val- 

lutionary biology. Perhaps the only aspect leys — was forever transformed. Tribos- 

of these proceedings not fully conveyed in phenic teeth, no longer the statuary mark- 

these published papers was a collegiality at ing therian beginnings, became but one 

the syinposium that spilled over into the expression of the functional continuum in 

open joviality of a reunion, an enthusiasm synapsid evolution. Not content with 

for science, and a delight in professional odontology explained. Fuzz's studies with 

friendships that all derive from the man his students and collaborators became ever 

we honored. more complex orchestrations of experi- 

Launching a career remarkable for ac- mental apparatus as he elucidated the in- 
complishment. Fuzz's publication record terrelationships of jaw and tongue move- 
began with a paper that appeared in Acta ments, the patterns of chewing and the 
Zoologica in 1953, "The development of mechanics of jaws, the millisecond events 
the chondrocranium of Spheniscus demer- of swallov^ng, and the neurobiology of 
sus with special reference to the columella that most primal of mammalian feeding 
auris of birds," a product of his graduate patterns, suckling. Marsupials, primates, 
work at Stellenbosch University. But it was insectivores, pigs, and goats were among 
with a second doctorate, completed in a his subjects, but in each case the real sub- 
stunningly short two years and under the ject was integrative biology, with results 
guidance of Francis Rex Parrington, FRS, that were invariably exemplary. This inte- 
at Cambridge University, that he redirect- grative approach was especially empha- 
ed his investigations into vertebrate pale- sized in Biology 21, Structure and Physi- 
ontology, and specifically into the mor- ology of the Vertebrates, a course offered 
phology and relationships of mammallike by Fuzz and the late Dick Taylor for al- 
reptiles, and eventually into the origins of most three decades. At the top of the scale 
mammals. A series of classic, oft-cited pa- in terms of the required workload, the ex- 
pers over the next two decades made ma- perience was consistently and enthusiasti- 
jor advances in our understanding of the cally rated by students as providing the 
intricacies of dental and cranial evolution highest of intellectual returns, and a 
among synapsids of the Mesozoic. source of genuine enjoyment. 

Fuzz's early interests in the tangible re- Unless we aspire to monographic 

cord of vertebrate evolution soon expand- lengths, we cannot adequately recount by 

ed into what is now generally known as way of gratitude all of Fuzz Crompton's 

functional morphology — understanding contributions to science, teaching, and 

the mechanics and other functional inte- promoting the careers of others. And yet 

grations of structural features. In a fun- there is one contribution, distinctively in- 

damental sense, as he himself so often tegrative, that must appear in this perma- 

said, he wanted to know how animals nent record of die Museum of Compara- 

worked. Although precedents for such an tive Zoology if we are to salute him at all. 

approach had previously been established During his career, which began as Curator 

in vertebrate paleontology, it was Fuzz's of Palaeontology Collections in the Na- 

signal contribution that he placed his anal- tional Museum in Bloemfontein, South Af- 

yses in the context of experimental work rica. Fuzz served as Director of three ma- 

on extant animals. Cineradiographic and jor museums: first at The South African 

Introduction 'Jenkins et al. 

Museum, then at Yale's Peabody Museum 
of Natural History, and finally at the Mu- 
seum of Comparative Zoology. During the 
course of this extensive experience he orig- 
inated the concept of Professor/Curator in 
face of the long-standing belief, reinforced 
by practice, that these were separate spe- 
cies with different territories. On the pre- 
mise that natural history museums ought 
not to be simply repositories, but are jus- 
tifiable to the extent that they promote re- 
search and knowledge of our natural 
world, the conclusion is inescapable that 
senior museum staff must be, first and 
foremost, scientists of distinction. In a uni- 
versity setting, as a consequence, positions 
supported by museum resources inust be 
professorial, with standards and expecta- 
tions no less than those held for every fac- 
ulty appointment. No longer would cura- 

tors exist as another class of citizenry apart 
from academic departments. Rather, pro- 
fessors with adininistrative appointments 
as curators would continue to ensure the 
museums growth and participation in ac- 
ademic research and instruction. The con- 
cepts of Professor and Curator, once a du- 
ality, became inseparably integrated. At 
Harvard, this legacy froin Fuzz Cromp- 
ton's visionary directorship persists today, 
to the intellectual enhancement of the 
MuseuiTj of Comparative Zoology, the De- 
partment of Organisinic and Evolutionary 
Biology, and the University. 

Parish A. Jenkins, Jr. 
Michael D. Shapiro 
Tomasz Owerkowicz 

Cambridge, Massachusetts 
10 April 2001 

Symposium participants, 1 5 May 1 999. From left to right, front row: Parish Jenkins, Jose Bonaparte, Neil Shubin, Kathleen Smith, 
Fuzz Crompton, Ken Dial, Nick Hotton. Second row: Jim Hopson, Christine Janis, Craig Wood, Zhexi Luo, Rebecca German, 
Allan Thexton, Tomasz Owerkowicz. Third row: Steve Gatesy, Dan Lieberman, Kurt Schwenk, Alex Werth, Beth Brainerd. Fourth 
row: Uri Oron, Mike Shapiro, Tom Roberts, Andy Biewener (photograph by Leon Claessens). 



Abstract, a new small cynodont from subzone B 
of the Cynognathus Assemblage Zone (earliest Mid- 
dle Triassic) of South Africa is described as Lumkuia 
ftizzi. It is represented by a nearly complete skull and 
lower jaw, a shoulder girdle and forelimb, and artic- 
ulated dorsal and caudal vertebrae. It is placed in the 
eucynodont clade Probainognathia on the basis of 
four unequivocal synapomorphies, including absence 
of a parietal foramen and expanded plates on the ribs 
and a secondary palate extending posteriorly to the 
level of the orbit. Lumkuia is the oldest and most 
primitive probainognathian represented by adequate 
material. A cladistic analysis strongly supports the 
monophyly of Cynodontia, Epicynodontia (a new tax- 
on including Galesaurus, Thrinaxodon, and eucyno- 
donts), and Eucvnodontia. The analysis also supports 
the eucynodont clades Probainognathia and Cynog- 
nathia, and Gomphodontia as a subgroup of the latter. 
Within Probainognathia, a chiniquodontid clade and 
a tritheledontid + mammaliaform clade are well sup- 
ported. Probainognathus is sister to the latter clade, 
but this node breaks down in trees two steps longer 
than the shortest tree. Tritylodontids are deeply nest- 
ed within the traversodont gomphodonts, with "Sca- 
lenodon" hirschoni weakly supported as their sister 


The Eucynodontia (Kemp, 1982, 1988), 
that is, those cynodonts more derived than 
the basal Triassic Thrinaxodon, have tra- 
ditionally been divided into a carnivorous 
line leading to maminals and a herbivo- 
rous, or gomphodont, line leading to the 
Jurassic tritylodontids (Crompton and El- 
lenberger, 1957; Crompton, 1972b; Hop- 
son and Kitching, 1972; Sues, 1985; Hop- 
son and Barghusen, 1986; Hopson, 1991b, 
1994). However, Kemp (1982, 1983, 1988) 

' Department of Organismal Biology and Anatomy, 
University of Chicago, 1027 East 57th Street, Chi- 
cago, Illinois 60637. 

' Bernard Price Institute for Palaeontological Re- 
search, University of the Witwatersrand, Johannes- 
burg, South Africa. 

noted that tritylodontids and mammals 
share many derived features that are ab- 
sent in Triassic cynodonts, which led him 
to suggest that tritylodontids should be 
separated from the herbivorous cynodonts 
and placed in the carnivorous line close to 
Mammalia; the herbivorous specializations 
of tritylodontids thus would be convergent 
on those of gomphodonts. Rowe (1986, 
1988, 1993) went still further in obliter- 
ating the distinction between the carnivo- 
rous and herbivorous lineages by inter- 
leaving Middle Triassic to Early Jurassic 
cynodonts in a paraphyletic series of car- 
nivorous and gomphodont taxa that lead to 
a terminal clade Mammaliamorpha, con- 
taining tritylodontids and traditionally de- 
fined mammals (termed Mammaliaformes 
by Rowe). The sister-group relationship of 
Tritylodontidae and Mammaliaformes has 
become widely accepted (Wible, 1991; Lu- 
cas and Luo, 1993; Martinez et al, 1996), 
although Sues (1985) and Hopson (1991b, 
1994) have argued against it. 

The senior author (Hopson, 1990, 
1991a,b, 1994) has summarized the results 
of his cladistic analyses of cynodont rela- 
tionships, although, to date, has not pub- 
lished the data on which they are based. 
Hopson recognizes a primarily herbivo- 
rous clade that includes tritylodontids, the 
Cynognathia of Hopson and Barghusen 
(1986), and a carnivorous clade that in- 
cludes mammals, which has been desig- 
nated Probainognathia (Hopson, 1990). A 
data matrix of synapsids as a whole was 
pubhshed by Sidor and Hopson (1998), 
but it lacks critical taxa and characters for 

Bull. Mus. Comp. ZooL, 156(1): 5-35, October, 2001 5 

6 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

resolving lower-level relationships within 
nonmammalian cynodonts. 

In May, 1988, while studying in the Ka- 
roo fossil collection of the Bernard Price 
Institute for Palaeontological Research 
(BPI) at the University of the Witwaters- 
rand, Hopson noted a small skull and par- 
tial skeleton (BP/1/2669) from the Cynog- 
nathus Assemblage Zone. The specimen 
was identified in the catalog as a juvenile 
Trirachodon, but its skull morphology 
more closely reseinbled that of Prohain- 
ognathus and the Chiniquodontidae, car- 
nivorous eucynodonts best known from 
the Middle and Late Triassic of South 
America. A notice of the specimen, with 
preliminary conclusions on its phylogenet- 
ic significance, was presented at the 48th 
annual meeting of the Society of Verte- 
brate Paleontology (Hopson and Kitching, 

This new cynodont is named and briefly 
described here. It is compared with Thri- 
naxodon, as a member of a more primitive 
cynodont grade; with Prohainognathus and 
chiniquodontids, as members of the Pro- 
bainognathia; and with Cijnognathus and 
early gomphodont genera, as members of 
the Cynognathia. One purpose of this pa- 
per is to justify the establishment of the 
eucynodont clades Cynognathia (sensu 
Hopson and Barghusen, 1986) and Pro- 
bainognathia {sensu Hopson, 1990, 1991a, 


Specimen BP/1/2669 had been partially 
prepared at the BPI so that portions of the 
skeleton were exposed on both sides of a 
small sandstone slab. The skull and lower 
jaws were subsequently removed from the 
slab and more fully prepared by Ms. Claire 
Vanderslice. Although portions of the ex- 
ternal surface of the skull are damaged, 
the palate, braincase, and medial surface 
of the lower jaw are beautifully preserved. 
Because the postcranial elements are, for 
the most part, heavily eroded, they have 
been further prepared only slightly. 

Comparisons with other cynodonts are 

based on specimens, stereophotographs, 
notes and drawings, and published ac- 
counts. The data matrix of cynodonts in- 
cludes characters published by Sidor and 
Hopson (1998), with many new characters 
added, particularly from the dentition. The 
matrix was analyzed using the 3.1 version 
of PAUP (Swofford, 1993). In the follow- 
ing section, phylogenetic definitions of a 
number of suprageneric taxa are given, 
with a distinction made between node- 
based and stem-based definitions, as rec- 
ommended by Sereno (1999). 

Therapsida Broom, 1905 

Cynodontia Owen, 1861 

Definition. The most inclusive clade in- 
cluding Mammalia and excluding Bauria. 
This clade and its sister group, the Ther- 
ocephalia (defined as the most inclusive 
clade including Bauria and excluding 
Mammalia), are stem-based members of a 
node-based Eutheriodontia (defined as the 
least inclusive clade including Mammalia 
and Bauria). (See Sereno [1999] for dis- 
cussion of node-stem triplets.) 

Epicynodontia new taxon 

Definition. The most inclusive clade in- 
cluding Mammalia and excluding Procy- 
nosuchus. This clade includes, among oth- 
ers, Galesaurus, Thrinaxodon, and eucy- 

Eucynodontia Kemp, 1982 

Definition. The least inclusive clade in- 
cluding Mammalia and Exaeretodon. This 
is a node-based taxon, with two stem- 
based subgroups: Cynognathia (defined as 
the most inclusive clade including Exaer- 
etodon and excluding Prohainognathus) 
and Probainognathia (defined below). 
Within Cynognathia is a major stem-based 
subgroup, the Gomphodontia (defined as 
the most inclusive clade including Exaer- 
etodon and excluding Cynognathus). 

Probainognathian Cynodont From South Africa • Hopson and Kitching 7 

Probainognathia Hopson, 1 990 orbital wall only to the level of the lacrimal 

Definition. The most inclusive clade in- ^ramina and an orbital process of the pal- 

cluding Probainognathus and excluding ^^"^^ ^^ ^^^^^^g- ^^ }^ "^°^^, primitive than 

Exaeretodon. other eucynodonts m that the dentaiy does 

not extend as tar posteriorly, resulting in a 

Family LUMKUIIDAE new family longer dorsal exposure of the surangular 

between the rear of the dentary and the 

Definition. The most inclusive clade in- articular. At the rear of primary palate is 

eluding Lumkuia and excluding Ecteni- ^n autapomorphic feature: the pterygoids 

^^^on. form a deep median depression with a 

, , . , . . . nearly vertical posterior wall, behind 

Livm/cty/a ftvzz/ new genus and species u- u 4-u r i. j- u 

^ ^ wiiicfi tJiey lorm a prominent median boss 

Etymology. The generic name is from anterior to the interpterygoidal vacuities, 

the Lumku Mission, near which the spec- The presence of interpterygoidal vacuities 

imen was found. The species name is in suggests that the type specimen may be a 

honor of A. W. "Fuzz" Crompton, in rec- subadult individual, 
ognition of his distinguished career as a 

student of cynodonts and early mammals. DESCRIPTION 

Holotype. BP/1/2669, partial skeleton, Qy,,A\ 
including skull with lower jaws; left sca- 

pulocoracoid and clavicle, interclavicle, In dorsal view (Fig. 1), the general ap- 

and proximal part of right clavicle; most of pearance of the skull o^ Lumkuia is similar 

left forelimb; and two articulated segments to that of Thrinaxodon, although the pre- 

of the axial skeleton; the latter consist of orbital region is shorter and the temporal 

10 dorsal vertebrae with associated ribs fossa longer. Thus, the center of the orbits 

and eight caudal vertebrae. lies anterior to the middle of the skull. 

Horizon and Locality. The specimen is whereas in Thrinaxodon the orbits are 

from the Burgersdorp Formation, in sub- centered exactly at midlength. The ptery- 

zone B of the Cynognathus Assemblage goid flanges, which in Thrinaxodon lie be- 

Zone. It was collected by Father Paul low the middle of the orbits, are visible in 

Reubsamen in the vicinity of the Lumku Lumkuia behind the postorbital bar (as 

Catholic Mission, near the town of Lady they commonly are in eucynodonts). As in 

Frere, Eastern Cape Province, South Af- Thrinaxodon, the sagittal crest terminates 

rica. above the occiput, so that the occipital 

Age. The Cynognathus Assemblage condyles are visible from above. This con- 
Zone in the region of Lady Frere is rep- trasts with Ecteninion and Probainogna- 
resented by subzone B of Hancox et al. thus (although not chiniquodontids), in 
(1995; B. S. Rubidge, personal communi- which the sagittal crest overhangs the oc- 
cation), which is considered to be of early ciput and covers the condyles. The lamb- 
Middle Triassic (Anisian) age (Hancox and doidal crests in Lumkuia, as in Thrinaxo- 
Rubidge, 1997). don, diverge at greater than 90 degrees 

Diagnosis. Lumkuia fiizzi is character- and extend posteriorly at their outer ends 

ized by a unique combination of primitive only a short distance beyond the occipital 

and derived features. It possesses the fol- condyles. In Probelesodon and Chiniquo- 

lowing probainognathian features: parietal don, and in gomphodonts, the sagittal crest 

foramen absent; rear of secondary palate terminates slightly in front of the condyles 

lies below anterior border of orbit; ex- but the lambdoidal crests diverge at an 

panded plates on ribs absent. It is more acute angle and extend back well beyond 

primitive than other probainognathians in the condyles. The zygomatic arches of Lw- 

that the frontal extends down the medial mkuia are more flared and rounded in pro- 

8 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


Probainognathian Cynodont From South Africa • Hopson and Kitching 9 

Figure 1. Skull of Lumkuia fuzzi (BP/1/2669) in dorsal view (on left enlarged x3). Scale bar = 10 mm. Abbreviations; e, 
epipterygoid; eo, exoccipital; f, frontal; fic, foramen of lacrimal canal; j, jugal; I, lacrimal; mx, maxilla; n, nasal; op, opisthotic; p, 
parietal; pi, palatine; pm, premaxilla; po, postorbital; pp, postparietal; pr, prootic; prf, prefrontal; pt, pterygoid; ptpf, pterygopar- 
occipital foramen; q, quadrate; qj, quadratojugal; sm, septomaxilla; sq, squamosal; V^j, trigeminal foramen. 

10 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Probainognathian Cynodont From South Africa • Hopson and Kitching 




O Q. 

c B 

0) o 

0) . 
- o 

£ °- 

<D i5 

B E 

S a? 

-D Q. 

^' E 

^" D. 

<" « 
,_- Q. 

CO . 


i/i 2 


_ cc 

J2 J_- 3 





to cr 
to . 

. I/) 
c . - 




ro E o 

-Q . CL 

0) X o 


X o i5 

to 3 

T3 — O) 

0) _- C 

O) CO 

^_ j_r ^— 

to to 3 

"F cn o 

«: - <fl 
cu . . 

~ -b' « 
c to c 

g o « 

.9? To -a 

> E 2 

"to '^ ^ 
Ir o D 

tU CO = 

to .^ £ 

---- <D to 

05 £ — 

to t »- 

CD CO p 

Si! oil- 
^ "^ <" 
Q. d §> 

Nj to CO 

•5 o to 
§2 2 

■~l EL'S 

o o 2- 
— o "-^ 

X — i 

<u cr 


r< O- 


CD — 
CM ;-iS 

^ 2 o 

3 to ■i= 
g) 0) O 

LL E Q. 

12 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Probainognathian Cynodont From South Africa • Hopson and Kitching 13 

Figure 3. Skull of Lumkuia fuzzi (BP/1/2669) in ventral view (on left enlarged x3). Scale bar = 10 mm. Abbreviations: a, 
angular; ar, articular; bo, basioccipital; d, dentary; e, epipterygoid; eo, exoccipital; hf, hypoglossal foramina; ic, internal carotid 
foramina; iptv, interpterygoidal vacuity; j, jugal; jf, jugular foramen; Ifpr, lateral flange of prootic; mpf, major palatine foramen; mx, 
maxilla; op, opisthotic; pa, pila antotica; pi, palatine; pm, premaxilla; pr, prootic; pra, prearticular; ps, parasphienoid; pt, pterygoid; 
ptb, pterygoid boss; ptpf, pterygoparoccipital foramen; q, quadrate; qj, quadratojugal; qre, quadrate ramus of epipterygoid; ref 
lam, reflected lamina; rps, parasphenoid rostrum; s, stapes; sa, surangular; sp, splenial; sq, squamosal; t, tabular; v, vomer. 

14 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

file than those of Thrinaxodon, being wid- described below, with only salient features 
est anterior to, rather than at, the level of noted. The facial portions of the premax- 
the jaw joint; in this Lumkuia resembles ilia and septomaxilla are missing. On the 
other Triassic probainognathians and dif- palate, backwardly pointing processes of 
fers from cynognathians. the premaxillae separate the elongate, slit- 
In lateral view (Fig. 2), the skull and like, incisive foramina. The premaxilla 
lower jaw of Lumkuia appear to be more forms all but the posteriormost parts of the 
robust than in Thrinaxodon (Fig. 5A), due lateral border of the incisive foramen and 
to the shorter snout, longer temporal re- the fossa for the lower canine, 
gion, and deeper dentary. Also, as in other The alveolar border of the maxilla is 
eucynodonts except Ctjnognathus (Fig. straight, turning up slightly at the level of 
6A), the jaw joint is located more anteri- the last tooth and passing smoothly into 
orly, so that the lambdoidal crests extend the suborbital bar where, a short distance 
well behind the articular region. The ca- behind the last postcanine, it contacts the 
nines are more robust and the posterior jugal. In the palate, the maxilla contributes 
cheek teeth proportionately larger than in to the rear margin of the incisive foramen 
Thrinaxodon and Probainognathus (Fig. and the posteriormost part of the lower ca- 
5C), although not in Ecteninion and chi- nine fossa. The maxilla forms the anterior 
niquodontids. The zygomatic arch appears two thirds of the secondary palate, extend- 
to be no more robust than that of ThH- ing as far back as the gap between the 
naxodon, except perhaps posteriorly, third and fourth postcanines. The inajor 
whereas that of Probainognathus, and es- palatine foramen opens anteroventrally on 
pecially of chiniquodontids and cynogna- the maxillary— palatine suture well lateral 
thians, is much deeper. to the midline. 

In ventral view (Fig. 3), the symphyseal The nasals have largely flaked off, leav- 

region is shorter than that of Thrinaxodon ing some bone only posterolaterally As 

and the shorter jaws diverge at a greater shown by impressions on the surface of 

angle. The secondary palate is only slightly the frontals, the nasals overlap the frontals 

inore developed, with a nearly straight and the nasofrontal suture lies a short dis- 

rather than concave posterior margin. Be- tance behind the anterior border of the or- 

hind the pterygoid flanges, the basicranial bit. 

axis is more transversely compressed than The eroded dorsal surface of the fron- 
in Thrinaxodon, so that the subtemporal tals preserves a slightly undulating midline 
fossa is proportionately wider. suture. The contact with the parietals on 
In occipital view (Fig. 4), the most dis- the skull roof is not preserved, but prob- 
tinctive difference from Thrinaxodon is in ably lay between the posteriormost part of 
the constriction of the base of the zygo- the temporal crests of the postorbitals. In 
matic arch and the separation of the zy- the medial wall of the orbit, a thin strip of 
goma from the more flaring lambdoidal frontal is exposed behind the large de- 
crest by a V-shaped notch. In noneucyno- scending flange of prefrontal, extending 
donts, such as Procynosuchus, Galesaurus, ventrally about to the level of the lacrimal 
and Thrinaxodon, the lambdoidal crest is foramina. 

continuous with the dorsal ridge on the zy- Within the orbit, the prefrontal overlies 
gomatic arch. Only in eucynodonts, with the frontal and is itself overlain by the lac- 
the exception of Cynognathus, is there a rimal; it extends ventrally nearly to the lev- 
distinct break between the two crests, with el of the palatine on the dorsal surface of 
the lambdoidal crest passing back poste- the palate. 

rior to the medial end of the dorsal zygo- The lacrimal has a short exposure on the 

matic ridge. face compared with that o{ Thrinaxodon or 

The individual skull bones are briefly Probainognathus. Within the orbit, it 

Probainognathian Cynodont From South Africa • Hopson and Kitching 15 

Figure 4. Skull of Lumkuia fuzzi (BP/1/2669) in occipital view (upper drawing enlarged x3). Scale bar = 10 mm. Abbreviations: 
bo, basioccipital; earn, external auditory meatus; eo, exoccipital; f, frontal; j, jugal; I, lacrimal; If, lacrimal foramina; op, opisthotic; 
p, parietal; po, postorbital; pp, postparietal; prf, prefrontal; pt, pterygoid; ptf, posttemporal foramen; q, quadrate; qj, quadratojugal; 
s, stapes; so, supraoccipital; sq, squamosal; t, tabular. 

forms the anterior half of the orbital wall 
and most of its floor. Paired lacrimal fo- 
ramina open forward inside the anterior 
rim of the orbit; a small forannen opens 
anterolaterally from the lower lacrimal ca- 
nal on to the facial portion of the lacrimal. 
A posterolaterally directed process of 
the postorbital forms the dorsal part of the 
postorbital bar; it extends down internal to 
the postorbital process of the jugal for an 
indeterminate distance. The upper, more 

horizontal, part of the postorbital bar is 
roughly triangular in cross section; its pos- 
terior face forms a flat vertical surface that 
is continuous posteromedially with a ver- 
tical lappet of postorbital that overlies the 
lateral surface of the parietal. The poste- 
rior parts of the paired postorbitals con- 
verge backwards as temporal crests and 
merge into the median sagittal crest on the 
parietals. These vertical surfaces on the 
postorbital mark the area of attachment of 

16 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 5. Skulls in lateral and ventral views of (A, B) Thrinaxodon liorhinus, and (C, D) Probainognathus jenseni. Scale bars 
= 20 mm. Abbreviations: a, angular; ar, articular; bo, basioccipital; d, dentary; e, epipterygoid; ec, ectopterygoid; eo, exoccipital; 
f, frontal; ic, internal carotid foramen; j, jugal; 1, lacrimal; mx, maxilla; n, nasal; op, opisthotic; p, parietal; pi, palatine; pm, 
premaxilla; po, postorbital; pr, prootic; prf, prefrontal; ps, parasphenoid; pt, pterygoid; q, quadrate; qj, quadratojugal; ref lam, 
reflected lamina; s, stapes; sa, surangular; sm, septomaxilla; sq, squamosal; t, tabular; v, vomer. 

Probainognathian Cynodont From South Africa • Hopson and Kitching 17 


ref lam 

op qj 

Figure 6. Skulls in lateral and ventral views of (A, B) Cynognathus crateronotus, and (C, D) Diademodon mastacus. Scale bar 
in A = 100 mm, in B = 90 mm. Abbreviations: a, angular; ar, articular; bo, basioccipital; d, dentary; e, epipterygoid; ec, ectop- 
terygoid; eo, exoccipital; f, frontal; j, jugal; I, lacrimal; mx, maxilla; n, nasal; op, opisthotic; p, parietal; pi, palatine; pm, premaxilla; 
po, postorbital; pr, prootic; prf, prefrontal; ps, parasphenoid; pt, pterygoid; q, quadrate; qj, quadratojugal; ref lam, reflected lamina; 
s, stapes; sm, septomaxilla; spj, suborbital process of jugal; sq, squamosal; t, tabular; v, vomer. 

18 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

the anteriormost portion of the temporahs zygomatic processes, separated at the level 

muscle. A medially extending horizontal of the V-shaped notch. The cranial process 

lappet of postorbital contacts the sliver of is a relatively flat, triangular plate extend- 

frontal exposed in the orbital wall. ing nearly to the apex of the posterior end 

The dorsal parts of the fused parietals of the sagittal crest. The cranial process 
are damaged, but enough is preserved to overhangs the anterior opening of the 
indicate that the sagittal crest was relative- posttemporal foramen and its flaring rear 
ly low anteriorly and increased only mod- border forms the lambdoidal crest. The V- 
erately in height posteriorly. No evidence shaped notch has an anteroposteriorly 
of a parietal foramen is preserved and it rounded dorsal surface that separates the 
was almost certainly absent. The ventral lambdoidal ridge from the dorsal ridge on 
margin of the parietal contacts the dorsal the zygomatic process. Directly below the 
edge of the orbitosphenoid, behind which notch is a triangular lappet of squamosal 
it is overlapped by the dorsal margin of the that on its medial surface bears a depres- 
epipterygoid back to its midlength. Be- sion for the incompletely ossified distal 
tween the parietal and the dorsal margins end of the paroccipital process. Antero- 
of the epipterygoid and prootic is an elon- medially, the lappet appears to contact the 
gate opening into the cranial cavity. Lead- lateral flange of the prootic. Laterally, it 
ing to this opening from behind, along the forms the medial wall of the recess for the 
prootic-parietal suture, is a deeply incised quadrate, which is open behind as an in- 
groove that begins at the anterior opening verted V-shaped emargination. The emar- 
of the posttemporal foramen. The groove gination is bounded laterally by a slender, 
presumably contained the supraorbital ra- pointed process that descends between the 
mus of the ramus superior of the stapedial upper ends of the quadrate and quadra- 
artery, with a meningeal branch entering tojugal. Further laterally, the zygomatic 
the cranial cavity through the elongate portion of the squamosal forms a descend- 
opening (Rougier et al., 1992; Wible and ing process behind the jugal that in life 
Hopson, 1995). The parietals broaden pos- presumably contacted the surangular (al- 
teriorly, where they are overlain by the though here a contact is absent because 
cranial process of the squamosal, and con- the lower jaw appears to have shifted 
tribute to the roof of the posttemporal fo- slightly forward). The zygomatic process 
ramen. curves foiward from this level extending 

The jugal is a relatively slender bone, over the jugal nearly to the level of the 

not unlike that of Thrinaxodon and Pro- postorbital bar. A shallow sulcus, the ex- 

bainognathus. In the zygomatic arch, the ternal auditory meatus, extends up and 

jugal is overlain dorsally by the zygomatic foi-ward from the distal end of the paroc- 

process of the squamosal and bounded be- cipital process on to the posterolateral sur- 

hind by a descending lappet of squamosal, face of the zygoma. 

A moderate-sized, anterolaterally directed The fused vomers form the center of 

foramen pierces the jugal below the orbit, the arched roof of the primary palate 

A short distance behind the last postca- above the secondaiy palate and roof the 

nine, the jugal passes medial to the rear of choanal trough to a point just behind the 

the maxilla to contact the anterolateral level of the last postcanine. 

margin of the pterygoid, and perhaps the The palatal plates of the palatines form 

palatine, at the anterior border of the sub- the posterior third of the short secondary 

temporal fossa. The jugal is exposed be- palate, underlying the posterior margins of 

hind the lacrimal in the posterior part of the maxillae. A small foramen pierces the 

the orbital floor. palatine a short distance posterointernal to 

The squamosal may be described as the major palatine foramen. The rear mar- 
consisting of two portions, the cranial and gin of the secondary palate is thickened 

Probainognathian Cynodont From South Africa • Hopson and Kitching 19 

and slightly rugose. As in other Triassic tween the ridges is divided by the long ros- 
probainognathians, the lateral margin of trum of the parasphenoid to form paired 
the secondary palate curves dorsally, so interpteiygoidal vacuities. Such vacuities 
that the palatine meets the maxilla in the are present in Dvinia and Procynosiichus 
floor of a narrow longitudinal trough in- and in juveniles of Thrinaxodon (Estes, 
temal to the posterior postcanines. This 1961), but are usually absent in postpro- 
trough continues back beyond the level of cynosuchid cynodonts. 
the secondary palate, where it is bounded The orbitosphenoid is roughly the shape 
medially by slender ridges that extend of an elongate half-cylinder, with a U- 
back nearly to die lateral margins of the shaped cross section. It lies on the midline 
pterygoid flanges. An ectopterygoid is not below the postorbitals and parietals in the 
present, so the posterolateralmost part of space between the postorbital bar and the 
the palatine contacts the pterygoid, and anterodorsal end of the epipteiygoid. 
perhaps the jugal, internal to the last post- The ascending lamina of the epiptery- 
canine. The palatine is here pierced by goid is extremely long fore to aft, being 
several small foramina, with a larger open- nearly twice the length of the prootic por- 
ing between its posterior margin and the tion of the braincase sidewalk This con- 
overlying pteiygoid. The palatines form trasts sharply with the condition in Pro- 
the lateral walls of the choanal trough, bainognathus and Ecteninion, in which the 
contacting the vomer and pterygoids me- ascending lamina tapers anterodorsally and 
dially and contributing to the anterior half is much shorter than the prootic. The as- 
of the more medial palatal ridges that cending lamina is suturally joined to the 
bound the posterior half of the trough, anterodorsal margin of the prootic above 
The palatine is exposed on the upper sur- the anterior border of the large trigeminal 
face of the primary palate as a broad plate foramen. The epipterygoid contacts the 
that lacks a dorsal orbital process. basicranial wing of the pterygoid ventrally 

The ectopterygoid is absent. Although and appears to have a short medial contact 

described in other Triassic probainogna- with the basipterygoid process. Its quad- 

thians (Romer, 1969, 1970; Martinez et ak, rate ramus is a shallow vertical lamina that 

1996), we believe its presence has not extends back below the trigeminal fora- 

been convincingly demonstrated. men to meet the lateral flange of the pro- 

The pteiygoids form the rear of the cho- otic. The epipterygoid continues back for 

anal trough, which is uniquely deep and is a short distance in contact with the lateral 

bordered behind by a near-vertical wall. At flange, terminating at the level of the an- 

the posterior end of the medial palatal terior border of the pteiygoparoccipital fo- 

ridges, where they converge at the rear of ramen. That portion of the epipteiygoid 

the choanal trough, is a prominent median behind the basipterygoid joint forms the 

boss; this feature appears to be unique lateral wall of a ventrally open space, the 

among cynodonts. Lateral to the anterior cavum epiptericum. 

end of the medial palatal ridge, adjacent The basisphenoid consists of slightly ex- 
to the suture with the palatine, are one or panded anterior basipterygoid processes 
more slitlike openings that pierce the pter- that contact the pterygoids and epipteiy- 
ygoid. Laterally, the deep, triangular pter- goids, a very narrow middle portion that 
ygoid flanges descend well down the inside underlies the sella turcica and is pierced 
of the lower jaws. The ridges forming their by paired carotid foramina, and an ex- 
rear margins converge posteriorly and ex- panded posterior part that contacts the 
tend on to the basipteiygoid rami of the prootic dorsally and the basioccipital pos- 
pteiygoids, nearly meeting where the lat- teriorly. The dermal parasphenoid is fused 
ter contact the basipteiygoid processes of to its ventral surface, forming a near-hor- 
the basisphenoid. An elongate gap be- izontal, triangular plate posteriorly that 

20 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

covers the basisphenoid-basiooccipital opisthotic with the space occupied by the 

contact. Further forward, the parasphen- inner ear. Deep within the jugular fora- 

oid passes between the carotid foramina men is a low ridge that extends a short 

and forms an elongate midline process, the distance into the opening from its postero- 

rostrum, that extends forward between the lateral wall. This ridge in more derived cy- 

pterygoids to separate the interpteiygoidal nodonts, such as Prohainognathus (Fig. 

vacuities. The anterior end of the paras- 5D) and Massetognathus (Rougier et al., 

phenoid is suturally joined to the ptery- 1992, figs. 7B, D), is a long fingerlike pro- 

goids immediately behind and dorsal to jection that extends toward the medial wall 

the median pterygoid boss. of the foramen. In tritheledontids, tritylo- 

The prootic portion of the ossified otic dontids, and mammaliaforms, the foramen 

capsule lacks a sutural separation from the is fully subdivided, with a true jugular fo- 

opisthotic portion, although the prootic ramen posteriorly (transmitting nerves and 

typically contributes to the anterior part of vessels from the cranial cavity), and a peri- 

the paroccipital process and rim of the fe- lymphatic foramen anteriorly (transmitting 

nestra ovalis. The lateral flange of the pro- the perilymphatic duct from the inner ear 

otic extends posterolaterally from behind cavity). 

the trigeminal foramen; although the distal The basioccipital is exposed midventral- 

end of the lateral flange is damaged, it un- ly behind the parasphenoid, with which it 

doubtedly contacted the squamosal in life, has an interdigitating transverse suture. It 

thus enclosing the large, oval, pterygopar- forms the midventral part of the foramen 

occipital foramen. The lateral surface of magnum, bearing a narrow transverse ar- 

the prootic bears a slight groove that ex- ticular facet for the atlas intercentiaim. 
tends between the latter opening and the The paired exoccipitals form the occip- 

trigeminal foramen. Such a groove is usual ital condyles, damaged here, which extend 

in cynodonts, although here it is unusually about one third of the distance up the lat- 

faint. The system of grooves and foramina eral sides of the foramen magnum. More 

in the lateral surface of the prootic are in- dorsally, they meet the supraoccipital, but 

terpreted as transmitting arteries and veins the sutural contact cannot be distin- 

(see Rougier et al., 1992; Wible and Hop- guished. The exoccipital contributes to the 

son, 1995). Deep to the outer margin of posteromedial wall of the jugular foramen, 

the trigeminal foramen, the ossified pila which bears a shallow depression in which 

antotica extends anterodorsally approxi- lie two hypoglossal foramina, a smaller an- 

mately to the level of the basipteiygoid terior one and a larger posterior one, 

joint. Just in front of the fenestra ovalis is which open into the cranial cavity shortly 

the small, posterolaterally directed fora- in front of the occipital condyle, 
men for the facial (Vllth) nei-ve. The median supraoccipital forms an in- 

The opisthotic forms the posterior half determinate part of the dorsal border of 

of the rim of the fenestra ovalis; most of the foramen magnum. The supraoccipital 

the paroccipital process; and the anterior, is overlain by the tabular laterally and the 

anteromedial, and lateral borders of the postparietal above. The postparietal occu- 

jugular foramen. The ventral surface of the pies the upper surface of the occiput 

paroccipital process slopes up and foi-ward above the supraoccipital and tabulars and 

from its rounded posteroventral margin, to between the flaring lambdoidal crests, 

form the posterodorsal wall of the middle Middorsally, the postparietal has a short, 

ear cavity (Hopson, 1966). The opisthotic pointed process that extends forward be- 

contacts the basioccipital medially and the tween the fused parietals. The tabulars oc- 

exoccipital posteromedially and posterolat- cupy the entire occiput lateral to the su- 

erally on the margins of the jugular fora- praoccipital, completely surrounding the 

men. This foramen is confluent within the small, circular posttemporal foramina. 

Probainognathian Cynodont From South Africa • Hopson and Kitching 21 

The quadrate is exposed on the right dentary is a slightly projecting pseudan- 
side, where it has shifted slightly forward gular process, above which the lower mar- 
from its contact with the squamosal. The gin of the bone curves up and back over 
transversely oriented articular condyle of the postdentary elements. A low out- 
the quadrate is about as wide as the total turned ridge overlies the surangular and 
bone is high. The flat posterior surface of angular and continues forward across the 
its ascending process is oriented obiquely masseteric fossa, fading into its surface be- 
to the transverse axis and fits against a low the last upper postcanine. The mas- 
matching surface on the anterior face of seteric fossa extends forward as a slight de- 
the squamosal. The posterolateral third or pression to the level of the fifth upper 
so of the quadrate is exposed from behind postcanine. The coronoid process rises 
in the inverted V-shaped emargination of slightly above the dorsal border of the zy- 
the squamosal. The lateral end of the gomatic arch just behind the postorbital 
quadrate condyle extends well beyond the bar. The lateral surface of the coronoid 
outer margin of the ascending process; its process forms a broad, slightly concave 
dorsal surface is clasped by the transverse- trough between out-turned anterodorsal 
ly expanded lower end of the quadratoju- and posteroventral borders. The slightly 
gal. convex posterior margin of the process 

The quadratojugal has a transversely slopes down to meet the surangular about 

compressed ascending process that fits into 5 mm anterior to the articular glenoid, 

a narrow groove in the squamosal behind As is usual in eucynodonts, the laterally 

the lateral part of the ascending process of exposed postdentary bones are much shal- 

the quadrate. The quadratojugal is separat- lower than in Thrinaxodon (Fig. 5A), with 

ed from the quadrate posteriorly by a thin their lower border sloping up and back, 

descending prong of squamosal. The lower The surangular has less exposure behind 

end of the bone is expanded transversely, the dentary than in Thrinaxodon, but more 

its medial portion overlying the lateral con- than in other eucynodonts, where the den- 

dyle of the quadrate and its lateral portion tary nearly reaches the articular (Figs. 5C, 

forming a free rounded process. 6A, C). On the medial surface of the jaw. 

An incomplete right stapes is preserved the surangular has a flat dorsal surface that 

nearly in situ, its oval footplate separated is buttressed by an overlying ridge on the 

slightly from the depression that houses dentary. The exposed part of the suran- 

the fenestra ovalis. The preserved poste- gular behind the dentary- has a transversely 

rior cms of the stapes extends anterolat- thickened upper margin. Anterolateral to 

erally toward the medial surface of the the articular glenoid is a slightly raised 

quadrate condyle. area that in life may have contacted the 

descending flange of the squamosal; how- 
Lower Jaw ever, it lacks the prominent articular boss 

The right lower jaw is essentially com- that contacts the squamosal in Cynogna- 

plete and well preserved. The large den- thus and Diademodon (Crompton, 1972a). 

tary consists of a deep tooth-bearing hor- The angular covers most of the suran- 

izontal ramus and a broad ascending pro- gular laterally and has a dorsal ridge that 

cess (for insertion of jaw-closing muscles), overhangs its concave outer surface. The 

each forming about one half of its length, reflected lamina is damaged, but it appears 

The short, deep symphysis is fused. The to be more slender than that of Thrinax- 

anteroventral surface of the fused dentar- odon. The articular is transversely narrow- 

ies bears numerous tiny foramina. Two er than in Thrinaxodon, more closely re- 

smafl mental foramina fie below the first sembling that of Probainognathus and 

and second postcanines. At the posterior Probelesodon. The remaining postdentary 

end of the convex lower margin of the elements are similar to those of Thrinax- 

22 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

odon, except that the splenials are fused in ence of an anterior accessory cusp cannot 

the rear of the symphysis. be determined. Both upper and lower po- 

stcanines appear to lack lingual cingula. 

The dental formula is: 14/3, Cl/1, Pc7/ 

Postcranial Skeleton 

5. The incisors are all small and closely The poorly preserved shoulder girdle, 

spaced. The canines are long and broad, forelimb, and caudal vertebrae show no 

with extremely robust roots. The canines unusual features and will not be described, 

have a rounded anterior surface and, in The dorsal vertebral series, although not 

the uppers at least, an unserrated ridge well preserved, merits description inas- 

posteriorly. much as it possesses features that distin- 

The upper postcanines increase in size guish probainognathians from cynognathi- 
from first to sixth, with the seventh being ans. The articulated section of the dorsal 
slightly smaller than the fifth. The first vertebral column contains 10 vertebrae ex- 
three have a slightly recurved main cusp posed in ventral view. On the partially ex- 
and a small posterior accessory cusp. The posed left side, the last two vertebrae show 
fourth is well preserved on the left, where a pair of articulating zygapophyses, thus 
it possesses a large recurved main cusp, a establishing directionality along the col- 
smaller accessory cusp behind it, and a umn. The last nine vertebrae preserve 
second, much smaller, posterior accessory ribs. Of these, the last two possess features 
cusp near the base of the crown. Anterior, that together characterize cynodont lum- 
and slightly internal, to the main cusp is a bar vertebrae (Jenkins, 1971): the rib at- 
very small accessory cusp; this cusp is ab- tachments are entirely on the vertebral 
sent on the right, perhaps obliterated by body, and these ribs are synostosed to the 
wear. Upper postcanine 5 is much larger vertebrae with a serrate suture. In the 
than Pc^, but is nearly identical in mor- more anterior ribs, the capitular articula- 
phology. Upper postcanine 6 has a small tion spans two adjacent centra; whether 
anterointernal cusp, a strongly recurved any are synostosed is uncertain, although 
main cusp, and a smaller recurved poste- the first rib, at least, appears to be free, 
rior accessory cusp. The rear of the crown The anterior four pairs of ribs are dam- 
is damaged, so the presence of a second aged distally, but they appear to be an- 
posterior cusp is uncertain. The damaged teroposteriorly compressed, thus resem- 
seventh postcanine has a recurved main bling typical thoracic ribs. The posterior 
cusp followed by an accessory cusp, but five sets of ribs appear to be short, because 
the presence of additional cusps is uncer- their more or less rounded ends retain 
tain. The teeth are set at a slight angle to some matrix distally. These ribs are per- 
the line of the tooth row, so that, where haps slightly broader than those preceding 
present, the posterior accessory cusp con- them, but they do not expand distally to 
tacts the succeeding crown lingual to its any noticeable degree. The last rib is 
anterior accessory cusp. broader than the preceding ones, as is the 

The lower postcanines are less fully ex- last (second) lumbar vertebra of Cijnog- 

posed. The first tooth is damaged, but the nathus illustrated by Jenkins (1971, fig. 

well-preserved second closely resembles 15A). Also as in Cijnogriathus, the last 

the fourth upper postcanine; both resemble three sets of ribs curve slightly forward, 

a typical lower postcanine of Thrinaxodon. However, at a comparable distance from 

The crowns of Vc^^, are exposed lingually; the proximal synostosis, the posterior ribs 

the third and fourth have an anterior ac- of Liinikuia show no trace of the distal ex- 

cessory cusp and at least one posterior ac- pansions seen in Galesaunis, Thrinaxodon, 

cessoiy cusp, whereas the fifth has two pos- Cynognathus, and Diademodon (Jenkins, 

terior accessory cusps, although tlie pres- 1971). Thus, they resemble the lumbar 

Probainognathian Cynodont From South Africa • Hopson and Kitching 23 

ribs of the probainognathians Probeleso- 
don and Probainognathus (Romer, 1973). 


A cladistic analysis of cynodonts was 
performed, with 23 cynodont taxa and the 
basal therocephalian Lycosuchus and a 
gorgonopsid as successive outgroups (see 
Appendix 2). Of 101 characters, 43 are 
from the skull, 9 from the lower jaw, 29 
from the dentition, and 20 from the post- 
cranial skeleton (see Appendix 1). The 
aims of most past phylogenetic analyses 
have been to order therapsid taxa with re- 
spect to mammals, hence only taxa and 
characters that served to do this were in- 
cluded. We have made a special effort to 
include a large sample of gomphodont taxa 
and to include characters, particularly 
from the postcanine dentition, that would 
specifically aid in resolving their interre- 
lationships. The data were analyzed using 
a random addition sequence with 10 rep- 
licates and the tree bisection— reconnection 
(TBR) algorithm of PAUP 3.1 (Swofford, 
1993), with the resulting character distri- 
bution optimized under delayed transfor- 
mation (DELTRAN). 

Although resolution of the phylogenetic 
relationships of tritylodontids with respect 
to tritheledontids and mammaliaforms is 
not the principal aim of this study, we have 
attempted to deteriuine where these three 
taxa are placed under different treatments 
of the characters. When all characters 
were run unordered, the tritheledontid 
Pachygenelus and the mammaliaform 
Morganucodon were the sister group of 
Tritylodontidae, nested deeply within the 
gomphodont clade (tree length = 233; 
Consistency Index (CI) = 0.58; Retention 
Index (RI) = 0.78; Rescaled CI = 0.45). 
When a minimum of four multistate char- 
acters (18, 22, 63, 73) were ordered, Pach- 
ygenelus and Morganucodon shifted to a 
probainognathian clade, where they re- 
mained under tests of cladogram robust- 
ness (see below). This is the cladogram il- 
lustrated here (Fig. 7). 

The analysis (with four ordered charac- 
ters) resulted in three most parsimonious 
trees of 238 steps (CI = 0.57; RI = 0.77; 
RC = 0.44). The trees differ only in the 
placement of the basal cynodonts Ovinia 
and Procynosuchus with respect to "high- 
er" cynodonts, either in a trichotomy with 
the latter, as their sister clade, or with Pro- 
cynosuchus and Ovinia as their successive 
outgroups. The cladogram (Fig. 7) shows 
the last (our preferred) alternative. 

Three near-basal clades are supported by 
large numbers of unequivocal synapomor- 
phies: Cynodontia by 26, Epicynodontia by 
14, and Eucynodontia by 11. A dichotomy 
witliin Eucynodontia includes a well-char- 
acterized Cynognatliia, witli eight unequiv- 
ocal synapomorphies, and a less well-char- 
acterized Probainognatliia, witli four un- 
equivocal (and two equivocal) synapomor- 
phies. Within Cynognathia, the 
Gomphodontia are characterized by five un- 
equivocal (and two equivocal) synapomor- 
phies. Witliin tlie latter clade, die paraphy- 
letic traversodonts (including Tritylodonti- 
dae as a derived subgroup) are chai'acterized 
by three unequivocal (and one equivocal) 
synapomorphies. Characters diagnosing 
each clade are listed in Appendix 3. 

Lunikuia is the basal member of the Pro- 
bainognatliia, although Ecteninion from the 
early Late Triassic is more derived in but a 
single feature: a frontal— palatine contact in 
the orbital wall. Prohelesodon, Chiniquo- 
don, and Aleodon represent a nionophyletic 
Chiniquodontidae, characterized by a very 
long secondary palate and a posterior an- 
gulation of the maxilla. Probainognathus is 
allied with the tritheledontid/mammali- 
aform clade by two synapomorphies: the 
presence of postcanine lingual cingula and 
a medial shift of die maxillary tooth rows. 
Pachygenehis and Morganucodon form an 
extremely robust clade, supported by 18 
unequivocal synapomorphies. Although this 
clade shares many character states with tri- 
tylodontids, the latter are deeply nested 
within tlie Gomphodontia on the basis of 
numerous cynognathian and gomphodont 

24 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


.r.v^ XO^ 

.^^/ ^A^y 












CYNODONTIA Tree Length: 238 

Consistency Index (CI): 0.57 
Retention Index (Rl): 0.77 
Rescaled CI: 0.44 

Figure 7. Cladogram of nonmammalian cynodonts. One of three shortest trees (238 steps in length), in which Ovinia is sister 
taxon to remaining cynodonts. "Traversodonts" refers to Pascualgnathus and more derived gomphodonts, usually designated 
as Traversodontidae. However, inclusion of Tritylodontidae in this "family" makes it paraphyletic, hence use of the informal term 

The robustness of this cladogram was 
tested by generating trees of incrementally 
greater length (up to six steps longer) to 
determine where specific nodes break 
down. At one step longer (239 steps), in 
the strict consensus of 59 trees, nearly all 
resolution within Eucynodontia breaks 
down, leaving only the grouping of the chi- 
niquodontids Aleodon + Chiniquodon and 
of Pachygenelus + Morganucodon. How- 
ever, the 50% majority-rule consensus tree 
has the same topology as the minimimi- 
length tree. At two steps longer, in the 
50% majority-rule consensus of 286 trees, 
the node between Probainognathus and 
Ecteninion breaks down, yielding a tri- 

chotomy with a chiniquodontid + Pachtj- 
genelus/Morganucodon clade. At three 
steps longer, in the 50% majority-rule con- 
sensus of 1,024 trees, the node between 
Tritylodontidae and "Scalenodon" hir- 
schoni breaks down. At four steps longer 
(3,480 trees), the node between Probele- 
sodon and the remaining chiniquodontids 
breaks down. Only in the 50% majority- 
rule consensus of 30,120 trees that are six 
steps longer than the minimum-length 
tree does the node between Lumkuia and 
the remaining probainognathians break 
down (Fig. 8). A probainognathian clade 
occurs in 75% of these trees and cynog- 
nathian and gomphodont clades both oc- 

Probainognathian Cynodont From South Africa • Hopson and Kitching 25 

Traversodonts' (77) 






Shortest Tree: 238 steps 

Number of steps to collapse node 

Percent of trees retaining node in 
Majority-Rule consensus of 
30,120 trees of < or = 244 steps 

Figure 8. Cladogram of nonmammalian cynodonts. Shortest tree (238 steps) is shown, with numbers In bold italics (2-6) 
indicating the number of steps required to collapse that node, and numbers in bold (54-100) indicating the percentage of trees 
retaining that node in a 50% majority-rule consensus of 30,120 trees of less than or equal to 244 steps (six steps longer than 
minimum-length tree). 

cur in 83% of the trees. The gomphodont 
genera retain the ordering seen in the 
minimum length tree in the great majority 
of trees that are six steps longer. Tritylo- 
dontids and "S. " hirschoni form a trichot- 
omy with the Exaeretodon/Gomphodonto- 
suchiis clade in 54% of these trees. 

In order to determine how parsimoni- 
ous our preferred tree is to that of Rowe 
(1993, fig. 10.2), we used MacClade (Mad- 
dison and Maddison, 1992) to order our 
19 eucynodont taxa in the most parsimo- 
nious tree in which the cynognathian— pro- 
bainognathian dichotomy is not recog- 
nized. This turned out to duplicate the or- 
der of the far fewer taxa in Rowe's clado- 

gram except that tritheledontids, not 
tritylodontids, form a clade with mammal- 
iaforms (Fig. 9A). This tree is 267 steps 
long, 29 steps longer than our preferred 
tree. In the comparison of this tree with 
our preferred tree (Fig. 9B), the distribu- 
tion of the internal carotid foramina is 
shown. Absence of these foramina in the 
basisphenoid is a synapomorphy of Cynog- 
nathia (Fig. 9B). When the cynognathian- 
probainognathian dichotomy is eliminated, 
the distribution of this character becomes 
extremely unparsimonious, with the fo- 
ramina lost, regained, then lost and sub- 
sequently regained again. 

26 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 





□ □ nnnnnnn 

Figure 9. Cladograms constructed using MacClade showing distribution of internal carotid foramina in basispiienoid (character 
26). Absence of carotid foramina is derived (black). (A) Cladogram in which 19 eucynodont taxa are ordered in the most 

Probainognathian Cynodont From South Africa • Hopson and Kitching 27 


Lumkuia fuzzi is a basal member of the 
eucynodont clade Probainognathia. The di- 
chotomy of eucynodonts into Probainog- 
nathia and Cynognathia is supported by this 
analysis, as is the placement of tlie Trity- 
lodontidae widiin the cynognathian sub- 
group Gomphodontia. The sister group re- 
lationship of Tridieledontidae and Mam- 
maliaformes is extremely well supported, 
aldiough their placement in Probainogna- 
thia, rather than as sister to tritylodontids 
widiin Cynognadiia, is less firm, requiring 
the ordering of four characters to achieve 
diis placement. This uncertainty results 
from the extraordinarily large number of 
derived (mammallike) features shared by 
diese three groups, features treated in odi- 
er recent analyses as synapomorphies but 
here shown more likely to be convergences. 
A definitive solution to this phylogenetic 
problem will be found when morphologic 
intermediates between typical Triassic cy- 
nodonts and these primarily Jurassic (and, 
in the case of mammaliaforms, later) taxa 
are incorporated into analyses. Within the 
gomphodonts, Exaeretodon has a very 
mammallike postcranial skeleton and helps 
bridge the morphologic gap between Mid- 
dle Triassic gomphodonts and tritylodon- 
tids. Bonaparte and Barberena (2001) de- 
scribe two Late Triassic carnivorous cyno- 
donts that are also very mammallike post- 
cranially, and that appear to bridge the gap 
between ProbainognatJms and tridieledon- 
tids/mammaliaforms in cranial and dental 
morphology. We believe diese newly de- 
scribed Late Triassic cynodonts will provide 
critical evidence supporting the probain- 
ognathian— cynognathian dichotomy and the 
occurrence of a truly extraordinary amount 
of homoplasy in eucynodont evolution. 


We thank Drs. M. A. Raath and B. S. 
Rubidge for the loan of material; Ms. Clai- 
re Vanderslice for preparation and illustra- 
tion of specimens; and Mr. C. A. Sidor and 
Drs. G. W Rougier, J. R. Wible, and J. A. 
Wilson for help in preparation of the man- 
uscript. Hopson's research was supported 
by National Science Foundation grants 
BSR 86-15016 and 89-06619. We are 
grateful to Dr. F. A. Jenkins, Jr., and his 
co-organizers of the symposium for pro- 
viding this opportunity to honor Fuzz 
Crompton. Hopson also \\dshes to express 
his pleasure at attending the symposium 
and enjoying the fellowship of good 
friends and colleagues. 


Bonaparte, J. R, and M. C. Barberena. 2001. On 
two advanced carnivorous cynodonts from tlie 
Late Triassic of Southern Brazil. Bulletin of tlie 
Museum of Comparative Zoology, 156: 59—80. 

Broom, R. 190.5. On the use of the term Anomodon- 
tia. Records of the Albany Museum, 1: 266-269. 

Crompton, A. W. 1972a. The evolution of the jaw 
articulation of cynodonts, pp. 231-253. In K. A. 
Joysey and T. S. Kemp (eds.), Studies in Verte- 
brate Evolution. Edinburgh: Oliver and Boyd. 
284 pp. 

. 1972b. Postcanine occlusion in cynodonts 

and tritylodontids. Bulletin of the British Muse- 
um (Natural Histoiy) Geology, 21: 29-71. 

Crompton, A. W., and F. Ellenberger. 1957. On 
a new cynodont from the Molteno Beds and the 
origin of the tritylodontids. Annals of the South 
African Museum, 44: 1-14. 

ESTES, R. 1961. Cranial anatomy of Thrinaxodon lior- 
hinus. Bulletin of the Museum of Comparative 
Zoolog>', Harvard, 125: 165-180. 

Hancox, R J., AND B. S. Rubidge. 1997. The role 
of fossils in interpreting die development of the 
Karoo Basin. Palaeontologia Africana, 33: 41-54. 

Hancox, R J., M. A. Shishkin, B. S. Rubidge, and 
J. W. Kitching. 1995. A threefold subdivision of 
the Cijnognathus Assemblage Zone (Beaufort 
Group, South Africa) and its palaeogeographical 

parsimonious tree that does not recognize a probainognathian-cynognathian dichotomy. Tree length is 267 steps, 29 steps 
longer than tree shown in (B), in which eucynodont taxa are ordered in the most parsimonious tree, determined from PAUP 
analysis (see Fig. 8). In this tree, the node-based Eucynodontia comprises the stem-based Probainognathia and Cynognathia. 
Absence of carotid foramina is a synapomorphy of Cynognathia. In tree A, CI = 0.51, Rl = 0.71, RC = 0.36. In tree B, CI = 
0.57, Rl = 0.77, RC = 0.44. 


Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

implications. South African Journal of Science, 
91: 143-144. 

HOPSON, J. A. 1966. The origin of the mammalian 
middle ear. American Zoologist, 6: 437—450. 

. 1990. Cladistic analysis of therapsid relation- 
ships. Journal of Vertebrate Paleontology, 10(3, 
Suppl.): 28A. 

. 1991a. Convergence in mammals, trithele- 

donts, and tritylodonts. Journal of Vertebrate Pa- 
leontology, 11(3, Suppl.): 36A. 

. 1991b. Systematics of the nonmammalian 

Synapsida and implications for patterns of evo- 
lution in synapsids, pp. 635—693. In H.-P Schul- 
tze and L. Trueb (eds.). Origins of the Higher 
Groups of Tetrapods: Controversy and Consen- 
sus. Ithaca, New York: Comstock Publishing As- 
sociates, a Division of Cornell University Press, 
xii + 724 pp. 

1994. Synapsid evolution and the radiation 

of non-therian mammals, pp. 190—219. In D. R. 
Prothero and R. M. Schoch (eds.). Major Fea- 
tures of Vertebrate Evolution, Short Courses in 
Paleontology, No. 7. Knoxville: University of 
Tennessee. 270 pp. 

analysis of therapsid relationships, pp. 83-106. In 
N. Hotton III, P. D. MacLean, J. J. Roth, and E. 
C. Roth (eds.). The Ecology and Biology of 
Mammal-Like Reptiles. Washington, D.C.: 
Smithsonian Institution Press, x + 326 pp. 

HOPSON, J. A., AND J. W. KiTCHING. 1972. A revised 
classification of cynodonts. Palaeontologia Afri- 
cana, 14: 71-85. 

. 1988. A Chiniquodon-like cynodont from the 

Early Triassic of South Africa and the phylogeny 
of advanced cynodonts. Journal of Vertebrate Pa- 
leontology, 8(3, Suppl.): 18A. 

Jenkins, F. A., Jr. 1971. The postcranial skeleton of 
African cynodonts. Bulletin of the Peabody Mu- 
seum, Yale University, 36: 1—216. 

Kemp, T S. 1982. Mammal-Like Reptiles and the Or- 
igin of Mammals. London: Academic Press, xiv 
+ 363 pp. 

. 1983. The relationships of mammals. Zoo- 
logical Journal of the Linnean Society, 77: 353— 

. 1988. Interrelationships of the Synapsida, pp. 

1-22. In M. J. Benton (ed.). The Phylogeny of 
the Tetrapods, Vol. 2: Mammals, Systematics As- 
sociation Special Volume 35B. Oxford: Claren- 
don Press. 323 pp. 

Lucas, S. G., and Z. Luo. 1993. Adelobasileus from 
the Upper Triassic of West Texas: the oldest 
mammal. Journal of Vertebrate Paleontology, 13: 

Maddison, W p., and D. R. Maddison. 1992. 
MacClade, Version 3.01. Sunderland, MA: com- 
puter program distributed by Sinauer Associates, 

Martinez, R. N., C. L. May, and C. A. Forster. 

1996. A new carnivorous cynodont from the Is- 
chigualasto Formation (Late Triassic, Argentina), 
with comments on eucynodont phylogeny. Jour- 
nal of Vertebrate Paleontology, 16: 271-284. 

Owen, R. 1861. Palaeontology, or a Systematic Sum- 
mary of Extinct Animals and Their Geological 
Relations, second edition. Edinburgh: Adam and 
Black, xvi + 463 pp. 

ROMER, A. S. 1969. The Chaiiares (Argentina) Tri- 
assic reptile fauna. V. A new chiniquodontid cy- 
nodont, Probelesodon lewisi — cynodont ancestry. 
Breviora, 333: 1-24. 

. 1970. The Chanares (Argentina) Triassic rep- 
tile fauna. VI. A chiniquodontid cynodont with 
an incipient squamosal-dentary jaw articulation. 
Breviora, 344: 1-18. 

. 1973. The Chaiiares (Argentina) Triassic rep- 
tile fauna. XIX. Postcranial materials of the cy- 
nodonts Probelesodon and Probainognathus. 
Breviora, 407: 1-26. 

1992. Reconstruction of the cranial vessels the 
early Cretaceous mammal Vincelestes neuqueni- 
aniis: implications for the evolution of the mam- 
malian cranial vascular system. Journal of Ver- 
tebrate Paleontology, 12: 188-216. 

ROWE, T. 1986. Osteological diagnosis of Mammalia, 
L. 1758, and its relationship to extinct Synapsida. 
Ph.D. dissertation. Berkeley: University of Cali- 
fornia. 446 pp. 

. 1988. Definition, diagnosis, and origin of 

Mammalia. Journal of Vertebrate Paleontology, 
8: 241-264. 

. 1993. Early mammal phylogenetic system- 
atics, pp. 129-145. In F. S. SzaJay, M. J. Novacek, 
and M. C. McKenna (eds.), Mammal Phylogeny: 
Mesozoic Differentiation, Multituberculates, 
Monotremes, Early Therians, and Marsupials. 
New York: Springer- Verlag. x -I- 249 pp. 

Sereno, p. C. 1999. Definitions in phylogenetic tax- 
onomy: critique and rationale. Systematic Biolo- 
gy, 48: 329-351. 

SidoR, C. a., and J. A. HOPSON. 1998. Ghost hne- 
ages and "mammalness": assessing the temporal 
pattern of character acquisition in the Synapsida. 
Paleobiology, 24: 254-273. 

Sues, H.-D. 1985. The relationships of the Tritylo- 
dontidae (Synapsida). Zoological Journal of the 
Linnean Society, 85: 205-217. 

Swofford, D. L. 1993. PAUP: Phylogenetic Analysis 
Using Parsimony, Version 3.1. Champaign, Illi- 
nois: computer software and documentation dis- 
tributed by Illinois Natural History Survey. 

WlBLE, J. R. 1991. Origin of Mammalia: the craniod- 
ental evidence reexamined. Journal of Vertebrate 
Paleontology, 11: 1-28. 

WlBLE, J. R., AND J. A. HOPSON. 1995. Homologies 
of the prootic canal in mammals and non-mam- 
malian cynodonts. Journal of Vertebrate Paleon- 
tology, 15: 331-356. 

Probainognathian Cynodont From South Africa • Hopson and Kitching 29 


States are denoted as (0) = primitive 
state; (1), (2), and (3) = derived states. 


1 . Premaxilla forms posterior border in- 
cisive foramen: absent (0), present 


2. Nasal-lacrimal contact: absent (0), 

present (1). 

3. Prefrontal: present (0), absent (1). 

4. Postfrontal: present (0), absent (1). 

5. Postorbital: present (0), absent (1). 

6. Prefrontal-postorbital contact: ab- 
sent (0), present (1). 

7. Parietal foramen: present (0), absent 


8. Vomer intemarial shape: broad plate 

(0), parallel-sided keel (1). 

9. Ectopterygoid: contacts maxilla (0), 
does not contact maxilla (1), absent 


10. Interpterygoid vacuity in adult be- 
tween pterygoid flanges: present (0), 
absent (1). 

11. Palatal exposure of maxifla behind 
canine greater than 20% distance 
from canine to posterior end of pal- 
atine: absent (0), present (1). 

12. Secondary palatal plate on maxilla: 
absent (0), present, does not reach 
midline (1), present, reaches midline 


13. Secondary palatal plate on palatine: 
absent or low ridge (0), present, ex- 
tends nearly to midline (1), present, 
reaches midline (2). 

14. Length secondary palate relative to 
toothrow: shorter (0), about equal 
(1), longer (2). 

15. Length secondary palate relative to 
anterior border of orbit: shorter (0), 
about equal (1), longer (2). 

16. Teeth on pterygoid flange: present 
(0), absent (1). 

17. Ventral surface of basisphenoid de- 
pressed below occipital condyles: less 
than Va occipital height (0), greater 
than Va occipital height (1). 

18. Zygomatic arch dorsoventral height: 
slender (0), moderately deep (1), 
very deep (2). 

19. Zygomatic arch dorsal extent: below 
middle of orbit (0), above middle of 
orbit (1). 

20. Jugal depth in zygomatic arch rela- 
tive to exposed squamosal depth: less 
than twice (0), greater than twice (1). 

21. Jugal suborbital process: absent (0), 
present (1). 

22. Squamosal groove for external audi- 
tory meatus: shallow (0), moderately 
deep (1), very deep (2). 

23. Frontal-palatine contact in orbit: ab- 
sent (0), present (1). 

24. Tabular extends around posttemporal 
foramen: absent (0), present (1). 

25. Descending flange of squamosal lat- 
eral to quadratojugal: absent (0), pre- 
sent not contacting surangular (1), 
present contacting surangular (2). 

26. Internal carotid foramina in basi- 
sphenoid: present (0), absent (1). 

27. Groove on prootic extending from 
pterygoparoccipital foramen to tri- 
geminal foramen: absent (0), present 
and open (1), present and enclosed 
as a canal (2). 

28. Trigeminal nerve exit: between pro- 
otic incisure and epipterygoid (0), via 
foramen between prootic and epip- 
terygoid (1), via two foramina (2). 

29. Quadrate contact: primarily squa- 
mosal (0), primarily crista parotica 


30. Quadrate ramus of pterygoid: pre- 
sent (0), absent (1). 

31. Quadrate posteroventral process in 
squamosal posterior notch: absent 
(0), present (1). 

32. Epipteiygoid ascending process at 
level of trigeminal foramen: rodlike 
(0), moderately expanded (1), greatly 
expanded (2). 

33. Epipterygoid-prootic overlap: absent 
(0), present (1). 

34. Lateral flange of prootic: absent (0), 
present (1). 

30 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

35. Epipteiygoid— frontal contact: absent 
(0), present (1). 

36. Separate foramina for vestibular and 
cochlear nerves: absent (0), present 


37. Double occipital condyles: absent 
(0), present (1). 

38. Stapedial foramen orientation: an- 
teroposterior (0), dorsoventral (1). 

39. Greatest width of zygomatic arches: 
near middle of arch (0), at posterior 
end of arch (1). 

40. Length of palatine relative to maxilla 
in secondary palate: shorter (0), 
about equal (1), longer (2). 

41. Posterolateral end of maxilla: passes 
obliquely posterodorsally into subor- 
bital bar (0), forms right angle ventral 
to jugal contact (1). 

42. Fenestra rotunda separation from 
jugular foramen: confluent (0), par- 
tially separated by fingerlike projec- 
tion from posterolateral wall of jug- 
ular foramen (1), completely separat- 
ed (2). 

43. V-shaped notch separating lambdoi- 
dal crest from zygomatic arch: absent 
(0), present (1). 

Lower Jaw 

44. Dentaiy symphysis: not fused (0), 
fused (1). 

45. Dentary masseteric fossa: absent (0), 
high on coronoid region (1), extends 
to lower border of dentary (2). 

46. Dentary overlap of dorsal surface of 
surangular: short (0). long (1). 

47. Dentaiy coronoid process height: be- 
low middle of orbit (0), above middle 
of orbit (1). 

48. Position of dentary— surangular dorsal 
contact relative to postorbital bar and 
jaw joint: closer to postorbital bar (0), 
midway between (1), closer to jaw 
joint (2). 

49. Postdentary rod height relative to ex- 
posed length (distance between base 
of reflected lamina and jaw joint): 
greater than V2 length (0), about Vi 
length (1), less than V2 length (2). 

50. Coronoid mediolaterally thickening: 
absent (0), present (1). 

51. Reflected lamina of angular posterior 
extent relative to distance from angle 
of dentary to jaw joint: greater than 
V2 the distance (0), less than V2 the 
distance (1). 

52. Reflected lamina of angular shape: 
deep corrugated plate (0), spoon- 
shaped plate (1), hook with depth 
greater than V2 length (2), hook with 
depth less than V2 length (3). 


53. Upper incisor number: five or more 
(0), four (1), three (2). 

54. Lower incisor number: four or more 
(0), three (1), two (2). 

55. Incisor cutting margins: serrated (0), 
smoothly ridged (1), denticulated (2). 

56. Incisor size: all small (0), some or all 
enlarged (1). 

57. Upper canine size: large (0), reduced 
in size (1), absent (2). 

58. Lower canine size: large (0), reduced 
in size (1), absent (2). 

59. Canine serrations: present (0), absent 


60. Postcanine shape: single point (0), 
two or more cusps in line (1). 

61. Upper postcanine buccal cingulum: 
absent (0), present (1). 

62. Postcanine lingual cingulum: absent 
(0), narrow (1), lingually expanded 

63. Number of upper cusps in transverse 
row: one (0), two (1), three or more 

64. Position of upper transverse cusp 
row on crown: on anterior half of 
crown (0), midcrown almost to pos- 
terior margin (1), at posterior margin 
(no posterior cingulum) (2). 

65. Central cusp of upper transverse 
row: absent (0), midway between 
buccal and lingual cusps (1), closer to 
lingual cusp (2). 

66. Longitudinal shear surface of main 
upper cusp: anterior and posterior 

Probainognathian Cynodont From South Africa • Hopson and Kitching 31 

(to transverse ridge) (0), posterior 
only (1), anterior only (2). 

67. Upper anterobuccal accessory cusp: 
present (0), absent (1). 

68. Upper posterobuccal accessory cusp: 
present (0), absent (1). 

69. Upper anterolingual accessory cusp: 
absent (0), present (1). 

70. Upper anterior transverse (cingulum) 
ridge: low (0), high (1). 

71. Upper lingual ridge: absent (0), pre- 
sent (1). 

72. Transverse axis of crown strongly 
oblique to midline axis: absent (0), 
present (1). 

73. Number of lower cusps in transverse 
row: one (0), two (1), three or more 

74. Lower anterior cingulum or cusp: ab- 
sent (0), present (1). 

75. Lower posterior basin: absent (0), 
present (1). 

76. Widest lower cusp in transverse row: 
lingual (0), buccal (1). 

77. Posterior portion maxillary tooth row 
inset from lateral margin of maxilla 
(cheek developed): absent (0), pre- 
sent (1). 

78. Axis of posterior part of maxillary 
tooth row: directed lateral to subtem- 
poral fossa (0), directed toward cen- 
ter of fossa (1), directed toward me- 
dial rim of fossa (2). 

79. Posterior portion of maxillary tooth 
row extends medial to temporal fos- 
sa: absent (0), present (1). 

80. Posteriormost postcanine(s) gompho- 
dont: absent (0), present (1). 

81. Postcanine replacement pattern in 
adult: "alternating" (0), widely 
spaced waves (three or more teetlV 
wave) (1), single wave (2). 


82. Expanded costal plates on ribs: ab- 
sent (0), present (1). 

83. Lumbar costal plates with ridge over- 
lapping preceding rib: absent (0), 
present (1). 

84. Scapula infraspinous fossa with out- 

turned anterior and posterior bor- 
ders: absent (0), present (1). 

85. Acromion process: absent (0), pre- 
sent (1). 

86. Scapular constriction below acromi- 
on: absent (0), present (1). 

87. Scapular elongation between acro- 
mion and glenoid: absent (0), present 


88. Procoracoid in glenoid: present (0), 
barely present or absent (1). 

89. Procoracoid contact with scapula: 
greater than coracoid contact (0), 
equal to or less than coracoid contact 


90. Humerus ectepicondylar foramen: 
present (0), absent (1). 

91. Ulna olecranon process: absent (un- 
ossified) (0), present (1). 

92. Manual digit III phalanx number: 
four (0), three (1). 

93. Manual digit IV phalanx number: five 
(0), four (1), three (2). 

94. Length of anterior process of ilium 
anterior to acetabulum (relative to 
diameter of acetabulum): less than 
1.0 (0), 1.0-1.5 (1), greater than 1.5 

95. Length of posterior process of ilium 
posterior to acetabulum: (relative to 
diameter of acetabulum): between 
0.5 and 1.0 (0), greater than 1.0 (I), 
less than 0.5 (2). 

96. Dorsal profile of ilium: strongly con- 
vex (0), flat to concave (1). 

97. Total length of pubis relative to ace- 
tabulum diameter: greater than 1.5 
(0), between 1.5 and 1.0 (1), less than 
1.0 (2). 

98. Greater trochanter separated from 
femoral head by distinct notch: ab- 
sent (0), present (1). 

99. Greater trochanter joined to femoral 
head by ridge: present (0), absent 


100. Lesser trochanter position: on ven- 
tromedial surface of femoral shaft 
(0), on medial surface of femoral 
shaft (1). 

101. Vertebral centra: amphicoelous (0), 
platycoelous (1). 

32 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

States are denoted as (primitive); and 1, 2, or 3 (derived). ? = state unknown. 








































































Scalenodon angustifrons 






''Scalenodon" hirschoni 














































































Mo rga n ucodon 







Numbers refer to characters in Appendix 
1. Numbers in parentheses refer to equiv- 
ocal synapomorphies under the Delayed 
Transformation (DELTRAN) option of 





Nasal— lacrimal contact. 
Rostfrontal absent. 
Rrefrontal-postorbital contact. 
Ectopterygoid does not contact max- 

Palatal exposure of maxilla behind 
canine greater than 20% distance 
from canine to posterior end of pal- 

Secondary palatal plate on maxilla. 
Secondary palatal plate on palatine. 



16. Teeth on pterygoid flange absent. 
24. Tabular extends around posttem- 

poral foramen. 

Groove on prootic extending from 

pterygoparoccipital foramen to tri- 
geminal forainen. 

Trigeminal nerve exit via foramen 

between prootic and epipterygoid. 
31. Posteroventral process on quadrate 

in posterior notch of squamosal. 
(32). Ascending process of epipterygoid 

greatly expanded. 

Epipterygoid— prootic overlap. 

Lateral flange of prootic. 

Epipterygoid— frontal contact. 

Double occipital condyles. 

Stapedial foramen with dorsoven- 

tral orientation. 

Dentaiy masseteric fossa present 

high on coronoid region. 



Probainognathian Cynodont From South Africa • Hop.son and Kitching 33 















1707777 70? 


012 77777?? 
7771772 777 

















Dentary overlap of surangular long. 
Reflected lamina of angular spoon- 
shaped plate. 

Incisor cutting margins smoothly 

Canine serrations absent. 
Postcanines with two or inore cusps 
in line. 

Lower anterior cingulum or cusp 

Length anterior process of ilium 
1.0-1.5 times diameter of acetabu- 

Length of pubis between 1.5 and 
1.0 times acetabular diameter. 


10. Interpterygoidal vacuity between 

pterygoid flanges absent in adult. 
18. Zygomatic arch moderately deep. 

22. Groove for external auditory meatus 
moderately deep. 

25. Descending flange of squamosal lat- 
eral to quadratojugal present. 

45. Masseteric fossa extends to lower 
border of dentary. 

47. Coronoid process of dentaiy ex- 
tends above middle of orbit. 

48. Dentary-surangular dorsal contact 
midway between postorbital bar 
and jaw joint. 

49. Height of postdentary rod about 
one half the length of the laterally 
exposed portion of the rod (distance 
between base of reflected lamina 
and jaw joint). 

52. Reflected lamina of angular hook- 
shaped, with depth greater than one 
half its length. 

53. Four upper incisors. 

34 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

54. Three lower incisors. 

82. Expanded plates on ribs. 

95. Lengdi of posterior process of ilium 
greater than diameter of acetabu- 

97. Length of pubis less than diameter 
of acetabulum. 


25. Descending flange of squamosal lat- 
eral to quadratojugal contacts sur- 

30. Quadrate ramus of pterygoid ab- 

44. Dentary symphysis fused. 

48. Dentary-surangular dorsal contact 
closer to jaw joint that to postorbital 

49. Height of postdentaiy rod less than 
one half the length of the laterally 
exposed portion of the rod. 

51. Reflected lamina of angular less 
than one half the distance from an- 
gle of the dentary to jaw joint. 

52. Reflected lamina of angular hook- 
shaped, with depth less than one 
half its length. 

81. Postcanine replacement pattern of 
widely spaced waves (three or more 
teeth per wave). 

85. Acromion process on scapula. 

92. Manual digit III with three phalan- 

93. Manual digit IV with three phalan- 


7. Parietal foramen absent. 
(9). Ectopterygoid absent. 
15. Rear of secondary palate lies below 
anterior border of orbit. 
(43). V-shaped notch separates lambdoi- 
dal crest from zygoma. 

82. Expanded plates on ribs absent. 
88. Procoracoid barely present in or 

absent from glenoid. 

Probainognathia Minus Lumkuia 

23. Frontal contacts palatine in orbital 

Probainognathia Minus Lumkuia, Ectini- 

(1). Premaxillae form posterior border 

of incisive foramina. 
14. Secondary palate about equal in 
length to tooth row. 

(42). Fenestra rotunda partially separat- 
ed from jugular foramen by finger 
like projection. 

(86). Scapula constricted below acromi- 
on process. 

(94). Anterior process of ilium anterior 
to acetabulum greater than 1.5 
times acetabular diameter. 

(96). Dorsal profile of ifium flat to con- 

Probainognathus, Pachygenelus, and Mor- 

62. Narrow postcanine lingual cingu- 

78. Axis of posterior part of maxillary 

tooth row directed toward center of 

temporal fossa. 

Pachygenelus and Morganucodon. 

3. Prefrontal absent. 

5. Postorbital absent. 

6. Prefrontal-postorbital contact ab- 

(14). Secondary palate longer than tooth- 
(15). Secondary palate extends posterior 
to anterior border of orbit. 

18. Zygomatic arch slender. 

22. Squamosal groove for external au- 
ditoiy meatus shallow. 

25. Descending flange of squamosal 
lateral to quadratojugal absent. 

36. Separate foramina in petrosal for 
vestibular and cochlear nerves. 

42. Fenestra rotunda completely sepa- 
rated from jugular foramen. 

44. Dentary symphysis not fused. 

50. Coronoid mediolaterally thickened. 

58. Lower canine reduced in size. 

61. Upper postcanines with buccal cin- 

Probainognathian Cynodont From South Africa • Hopson and Kitching 35 

87. Scapula elongated between acro- 
mion and glenoid. 
(89). Procoracoid contact with scapula 
equal to or less than coracoid con- 

91. Ulnar olecranon process present. 

95. Posterior process on ilium less than 
one half diameter of acetabulum. 

98. Greater trochanter separated from 
femoral head by deep notch. 

100. Lesser trochanter on medial sur- 
face of femoral shaft. 

101. Vertebral centra platycoelous. 


(15). Secondary palate extends posterior 
to anterior border of orbit. 
41. Posterolateral end of maxilla forms 
right angle ventral to jugal contact. 


18. Zygomatic arch very deep. 

19. Zygomatic arch extends above mid- 
dle of orbit. 

21. Suborbital process on jugal. 

22. Groove for external auditoiy mea- 
tus very deep. 

26. Basisphenoid internal carotid fo- 
ramina absent. 

39. Greatest width of zygomatic arches 
at posterior end of arch. 

59. Canine serrations present. 

83. Lumbar costal plates with ridge 
overlapping preceding rib. 


20. Depth of jugal in zygomatic arch 
greater than twice that of exposed 
part of squamosal. 

28. Trigeminal nerve exit via two fo- 

(43). V-shaped notch separates lambdoi- 
dal crest from zygoma. 

62. Upper postcanine internal cingu- 
luin lingually expanded. 

63. Upper postcanines with three cusps 
in transverse row. 

73. Lower postcanines with two cusps 
in transverse row. 
(96). Dorsal profile of ilium flat to con- 

Traversodonts (Incl. Tritylodontidae) 

75. Posterior basin on lower postcani- 

(78). Axis of posterior part of maxillary 
tooth row directed toward medial 
rim of subtemporal fossa. 

81. Adult postcanine replacement pat- 
tern consists of single wave. 

86. Scapula constricted below acromi- 
on process. 

''Scalenodon" hirschoni + Tritylodontidae. 

(53). Three upper incisors. 

54. Two lower incisors. 
(56). Some or all incisors enlarged. 

Gomphodontosuchus + Exaeretodon 

63. Two cusps in transverse row on up- 
per postcanines. 

65. Central cusp of upper transverse 
row absent. 

70. High anterior transverse ridge on 
upper postcanines. 

72. Transverse axis of postcanine 
crowns strongly oblique to midline 
(74). Lower anterior cingulum or cusp 

76. Widest lower cusp in transverse 
row buccal. 



Abstract. Three incomplete dentaries with teeth 
and several isolated postcanine teeth of a small cy- 
nodont synapsid from tlie Upper Triassic (Camian) 
Tomahawk Member of the Vinita Fonnation of the 
Richmond basin (Newark Supergroup) in Virginia are 
referable to Microconodon tenuirostris Osbom, 1886. 
This taxon was previously known only from a single 
specimen, an incomplete right dentary with four post- 
canine teeth from the Upper Triassic (Camian) Cum- 
nock Formation of North Carolina. Once considered 
one of the earliest and most primitive mammals, Mi- 
crocoiiodon is a derived eucynodont of uncertain af- 
finities. Its more posterior postcanine teeth have 
three or four anteroposteriorly aligned cusps, lack 
cingula, and the roots of some postcanines are incip- 
iendy divided. 


Emmons (1857) named Dromatherium 
sylvestre on the basis of three small tooth- 
bearing jaws from Late Triassic coals (ini- 
tially thought to be Permian in age) in the 
Chatham coal field of central North Car- 
olina. He interpreted these fossils as the 
oldest known mammalian remains. Dro- 
rtiatheriuni quickly became widely estab- 
lished as the first reputed American Me- 
sozoic mammal and as the oldest mammal 
known at that time (e.g., Owen, 1871). Os- 
bom (1886a) restudied two of the jaws; he 
could not trace the repository for the third 
specimen mentioned by Emmons, which 
was presumed lost. Osbom recognized the 
distinctive nature of one of the dentaries, 
housed in the collections of the Academy 
of Natural Sciences of Philadelphia (ANSP 
10248), and made it the holotype of a new 
genus and species, Microconodon teniii- 

'- Department of Palaeobiology, Royal Ontario Mu- 
seum, 100 Queens Park, Toronto, Ontario M5S 2C6, 

rostris (see also Osbom, 1886b, 1887). Un- 
aware of Osbom s work, Gillette (1978) 
still listed the holotype of M. tenuirostris 
as a "syntype" of D. sylvestre. 

During his survey of all Mesozoic mam- 
mals then known, Simpson (1926a,b) re- 
examined the holotypes of D. sylvestre and 
M. tenuirostris. He removed both taxa 
from the Mammalia and referred them to 
the Cynodontia, as had first been suggest- 
ed by Seeley (1895). In support of his re- 
assessinent, Simpson cited evidence for 
the presence of more than one bone in the 
lower jaw, the undivided roots of the lower 
postcanine teeth, and the resemblance of 
the crowns of these teeth to those of cer- 
tain nonmammalian cynodonts. The first 
and third features are phylogenetically un- 
informative because they are plesiomorph- 
ic features. The second character is of 
questionable significance because, as 
Simpson (1926b) himself observed, the 
roots of the more posterior postcanines in 
the holotype of M. tenuirostris are incipi- 
ently divided. Since Simpson's redescrip- 
tion, little attention has been paid to these 
fossils, although doubts concerning their 
phylogenetic position have persisted to the 
present day. Hopson and Kitching (1972) 
classified Dromatherium and Microcono- 
don as Cynodontia incertae sedis, but ex- 
plicitly noted possible mainmalian affini- 
ties for both taxa. Most recently, Hahn et 
al. (1994) placed both forms, together with 
several other problematic taxa of Late Tri- 
assic nonmammalian cynodonts in a family 
Dromatheriidae, which they considered 
the sister-taxon of Mainmalia. 

The holotypes of D. sylvestre and M. 

Bull. Mus. Comp. ZooL, 156(1): 37-48, October, 2001 37 

38 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

tenuirostris were collected from coal 
seams near the base of the Cumnock For- 
mation, of late Carnian age (Litwin and 
Ash, 1993), in a now abandoned mine near 
Egypt, Chatham County, North Carolina. 
The strata belong to the Sanford subbasin 
of the Deep River basin of the Newark 
Supergroup. Simpson (1926b) provided a 
meticulous description of these specimens, 
and little can be added to his account. 
Both fossils have been adversely affected 
by early attempts at preparation and con- 
servation. The holotype of D. sylvestre was 
originally housed in the Museum of Wil- 
liams College but has recently been per- 
manently transferred to the collections of 
the National Museum of Natural History. 
The postcanine teeth of this specimen 
have been badly damaged since the time 
of Simpson s study, and most details of his 
account can no longer be verified. Dro- 
matheriinn sylvestre is quite different from 
M. tenuirostris in most comparable fea- 
tures (Simpson, 1926a). The holotype of 
M. tenuirostris (ANSP 10248) holds fewer 
teeth than the holotype of D. sylvestre, but 
the postcanines of M. tenuirostris are bet- 
ter presei'ved. 

Three partial dentaries with teeth and 
several isolated postcanine teeth of a small 
cynodont from the Tomahawk Member of 
the Vinita Formation (Turkey Branch For- 
mation sensu Cornet and Olsen [1990]) in 
the Richmond basin (Newark Supergroup) 
of Virginia (Sues and Olsen, 1990; Sues et 
al., 1994) closely resemble the holotype of 
M. tenuirostris in all comparable charac- 
ters. They appear to be referable to the 
same taxon and exliibit significant addi- 
tional anatomical detail. The new dentaries 
also represent different ontogenetic stages. 
The purpose of this paper is to describe 
and illustrate this material and to assess 
the affinities of M. tenuirostris within the 
phylogenetic framework provided by re- 
cent hypotheses of cynodont interrelation- 
ships (Rowe, 1988; Battail, 1991; Hopson, 
1991). I will also briefly review the status 
of several possibly related but poorly 
known cynodont taxa from the Upper Tri- 

Table 1. Measurements (in mm) for the an- 





USNM 437637 



ROM 44300 










ROM 44301 










assic of central and western Europe (Hu- 
ene, 1933; Peyer, 1956; Clemens, 1980; 
Hahn et al., 1984, 1987, 1994; Godefroit 
and Battail, 1997), New Mexico (Lucas 
and Oakes, 1988), and Brazil (Bonaparte 
and Barberena, 1975, 2001). 

The following abbreviations for institu- 
tional names preceding catalogue numbers 
are used in this paper: ANSP, Academy of 
Natural Sciences of Philadelphia; ROM, 
Royal Ontario Museuin, Toronto; USNM, 
National Museum of Natural History (for- 
merly United States National Museum), 
Washington, D.C. 

All dental measurements (Table 1) were 
made with a graded ocular scale on a Ni- 
kon SMZU stereoscopic microscope; each 
measurement was repeated three times. 


Monophtjletic Hierarchy. Amniota: Syn- 
apsida: Therapsida: Cynodontia: 
Eucynodontia incertae sedis. 

Genus M icroconodon Osborn, 1886 

Dromatherium Emmons, 1857: 93 (in part) 
Microconodon Osbom, 1886a: 540 
Ttjtthoconiis Palmer, 1903: 873 (objective junior syn- 

Type Species. Microconodon tenuirostris 
Osborn, 1886 (by monotypy). 

Diagnosis. Dentary with very slender 
horizontal ramus. Angular region of den- 
tary without distinct process. Posterior 
postcanine teeth with three or four an- 
teroposteriorly aligned cusps. Postcanines 


Figure 1. Microconodon tenuirostris, USNM 437637, left dentary (with splenial and attached symphyseal fragment of right 
dentary) in lingual view. Scale bar = 2 mm. Abbreviations: an, angle of dentary; ar.p, articular process of dentary; c, canine 
alveolus; co.p, coronoid process; co.r, coronoid ridge; f.sp., articular facet for splenial; g.l, groove for dental lamina; i1-3, alveoli 
for incisors 1-3; i1r, alveolus for right first incisor; i.g, internal mandibular groove; m, mental foramen; p1-8, postcanine 1-8 
(tooth or alveolus); r, pit for replacement tooth; sp, splenial; t?, tooth fragment. 

without cingula. Root of some postcanine 
teeth constricted, with figure-eight shape 
in transverse section. 

Comment. Pahner (1903) regarded Mi- 
croconodon Osbom, 1886 as preoccupied 
by Microconodiis Traquair, 1877 and pro- 
posed Tytthoconus as a replacement name. 
Pahner's action is invalid under the rules 
of the International Code of Zoological 
Nomenclature, and Tytthoconus Palmer, 
1903 is an objective junior synonym of Mi- 
croconodon Osbom, 1886. 

Microconodon tenuirostris Osbom, 1886 

Dromatherium sijlvestre Emmons, 1857: 93 (in part) 
Microconodon tenuirostris Osbom, 1886a: 540 

Holotype. ANSP 10248, right dentary 
with four preserved postcanine teeth, lack- 
ing the articular process and part of the 
coronoid process (Simpson, 1926b). The 
specimen is preserved on a slab of coal, 
and preservation of most structural details 
is indistinct. 

Type Horizon and Locality. Basal coals 
of the Cumnock Formation, Sanford sub- 
basin of the Deep River basin, Newark Su- 
pergroup; coal mine (now abandoned) at 
Egypt, Chatham County, North Carolina. 
Age: Late Triassic (late Camian; Litwin 
and Ash, 1993). 

Newly Referred Material. USNM 

437637, incomplete left dentary lacking 
much of the articular and coronoid pro- 
cesses, with alveoli for three incisors and 
the canine, seven mostly broken postca- 
nine teeth, and attached left splenial as 
well as fragment of the symphyseal end of 
the right dentary (Fig. 1). ROM 44300, in- 
complete left dentary with alveoli for one 
incisor and the canine, three postcanine 
teeth, and basal portions of four postcan- 
ines; mandibular symphysis for the most 
part preserved only as an impression (filled 
in with epoxy resin during preparation) 
and posterior portion of dentary broken 
and displaced anterolaterally (Fig. 2). 
ROM 44301, anterior portion of right den- 
tary with alveolus for canine, five pre- 
serN^ed postcanine teeth, and alveoli for 
three postcanines (Fig. 3). USNM 448579, 
isolated complete postcanine tooth. 
USNM 448600 (Sues et al., 1994, fig. 8.4) 
and ROM 44302, isolated postcanine teeth 
with most of the root broken off. 

Horizon and Locality of Newly Referred 
Material. Tomahawk Member of Vinita 
Formation (Turkey Branch Formation sen- 
sii Comet and Olsen [1990]), Richmond 
basin, Newark Supergroup; USNM locality 
39981, 0.16 km (0.1 miles) east of the east- 
em branch of Little Tomaliawk Creek 
along the former course of VA 652 (Old 

40 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

9 _J3 

Figure 2. Microconodon tenuirostris, ROM 44300, left dentary in lingual (top) and buccal (bottom) views. Unshaded areas 
represent impressions in the matrix that were filled in with colored epoxy resin during preparation. Abbreviations as in Figure 1. 
Scale bar = 2 mm. 

Hundred Road), near Midlothian, Ches- 
terfield County, Virginia. Latitude 
77°40'17"N, longitude 37°27'50"W, Halls- 
boro 7.5 Minute Quadrangle. Age: Late 
Triassic (late Carnian according to Lucas 

Diagnosis. Type and only known species 
of genus, as diagnosed above. 

Discussion. The new material from the 

Richmond basin closely resembles ANSP 
10248, the holotype of Microconodon ten- 
uirostris Osbom, 1886, in most compara- 
ble features, particularly in the structure 
of the postcanine teeth. The only feature 
showing variation is the course of the in- 
ternal mandibular groove, which ap- 
proaches the ventral margin of the dentary 
in ROM 44301 but extends parallel to it in 

Figure 3. Microconodon tenuirostris, ROM 44301, anterior portion of right dentary in lingual view. Unshaded areas represent 
impressions in the mathx that were filled in with colored epoxy resin during preparation. Abbreviations as in Figure 1. Scale bar 
= 2 mm. 


ROM 44300 and USNM 437637. This dif- teeth. A narrow groove, which presumably 

ference may be ontogenetic in nature, and, housed the dental lamina in life (Cromp- 

in the absence of other differences, I refer ton, 1963), extends lingually just below 

all specimens to the same taxon. and parallel to the alveolar margin and an- 
teriorly up to the canine alveolus. The in- 

DESCRIPTION temal mandibular groove (sulcus primor- 

Dpntarv dialis) is developed on the lingual surface 

close to the ventral margin of the horizon- 

The long horizontal ramus of the den- tal ramus of the dentary. This groove ex- 

tary is slender throughout its entire length tends just above, and anteriorly approach- 

(Figs. 1-3). Behind the mandibular sym- es, the ventral margin on the small dentary 

physis, the alveolar (dorsal) and ventral ROM 44301 (Fig. 3), but its course is 

margins of the ramus are rather straight more or less parallel to and well above the 

and extend more or less parallel to each margin in USNM 437637 (Fig. 1) and 

other. The alveolar margin abruptly rises ROM 44300 (Fig. 2) where it reaches the 

toward the canine alveolus in USNM posterior end of the symphysis. The inter- 

437637; this rise is accompanied by a lat- nal groove forms the anterior continuation 

eral bulging of the dentary. Part of a men- of the well-developed posterior trough for 

tal foramen is visible on the lateral surface the reception of the postdentary bones, 

behind the canine alveolus in ROM 44300. The angular region of the dentaiy does not 

Anteriorly, the ventral margin of the den- form a distinct process, unlike in many 

tary forms a slight projection below the ca- other cynodonts; the ventral margin of the 

nine in ROM 44300 and USNM 437637 dentary curves gently upward and back- 

and then curves forward and upward in all ward toward the articular process. Lateral 

specimens. The dentary is gently convex to the last postcanine tooth, the low but 

beneath the incisors. The ventral edge is anteroposteriorly broad coronoid process 

rounded and thickened back to the region smoothly rises posterodorsally at an angle 

of the angle where it becomes sharper, of about 45° relative to the long axis of the 

The robust symphyseal portion of the den- dentary. The anterior margin of the pro- 

tary holds alveoli for a large canine and cess forms a coronoid ridge, which is most 

three apparently slightly procumbent in- pronounced anteriorly, before it merges 

cisors. The dentaries are fused along their into the horizontal ramus. The masseteric 

long, sloping symphysis. In USNM fossa is weakly defined. No distinct facet 

437637, a fragment of the symphyseal por- for the coronoid bone is apparent on the 

tion of the right dentary, containing the al- medial aspect of the coronoid process of 

veolus for the first lower incisor, still ad- the dentary. The lateral ridge on the artic- 

heres to the left element (Fig. 1). The ular process of the dentaiy is expanded, 

mandibular symphysis extends back to the especially more posteriorly, but there is no 

level of the posterior margin of the alve- indication that it contributed to a mam- 

olus for the canine. The buccal surface of mallike condyle posteriorly, 
the horizontal ramus of the dentary is con- 

vex dorsoventrally. A faint buccal groove i^pieniai 

on the horizontal ramus of the dentary A featureless, elongate-triangular bone 

ANSP 10248 noted by Osbom (1886b, in the posterior portion of the internal 

1887) appears to be the result of postmor- groove of the dentary in USNM 437637 

tern crushing and is not present on any of (Fig. 1) represents a partial left splenial. 

the specimens from the Richmond basin. Articular facets for the splenial on the den- 

The lingual surface of the horizontal ra- tary ROM 44300 (Fig. 2) indicate that this 

mus is flat near its symphyseal end but be- bone may have entered into the mandib- 

comes gently convex below the postcanine ular symphysis anteriorly. 

42 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


The incisors and canine are document- 
ed only by their alveoli. Three alveoli for 
incisors are present in USNM 437637. 
They indicate that the incisors were similar 
in size to each other and that they were 
slightly procumbent. The large alveolus for 
the canine forms an elongate oval in out- 
line. The count of three incisors and one 
canine is in agreement with Simpson's 
(1926b) estimate based on a series of de- 
pressions on the lateral aspect of the man- 
dibular symphysis of ANSP 10248, which 
he correctly interpreted as crushed alveoli. 

In occlusal view^, the crowns of the post- 
canine teeth are elliptical in outline and 
buccolingually narrow (Sues et al., 1994, 
fig. 8.4). Those of the tricuspid and tetra- 
cuspid postcanines are more than twice as 
long anteroposteriorly as wide buccolin- 
gually. No wear facets indicative of cusp 
contact resulting from tooth-to-tooth oc- 
clusion are present. The apex of the prin- 
cipal cusp is slightly blunted by abrasion 
on most teeth. The postcanine teeth pre- 
served in situ confirm the anatomical ori- 
entations for isolated tooth crowns of pos- 
sibly related cynodont taxa proposed by 
Peyer (1956) and Hahn et al. (1984). In 
buccal or lingual view, the slightly more 
convex cutting edge of the principal cusp 
faces anteriorly (mesially); in anterior or 
posterior view, the vertical curvature of the 
buccal surface is slightly more convex than 
that of the lingual surface. 

The specimen USNM 437637 has seven 
postcanine tooth positions and ROM 
44300 has eight (the last alveolus being in- 
complete). The specimen ROM 44301 
preserves eight postcanine teeth or alveoli, 
but the posterior end of the tooth-bearing 
ramus of the dentary is not preserved. 

The crowns of the anterior two postcan- 
ines of ANSP 10248 and of the anterior 
two preserved postcanines (positions 2 and 
3) in ROM 44301 are simple cones. In 
ANSP 10248 they show a slight posterior 
swelling. This single-cusped type of post- 

canine is absent in the largest known spec- 
imen (USNM 437637). 

The crowns of most of the other post- n 
canine teeth are tricuspid. The only excep- II 
tions in the sample described here are the 
more posterior of the two multicuspid po- 
stcanines preserved in ANSP 10248 and 
the seventh postcanine in ROM 44300, 
both of which have one anterior and two 
posterior accessory cusps. The specimen 
ROM 44300 shows some differentiation 
among its multicuspid postcanine teeth: 
compared to the crowns of the sixth and 
seventh postcanines, that of the fourth is 
shorter anteroposteriorly and taller. The 
fourth postcanine also has relatively small- 
er anterior and posterior accessory cusps 
that are placed closer to the base of the 
crown. The crowns of the multicuspid 
postcanines progressively increase in 
length toward the posterior end of the 
tooth row. The buccolingually narrow 
cusps are aligned behind one another in a 
straight line and are clearly separated from 
each other. The median or principal cusp 
is much larger than the anterior (mesial) || 
and posterior (distal) accessory cusps and 
dominates the tooth crown. This cusp is 
broad anteroposteriorly and somewhat 
compressed buccolingually. The subequal 
accessory cusps are symmetrically posi- 
tioned in front and behind the principal 
cusp and are separated from the latter by I 
distinct V-shaped notches. A distinct cut- 
ting edge extends down from the apex of 
the principal cusp along both the anterior 
and posterior faces; the anterior cutting 
edge is slightly more convex than the pos- 
terior one. Each accessory cusp bears a 
shaip cutting edge only on the side facing 
the central cusp. The accessory cusps pro- 
ject slightly away from the principal cusp. 
Neither the principal nor the accessory 
cusps are recurved. Cingula are absent. 
The enamel is completely smooth on most 
teeth. Fine, stained lines on the enamel of 
the isolated postcanine tooth USNM 
448579 presumably represent postmortem 
fracturing (see Hahn et al., 1984). The an- 
teroposteriorly broad and buccolingually 



flattened root is not set off from the crown 
by an annular constriction. The specimen 
USNM 448579 shows a pronounced me- 
dian longitudinal constriction of its root, 
the apical portion of which is broken. This 
incipient division of the root is not evident 
on any of the teeth preserved in situ in the 
referred dentaries. However, this division 
is visible on the two multicuspid postca- 
nine teeth preserved in ANSP 10248. In 
ROM 44300 and ROM 44301, a ring of 
bone connects the roots of at least soine 
functional teeth to the alveolar margin; de- 
tails are not clearly visible for all tooth po- 
sitions. This ankylosis is also present in 
basal eucynodonts such as Thrinaxodon 
(Crompton, 1963). 

In ROM 44301, small pits for develop- 
ing replacement teeth are present in the 
groove for the dental lamina anterolingual 
to postcanine positions 4 and 6 and lingual 
to tooth 7 (Fig. 3). Comparison of the four 
known dentaries indicates that the simple 
anterior postcanines were lost without re- 
placement during growth, resulting in a 
progressively longer diastema in larger 
specimens, as in many nonmammalian cy- 
nodonts (Crompton, 1963) and in the 
mammaliamorph Sinoconodon (Crompton 
and Luo, 1993). However, in the latter, 
new teeth are added only at the posterior 
end of the postcanine series. The small 
dentaries ANSP 10248 (length: 16.5 mm) 
and ROM 44301 (length of preserved por- 
tion: 10.3 mm) both lack a diastema be- 
tween the canine and postcanine teeth, 
and the postcanine tooth row begins im- 
mediately behind the canine position. On 
the large dentary USNM 437637, an ex- 
tensive diastema separates the alveolus for 
the canine from the postcanine teeth (Fig. 
1); a tooth fragment attached to the bone 
about midway probably represents a dis- 
placed fraginent. The configuration of the 
tooth row is consistent with evidence from 
the pattern of bone grain for the imma- 
turity of ROM 44300 and ROM 44301 
(Figs. 2, 3); the grain on the latter two 
dentaries coinprises fine longitudinal 
grooves and pores typical of immature 

bone (Enlow, 1969). The specimen USNM 
437637 does not show this type of bone 
grain although it was almost identical in 
length to ROM 44300 (estimated lengths 
of 33 mm and 34 mm, respectively). The 
surface of ANSP 10248 is too poorly pre- 
served to show details of texture. 


Taxa Possibly Related to Microconodon 

Lees and Mills (1983: 179) observed 
that "[m]any of the later small carnivorous 
cynodonts and early mammals had molars 
consisting, more or less, of a single main 
cusp, flattened bucco-lingually, with mesial 
and distal cuspules" [accessoiy cusps in 
the present paper]. This statement aptly 
characterizes the postcanine teeth of Mi- 

Many isolated teeth resembling those of 
Microconodon as well as jaw fragments 
containing such teeth have been reported 
from the Upper Triassic of central and 
western Europe (Peyer, 1956; Clemens, 
1980; Hahn et al, 1984, 1987, 1994; Sig- 
ogneau-Russell and Hahn, 1994; Godefroit 
and Battail, 1997) and New Mexico (Lucas 
and Oakes, 1988). These fossils have been 
variously referred to nonmammalian cy- 
nodonts or mammals, but, in some cases, 
even their synapsid affinities remain yet to 
be established (see below). The situation 
was complicated by the discovery of tri- 
cuspid teeth in the Late Triassic "rham- 
phorhynchoid" pterosaur Eiidimorphodon 
(Wfld, 1978), although Halm et al. (1984) 
have provided structural criteria for distin- 
guishing between teeth of Eiidimorphodon 
and those of nonmammalian cynodonts. 

Tricuspes tuebingensis Huene, 1933 
from the Rhaeto— Liassic bonebed at Gais- 
brunnen (Baden-Wiirttemberg, Germany) 
as well as Tricuspes sigogneauae Hahn et 
al., 1994 and Tricuspes tapeinodon Gode- 
froit et Battail, 1997 from the Upper Tri- 
assic (Norian) of Saint-Nicolas-de-Port 
(France) and Hallau (Switzerland) are 
known only from isolated postcanine teeth 
(Clemens, 1980; Hahn et al., 1994; God- 

44 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

efroit and Battail, 1997). Clemens (1980) cynodont taxa based on isolated postcanine 

tentatively placed Tricuspes in the Mam- teeth recovered from a Rhaetian bonebed 

malia whereas Hahn et al. (1994) referred (Sables de Mortinsart) at Gaume, southern 

it to the nonmammalian cynodont taxon Belgium: Lepagia gauniensis, Gaumia lon- 

Dromatheriidae. The tricuspid and tetra- giradicata, and ?Gaumia incisa. They also 

cuspid postcanines of Tricuspes closely re- referred some postcanines from the Hallau 

semble those of Microconodon in most bonebed to L. gaumensis and ?G. incisa. 

features, including the incipient division of Hahn et al. (1987) assigned Lepagia to the 

the root. However, in occlusal view, the Chiniquodontidae, and Sigogneau-Russell 

apex of the principal cusp in Tricuspes is and Hahn (1994) referred it to the Chi- 

slightly displaced so that the cusps are ar- niquodontidae or Probainognathidae on 

ranged in a broadly V-shaped pattern. Fur- the basis of similarities in the position and 

thermore, the holotype of T tuebingensis, shape of the splenial between Lepagia and 

which was identified as a right lower post- Probainognathus. However, the phyloge- 

canine by Clemens (1980) and Godefroit netic significance of those features is un- 

and Battail (1997) but as a left lower post- certain (see Battail, 1991). Tooth crowns 

canine by Hahn et al. (1994), bears an ac- of Lepagia are asymmetrical in side view 

cessory cuspule (Hahn et al., 1994, fig. 3, and have a principal cusp and one or two 

m) in a posterobuccal (according to Clem- anterior and posterior accessory cusps, 

ens) or posterolingual (according to Hahn Cingula are absent. The undivided root is 

et al.) position. This cuspule is absent on separated from the crown by a distinct an- 

the teeth of T sigogneauae and T tapei- nular constriction. Hahn et al. (1987) left 

nodon from Saint-Nicolas-de-Port (Hahn the systematic position of Gaumia unre- 

et al., 1994; Godefroit and Battail, 1997). solved, and Sigogneau-Russell and Hahn 

Hahn et al. (1984) named Pseudotricon- (1994) considered it a chiniquodontoid of 
odon wildi on the basis of isolated post- uncertain affinities. Teeth referable to 
canine teeth from a mid-Norian bonebed Gaumia are distinguished mainly by the 
in Luxembourg. Some teeth referred to P. great length of the undivided, distally ta- 
wildi have three cusps whereas others pering root, which is set off from the 
have four or five. The tooth crowns are crown by a slight constriction, 
devoid of cingula and closely resemble Lucas and Oakes (1988) described 
those of Microconodon in their overall ap- "P.seudotriconodon" chatterjeei on the ba- 
pearance. As in Microconodon, the cusps sis of a tiny tooth-bearing jaw fragment 
are aligned directly behind one another, and two isolated teeth from the Bull Can- 
Both the principal and accessory cusps yon Formation (Upper Triassic: lower No- 
form distinct cutting edges anteriorly and rian) of New Mexico. "Pseudotriconodon" 
posteriorly, unlike the condition in Micro- chatterjeei differs from P. wildi in the 
conodon. Incipient root division appears to structure of its teeth, and indeed reference 
be restricted to the apical portion of the of this material to the Cynodontia remains 
root on some postcanines (Hahn et al., to be confirmed by additional specimens. 
1984, pi. 2, fig. 6). As in Microconodon, The cusps of "P." chatterjeei bear numer- 
the root is not set off from the crown by ous prominent vertical ridges buccally and 
an annular constriction (Hahn et al., 1984, lingually; similar striations are present on 
pi. 3, figs, la, 2a, 6d). Pseudotriconodon the teeth of the Late Triassic pterosaur 
possibly differs from Microconodon in that Eudimorphodon (Wild, 1978; Hahn et al., 
tetracuspid and pentacuspid postcanines 1984). 

are as common as or more common than Therioherpeton cargnini from the Up- 

tricuspid teeth in the currently available per Triassic Santa Maria Formation of 

samples (Hahn et al., 1984). southern Brazil is based on an incomplete 

Hahn et al. (1987) described three new skull and a partial postcranial skeleton 


(Bonaparte and Barberena, 1975). Cranial odon is referable to the Eucynodontia as 

features shared with both Tritheledontidae diagnosed by Hopson (1991; see also Mar- 

and Mammaliaformes include the absence tinez et al. [1996]) based on the possession 

of the prefrontal and postorbital (and post- of the following apomorphies: dentary 

orbital bar). The 2ygomatic arch is very considerably enlarged and fused mandib- 

slender along its entire length. The post- ular symphysis. Microconodon also shares 

canine teeth lack cingula and have four with other eucynodonts a splenial reduced 

cusps that are aligned directly behind one to a slender, thin bone covering the inter- 

another The root of at least one postca- nal groove of the dentary and postdentary 

nine shows a median longitudinal constric- bones (articular, prearticular, and suran- 

tion of the anteroposteriorly broad root, gular) forming a rodlike complex that is 

Kemp (1982) referred Therioherpeton to lodged in a posterior trough on the lingual 

the Tritheledontidae, but it lacks the den- surface of the dentary (inferred from the 

tal features diagnostic for the latter taxon shape of the trough). The material cur- 

(Gow, 1980; Shubin et al., 1991). rently referable to Microconodon is insuf- 

Lucas and Luo (1993) suggested that ficient to permit more precise determina- 

the possible basal mammaliaform Adelo- tion of the phylogenetic relationships of 

hasileus cromptoni, known only from an this taxon. 

isolated braincase from the Tecovas Mem- Hahn et al. (1984) redefined Dromath- 
berof the Dockum Formation (Upper Tri- eriidae Gill, 1872, to include Dromather- 
assic: upper Gamian) of Texas, might iurn, Microconodon, Pseudotriconodon, 
prove referable to Microconodon. How- and Therioherpeton. Battail (1991) exclud- 
ever, in the absence of associated jaws and ed Therioherpeton from the Dromatheri- 
(or) postcanine teeth, this association re- idae, but hypothesized a sister-group re- 
mains untestable. lationship between the two taxa based on 

With the exception of Therioherpeton, the shared absence of cingula on the post- 
the aforementioned taxa, including Micro- canine teeth. However, Hopson (1991) 
conodon, are poorly represented by skel- noted that cingula are also absent on the 
etal remains. Although the similarity in the postcanines of Prohelesodon from the 
structure of the postcanine teeth in all Middle Triassic Ghaiiares Formation of 
these forms may prove to be phylogeneti- Argentina, and this character-state may ei- 
cally significant, it is more prudent to con- ther diagnose a more inclusive grouping or 
sider them incertae sedis among the Eu- may have developed more than once. In- 
cynodontia until more complete speci- deed, the distribution of cingula is appar- 
mens become available for study. Therio- ently variable among derived cynodont 
herpeton shares some apomorphic cranial synapsids: the lower postcanine teeth of 
features with both Tritheledontidae and the mammaliamorph Sinoconodon have 
Mammaliaformes, but the distribution of weakly developed posterolingual cingula 
those characters in other derived cyno- (Grompton and Luo, 1993), whereas those 
donts has yet to be fully documented. I of the tritheledontid Pachygenelus (Gow, 
believe that it is premature to use these 1980; Shubin et al., 1991) and the mam- 
features to diagnose a family-level taxon maliaform Morganucodon (Mills, 1971; 
Dromatheriidae, as has been proposed by Grompton and Luo, 1993) have well-de- 
Hahn et al. (1994). veloped fingual cingula. Battail (1991) pro- 
posed a clade incTuding Dromatheriidae 
Phylogenetic position of Microconodon ^^d Therioherpeton, which he placed as 

Assessing the phylogenetic position of the sister-taxon of Tritheledontidae (and 

Microconodon is difficult because of the possibly Mammaliaformes) on the basis of 

very limited set of character-states observ- several cranial and dental characters. The 

able in the available specimens. Microcon- former cannot be determined in the 


Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

known material referable to Microcono- 
don. One of the dental characters cited by 
Battail, the more or less oblique implan- 
tation of the postcanine teeth relative to 
the long axis of the tooth row, is absent in 
Microconodon. Most recently, Hahn et al. 
(1994) have redefined the Dromatheriidae 
to include Tricuspes and Meurthodon Sig- 
ogneau-Russell et Hahn, 1994, from the 
Upper Triassic of Saint-Nicolas-de-Port 
(France). They also explicitly considered 
Dromatheriidae the sister-taxon of "Mam- 
malia" (Mammaliaformes sensu Rowe 
[1988]). The currently available material is 
insufficient for a rigorous test of this in- 
triguing hypothesis. 

The postcanine teeth of Microconodon, 
Pseudotriconodon, and Therioherpeton 
differ from those of other nonmammalian 
cynodonts (Cynognathus, Frobainogna- 
thus, Probelesodon, and Pachygenelus as 
well as the sectorial teeth of Diademodon 
and other basal gomphodont cynodonts) in 
the absence of the backward curvature of 
the principal cusp and (with the exception 
of Probelesodon) in the lack of cingula. 
They also differ from the postcanines of 
Cynognathus and Probelesodon in the ab- 
sence of serrations on the cutting edges of 
individual cusps. Furthermore, the post- 
canine teeth of Microconodon and Ther- 
ioheiyeton share incipient division of the 
roots, resulting in a figure-eight shape in 
transverse section. Although this feature is 
also present in the tritheledontid Pachy- 
genelus (Shubin et al, 1991), it may well 
prove diagnostic for those probainognathi- 
an eucynodonts closest to Tritheledontidae 
+ Mammaliaformes (Hopson, personal 
communication ) . 

Microconodon tenuirostris is a derived 
eucynodont, but the currently available 
material does not permit a more precise 
placement. This uncertainty also reflects 
the still inadequate fossil record of small 
cynodonts from Mid- to Late Triassic con- 
tinental strata. 


I am indebted to R A. Kroehler (Na- 
tional Museum of Natural History), P. E. 

Olsen (Lamont-Doherty Earth Observa- 
tory, Columbia University), and especially 
E. B. Sues for their enthusiastic help in 
the field. W. W. Amaral (Harvard Univer- 
sity) assisted in the preparation of the new 
specimens reported in this paper. E. R. 
Daeschler (Academy of Natural Sciences 
of Philadelphia) arranged for the extended 
loan of the holotype of Microconodon ten- 
uirostris. S. G. Lucas (New Mexico Mu- 
seum of Natural History) kindly provided 
a cast of the holotype of '^ Pseudotricono- 
don" chatterjeei for comparisons. D. M. 
Scott prepared the illustrations with her 
customary skill. J. A. Hopson (University 
of Chicago) and Z. Luo (Carnegie Muse- 
um of Natural History) offered construc- 
tive comments on a draft of the manu- 
script. I gratefully acknowledge financial 
support from the National Geographic So- 
ciety (grants 3592-88 and 4232-89), Na- 
tional Science Foundation (NSF EAR- 
9016677 to H.-D. S. and R E. Olsen), 
Smithsonian Institution, and the Natural 
Sciences and Engineering Research Coun- 
cil of Canada. I dedicate this paper to 
Fuzz Crompton who introduced me to 
nonmammalian cynodonts and who has 
made so many important contributions to 
our knowledge of these animals during his 
long and distinguished career. 


Battail, B. 1991. Les Cynodontes (Reptilia, Ther- 
apsida): une phylogenie. Bulletin du Museum 
National d'Histoire Naturelle, Serie 4, Section C, 
13: 17-105. 

BONAPARTE, J. F., AND M. C. Barberena. 1975. A 
possible mammalian ancestor from the Middle 
Triassic of Brazil (Therapsida-Cynodontia). Jour- 
nal of Paleontology, 49: 931-936. 

Bonaparte, J. R, and M. C. Barberena. 2001. On 
two advanced cynodonts form the Late Triassic 
of Southern Brazil. Bulletin of the Museum of 
Comparative Zoology, 156: 59-80. 

Clemens, W. A. 1980. Rhaeto-Liassic mammals 
from Switzerland and West Germany. Zitteliana, 
5: 51-92. 

Cornet, B., and R E. Olsen. 1990. Early to Middle 
Carnian (Triassic) Flora and Fauna of the Rich- 
mond and Taylorsville Basins, Virginia and Mary- 
land, U.S.A. Virginia Museum of Natural History 



Guidebook Number 1. Martinsville, Virginia. 87 

Crompton, a. W. 1963. Tooth replacement in the 
cynodont Thrinaxodon liorhinus Seeley. Annals 
of tlie South African Museum, 46: 479—521. 

Crompton, A. W., and Z. Luo. 1993. Relationships 
of the Liassic mammals, Sinoconoclon, Morgan- 
iicodon oehleri, and Dinnetherium, pp. 30^4. In 
F. S. Szalay, M. J. Novacek, and M. C. McKenna 
(eds.). Mammal Phylogenv: Mesozoic Differen- 
tiation, Multituberculates, Monotremes, Early 
Therians, and Marsupials. Berlin: Springer-Ver- 
lag. X + 249 pp. 

Emmons, E. 1857. American Geology. Part VI. Al- 
bany, New York: Sprague and Co. x + 152 pp. 

Enlow, D. H. 1969. The" bone of reptiles, pp. 45- 
80. In C. Cans, A. d"A. Bellairs, and T S. Parsons 
(eds.). Biology of the Reptilia. Vol. 1: Morphol- 
og\' A. New York: Academic Press, x-v + 373 pp. 

Gill, T 1872. Arrangement of the families of mam- 
mals and synoptical tables of characters of the 
subchvisions of mammals. Smithsonian Miscella- 
neous Collections, 230: 1-98. 

Gillette, D. D. 1978. Catalogue of type specimens 
of fossil vertebrates, Academv of Natural Scienc- 
es, Philadelphia. Part IV: Reptilia, Amphibia, and 
tracks. Proceedings of the Academy of Natural 
Sciences of Philadelphia, 129: 101-111. 

GODEFROIT, P., AND B. Battail. 1997. Late Triassic 
cynodonts from Saint-Nicolas-de-Port (north- 
eastern France). Geodiversitas, 19: 567-631. 

Cow, C. E. 1980. The dentitions of the Tritheledon- 
tidae (Therapsida: Cynodontia). Proceedings of 
the Royal Society of London, Series B, 208: 461- 

Hahn, G., R. Hahn, and p. Godefroit. 1994. Zur 
Stellung der Dromatheriidae (Ober-Trias) 
zwischen den C\iiodontia und den Mammalia. 
Geologica et Palaeontologica, 28: 141-159. 

Hahn, G., J. C. Lepage, and G. Wouters. 1984. 
CvTiodontier-Zahne aus der Ober-Trias von Med- 
emach, Grossherzogtum Luxemburg. Bulletin de 
la Societe beige de Geologic, 93: 357-373. 

Hahn, G., R. Wild, and G. Wouters. 1987. Gy- 
nodontier-Zaline aus der Ober-Trias von Gaume 
(S-Belgien). Memoires pour servir a I'Explication 
des Cartes Geologiques et Minieres de la Bel- 
gique. 24: 1-33. 

HOPSON, J. A. 1991. Systematics of nonmammalian 
Synapsida and implications for patterns of evo- 
lution in synapsids, pp. 635—69.3. In H.-P Schul- 
tze and L. Trueb (eds.). Origins of the Higher 
Groups of Tetrapods: Controversy and Consen- 
sus. Ithaca, New York: Comstock Publishing As- 
sociates, xii + 724 pp. 

HOPSON, J. A., AND J. W. KiTCHING. 1972. A revised 
classification of cynodonts (Reptilia; Therapsida). 
Palaeontologia Africana, 14: 71—85. 

Huene, E. \0N. 1933. Zur Kenntnis des Wiirttem- 
bergischen Ratbonebeds mit Zahnfunden neuer 
Sanger und stiugeralmlicher Reptilien. Jahres- 

hefte des Vereins fiir vaterlandische Naturkunde 
in Wilrttemberg, 89: 65-128. 

Kemp, T. S. 1982. Mammal-like Reptiles and the Or- 
igin of Mammals. London: Academic Press. 363 

Lees, P. M., and R. Mills. 1983. A quasi-mammal 
from Lesotho. Acta Palaeontologica Polonica, 
28: 171-180. 

Litwtn, R. J., AND S. Ash. 1993. Revision of the bio- 
stratigraphy of the Chatham Group (Upper Tri- 
assic), Deep River basin, NortJi Carolina, U.S.A. 
Review of Palaeobotany and Pal^nology, 77: 75— 

Lucas, S. G. 1998. Placerias (Reptilia, Dicyiiodontia) 
from the Upper Triassic of the Newark Super- 
group, North Carolina, USA, and its biochron- 
ological significance. Neues Jalirbuch fiir Geo- 
logic und Palaontologie, Monatshefte, 1998: 

Lucas, S. G., and Z. Luo. 1993. Adelohasileus from 
the Upper Triassic of west Texas: the oldest 
mammal. Journal of Vertebrate Paleontology', 13: 

Lucas, S. G., and W. Oakes. 1988. A Late Triassic 
cynodont from the American South-West. Pa- 
laeontologv', 31: 445^49. 

Martinez, R. N., C. L. M.\y, and C. A. Forster. 
1996. A new cami\orous cynodont from the 
Ischigualasto Formation (Late Triassic, Argenti- 
na), with comments on euc^nodont phylogeny. 
Journal of \^ertebrate Paleontology', 16: 271-284. 

Mills, J. R. E. 1971. The dentition of Morganuco- 
don, pp. 29-63. //; D. M. Kermack and K. A. 
Kermack (eds.). Early Mammals. Zoological 
Journal of the Linnean Society, 50(Suppl. 1). .xiv 
+ 203 pp. 

OSBORN, H. F. 1886a. A new mammal from the 
American Triassic. Science, 8: 540. 

. 1886b. Obser\ations on die Upper Triassic 

mammals, Droniatfierium and Microconodon. 
Proceedings of the Academy of Natural Sciences 
of Philadelphia. 37: 359-363. 

. 1887. The Triassic mammals, Dromathciium 

and Microconodon. ProceecUngs of the American 

Philosophical Society, 24: 109-111. 
0\\t:n, R. 1871. Monograph of the Fossil Mammalia 

from the Mesozoic Formations. London: Pa- 

laeontographical Societ)'. xi + 115 pp. 
Palmer, T. S. 1903. Some new generic names of 

mammals. Science, 17: 873. 
PEYER, B. 1956. Uber Zalme von Haramiyden [sic], 

von Triconodonten und von wtilirscheinlich svii- 

apsiden Reptilien aus deni Rhat von Hallau Kt. 

Schaffhausen, Schweiz. Schweizerische Palaon- 

tologische Abhandlungen, 72: 1—72. 
Ro\VE, T. 1988. Definition, diagnosis, and origin of 

Mammalia. JouiTial of Vertebrate Paleontology, 

8: 241-264. 
Seeley, H. G. 1895. Researches on tlie structure, 

organization, and classification of the fossil Rep- 

tiha. Part IX, Section 5. On the skeleton in new 

48 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Cynodontia from the Karroo rocks. Philosophical 
Transactions of the Royal Society of London, Se- 
ries B, 186: 59-148. 

Shubin, N. H., a. W. Crompton, H.-D. Sues, and 
P. E. Olsen. 1991. New fossil evidence on the 
sister-group of mammals and early Mesozoic fau- 
nal distributions. Science, 251: 1063-1065. 

SiGOGNEAU-RUSSELL, D., AND G. Hahn. 1994. Late 
Triassic microvertebrates from central Europe, 
pp. 197-213. In N. C. Eraser and H.-D. Sues 
(eds.). In the Shadow of the Dinosaurs: Early 
Mesozoic Tetrapods. Cambridge, United King- 
dom: Cambridge University' Press, x + 435 pp. 

Simpson, G. G. 1926a. Are Dromatherium and Mi- 
croconodon mammals? Science, 63: 548—549. 

. 1926b. Mesozoic Mammalia. V. Dromather- 
ium and Microconodon. American Journal of Sci- 
ence, 12: 87-108. 

Sues, H.-D., and P. E. Olsen. 1990. Triassic verte- 
brates of Gondwanan aspect from the Richmond 
basin of Virginia. Science, 249: 1020-1023. 

Sues, H.-D., P. E. Olsen, and P. A. Kroehler. 
1994. Early Late Triassic small tetrapods from 
the Richmond basin of Virginia, pp. 161-170. In 
N. C. Eraser and H.-D, Sues (eds.). In the Shad- 
ow of the Dinosaurs: Early Mesozoic Tetrapods. 
Cambridge, United Kingdom: Cambridge Uni- 
versity Press. X + 435 pp. 

Traquair, R. H. 1877. The Ganoid Fishes of the 
British Carboniferous Formations. Part I. Pa- 
laeoniscidae. London: Palaeontographical Socie- 
ty. 60 pp. [first installment only]. 

Wild, R. 1978. Die Flugsaurier (Reptilia, Pterosau- 
ria) aus der Oberen Trias von Gene bei Bergamo, 
Italien. Bollettino della Societa Palaeontologica 
Italiana, 17: 176-256. 



Abstract. A new genus and species of cynodont 
from the Upper Triassic Fleming Fjord Formation of 
East Greenland possesses double-rooted postcanine 
teeth and a nonaltemate pattern of tooth replace- 
ment. The specimen represents an addition to the 
known diversity of Early Mesozoic taxa with multi- 
rooted dentitions (tritylodontids, Sinoconodon sp., 
haramiyids, morganucodontids, Meurthodon galli- 
cus), and casts doubt on traditional interpretations of 
the interdependency of reduced tooth replacement 
patterns and teeth with multiple roots. 


The Upper Triassic Fleming Fjord For- 
mation of Jameson Land, East Greenland, 
preserves a diverse fossil vertebrate fauna 
that includes mammals, theropod and pro- 
sauropod dinosaurs, plagiosaurid and cy- 
clotosaurid amphibians, turtles, aetosaurs, 
phytosaurs, and pterosaurs (Jenkins et al., 
1994, 2001). Mainmals are represented 
primarily from the upper Tait Bjerg Beds 
and include Kuehneotlieriuin, cf Brachij- 
zostrodon, and the haramiyid Haramijavia 
clemmenseni (Jenkins et al., 1994, 1997). 
We describe here an additional compo- 
nent of the fauna, a cynodont that bears 
double-rooted teeth, the only known spec- 
imen of this taxon. A comparable form of 
Late Triassic age is Meurthodon gallicus 
(Russell et al., 1976; Sigogneau-Russell 
and Hahn, 1994; Godefroit and Battail, 
1997), represented by isolated teeth from 
Rhaetic deposits in France, but this taxon 
differs in significant details. 

The following abbreviations of institu- 
tional names are used: IRSNB, Institut 

royal des Sciences naturelles de Belgique, 
Brussels; MCZ, Museum of Comparative 
Zoology, HaiA/ard University, Cambridge, 
Massachusetts; MGUH, Geological Mu- 
seum, University of Copenhagen; and 
MNHP, Institut de Paleontologie, Muse- 
um National d'Histoire Naturelle, Paris. 

Order Therapsida Broom, 1905 
Infraorder Cynodontia Owen, 1861 

' Department of Organismic and Evolutionary Bi- 
ology, and Museum of Comparative Zoology, Harvard 
University, Cambridge, Massachusetts 02138. 

Family incertae sedis 

Mitredon cromptoni new genus and 

Etymology. The generic term refers to 
the highly peaked primaiy cusps, a com- 
bination of English mitre, the high-peaked 
ecclesiastical headdress, from Greek mitra, 
turban, and Greek odous {odon), tooth. 
The specific name honors A. W. Crompton 
for his important contributions to our un- 
derstanding of the paleobiology and evo- 
lution of c)Tiodonts. 

Holotijpe. MGUH VP 3392, MCZ field 
number 11/G95 (Figs. lA, B), a partial left 
dentary bearing an incomplete alveolus 
mesially, three unerupted postcanine 
teeth, roots of four other (erupted) po- 
stcanines, and an empty tooth ciypt dis- 

Horizon. Uppermost dolostone of Tait 
Bjerg Beds, 0rsted Dal Member of the 
Fleming Fjord Formation. 

Locality. 71°32.929'N, 22°55.450'W, 
north of y^renprisdal at its confluence 
with Pingel Dal, Jameson Land, East 

Bull. Mus. Comp. Zool., 156(1): 49-58, October, 2001 49 

50 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


2 mm 

Figure 1. Left dentary of Mitredon cromptoni, MGUH VP 3392, in (A) lingual view and (B) occlusolabial view. In occlusolabial 
view, tfie mandibular canal is visible distally as a filled cast through a window cut in the labial aspect of the jaw; the canal 
obscures the basal crown of Pc„. 

Age. Late Triassic (PNorian— Rhaetic; 
Jenkins et al., 1994). 

Diagnosis. Cingular cusps on postcanine 
teeth absent or lost, a derived character 
shared with galesaurid cynodonts, Ci/nog- 
nathiis, and Frobelesodon (Hopson and 
Kitching, 1972; Hopson and Barghusen, 
1986). Differs from Therioherpetidae 
(Bonaparte and Barberena, 1975) and oth- 
er nonmanimalian cynodonts in possessing 
bifurcate postcanine tooth roots (hkely 
convergent with multirooted tritylodon- 
tids) and lacking alternate tooth replace- 
ment. Characters shared with "chiniquo- 
donts" are crowns of lower postcanines 

with laterally compressed cusps arranged 
in a longitudinal row and clearly separated 
from each other (Bonaparte and Barber- 
ena, 1975; Sigogneau-Russell and Hahn, 
1994). A feature shared with Meui-thodon 
gallicus (Russell et al., 1976; Sigogneau- 
Russell and Hahn, 1994) and some "chi- 
niquodonts" (Kemp, 1982) is a recuived 
cusp a (following the nomenclature of 
Crompton and Jenkins, 1968); that is, the 
mesial crest is longer and more horizontal 
than the distal crest, which is shorter and 
more vertically oriented. A derived feature 
shared with Meurthodon, Sinoconodon, 
and Mammaliaformes (sensu Wible, 1991) 

Tooth Replacement and Double-Rootedness • Shapiro and Jenkins 51 


MGUH VP 3392 





aligned at "gumline" 

aligned at cusps 

Figure 2. Comparison between (A) PCs of Mitredon cromptoni, MGUH VP 3392, (B) tlie type specimen of Meurthodon gallicus, 
and (C, D) two isolated teeth referred to Meurthodon gallicus. Specimens MNHP SNP210W and IRSNB R163 were selected for 
comparison in addition to ihe type specimen for their gross similarity to MGUH VP 3392. In each column (B, 0, and D), a 
Meurthodon tooth (shaded) is compared to PCs of Mitredon (outline) in two ways to minimize the effects of size and completeness 
on morphologic comparison between specimens. First, the middle row depicts comparisons based on alignment of two teeth at 
the inferred gumline. Second, in the bottom row, cusps are aligned as closely as possible. Although the alignment of MGUH VP 
3392 at the "gumline" with IRSNP R163 and with the cusps of MNHP SNP1W reveal overall similarities in shape, all specimens 
referred to Meurthodon lack a cusp e. Cusp b of PCj in Mitredon is also substantially smaller than the corresponding structure 
in Meurthodon. Cusp designations in (A) after Crompton and Jenkins (1968). MNHP SNP1W redrawn from Sigogneau-Russell 
and Hahn (1994); MNHP SNP210W and IRSNB R163 redrawn from Godefroit and Battail (1997). For top row, scale bar = 
1 mm. 

is roots of lower postcanine teeth bifur- 
cate, a character once considered diagnos- 
tic of Mammalia (e.g., Crompton and Jen- 
kins, 1979). Distinguished from M. gallicus 
by the presence of a cuspule mesial to 
cusp b (on Pcj); a mesiodistally longer cusp 
a (if the type specimen of M. gallicus and 
Pcg o£ Mitredon cromptoni, new genus and 
species, are scaled to the same mesiodistal 
length at the crown— cervical junction, the 
mesiodistal length of cusp a is approxi- 
mately 60% total mesiodistal crown length 

in the latter, and only 40% in the former; 
Fig. 2); and the absolute mesiodistal 
length of Pc, is 20% smaller than that of 
the type specimen of M. gallicus (3.5 mm 
versus 4.2 mm). None of the three lower 
teeth of MGUH VP 3392 closely resem- 
bles the single isolated tooth of the type 
specimen of M. gallicus (Russell et al., 
1976; Signoneau-Russell and Hahn, 1994) 
nor other isolated teeth that Godefroit and 
Battail (1997) subsequently referred to 
that taxon (Fig. 2). 

52 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


_,(PC,) PC, PC, PC, PC3 PC, PC, 


1 cm 

Figure 3. Mitredon cromptoni and contemporaneous Greenlandic mammals. (A) Schematic reconstruction of the left lower jaw 
of Mitredon cromptoni, MGUH VP 3392. Lower jaws of the Late Triassic mammals (B) Haramiyavia (redrawn from Jenkins et 
al. [1997]) and (C) Morganucodon (redrawn from Hopson [1994], in Bonaparte and Crompton [1994]). Abbreviations: Pc, post- 
canine tooth;, internal dentary groove;, mental foramen;, mandibular canal. 


Lower Jaw 

The specimen is a partial left dentary, 
slightly convex along its ventral margin, 
and missing the anterior and posterior 
ends of the ramus; erupted teeth had bro- 
ken off postinortem, but several unerupt- 
ed teeth are preserved within the ramus. 
Although the total number of postcanine 
teeth is uncertain, the seven tooth posi- 
tions are here referred to as Pcj (most me- 
sial) through Pcy (most distal) (Fig. 3). The 
mental foramen is situated ventral and la- 

bial to Pc2. An internal groove extends lon- 
gitudinally along the inferior, lingual as- 
pect of the dentaiy, indicating the pres- 
ence of postdentary bones. The mandibu- 
lar canal is exposed through breakage on 
the lingual aspect of the jaw between Pc, 
and PC4 (Fig. lA). The canal passes to the 
labial side of Pcg and is preserved in cross 
section at the break across the posterior 
end of the specimen (Fig. IB). 


Lower postcanine 1 is indicated by a 
partially preserved alveolus. In the next 

Tooth Replacement and Double-Rootedness • Shapiro and Jenkins 53 

tooth position, Pcj, is an unerupted tooth bears an elongate, tapered cusp a, the apex 
of which the apex of cusp a and most of of which is directed between the roots of 
the hngual half of the crown are preserved; the predecessor tooth (Fig. 3). Cusp c is 
the apex of cusp a abuts a root fragment rounded and without a pointed apex, un- 
of the eiaipted tooth that was in the pro- like the other c cusps preserved in this 
cess of being replaced. Cusp a is laterally specimen. Much of this cusp lies on the 
compressed and, unlike the recurved, lingual side of the mandibular canal and is 
asymmetrical cusp a of the unerupted best observed occlusolabially. Lower post- 
tooth at PC5, appears to be nearly sym- canine 6 is smaller than Pcj (Fig. 3), sug- 
metrical. Two successively smaller cusps gesting that Pcr inay be the ultimate tooth 
lie distal to cusp a; their apices are direct- in die lower dentition; however, a cryptlike 
ed slightly distally, comparable to the distal depression distal to Pc^, as well as an anal- 
cusps of Pc, but unlike the strictly dorsally ysis of tooth replacement, appears to in- 
directed orientation of cusp a. The mesial dicate that a more distal tooth position 
end of the tooth is not presei^ved and (Pcj) may have been present, 
therefore the presence or absence of me- 
sial cusps is not possible to determine. The DISCUSSION 

remains of the crown of Pc, are sufficient phy|ogenetic Affinities 

to determme tliat tlie tooth is moi"pliolog- 

ically distinct from those at Pc, and Pcg, Based on dental morphology, Mitredon 

the only other intact teeth in the jaw. croniptoni is likely to be closely related to 
Lower postcanine 3 and PC4 are fully Meiirthodon gallicus. Although Sigogneau- 
erupted teeth, but the crowns are lost and Russell and Hahn (1994) interpreted M. 
only the roots remain. gallicus is most closely related to Therioh- 
Lower postcanine 5, in the process of erpeton cargnini, we believe that M. 
erupting from its crypt, was exposed by crotnptoni should be excluded from the 
preparation (Figs. lA, B). The tooth con- Therioherpetidae (as originally diagnosed 
sists of a nearly complete crown bearing by Bonaparte and Barberena, 1975) on the 
five cusps, and the upper parts of two basis of the bifurcate postcanine tooth 
roots. Mesial cusps b and e are the smallest roots and the nonalternate pattern of tooth 
and cusp a is the largest, the latter rising replacement (but see below discussion of 
2.55 mm from the base of the crown to root variability in early Mesozoic cyno- 
the point at which the apex is broken, donts). Inasmuch as M. gallicus also has 
Cusps c and d are successively smaller and fully bifurcate postcanine tooth roots (un- 
more distally directed than a. Cusp d sup- like the incipiently double-rooted condi- 
ports a distinct cuspule on its lingual sur- tion of T! cargnini, in which the cross sec- 
face; in an examination of a cast of Meur- tion of the single root is in the shape of a 
thodon gallicus, J. A. Hopson (personal figure 8; Bonaparte and Barberena, 1975), 
communication) observed a "very faint the inclusion of this taxon in the Therioh- 
swelling" in a similar position. erpetidae is questionable as well. 

Both a functional and a replacement Mitredon cromptoni might be consid- 

tooth are present at the Pc^ position. The ered a chiniquodontid cynodont, but un- 

functional tooth is preserved only by a pair resolved taxonomic issues at the familial 

of roots that straddle cusp a of a replac- and suprafamilial levels, as well as the in- 

ment tooth beneath. The fully divided completeness of the present specimen, 

roots are visible as ovoid cross sections at make such an assignment problematic, 

their broken surfaces in occlusal view. The Most taxa referred to "chiniquodonts" 

replacement tooth, ex-posed by preparation (Chiniquodontidae or Chiniquodontoidea) 

but partially obscured in labial view by a exhibit alternate tooth replacement and 

cast of the mandibular canal (Fig. IB), postcanines with three to four cusps that 

54 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

--. (PC,) 

1 cm 

Figure 4. Schematic reconstruction of the left lower jaw of Mitredon cromptoni, MGUH VP 3392. Most reptiles and many 
cynodonts exhibit a tooth replacement pattern characterized by tooth eruption in every second tooth position. This alternate 
replacement pattern results from narrow spacing between Zahnreihen (teeth in a developmental replacement series; Hopson, 
1980). As spacing between Zahnreihen increases, two or more teeth in a single Zahnreihe may be fully erupted and functional 
in the tooth row. Each tooth in a single Zahnreihe potentially may have a different crown morphology, and thus increased spacing 
between Zahnreihen may yield a heterogeneous tooth row. The dentition of M. cromptoni is represented by three Zahnreihen 
(heavy, shaded lines). The first Zahnreihe is comprised of erupting PCa and probably Pc, (represented by an empty alveolus). 
The second is comprised of the erupted PC2 (root fragment), functional PCj and PC4 (pairs of roots), and erupting PCj and PCe. 
The roots of PCg are part of the most distal Zahnreihe. 

are labiolingually coinpressed and inesio- 
distally aligned, features that are either 
primitive for cynodonts or widely distrib- 
uted (e.g., in galesaurids and cynognathids; 
Bonaparte and Barberena, 1975). Al- 
though Sigogneau-Russell and Hahn 
(1994: 204) assert that the teeth of "chi- 
niquodonts" are easily distinguished from 
those of mainmals and other therapsids, 
they also note that "chiniquodont" denti- 
tions have yet to be studied in depth and 
therefore the "problem of subdivision of 
the Chiniquodontoidea into families has 
yet to be resolved." 

Mitredon cromptoni also shares some 
similarities with the Early Jurassic Sino- 
conodon, which also has double-rooted 
teeth and likely replaced the molariform 
teeth (Crompton and Sun, 1985; Cromp- 
ton and Luo, 1993). The postcanine teeth 
of Sinoconodon, which have four mesio- 
distally aligned cusps, resemble the teeth 
of M. cromptoni in lateral profile. In Sin- 
oconodon postcanine teeth do not occlude, 
nor do upper and lower postcanines have 
a consistent relationship to one another; 
the single jaw of M. cromptoni, from 

which the erupted, functional teeth have 
been lost postmortem, does not permit an 
assessment of these features. In view of 
these uncertainties, we are reluctant to at- 
tempt a more precise taxonomic place- 
ment of M. cromptoni. 

Tooth Replacement 

Mitredon cromptoni possesses a lower 
postcanine dentition with at least three 
variants of crowii stioicture. Differences in 
the teeth of M. cromptoni appear to rep- 
resent different tooth replacement fami- 
lies, or Zahnreihen, comparable to those 
described for the cynodont Thrinaxodon 
liorhinus (Parrington, 1936; Crompton, 
1963c; Osborn and Crompton, 1973; Fig. 
4). However, tooth replacement in M. 
cromptoni is not comparable to the alter- 
nate pattern seen in T. liorhinus and allied 
forms. Furthermore, new generations of 
teeth erupt in the same sagittal plane as 
do previous ones (Pc^ and Pcv- eiiipt di- 
rectly below the intact roots of preceding 
teeth; Figs. 1, 4), not in a more lingual 
plane as in T. liorhinus. 

Although the number of crown variants 

Tooth Replacement and Double-Rootedness • Shapiro and Jenkins 55 

in each replacement series cannot be de- 
tennined, we would interpret tooth re- 
placement and variation in Mitredon 
cromptoni as representing three ZaJinreihe 
(Fig. 4). The most mesial Zahnreihe con- 
sists of Pci and the erupting Pc^. The next 
Zahnreihe begins mesially with the root 
remnant of the functional tooth at the Pco 
position. The functional teeth at positions 
Pcr^ (represented only by roots) are part 
of this second Zahnreihe, as probably also 
are tlie replacement teetli at positions Pc5_6. 
The last tooth in the most distal Zahnreihe 
(and thus the most distal tooth in the den- 
tition) always eiiapts de novo, in a position 
not previously occupied by another tooth. 
Lower postcanine 6, which is undergoing 
replacement, therefore cannot be the ul- 
timate tooth in the lower dentition of M. 
cromptoni and we would expect to see an- 
other erupting distal tooth in a more com- 
plete (and perhaps ontogenetically older) 
speciinen. Indeed, a shallow ciypt occurs 
distal to the roots of the functional tooth 
at PCfj, evidence of yet another tooth po- 
sition. The erupted Pcg and the potential 
tooth distal to it represent the most distal 

Functional Stability of the Tooth Row and 
Multirooted Teeth 

Before the discovery of Mitredon 
cromptoni, the fossil record appeared to 
provide evidence for the coevolution of 
stable patterns of occlusion and multiroot- 
ed teeth, with the possible implication of 
a functional relationship. In the primitive 
condition, exemplified by Thrinaxodon 
and many other cynodonts, alternate re- 
placement of single-rooted teeth resulted 
in continual disruption of the tooth row, 
and in any case there was little, if any, oc- 
clusion in the strict sense (tooth-to-tooth 
contact). More derived lineages (e.g., tri- 
tylodontids, Sinoconodon, morganucodon- 
tids, and haramiyids) developed replace- 
ment strategies to promote stability of the 
tooth row and, in most cases, possessed 
double- or multirooted postcanine teeth. 
An exception is the gomphodont cynodont 

Diademodon, which maintained single- 
rooted teeth but promoted stabilit)' within 
the tooth row by losing teeth mesially and 
adding teeth distally (Crompton, 1963a; 
Hopson, 1971). Tritylodontids (with up to 
six roots on postcanine teeth in Oligoky- 
phiis) and Sinoconodon (in which postcan- 
ines may be single- or double-rooted) pos- 
sessed tooth replacement patterns com- 
parable to that of Diademodon, but with a 
reduced number of teeth in each Zahn- 
reihe. Tritylodontids did not replace mesial 
teeth but instead added nonreplacing, 
"gomphodont" teeth de novo at the distal 
end of the row (Hopson, 1971). Similarly, 
Sinoconodon lost anterior postcanines and 
added sinall distal teeth, which were sub- 
sequently replaced by a second generation 
of larger ones as jaw size increased 
(Crompton and Luo, 1993). Thus, Sino- 
conodon neither followed the typical 
"mammalian" diphyodont tooth replace- 
ment pattern nor possessed true molars 
(that is, Sinoconodon did not bear teeth 
that erupted de novo distally in the tooth 
row and were not replaced by subsequent 
generations of teeth). 

Morganucodon was among the first 
inammals to possess a dentition that in- 
cluded true molars. Available fossils do not 
reveal how many times (or in what order) 
Morganucodon replaced generations of 
deciduous teeth, but dental wear patterns 
suggest that the positional relationships 
between upper and lower postcanine teeth 
were relatively consistent (Crompton and 
Jenkins, 1968). Fixed dental relationships 
were also promoted by interlocking ante- 
rior and posterior accessory cusps, main- 
taining alignment of the molars. Similarly, 
the occlusal interlocking of cusps and ba- 
sins on upper and lower molariforms of 
the haramiyid Haramiyavia clemmenseni 
(Jenkins et al., 1997) would have required 
ontogenetic stasis of the tooth row. Al- 
though the exact sequence of tooth re- 
placement cannot be ascertained for either 
of these mammalian taxa, their occlusal 
configurations are evidence that relatively 
precise relations were maintained between 

56 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

upper and lower dentitions throughout on- dence that this character evolved indepen- 
togeny. dently more times than previously recog- 
However, Mitredon cromptoni has a nized. In Morganiicodon and Ktiehneoth- 
double-rooted postcanine dentition but erium, premolar and molar roots range 
lacks ontogenetic stasis in the lower tooth from incompletely divided to widely diver- 
row. The tooth replacement pattern of M. gent; root shape also varies, froni straight 
cromptoni clearly does not retain the al- with a blunt or bulbous terminus, to those 
temate pattern of Thrinaxodon, nor are that are curved and tapered (Parrington, 
teeth exclusively lost mesially and added 1971, 1978). Root morphology in Sinocon- 
distally, as in gomphodonts, tritylodontids, odo7i also varies: some premolars (Zhang 
and Sinoconodon. Instead, M. cromptoni et al., 1998) and molars of Sinoconodon 
replaces mesial and distal postcanines si- are fully divided, but at least one example 
multaneously and thus compromises the is known w^here the two roots are conflu- 
stability of the tooth row. Of the teeth rep- ent beneath the crown, and are only nar- 
resented in MGUH VP 3392, the second rowly separated distally (Luo, 1994, fig. 
and fifth were being replaced and likely 6.6). Cui and Sun (1987) document exten- 
lacked functional predecessors at the time sive variability among tritylodontids, which 
of death. Thus, of the five tooth positions until recently were the only known Early 
represented by replacement teeth or func- Mesozoic cynodont clade (other than 
tional roots, only the third, fourth, and mammals) with multirooted teeth. In some 
sixth postcanines were occupied by func- taxa (e.g., Yiinnanodon) the roots are com- 
tional teeth. The low number of functional pletely separated, whereas in others {Oli- 
postcanines in MGUH VP 3392 contrasts gokijphiis and Lufengia) transverse sheets 
sharply with the "uninterrupted molari- of dentine connect root pairs. Bienother- 
form series" of mammals and gomphodont iinn exliibits differences in the degree of 
cynodonts (Hopson, 1971: 17). root separation along the upper postcanine 
^, ^ . . , . . , . . -,- row, with unseparated roots mesially and 
The Origins of Multirooted Teeth f^jl^ separated roots distally In general. 

Teeth with multiple roots have tradi- the number of roots in tritylodont teeth 

tionally been regarded as part of the func- varies with the number of cusps (Cui and 

tionally interdependent, coevolved com- Sun, 1987). Finally, to this spectrum of 

plex of the mammalian masticatoiy system, variation may be added Mitredon cromp- 

which includes such diagnostic features as toni. With a sectorial crown inoi"phology 

a dentary— squamosal jaw joint and unilat- and a tooth replacement pattern that is 

eral mastication (e.g., Crompton, 1963a, b; comparable to no known mammal, M. 

Barghusen and Hopson, 1970; Kiihne, cromptoni is best inteipreted as a cyno- 

1973; Crompton and Parker, 1978; dont with double-rooted teeth. 
Crompton, 1989). However, several fossil 

taxa challenge the notion that "mammaU- CONCLUSIONS 

an" characters are limited to the Mam- Advanced cynodonts may potentially ex- 
malia. For example, Shubin et al. (1991) hibit a mosaic of "mammalian" and "non- 
noted that unilateral occlusion may have mammalian" characteristics. Mitredon 
originated not with the Mammaliaformes cromptoni, Meurthodon gallicus, tritylo- 
but with the common ancestor of trithel- dontids, and mammals all possess postca- 
odontids and mammaliamoiphs, or may nine teeth with multiple roots. Previous 
have evolved independently in the Mam- analyses (e.g.. Sues, 1985; supported by 
maliaformes. Likewise, the presently Wible, 1991) point to an independent evo- 
known structural diversity of roots within lution of this character in tritylodontids. 
taxa, as well as the phylogenetic distribu- Hence, if M. cromptoni and Meurthodon 
tion of multirooted teeth, provide evi- gallicus are indeed "chiniquodonts" (no 

Tooth Replacement and Double-Rootedness • Shapiro and Jenkins 57 

recent analyses place chiniquodonts as the 
sister taxon of mammals), then multiple- 
rooted postcanines may have evolved up to 
three separate times in cynodonts (follow- 
ing the phylogenetic hypotheses of Hop- 
son, 1994; Hopson and Kitching, 2001). 
This character would no longer be useful 
in the diagnosis of Mammaliamorpha 
(Rowe, 1988) or Mammaliaformes (Wible, 
1991). Alternately, double-rootedness may 
have evolved only once, in the Mammali- 
aformes (Wible, 1991; Hopson and Kitch- 
ing, 2001). In this scenario, Mitredon and 
Meurthodon would fall within the Trithel- 
odontidae + Mammaliaformes clade, clos- 
er to mammaliaforms than to trithelodon- 

Double-rootedness did not necessarily 
evolve in concert with tooth row stasis dur- 
ing synapsid evolution. Mitredon cromp- 
toni has double-rooted teeth but retains a 
tooth replacement pattern uncharacteristic 
of taxa with precise occlusion and a func- 
tionally uninterrupted postcanine tooth se- 



We thank W W. Amaral, L. B. Clem- 
mensen, W. R. Downs, S. M. Gatesy, H. 
E. Jenkins II, D. V. Kent, D. C. Roberts, 
and N. H. Shubin for their spirited collab- 
oration in fieldwork; W W Amaral for his 
detailed preparation of the specimen; K. 
Brown-Wing for the precision of her ren- 
derings in Figure 1; and S. M. Gatesy and 
J. A. Hopson for helpful discussions. J. A. 
Hopson generously shared his unpub- 
lished drawings and observations oi Meur- 
thodon gallicus and provided useful in- 
sights in his review of the manuscript. We 
also thank Zhexi Luo for his thoughtful re- 
view. This work was supported by grants 
from the National Science Foundation, the 
Carlsberg Foundation, and the Putnam 
Expeditionary Fund of the MCZ. 


Barghusen, H. R., and J. A. Hopson. 1970. Den- 
tary-squamosal joint and tlie origin of mammals. 
Science, 168: 573-575. 

Bonaparte, J. F., and M. C. Barberena. 1975. A 
possible mammalian ancestor from the Middle 
Triassic of Brazil (Therapsida— Cynodontia). Jour- 
nal of Paleontology, 49: 931-936. 

Bonaparte, J. F., and A. W. Crompton. 1994. A 
juvenile probainognathid cynodont skull from 
the Ischigualasto Formation and the origin of 
mammals. Revista del Museo Argentino de Cien- 
cias Naturales "Bernardino Rivadavia," Palento- 
logi'a, 5: 1-12. 

Crompton, A. W. 1963a. The evolution of the mam- 
malian jaw. Evolution, 17: 431—439. 

. 1963b. On the lower jaw of Diarthrognathus 

and the origin of the mammalian lower jaw. Pro- 
ceedings of the Zoological Society of London, 
140: 697-750. 

. 1963c. Tooth replacement in the cynodont 

Thrinaxodon liorJiinus Seelev. Annals of the 
South African Museum, 46: 479-521. 

. 1989. The evolution of mammalian mastica- 

tion, pp. 23-40. In D. B. Wake and G. Roth 
(eds.). Complex Organismal Functions: Integra- 
tion and Evolution in Vertebrates. Chichester: 
John Wiley & Sons, viii + 451 pp. 

Crompton, A. W, and F. A. Jenkins, Jr. 1968. Mo- 
lar occlusion in Late Triassic mammals. Biologi- 
cal Reviews of the Cambridge Philosophical So- 
ciety, 43: 427^58. 

. 1979. Origin of mammals, pp. 59-73. In J. 

A. Lillegraven, Z. Kielan-Jaworowska, and W. A. 
Clemens (eds.), Mesozoic Mammals: The First 
Two-Thirds of Mammalian History. Berkeley: 
University of California Press, x + 311 pp. 

Crompton, A. W, and Z. Luo. 1993. Relationships 
of tlie Liassic mammals Sinoconodon, Morganu- 
codon oehleri, and Dinnetherium, pp. 30-44. In 
F. S. Szalay, M. J. Novacek, and M. C. McKenna 
(eds.). Mammal Phylogeny: Mesozoic Differen- 
tiation, Multituberculates, Monotremes, Early 
Therians, and Marsupials. New York: Springer- 
Verlag. x + 249 pp. 

Crompton, A. W, and P. Parker. 1978. Evolution 
of the mammalian masticatory apparatus. Amer- 
ican Scientist, 66: 192-201. 

Crompton, A. W, and A.-L. Sun. 1985. Cranial 
structure and relationships of the Liassic mam- 
mal Sinoconodon. Zoological Journal of the Lin- 
nean Society, 85: 99-119. 

CUI, C, AND A. Sun. 1987. Postcanine root system 
in tritylodontids. Vertebrata PalAsiatica, 25: 245- 

Godefroit, p., and B. Battail. 1997. Late Triassic 
c)niodonts from Saint-Nicolas-de-Port (north- 
eastern France). Geodiversitas, 19: 567-631. 

Hopson, J. A. 1971. Postcanine replacement in the 
gomphodont cynodont Diademodon, pp. 1-21. 
In D. M. Kermack and K. A. Kermack (eds.). 
Early Mammals. New York: Academic Press, xiv 
+ 203 pp. 

. 1980. Tooth function and replacement in 

58 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

early Mesozoic ornithischian dinosaurs: implica- 
tions for aestivation. Lethaia, 13: 93-105. 

1994. Synapsid evolution and the radiation 

of non-eutherian mammals, pp. 190-219. In D. 
R. Prothero and R. M. Schoch (eds.). Major Fea- 
tures of Vertebrate Evolution. Pittsburgh: The 
Paleontological Society. 270 pp. 

HOPSON, J. A., AND H. R. Barghusen. 1986. An 
analysis of therapsid relationships, pp. 83-106. In 
N. Hotton III, P D. Maclean, J. J. Roth, and C. 
Roth (eds.). The Ecology and Biology of Mam- 
mal-like Reptiles. Washington, D.C.: Smithsoni- 
an Institution Press, x + 326 pp. 

HOPSON, J. A., AND J. W. KiTCHlNG. 1972. A revised 
classification of cynodonts (ReptiHa: Therapsida). 
Palaeontologia Africana, 14: 71-85. 

. 2001. A probainognathian cynodont from 

South Africa and the phylogeny of nonmammal- 
ian cynodonts. Bulletin of the Museum of Com- 
parative Zoology, 156: 5-35. 

Jenkins, F. A., Jr., S. M. Gatesy, N. H. Shubin, and 
W. W. Amaral. 1997. Haramiyids and Triassic 
mammalian evolution. Nature, 385: 715-718. 

Jenkins, F. A., Jr., N. H. Shubin, W. W. Amaral, S. 
M. Gatesy, G. R. Schaff, L. B. Glemmensen, 
W. R. Downs, A. R. Davidson, N. Bonde, and 
F. OSBaeCK. 1994. Late Triassic continental ver- 
tebrates and depositional environments of the 
Fleming Fjord Formation, Jameson Land, East 
Greenland. Meddelelser om Gr0nland, Geosci- 
ence, 32: 1-25. 

Jenkins, F. A., Jr., N. H. Shubin, S. M Gatesy, and 
K. PadIAN. 2001. A diminutive pterosaur (Pter- 
osauria: Eudimoqohodontidae) from the Green- 
landic Triassic. Bulletin of the Museum of Com- 
parative Zoology, 156: 151-170. 

Kemp, T. S. 1982. Mammal-like Reptiles and the Or- 
igin of Mammals. London: Academic Press, xiv 
+ 363 pp. 

KUHNE, W. G. 1973. The evolution of a synorgan: 
nineteen stages concerning teeth and dentition 
from the pelycosaur to the mammalian condition. 
Bulletin dvi Groupement International pour la 

Recherche Scientifique en Stomatologie, 16: 

LUO, Z. 1994. Sister-group relationships of mammals 
and the transformations of diagnostic mammali- 
an characters, pp. 98-128. In N. C. Eraser and 
H.-D. Sues (eds.). In the Shadow of the Dino- 
saurs. Cambridge: University Press, x + 435 pp. 

OSBORN, J. W, AND A. W. CROMPTON. 1973. The 
evolution of mammalian from reptilian denti- 
tions. Breviora, 399: 1-18. 

Parrington, F. R. 1936. On the tooth-replacement 
in theriodont reptiles. Philosophical Transactions 
of the Royal Society of London, B, Biological 
Sciences, 226: 121-142. 

. 1971. On the Upper Triassic mammals. Phil- 
osophical Transactions of the Royal Society of 
London, B, Biological Sciences, 261: 231-272. 
1978. A further account of the Triassic mam- 

mals. Philosophical Transactions of the Royal So- 
ciety of London, B, Biological Sciences, 282: 

ROWE, T. 1988. Definition, diagnosis, and origin of 
Mammalia. Journal of Vertebrate Paleontology, 
8: 241-264. 

RUSSELL, D., D. Russell, and G. Wouters. 1976. 
Une dent d'aspect mammalien en provenance du 
Rhetien frangais. Geobios, 9: 377-392. 

Shubin, N. H., A. W. Crompton, H.-D. Sues, and 
P. E. Olsen. 1991. New fossil evidence on the 
sister-group of mammals and early Mesozoic fau- 
nal distributions. Science, 251: 1063-1065. 

Sigogneau-Russell, D., and G. Hahn. 1994. Late 
Triassic microvertebrates from central Europe, 
pp. 197-213. In N. C. Eraser and H.-D. Sues 
(eds.). In the Shadow of the Dinosaurs. Gam- 
bridge: University Press, x + 435 pp. 

Sues, H. D. 1985. The relationships of the Tritylo- 
dontidae (Synapsida). Zoological Journal of the 
Linnean Society, 85: 205-217. 

Wible, J. R. 1991. Origin of Mammalia: the craniod- 
ental evidence reexamined. Journal of Vertebrate 
Paleontology, 11: 1-28. 

Zhang, F K., A. W. Crompton, Z. Luo, and C. R. 
SGHAFF. 1998. Pattern of dental replacement of 
Sinoconodon and its imphcations for evolution of 
mammals. Vertebrata PalAsiatica, 36: 197-217. 



Abstract. Cranial and postcranial remains of the 
cynodonts Therioherpeton cargnini (Therioherpeti- 
dae) and a taxon originally referred to as Thrinaxodon 
brasiliensis (of indeterminate familial status) are de- 
scribed and compared wdth other advanced cyno- 
donts and morganucodontids. Our study provides ev- 
idence that these earlv Late Triassic cvnodonts, which 
possessed primitive carnivorous dentitions of tlie 
Thrinaxodon liorhiniis type, evolved derived charac- 
ters of the skull and postcranium that approximated 
the mammtilian level of organization as represented 
in morganucodontids. 


Therioherpeton cargnini (Therioher- 
petidae) is represented by an incoinplete 
skull and lower jaw, and most of the post- 
cranial skeleton, but only the skull and a 
fragment of the lower jaw have been de- 
scribed (Bonaparte and Barberena, 
1975). The genus Therioherpeton was re- 
ferred to the fainily "Therioherpetonti- 
dae" by Bonaparte and Barberena 
(1975), but Battail (1991) corrected the 
familial name to Therioherpetidae. Like- 
wise, Thrinaxodon brasiliensis (Barber- 
ena et al., 1987), of indeterminate famil- 
ial status, is also represented by an in- 
complete skull, lower jaws, and some 
postcranial bones, but only the skull and 
jaws were described. In this paper the 
complete material known from both spe- 
cies is analyzed and compared with other 
advanced Triassic cynodonts and mor- 

' Museo Argentino de Ciencias Naturales, Avenida 
Angel Gallardo 470, 1405 Buenos Aires, Argentina. 

- Institute de Geociencias, Universidade Federal 
de Rio Grande do Sul, Avenida Bento Gon^alves 
9500, 91509-900 Porto Alegre, Rio Grande do Sul, 

ganucodontids, and T. brasiliensis is as- 
signed to a new genus. The advanced an- 
atomical features of both taxa suggest a 
close proximity to the mammalian con- 
dition, not very different from that 
evolved by tritylodontids and tritheledon- 

Anatomical and phylogenetic issues 
concerning advanced cynodonts and 
primitive mammals have been widely an- 
alyzed (Kermack et al., 1981; Kemp, 
1982; Jenkins, 1984; Crompton and Sun, 
1985; Sues, 1985; Hopson and Barghu- 
sen, 1986; Battail, 1991; Hopson, 1991; 
Shubin et al., 1991; Crompton and Luo, 
1993; Luo, 1994), but the subject is far 
from well understood because, as Cromp- 
ton and Luo (1993: 30) remarked: "A lack 
of moi-phological information makes it 
difficult to intei-pret the relationships of 
advanced cynodonts and early mammals." 
As yet undescribed carnivorous cynodonts 
(recently discovered in southern Brazil in 
beds of early Late Triassic age), and a re- 
study of the advanced cynodonts Therio- 
herpeton cargnini (Bonaparte and Barber- 
ena, 1975) and Thrinaxodon brasiliensis 
(Barberena et al., 1987), may advance our 
understanding of the complex sequence of 
anatomical changes that occurred in cy- 
nodonts and that foreshadow the level of 
inammalian organization represented by 
morganucodontids (Kermack et al., 1973, 
1981; Crompton, 1974; Jenkins and Par- 
rington, 1976; Cow, 1986). Unfortunately, 
available cranial material of neither Thri- 
naxodon brasiliensis nor Therioherpeton 
cargnini includes critical anatomical data 
from the basicranial region. 

Bull. Mus. Comp. ZooL, 156(1): 59-80, October, 2001 59 


Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


Therapsida Broom, 1905 

Cynodontia Owen, 1861 

Family Therioherpetidae Bonaparte and 
Barberena, 1975 

Diagnosis. Differs from the Thrinaxo- 
dontidae by a more extensive secondary 
palate. Differs from Thrinaxodontidae, Cy- 
nognathidae, Chiniquodontidae, and Pro- 
bainognathidae in the following featui-es: 
triangular cross section of the 2ygoinatic 
arch; absence of a postorbital bar; frontal 
borders the orbit and bears an anterolat- 
eral projection that contacts a lateral por- 
tion of the nasal; interorbital wall more 
highly ossified; larger size of the neural ca- 
nal in the cervical and dorsal vertebrae; 
parallel dorsal and ventral borders of the 
ilium; convex lateral surface of ilium; a 
narrow, elongate neck of the ischium; ob- 
turator foramen relatively large; greater 
trochanter of the femur extends to the 
same proximal level as the femoral head. 
Therioherpetidae differs from Trithele- 
dontidae in the transverse narrowness of 
the upper postcanines. Therioherpetidae 
differs from gomphodont cynodonts, in- 
cluding Tritylodontidae, in possessing tri- 
conodont upper and lower postcanines, 
and in the triangular cross section of the 
zygomatic arch. Therioherpetids differ 
from basal mammals in lacking an articular 
condyle of the dentary and fully bifurcated 
roots on postcanine teeth (but see Shapiro 
and Jenkins, 2001). 

Genus Therioherpeton Bonaparte and 
Barberena, 1975 

Type Species Therioherpeton cargnini 
Bonaparte and Barberena, 1975 

Holotype. An unnumbered specimen in 
the private collection of the Patronato Al- 
ves Ramos, Santa Maria City, State of Rio 
Grande do Sul, Brazil. An incomplete skull 
lacking the basicranium, and preserving 
only the right upper fifth postcanine; a 
fragment of the right dentary; an isolated 

lower postcanine (the fifth or sixth, at- 
tached to lingual surface of Pc^); 29 artic- 
ulated, incomplete vertebrae including 4 
cervicals, 15 dorsals and (separated by a 
gap) 4 sacral and 6 caudal vertebrae; as- 
sociated, incomplete ribs; left scapular 
blade; distal half of the right humerus; in- 
complete right radius and ulna; incom- 
plete ilia; complete pubis; right ischium; 
complete left and incomplete right femur; 
fragments of tibiae and fibulae; and in- 
complete feet. 

Revised Generic and Specific Diagnosis. 
The triconodont upper and lower postca- 
nines are without cingula. The secondary 
bony palate extends nearly to the level of 
the last postcanine, as in chiniquodontids 
and Probainognathus. As in Morganuco- 
don, the frontals have an extensive poste- 
rior projection, anteriorly contact the lat- 
eral aspect of the nasals, and anterolater- 
ally project to the lacrimals. Prefrontals 
and postorbitals are absent. The cervical 
vertebrae are craniocaudally short, trans- 
versely wide, dorsoventrally low, and ex- 
hibit a very large neural canal. The iliac 
blade lacks a posterior projection, the lat- 
eral surface is convex, and the dorsal and 
ventral borders are subparallel. The obtu- 
rator fenestra is large. The greater tro- 
chanter extends proximally to the level of 
the femoral head, with which it is con- 
nected by a sheet of bone. 

Horizon and Locality. Upper Santa Ma- 
ria Formation. A road cut on the BR-216 
highway (outcrop BR- 14 in Bortoluzzi and 
Barberena, 1967), 200 m northwest of 
Cerriquito, Township of Santa Maria, Rio 
Grande do Sul, Brazil. 

Age. Probably early Late Triassic. 


Skull (Figs. 1—4). Only the salient char- 
acters of the type and only known skull of 
this species, originally described by Bon- 
aparte and Barberena (1975), may be not- 
ed here. The more significant characters 
are the absence of the prefrontal and post- 
orbital, the large lacrimal, the frontal bor- 
dering the orbit with a long posterior pro- 

Advanced Triassic Cynodonts from Brazil • Bonaparte and Barberena 61 


8 mm 


8 mm 


Figure 1 . Therioherpeton cargnini. Skull and dentary fragment in (A) right lateral and (B) dorsal views. (C) Incomplete interorbital 
wall viewed from the left side. (D) Upper postcanine 5 in occlusal, lingual, and buccal views, and Pc5o,6 in buccal and occlusal 
views. Abbreviations: AL, alisphenoid; D, dentary; FM, maxillary foramina; FR, frontal; J, jugal; L, lacrimal; MX, maxilla; N, nasal; 
PAL, palatine; P, parietal; PP, prearticular process; PR, prootic; SQ, squamosal. 

cess, and an anterolateral projection of the 
frontal in dorsal view. The parietal crest is 
low (Fig. lA) and the dorsal area of the 
braincase is large (Fig. 3); the z)'gomatic 
arch is slender with a triangular cross sec- 
tion (Figs. lA, 4). The lateral surface of 
the maxilla bears three large foramina 
(Fig. lA). The upper and lower postcani- 
nes (Fig. ID) are of the triconodont type 
and lack cingula, and the upper teeth show 
clear indications of an incipient bifurcation 
of the roots. 

The secondary bony palate of Therio- 

herpeton was misinterpreted by Bonaparte 
and Barberena (1975) because an unossi- 
fied, or damaged, area of the right palatine 
was considered as part of the internal na- 
res. Restudy of the palatine revealed that 
its posterior margin is in fact complete, 
and thus represents the posterior border 
of the secondaiy bony palate, which is in 
line with the penultimate postcanine (Fig. 

Postcranial Skeleton (Figs. 5-8). The as- 
sociated postcranium was found in nearly 
articulated condition lying on the external 

62 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 2. Therioherpeton cargnini. Ventral view of the incomplete skull of the holotype. The secondary osseous palate shows 
an unossified area of the palatine. However, the posterior border of the right palatine is well preserved. Most of the ventral side 
of the skull roof from nasals to parietals is shown. 

side of the rib cage of a rhynchosaur, prob- 
ably Scaphonijx sp. (Schultz, 1986). All but 
six vertebrae are incomplete. Three cei"vi- 
cals are articulated in a small, isolated 
block, associated with the blade of the 
right scapula. The centra of these verte- 
brae are anteroposteriorly short, trans- 
versely wide, and dorsoventrally low, and 
have a wide neural canal. The inferred last 
cervical, articulated with a series of 15 dor- 
sals (Fig. 5A), also has a short, wide, and 
dorsoventrally low centrum; the neural ca- 

nal is wider than the centrum because the 
pedicles of the neural arch project dorso- 
late rally. 

The anteroposterior lengths of the seven 
anterior dorsals gradually increase; neural 
canals remain very large (Fig. 5A). The 8th 
through 10th dorsals preserve the neural 
spines, which are posterodorsally inclined. 
The centra of the 11th through 14th dor- 
sals are longer than those of preceding 
vertebrae; in these vertebrae, the large 
neural canal is formed in part by the ven- 

Figure 3. Therioherpeton cargnini. Dorsal view of the incomplete skull of the holotype, showing the anterolateral projection of 
the frontals as well as their extensive, wedge-shaped posterior projection. 

Advanced Triassic Cynodonts from Brazil • Bonaparte and Barberena 63 

Figure 4. Therioherpeton cargnini. Right lateral view of the incomplete skull and jaw of the holotype. Note the slender zygomatic 
section of the jugal, and the incipiently bifurcated root of Pc^ 

tromedial surfaces of the neural pedicles 
that contribute to the floor of the neural 
canal. The number of dorsal vertebrae is 
uncertain because of a gap between the 
fifteenth dorsal and the next group of ver- 
tebrae. The zygapophyses of the last dorsal 
vertebrae are anteroposteriorly robust and 
oriented almost horizontally, with little in- 
clination toward the median plane. The 
vertebrae forming the sacrum are difficult 
to discern individually, although one bears 
fragments of sacral ribs. We infer that 
there might be three or possibly four sa- 
cral vertebrae. All are rather robust, and 
the neural canals are as wide as the centra. 

Subcylindrical fragments (and many 
molds) preserve 15 incomplete ribs on the 
right side and 17 on the left side (Figs. 5A, 
6). There is no evidence of overlapping 
uncinate processes as are known in many 
cynodonts. Cervical ribs, which are short, 
thick, and posteriorly deflected, articulate 
behind the anterior margins of the centra, 
not between adjacent centra as in Thri- 
naxodon liorhinus (Jenkins, 1971). The 
dorsal ribs, which in cross section are fig- 
ure 8-shaped, exliibit no clear indication 
of a lumbar region, except that the poste- 
rior dorsal ribs gradually decrease in 

The pectoral girdle is represented only 

by the blade of the right scapula (not fully 
prepared at present). The anterior and 
posterior borders of the blade project lat- 
erally, forming a deep sulcus for musculus 
supracoracoideus, similar to the condition 
in Thrinaxodon liorhinus (Jenkins, 1971). 
The distal half of the right humerus is pre- 
served, and appears to be relatively prim- 
itive by virtue of its great distal width and 
the presence of an ectepicondylar fora- 
men. The Pright radius and ulna are in- 
completely preserved and reveal no diag- 
nostic characters. 

Of the two incomplete ilia, the left ilium 
(Fig. 5B) preserves part of the blade, 
which has almost parallel dorsal and ven- 
tral borders and lacks a posterior process. 
The lateral aspect of the blade is dorso- 
ventrally convex. The pubic pedicle is 
more developed and stronger than the 
area of ischial contact. The ischium (Fig. 
5D) has an elongate, narrow neck, and 
forms wdth the pubis a large obturator fe- 
nestra. The pubis (Fig. 5C) exliibits a wefl- 
defined, "twisted" neck below the acetab- 
ulum. A thickening of the anterior margin 
of the pubis ventral to the neck represents 
a pubic process. The inferior border of the 
pubic ramus is rather straight, whereas the 
superior border is concave and forms part 

64 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

5 mm 

Figure 5. Therioherpeton cargnini. (A) Sequence of articulated incomplete vertebrae and ribs from the putative last cervical (at 
the bottom) to the 10th dorsal. (B) Incomplete left ilium in lateral (left) and ventral (right) views. (C) Both pubes in dorsal view 
as preserved, and the left pubis in lateral view. (D) Right ischium in lateral (left) and medial (right) views. (E) Reconstruction of 
the right half of the pelvis in lateral view. Abbreviations: AA, acetabular area; ANP, area for neural pedicles; CC, cervical centrum; 
CSA, crista supracetabularis; IP, ischial plate; IT, ischial tuberosity; NC, neural canal; NF, neural fossa; Nl, neck of the ischium; 
NSP, neural spine; OBT, obturator foramen; PP, pubic pedicle; PPR, pubic process; R, rib; SA, symphysial area. 

of the margin of the large obturator fe- 

The left femur is nearly complete (Figs. 
7A, 8). The proximal end gradually ex- 
pands mediolaterally, and the trochanters 
are less defined than in Oligokijphus (Kiih- 
ne, 1956) and Morganiicodon (Jenkins and 
Parrington, 1976). The femoral head is an- 
teromedially and somewhat dorsally di- 

rected, although the precise orientation is 
obscured by slight deformation. The prox- 
imally positioned lesser trochanter is sim- 
ilar to that of Oligokijphus and Morganii- 
codon, and different from the more ven- 
trally placed trochanters of Probelesodon 
(Romer and Lewis, 1973), Massetognatlins 
(Jenkins, 1970), and Exaeretodon (Bona- 
parte, 1963). The greater trochanter is 

Advanced Triassic Cynodonts from Brazil • Bonaparte and Barberena 65 

Figure 6. Therioherpeton cargnini. Part of the presacral vertebral column showing broad neural arches, posterodorsally directed 
neural spines, and ribs without overlapping processes. 


5 mm 

Figure 7. Therioherpeton cargnini. (A) Left femur in anterior and lateral views. (B) Distal portions of right tibia and fibula 
articulated with the incomplete foot in plantar view. Abbreviations: AST, astragalus; CAL, calcaneum; ENC, entocuneiform; F, 
fibula; FH, femoral head; GT, greater trochanter; LC, lateral condyle; LT, lesser trochanter; NAV, navicular; SCG, supracondylar 
groove; T, tibia; ll-V, metatarsals II through V. 


Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 8. Therioherpeton cargnini. On the left is a series of incomplete sacrocaudal vertebrae in dorsal view. Also present are 
the left femur, left pubis, incomplete left ilium, and incomplete left foot. The ilium rests on a rhynchosaur {Scaphonyx sp.) rib. 

proximally positioned and is united with 
the femoral head by a thin lamina of bone. 
Morphologically, the greater trochanteric 
structure is similar to that of Oligokijphus 
and Morganiicodon, and very different 
from the condition in Cijnognathus (Jen- 
kins, 1971), Prohelesodon (Romer and 
Lewis, 1973), Massetognathiis (Jenkins, 
1970), and Exaeretodon (Bonaparte, 
1963). The lateral femoral condyle has a 
larger radius of curvature than the medial; 
a shallow supracondylar groove is present 
on dorsal surface of the distal end. 

The astragalus, which is larger than the 
calcaneum, is only partially superimposed 
on the latter (Fig. 7B). Inasmuch as the 
calcaneum is damaged along its proximal 
margin, the presence or absence of a cal- 
caneal tuber cannot be determined. The 
relatively elongate, slender proportions of 
the metatarsals resemble those of Oligo- 
kijphus (Kiihne, 1956), rather than those 
of Exaeretodon (Bonaparte, 1963) or 
"PAleodon/?Scalenodon" (Jenkins, 1971). 

A Stimmanj of the Distinctive Features 
of Therioheipeton. Although Therioher- 
peton possesses many features that are 
found in other Triassic cynodont families, 
the following assemblage is characteristic 
only of Therioherpeton and represents the 

degree to which this taxon approaches the 
morganucodontid level of organization. 

1) Absence of prefrontal and postorbital 
bones as well as a postorbital bar (also 
in tritheledontids and tritylodontids). 

2) Frontal bordering the orbit (also in 
tritheledontids and tritylodontids). 

3) Frontal with anterolateral projection 
in dorsal view, and an elongate pos- 
terior projection. 

4) Large lacrimal (also in tritheledontids 
and tritylodontids). 

Comment. The first three characters 
listed above differentiate Therioherpeton 
from the Thrinaxodontidae, Cynognathi- 
dae, Chiniquodontidae, and Probaino- 
gnathidae. Some of the characters are 
shared with the Tritylodontidae (Kiihne, 
1956; Sun, 1984; Sues, 1985) and Trithe- 
ledontidae (Crompton, 1958; Bonaparte, 
1980), but in tritylodontids the zygomatic 
arch is deep and the dentition very de- 
rived, whereas in tritheledontids the spe- 
cialized incisors, reduced canines, and bul- 
bous upper postcanines differ from the 
corresponding features in Therioherpeton. 
The upper postcanines of the tritheledon- 
tids are transversely expanded, whereas in 
Therioherpeton they are narrow, with the 

Advanced Triassic Cynodonts from Brazil • Bonaparte and Barherena 6' 

cusps in line. The structure of the frontal 
resembles that of Morganucodon (Ker- 
mack et al., 1981) in the type of contact 
^^dth the nasal, a possible result of the ab- 
! sence of the prefrontal, as well as in the 
j long, tapering contact with the parietals. 

i 5) Contact between the ventral process 
of frontal and dorsal process of pala- 
tine (also in tritheledontids and trity- 
lodontids; and in Probainognathus 
and chiniquodontids; J. A. Hopson, 
personal communication). 

6) Large infraorbital foramen and two 
well-defined foramina for the trigem- 
inal nei^ve in the maxilla (also in tri- 
theledontids and tritylodontids). 

7) Carnivorous— insectivorous dentition, 
similar to that of Thrinaxodon and 
Morganucodon, but with upper post- 
canines without cingula and possess- 
ing incipiently bifurcated roots. 

8) Articular process of the dentary pos- 
teriorly and transversely expanded, 
without indication of a condyle, and 
set at a higher level than the alveoli (a 
common feature in derived cyno- 
donts; J. A. Hopson, personal com- 

9) Cervical centra anteroposteriorly 
short, transversely wide and dorsoven- 
trally low (also in Oligokijphus, other 
tritylodontids, and Morganucodon) . 

10) Neural canal of presacral vertebrae 
wider than the centi-um. 

Comment. The neural canal in Therio- 
herpeton is proportionally larger than that 
in most cynodonts wdth which we were 
able to make a comparison. Neural canal 
size in the cervical, thoracic, sacral, and 
proximal caudal vertebrae of Oligokijphus 
(Kiihne, 1956) approaches that of Therio- 
het-peton, but is nonetheless proportionally 
smaller. The neural canal of Therioherpe- 
ton is in fact almost identical in propor- 
tions to that in Morganucodon (Jenkins 
and Parrington, 1976). 

11) Absence of anapophyses. 

12) Ribs without expanded processes (also 

in Exaeretodon, chiniquodontids, Pro- 
bainognathus, and tritylodontids; and 
in tritheledontids as well; J. A. Hop- 
son, personal communication). 

13) Neural spines of presacral vertebrae 
posterodorsally directed (also in trity- 

Comment. Short, posterodorsally direct- 
ed neural spines in the posterior dorsals 
are known only in Oligokijphus and Ther- 
ioherpeton. In Morganucodon, the poste- 
rior dorsals bear vertical neural spines, 
with fully differentiated lumbar vertebrae. 
The similarities between some derived ax- 
ial characters in Oligokijphus and Therio- 
herpeton suggest the probability of parallel 

14) Iliac blade with dorsal and ventral 
borders subparallel, without posterior 
process (also in some tritylodontids; 
and tritheledontids; J. A. Hopson, per- 
sonal communication). 

15) Lateral side of the iliac blade dorso- 
ventrally convex (also in tritylodontids 
and tritheledontids; J. A. Hopson, per- 
sonal communication). 

16) Ischium with narrow neck posterior to 
the acetabulum, and a concave dorsal 
border (also in Oligokijphus and tri- 
theledontids; J. A. Hopson, personal 
communication ) . 

17) Large obturator fenestra (also in tri- 
tylodontids and tritheledontids; J. A. 
Hopson, personal communication). 

18) Pubis narrow with reduced distal con- 
tact with the ischium (also in trithe- 
ledontids; J. A. Hopson, personal 

Comment. The available parts of the 
ilia, the complete pubes, and the ischium 
show that the pelvis of Theriohetyeton is 
more derived than that in any known cy- 
nodont family except the Tritylodontidae 
(Kuhne, 1956) and Trithelodontidae (J. A. 
Hopson, personal communication). 

19) Greater trochanter at the same level 
as the femoral head (also in Oligoky- 


Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

phiis and tritheledontids; J. A. Hop- 
son, personal communication). 
20) Elongate, slender metatarsals (also in 
Oligokyplms) . 

Therapsida Broom, 1905 
Cynodontia Owen, 1861 
Family Incertae Sedis 
Genus Prozostrodon New Genus 

Type species, Prozostrodon brasiliensls (Barberena, 
Bonaparte, and Sa Teixeira, 1987). 

Synonymy Thrinaxodon brasiliensis Barberena, Bon- 
aparte, and Sa Teixeira, 1987. 

Holotype. PV 0248T, Department of Pa- 
leontology and Stratigraphy, Universidade 
Federal de Rio Grande do Sul, Brazil. 

Etymology. The generic designation al- 
ludes to the cingula of lower postcanine 
teeth (Greek, zoster, a girdle or band), in 
combination with Latin, pro-, before, in 
reference to the hypothesized antecedant 
phylogenetic position of the taxon. 

Revised Generic and Specific Diagnosis. 
Reduced prefrontal and postorbital; lacri- 
mal with large dorsal exposure; pro- 
nounced posterodorsal process of the pre- 
maxilla between septomaxilla and maxilla. 
Posterior projection of the frontal shorter 
than in Therioherpeton; anterolateral pro- 
cess of the frontal shorter than in Therioh- 
erpeton, and contacts the posterolateral 
border of the nasal, as in Therioherpeton 
and Morganiicodon. Frontal, palatine, and 
orbitosphenoid extensively contact one an- 
other in the orbital wall. Five conical up- 
per incisors; four lower incisors slightly 
spatulate and procumbent (as in Morgan- 
iicodon). Triconodontlike postcanines 
without well-defined cingula on the uppers 
(except for an incipient buccal cingulum 
on the distal upper postcanine, as in Thri- 
naxodon liorhinus and chiniquodontids); 
lingual cingula on lower postcanines bear 
up to nine small cusps (as in Thrinaxodon 
liorhinus). Length of lower tooth row 
more than half the length of the dentary 
(as in Morganiicodon). Secondary bony 

palate extends posteriorly beyond the last 
upper postcanine, as in chiniquodontids 
and tritheledontids. Neural canal of the 
presacral vertebrae large (as in Oligopky- 
phiis), but smaller than in Therioherpeton. 
Neural spines posterodorsally inclined (as 
in tritylodontids). Zygapophyses of poste- 
rior dorsal vertebrae anteroposteriorh 
elongated, with anterior and posterior fac- 
ets that are transversely concave and con- 
vex, respectively. Ribs without expanded 
processes. Iliac blade with a vestigial pos- 
terior process and a convex lateral surface 
(as in Therioherpeton and tritylodontids; 
also in tritheledontids; J. A. Hopson, per- 
sonal communication). 

Horizon and Locality. Facies Alemoa of 
the Santa Maria Formation, 200 m north- 
west of the hill Cerriquito, in a road cut 
of route BR-216, Municipio of Santa Ma- 
ria, State of Rio Grande do Sul, southern 

Age. Early Late Triassic. 

Material. An incomplete skull lacking 
most of the parietal crest, the braincase, 
and zygomatic arch. The orbital, preorbit- 
al, and infraorbital regions, secondary 
bony palate, and upper dentition are near- 
ly complete. The right dentary and denti- 
tion are complete; the left dentary also has 
a complete dentition but lacks the ascend- 
ing ramus and articular process. The post- 
cranium is represented by three incom- 
plete presacral vertebrae, 14 dorsal centra, 
seven dorsal neural arches, several frag- 
mentary ribs, interclavicle, incomplete 
right humerus, proximal half of the left hu- 
merus, incomplete right ilium, distal 
halves of both femora, and a disarticulated 
right foot. 

Comments. Significant features of the 
skull have been more clearly exposed 
through recent preparation. The skull 
shows some postmortem cracking and dis- 


Skull {Figs. 9-11). The prefrontal and 
postorbital are reduced, and there is no 
indication of a postorbital bar (Fig. 9). The 

Advanced Triassic Cynodonts from Brazil • Bonaparte and Barberena 69 



Figure 9. Prozostrodon braslliensis, new genus. Incomplete skull in (A) dorsal, (B) ventral, and (C) left lateral views. (D) 
Incomplete right lower jaw in medial view. Abbreviations: Al, alveolus for incisor; ALS, alisphenoid; CO, coronoid; D, dentary; 
FE, ethmoidal foramen; FM, maxillary foramina; FPB, fossa for postdentary bones; FR, frontal; GDL, groove of dental lamina; 
L, lacrimal; MX, maxilla; N, nasal; OBS, orbitosphenoid; P, parietal; PA, prearticular; PAL, palatine; PAP, prearticular process; 
PRF, prefrontal; PMX, premaxilla; PO, postorbital; PT, pterygoid; RPC, replacing postcanine; SF, symphysial foramen; SGR, 
symphysial groove; SMX, septomaxilla; SPL, splenial; SPT F, sphenopalatine foramen. 

70 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 10. 
is sliown. 

Prozostrodon brasillensis, new genus. Dorsal view of the holotype sl<ull. The anterolateral projection of the frontal I 

frontal borders the orbit and extends an- 
terolaterally to contact the lateral, poste- 
rior portion of the nasal, as in Therioher- 
peton. The orbital process of the frontal 
contacts the dorsal process of the palatine 
and the orbitosphenoid, similar to that of 
Morganucodon (Kermack et al., 1981, fig. 
96). The sphenopalatine foramen in Mor- 
ganucodon is bordered by the palatine and 
orbitosphenoid (Kermack et al., 1981), 
whereas in Prozostrodon it is entirely en- 
closed by the palatine, possibly the prim- 
itive condition, with the orbitosphenoid in 

a more posterior position. Despite this dif- 
ference, it is interesting to note the struc- 
tural similarities of the interorbital regions 
of Morganucodon and Prozostrodon, with 
the exception that the distance between 
the ethmoidal foramen (Fig. 9C) and the 
anterior border of the orbit is proportion- 
ally shorter in Morganucodon, possibly 
representing a derived condition. 

A small prefrontal is present (Fig. 9A), 
anteriorly reduced by the posterior expan- 
sion of the lacrimal. A small orbital process 
of the prefrontal is possibly present, con- 

Figure 1 1 . Prozostrodon brasiliensis, new genus. Ventral view of the holotype skull. 

Advanced Triassic Cynodonts from Brazil • Bonaparte and Barberena 71 

tacting a similar process of the frontal, but comprises five incisors, one canine, and 

it cannot be positively identified because seven postcanines (Fig. 9C). The incisors 

the sutures are not clear. are slender and posteriorly recurved. All 

The postorbital is represented only on five incisors, which are slender and slightly 
the left side (Fig. 9C) as a small plate over- recurved, are present on the right; four in- 
lapping the frontal and parietal on their cisors and an alveolus for the fifth incisor 
dorsolateral surfaces. The bone has some are present on the left, 
external sculpturing, but there is no indi- In right upper postcanine row, the third 
cation of the postorbital bar. and seventh (last) teeth were in the pro- 

The premaxilla has a narial process con- cess of erupting. Postcanine crowns are of 
tacting the nasal (Fig. 9C). The lateral pos- the "triconodont" type, with cusps aligned 
terodorsal process contacting the maxilla is mesiodistally. In the three posterior post- 
slender and is partially covered by the canines, four cusps (A, B, C, and D) are 
maxilla. The premaxilla has a long contact present. The labial side is mesiodistally 
with the septomaxilla and does not reach convex, whereas the lingual face is rather 
the nasal. concave or flat, with some ill-defined wear; 

The slender septomaxilla of Prozostro- the lack of well-defined facets suggests 

don (Fig. 9C) appears similar to that of that no precise occlusion was present. The 

Sinoconodon (Crompton and Luo, 1993), mesiolingual corner of the last left post- 

although it is uncertain whether a septo- canine bears a poorly defined, low cusp in 

maxillary foramen is present or not. a position that might be expected of an 

The rather large maxilla projects later- incipient lingual cingulum. Conversely, the 
ally over the premaxilla up to the level of last right postcanine (in the process of 
the third incisor (Fig. 9C). A posterior pro- erupting) bears a small buccal cusp distal- 
cess of the maxilla forms the anterior por- ly; there is no indication of buccal cingular 
tion of the zygomatic arch. A large infra- development in the functional postca- 
orbital foramen lies below the anterior nines. The occurrence of cingular cuspules 
margin of the lacrimal, and two well-de- only on the last upper postcanine in Fro- 
fined maxillary foramina are present above zostrodon is perhaps comparable to the 
the anterior border of the first postcanine. condition in Thrinaxodon liorhimis (Os- 

The anteroposteriorly elongate palatines bom and Crompton, 1973; also suggested 
of Frosos^roffon (Fig. 9B) resemble a sim- for Pachygenelus; Cow, 1980) in which 
ilar configuration in the bony palate of chi- morphological complexity increases to- 
niquodontids and morganucodontids. Al- wards the distal end of the tooth row. The 
though the posterior border of the right roots of the upper postcanines show evi- 
palatine is broken, the left side is com- dence of incipient bifurcation, as originally 
pletely preserved and extends posterior to indicated by Barberena et al. (1987). 
the last postcanine; this condition is similar Lower Jaw and Dentition (Figs. 9D, 12). 
to that in Probelesodon kitchingi (Sa Teix- The dorsal and ventral margins of the 
eira, 1979) and trithelodontids, and is a lit- elongate body of the mandible (Fig. 9D) 
tie more extensive than in Therioher})efon. are subparallel as far forward as the pos- 
The posterior half of the secondary bony terior border of the symphysis. Anterior to 
palate is widest posteriorly where the tooth this point, the dentary extends anterodor- 
row diverges posterolaterally. There is a sally, elevating the alveolar plane of the in- 
deep sulcus along the palatine-maxilla su- cisors and canine above that of the postca- 
ture to accommodate the crowns of the nines. 

lower postcanines when the jaw is closed. The articular process is transversely ex- 
Greater and lesser palatine foramina are panded, mostly medially, without evidence 
present. of an articular condyle. The postero ventral 
Upper Dentition. The upper dentition angle of the dentary is broadly convex. 

72 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 12. Prozostrodon brasiliensis, new genus. Right lower jaw of the holotype in nnedial view. Note the incipiently bifurcatec 
postcanine roots. 

lacking a posteriorly projecting angular 
process. A coronoid is probably present, 
largely fused to the medial side of the as- 
cending ramus. The unfused symphysis ex- 
hibits rugosity indicative of ligamentous 
insertion, and a symphysial fossa and fo- 
ramen as in Cynognathus (Kermack et al., 

The lower dentition consists of four in- 
cisors, one canine, and 10 postcanines 
(Barberena et al., 1987). The incisors are 
slightly procumbent and recurved, with 
some mesiodistal expansion of the crown 
in the first three. The fourth incisor is 
shorter, mesiodistally narrow, and separat- 
ed from the canine by a short diastema. 

The well-preserved postcanines become 
more complex toward the back (as is com- 
mon in carnivorous— insectivorous cyno- 
donts; Osborn and Crompton, 1973) as 
well as increase in size. The following ac- 
count is based on the right postcanines. 
Lower postcanine 1 is small, conical, and 
without accessory cusps. Lower postcanine 
2 bears a large main cusp a with smaller 
cusps b (mesially) and c (distally). A slight 
lingual cingulum is present. On Pcg, cusp 
a has a sharp distal edge, and cusps b and 
c are positioned on the buccal half of the 
tooth. Incipient division of the root is ap- 
parent in buccal aspect. Lower postcanine 

4, substantially larger than preceding post- 
canines and also exhibiting incipient root 
division, possesses a (broken) lingual cin- 
gulum and a mesial and distal lingual cus- 
pule. Lower postcanine 5 is mesiodistally 
longer than Pc^ and has larger accessory 
cusps b and c. The lingual cingulum is not 
continuous mesiodistally, but has anterior 
and posterior sections, each bearing a cus- 
pule. The base of the crown is larger than 
the root, which is well differentiated from 
the crown and exliibits a vertical groove. 
Lower postcanine 6, which is almost fully 
erupted, bears a conical, recurved cusp a, 
and a cusp d on the distal margin. Cusp c 
is larger than cusp b. The cingulum, which 
is more developed than in the preceding 
postcanines, extends along most of the lin- 
gual side and bears six cuspules. Cusp a is 
broken on Pc-, but cusps b, c, and d are 
present and aligned with the base of cusp 
a. The lingual cingulum is continuous. The 
occlusal surface of the cingulum and the 
surface above it suggest abrasion by food 
while chewing. The root is deeply grooved. 
Lower postcanine 8 is complete, with ba- 
sically the same features as in Pc-^. Lower 
postcanine 9, mesiodistally the longest 
postcanine, has a proportionally lower 
crown as well as the lowest cusp a (on the 
left side, Pcg appears to be in the final 

Advanced Triassic Cynodonts from Brazil • Bonapane and Barberena 73 

5 mm 

10 mm 

Figure 13. Prozostrodon brasiliensis, new genus. (A) Cervical centrum in dorsal view. (B) Anterior dorsal vertebra in lateral 
view. (C, D, E) Lateral views of three neural arcties from the dorsal series. (F) Incomplete ribs. Abbreviations: DP, diapophysis; 
?DP, ?diapophysis; NP, contact for neural pedicle; NSP, neural spine; PP, parapophysis; PRZ, prezygapophysis; PZ, postzy- 
gapophysis; PRZP, prezygapophysial process. 

Stage of eruption). Seven small cusps form jected dorsolaterally, the neural canal 

the lingual cingulum. The degree of incip- 
ient root bifurcation is greater than in pre- 
ceding postcanines, representing a mor- 
phological gradient that is also expressed 
in the increasing complexity of the cingula. 
The right Pc^o is unei"upted; cusp a and 

would have been relatively wide. The par- 
apophyseal facets are located entirely on 
the anteroventral margin of the centnjm; 
there is no indication that rib heads 
spanned adjacent vertebrae. 

A nearly complete anterior dorsal ver- 

part of c are exposed just below the groove tebra (Fig. 13B) exliibits a parapophysis on 

for the dental lamina. The left Pci,, is just the dorsolateral region of the centrum, a 

beginning to eiaipt; the position of this large diapophysis that projects laterally 

tooth distal to the end of the alveolar row from the middle of the pedicle, and a neu- 

is evidence that the individual was still ral spine that is directed posterodorsally 

growing. The centrrmi is moderately amphicoelous. 

Postcraniiim. One centrum (Fig. 13 A) is The neural canal is large but narrower 

tentatively considered to be cervical be- than the centnnn, differing in this respect 

cause it is transversely wide and antero- from Therioherpeton (and possibly reflec- 

posteriorly short, as are the cervicals of 
Therioherpeton. The facets for the pedi- 
cles of the neural arch are in a dorsolateral 
position; if the pedicles were to have pro- 

tive of the more adult stage of Prozostro- 
don; J. A. Hopson, personal communica- 

The dorsolumbar region is further rep- 

74 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 




10 mm 

Figure 14. Prozostrodon brasiliensis, new genus. (A) ?Right clavicle in ?posterior view. (B) Interclavicle in dorsal view. (C) 
Right humerus in ventral view. Abbreviations: AAC, attachment area for right clavicle; CAP, contact with acromial process; CI, 
contact with interclavicle; DC, deltopectoral crest; ECT, ectepicondyle; ECR, ectepicondylar ridge; ENF, entepicondylar foramen; 
ENR, entepicondylar ridge; ENT, entepicondyle; GT, greater trochanter; HC, humeral capitulum; PP, posterior projection. 

resented by an incomplete vertebra and 
four incomplete neural arches. The former 
has a massive centrum with marked am- 
phicoely. The transverse processes are di- 
rected laterally and the broken base of the 
neural spine indicates that the spine was 
directed posterodorsally. The neural canal 
is large but narrower than the centrum. 
StiTictural variations in the spinous pro- 
cesses and zygapophyses are evidence of 
regional variation along the dorsolumbar 

In one neural arch (Fig. 13C), the ro- 
bust prezygapophyses are positioned more 
ventrally than the postzygapophyses; 
postzygapophyseal facets are anteroposte- 
riorly elongate and transversely convex, 
possibly representing a specialization for 
dorsoventral flexure of the vertebral col- 
umn (J. A. Hopson, personal communica- 
tion). The neural spine, with almost par- 
allel anterior and posterior borders, is pos- 
terodorsally directed, and more steeply in- 
clined than in Exaeretodon (Bonaparte, 
1963). Another neural arch (Fig. 13E) has 

a similarly elongate set of postzygapophys- 
es with convex facets, and prezygapophys- 
es with transversely concave facets. Jenkins 
and Parrington (1976, figs. 2A-D) illus- 
trated a neural arch in Morganucodon with 
zygapophyses and a neural spine of com- 
parable structure. 

A different type of dorsal vertebra (Fig. 
13D), represented by two neural arches, 
possibly derives from a position posterior 
to those described above. The zygapophy- 
seal facets have less curvature and the 
damaged neural spine appears to be short- 
er than in the previously described neural 

Ribs (Fig. 13F). Eight to 10 ribs are rep- 
resented by 20 fraginents, and none show 
any indication of processes or expansions. 
Most are figure 8— shaped in cross section. 

Clavicle (Fig. 14 A). Most of the Pright 
clavicle is present. On the proximal end 
are furrowlike rugosities representing the 
area of attachment with the interclavicle. 
The wide distal end is complexly config- 
ured with surfaces representing the aero- 

Advanced Triassic Cynodonts from Brazil • Bonaparte and Barberena 75 

mial attachment. The greatest curvature of 
the shaft is closer to the distal end. 

Interclavicle {Fig. 14B). This Y-shaped 
eleinent, with broad anterolateral exten- 
sions for reception of the clavicles, is an- 
teroposteriorly shorter (19 mm) than wide 
(23 mm). The ventral surface is not ex- 

Scapulocoracoid. These bones are rep- 
resented only by small fragments that are 
too fragmentary to provide useful infor- 
mation on the glenoid and other features 
of interest. 

Huiiienis (Fig. 14C). The left humerus is 
represented by the proximal half, including 
the deltopectoral crest and the humeral 
head. The right humenis is almost com- 
plete, lacking only tlie humeral head and 
part of the proximal end, and a small por- 
tion of the distal end including the articular 
surface for the ulna. The humerus is struc- 
turally no more derived than that of Exaer- 
etodon, and is similar to that of Probeleso- 
don lewisi (Romer and Lewis, 1973). The 
entepicondylar foramen is very large, and 
opens distally into a deep sulcus that con- 
tinues to the entepicondylar terminus. The 
entepicondyle, which is larger than the ec- 
tepicondyle, terminates in a distinct projec- 
tion. Near the anterior margin that extends 
proximally from the ectepicondyle is a small 
ectepicondylar foramen; this margin ex- 
tends farther proximally onto the diaphysis 
than does the comparable margin from the 

Ilium (Fig. 15A). The right ilium is al- 
most complete except for the most anterior 
end of the iliac blade. The neck above the 
acetabulum is well defined, and the poste- 
rior end of the iliac blade bears a veiy short 
process. Both of these features are derived 
compared with the structure seen in TJiri- 
naxodon and CynognatJms (Jenkins, 1971) 
and in Exaeretodon (Bonaparte, 1963). The 
outer surface of the iHac blade is largely 
convex dorsoventrally, not concave as in the 
above cited cynodonts, and its dorsal and 
ventral borders are nearly parallel to one 

Foot (Figs: 15B, C). The right hind foot 

is complete, except for the lateral part of 
the calcaneum. The superposition of as- 
tragalus and calcaneum appears to be of 
the type present in Exaeretodon (Bona- 
parte, 1963), defined by Jenkins (1971) as 
the "therapsid type of plantigrady" How- 
ever, the metatarsals and phalanges are 
proportionally longer than in Exaeretodon. 
The phalangeal formula is 2-3-3-3-3. 

Comparison of Prozostrodon with other 
Cynodonts. The referral of PV 0248T to 
the genus Thrinaxodon by Barberena et al. 
(1987) is untenable, as pointed out by Bat- 
tail (1991), who interpreted the specimen 
as a chiniquodontid. The following derived 
characters present in the holotype of Pro- 
zostrodon brasiliensis are not found in 
specimens of Thrinaxodon from the Early 
Triassic of South Africa (Parrington, 1946; 
Estes, 1961) and Antarctica (Colbert and 
Kitching, 1977). 

1) Frontal bordering the orbit (also in 
Therioherpeton, tritheledontids, and 

2) Contact between the ventral process of 
frontal and dorsal process of palatine 
(also in tritheledontids and tritylodon- 
tids; additionally present in Probainog- 
nathtts, Ecteninion, and chiniquodon- 
tids; J. A. Hopson, personal communi- 

3) Presence of an orbitosphenoid contact- 
ing frontal and palatine, and medially 
placed relative to the dorsal process of 
the palatine (also in Morganucodon; 
Kermack et al, 1981). 

4) Secondary bony palate with large pal- 
atines that extend to the level of the last 

5) Incipient bifurcation of the roots in the 
upper and lower postcanines. 

6) Small postorbital and prefrontal. 

7) Absence of postorbital bar (also in 
Therioherpeton, tritheledontids, and 

8) Large infraorbital and two well-defined 
foramina for the trigeminal nerve in the 
maxilla (also in tritheledontids and tri- 

76 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 15. Prozostrodon brasiliensis, new genus. (A) Incomplete right ilium in lateral view. (B) Right foot as preserved. (C) 
Reconstruction of right foot in dorsal view. Abbreviations: A, acetabular area; AST, astragalus; CAL, calcaneum; C, cuneiform; 
EC, ectocuneiform; ENC, entocuneiform; IC, ischial contact; IB, iliac blade; MC, mesocuneiform; NAV, navicular; PC, pubis 
contact; PP, posterior process of the ilium; l-V, metatarsals I through V. 

This combination of derived characters dontidae, and Probainognathidae. The lack 

precludes referral of Prozostrodon brasi- of gliriform specialization in the incisors 

liensis not only to the Thrinaxodontidae, and the transversely narrow postcanines 

but also to the Cynognathidae, Chiniquo- preclude referral of the specimen to the 

Advanced Triassic Cynodonts from Brazil • Bonaparte and Barberena 

1 1 

Tritheledontidae or to any gomphodont 
family, including the Tritylodontidae. 

A Summary of the Distinctive Features 
of Prozostrodon. In addition to the char- 
acters listed above as well as in the diag- 
nosis, Prozostrodon possesses other dis- 
tinctive features. Although many of these 
features may be found among various oth- 
er taxa, only in Prozostrodon, as far as we 
are aware, do they occur together. The 
suite is as follows. 

1) Reduced unossified area between 
frontal, orbitosphenoid, and alisphe- 
noid (also in tritylodontids and Ecten- 
inion; Martinez et al., 1996). 

2) Sphenopalatine foramen within the 
posterior portion of the dorsal process 
of palatine. 

3) Anteroposteriorly long palatines in the 
secondary bony palate (also in chini- 
quodontids, tritheledontids, and some 

4) Root of the zygomatic arch distinctly 
offset from the posterior margin of 
the maxilla (also in Probainognathus, 
chiniquodontids, tritylodontids, trav- 
ersodontids, and diademodontids). 

5) Carnivorous— insectivorous dentition 
comparable in general moi-phology to 
that of Thrinaxodon and Morganuco- 
don, but with cingula on the upper 
postcanines that are either poorly de- 
veloped or absent altogether 

6) Posterior portion of the upper tooth 
row inset from the lateral border of 
the maxilla (also in traversodontids, 
tritylodontids, tritheledontids, and 
Probainognathus ) . 

7) Incipient bifurcation of the upper 
postcanine roots. 

8) Incipient bifurcation of the lower 
postcanine roots (also in Pachygenehis 
and Microconodon; H.-D. Sues, per- 
sonal communication; and in Therio- 
herpeton) . 

9) Mandibular symphysis unfused, an- 
teroposteriorly elongated, dorsoven- 
trally narrow (as in Thrinaxodon 
[primitive], tritheledontids, and trity- 

lodontids [derived]; J. A. Hopson, per- 
sonal communication). 

10) Presence of symphysial fossa and fo- 
ramen in the lower jaw (also in Cy- 
nognathus; Kermack et al., 1973). 

11) Articular process of the dentary ex- 
tended posteriorly and expanded 
transversely, without indication of a 
condyle, and set above the level of the 
postcanine teeth (a feature of derived 
cynodonts; J. A. Hopson personal 

12) Cervical centra anteroposteriorly 
short, transversely wide, and dorso- 
ventrally low (also in tritylodontids 
and Morganucodon) . 


The specimens of Therioherpeton (skull 
length, 38 mm) and Prozostrodon (skull 
length estimated to be 67 mm) are rela- 
tively small, and consideration should be 
given to whether the smaller of the two 
represents a juvenile individual. Four fea- 
tures of Therioherpeton, in comparison to 
those in Prozostrodon, might be interpret- 
ed as evidence of immaturity: the frontals 
extend further posteriorly; the anterior 
portion of the braincase is proportionally 
wider; the postcanines lack cingula; the 
neural canals of presacral vertebrae are 
proportionally larger. However, the possi- 
bility that these are juvenile characters is 
contradicted by the state of ossification in 
the postcranium of Therioherpeton. With 
the exception of the pelvis, in which the 
three elements are not synostosed, we find 
no evidence of a subadult condition. Fur- 
thermore, Therioherpeton does not exliibit 
the proportionately large orbital size that 
is characteristic of juvenile individuals. A 
juvenile skull of cf. Probainognathus (Bon- 
aparte and Crompton, 1994, figs. 1, 2) of 
comparable length (39 mm) to that of 
Therioherpeton exliibits a skull to orbit ra- 
tio of 3.5, versus an estimated 5.0 for Ther- 
ioherpeton. In a putatively inature skull of 
Probainognathus (Romer, 1970, fig. 2) the 
ratio of skull to orbital length is 6, only 
slightly greater than in Therioherpeton. Al- 

78 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 
















Figure 16. The hypothesized phylogenetic positions of Ther- 
ioherpeton and Prozostrodon. The cladogram was generated 
by Dr. J. A. Hopson, who incorporated information from the 
Brazilian genera described in this paper into his data set of 
cynodont characters. 

though we do not beheve that the type of 
Therioheiyeton represents a juvenile, we 
remain open to the possibiUt^' that the four 
characters cited above eventually might be 
shown to be juvenile characters retained 
in adulthood. 

The hypothesized relationships of Ther- 
ioJierpeton and Prozostrodon are depicted 
in Figure 16. The relative positions in the 
cladogram of the two genera seem due to 
the persistence of a prefrontal and post- 
orbital and the shorter posterior projection 
of the frontal in Prozostrodon, and to the 
more derived characters present in the 
pelvis and proxiinal feinur of Therioher- 

The materials of Therioheiyeton and 
Prozostrodon, although incomplete, pro- 
vide new anatomical information to fur- 
ther elucidate the complex transition be- 
tween cynodonts and primitive mammals. 
The taxa described here support the hy- 
pothesis that the ancestry of mammals lay 
among the clade of carnivorous— insectiv- 
orous cynodonts (Hopson, 1991, 1994; 
Hopson and Kitching, 2001), rather than 
among goinphodont cynodonts, and in 

particular the Tritylodontidae (Kemp, 
1982, 1983). Sues (1985) reviewed in de- 
tail most of the synapomorphies that 
Kemp (1982, 1983) proposed in support of 
his interpretation that tritylodontids and 
primitive mammals were closely related, 
and concluded that many of these pur- 
portedly shared derived features are either 
superficial similarities, are symplesio- 
morphic, or are the result of parallel de- 
velopment but which, nonetheless, evi- 
dence structural differences. 

We regard the dentition o{ Prozostrodon 
as iTioi"phologically intermediate between 
that of Thrinaxodon liorhinus from the 
Early Triassic and those evolved among 
Late Triassic and Early Jurassic morganu- 
codontids. Although tritheledontids such 
as Pachygenelus have been considered as 
closely related to mammals (Hopson and 
Barghusen, 1986), and are represented 
cladistically as the sister taxon to mainmals 
(Fig. 16; see also Shubin et al., 1991), we 
interpret the tritheledontid dentition as 
derived, particularly with respect to the 
buccolingual expansion of the upper post- 
canines. In contrast, the entire dentition of 
Prozostrodon (and what is known of that 
of Therioherpeton) is substantially similar 
to that in morganucodontids, with only 
slight differences in the development of 
upper postcanine cingula. During a span 
of time that witnessed major modifications 
in cranial and postcranial morphology, 
dental patterns were fundamentally con- 


Drafts of this paper were critically re- 
viewed by Drs. Z. Kielan-Jaworowska, G. 
W. Rougier, H.-D. Sues, and J. A. Hopson. 
Very special thanks are accorded Dr. J. A. 
Hopson for his generous help and advice 
on improvements, and for the use of his 
database in the cladogram presented here. 
We also thank Drs. F. A. Jenkins, Jr., and 
M. D. Shapiro for editorial assistance. The 
senior author expresses gratitude to two 
colleagues at the Fundacao Zoobotanica 
de Porto Alegre, Brazil: Dr. ]. Ferigolo for 

Advanced Triassic Cynodonts from Brazil • Bonaparte and Barberena 79 

interesting discussions on the subject; and 
Ana M. Riveiro for field and laboratory as- 
sistance. The senior author is also grateful 
to the CNPq. of Brazil, and to the Museo 
Argentino de Ciencias Naturales, Buenos 
Aires, for support in developing most of 
this research in the Museu de Ciencias 
Naturais de Porto Alegre, Brazil. 


Barberena, M. C, J. F. Bonaparte, and A. M. SA 
Teixeira. 1987. Thrinaxodon brasiliensis sp. 
nov., a primeira ocorrencia de cinodontes gales - 
sauros para o Triasico do Rio Grande do Sul. 
Anais do X Congresso Brasileiro de Paleontolo- 
gia, Rio de Janeiro, Brazil, pp. 67-76. 

BatTAIL, B. 1991. Les Cvnodontes (Reptilia, Ther- 
apsida): une phylogenie. Bulletin du Museum 
Nationale d'Histoire Naturelle. Section C, Sci- 
ences de la Terre Paleontolgie, Geologie, Miner- 
ologie, 13: 17-105. 

Bonaparte, J. F. 1963. Descripcion del esqueleto 
postcraneano de Exaeretodon sp. Acta Geologica 
Lilloana, 4: 5-52 

. 1980. El primer ictidosaurio (Reptilia— Ther- 

apsida) de America del Sur, Chaliminia muste- 
loides, del Triasico Superior de La Rioja, Repiib- 
lica Argentina. Actas del II Congreso Argentino 
de Paleontologfa y Bioestratigrafia, y I Congreso 
Latinamericano de Paleontologia, 1: 123-133. 

Bon.^parte, J. F., and M. C. Barberena. 1975. A 
possible mammalian ancestor from the Middle 
Triassic of Brazil (Therapsida-Cynodontia). Jour- 
nal of Paleontolog)', 49: 931-936. 

Bonaparte, J. F, and A. W. Crompton. 1994. A 
juvenile probainognathid cynodont skull from 
the Ischigualasto Formation and the origin of 
mammals. Revista Museo Argentino de Ciencias 
Naturales, Paleontologia, 5: 1—12. 

Bortoluzzi, C. a., and M. C. Barberena. 1967. 
The Santa Maria beds in Rio Grande do Sul 
(Brazil), pp. 169-195. In J. J. Bigarella, R. D. 
Becker, and I. D. Pinto (eds.). Problems in Bra- 
zilian Gondwana Geology. 1st International Sym- 
posium on Gondwana Stratigraphy and Paleon- 
tology. Curitiba, Brazil: Roesner Ltd. xviii + .344 

Colbert, E. H., and J. Kitching. 1977. Triassic cy- 
nodont reptiles from Antarctica. American Mu- 
seum Novitates, 2611: 1-30. 

Crompton, A. W. 1958. The cranial morphology of 
a new genus and species of ictidosaurian. Pro- 
ceedings of the Zoological Society of London, 
130: 18.3-216. 

. 1974. The dentitions and relationships of the 

Southern African Triassic mammals Enjthroth- 
eriiim parringtoni and Megazostrodon nidnerae. 
Bulletin of the British Museum (Natural Histo- 
ry), Geology, 24: 399-437. 

Crompton, a. W, and Z. Luo. 1993. Relationships 
of the Liassic mammals Sinoconodon, Mormnu- 
codon oehleri and Dinnetherium, pp. 30—44. In 
F S. Szalay, M. J. Novacek, and M. C. McKenna 
(eds.). Mammal Phylogeny: Mesozoic Differen- 
tiation, Multituberculates, Monotremes, Early 
Therians and Marsupials. New York: Springer- 
Verlag. x + 249 pp. 

Crompton, A. W., and A.-L. Sun. 1985. Cranial 
structure and relationships of the Liassic mam- 
mal Sinoconodon. Journal of the Linnean Society 
of London, 85: 99-119. 

ESTES, R. 1961. Cranial anatomy of the cynodont 
reptile Thrinaxodon Uorhinus. Bulletin of the 
Museum of Comparative Zoology, 125: 165-180. 

GOW, C. 1980. The dentitions of the Tritheledontidae 
(Therapsida: Cynodontia). Proceedings of the 
Royal Society of London, B, 208: 461^81. 

. 1986. A new skull oi Megazosfrodon (Mam- 
malia: Triconodonta) from the Elliot Formation 
(Lower Jurassic) of southern Africa. Palaeonto- 
logia Africana, 26: 1.3-23. 

HOPSON, J. A. 1991. Systematics of the non-mam- 
malian Synapsida and implications for patterns of 
evolution in svnapsids, pp. 635—693. In H.-P. 
Schultze and L. Trueb (eds.). Origins of the 
Higher Groups of Tetrapods. Ithaca, New York: 
Cornell University Press, xii + 724 pp. 

. 1994. Synapsid evolution and the radiation 

of nontherian mammals, pp. 190-219. In D. R. 
Prothero and R. M. Schoch (eds.). Major Fea- 
tures of Vertebrate Evolution, Short Courses in 
Paleontology, No. 7. Knoxville, Tennessee: Uni- 
versity of Tennesee. 270 pp. 

HoPSON, J. A., AND H. R. Barghusen. 1986. An 
analysis of therapsid relationships, pp. 83-106. In 
N. Hotton III, P D. McLean, J. J. Roth, and E. 
C. Roth (eds.). The Ecology and Biology of 
Mammal-like Reptiles. Washington, DC: Smith- 
sonian Institution Press, x -I- 326 pp. 

Hopson, J. A., AND J. W. Kitching. 2001. A pro- 
bainognathian cynodont from South Africa and 
the phylogeny of nonmammalian c\Tiodonts. Bul- 
letin of the Museum of Comparative Zoolog)', 
156: 5^5. 

Jenkins, F A., Jr. 1970. The Chafiares (Argentina) 
Triassic reptile fauna. VTI. The postcranial skel- 
eton of the traversodontid Massetognathus pas- 
cuali (Therapsida, Cynodontia). Breviora, 352: 

. 1971. The postcranial skeleton of African cy- 
nodonts. Bulletin of the Peabody Museum of 
Natural History, 36: 1-216. 

1984. A survey of mammalian origins, pp. 

32^7. In P D. Gingerich and C. D. Badgley 
(eds.). Mammals: Notes for a Short Course. 
Knoxvdlle, Tennessee: University of Tennesee 
Department of Geological Sciences Studies in 
Geologv' 8. 234 pp. 
Jenkins, F. A., Jr., and F. R. Parrington. 1976. 
The postcranial skeletons of the Triassic mam- 


Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

mals Eozo.strodon, Megazostrodon and Ery- 
throtherium. Philosophical Transactions of the 
Royal Society of London B, Biological Sciences, 
273: 387^31. 

Kemp, T. S. 1982. Mammal-like Reptiles and the Or- 
igin of Mammals. London: Academic Press, xiv 
-I- 363 pp. 

. 1983. The interrelationships of mammals. 

Zoological Journal of the Linnean Society, 77: 

Kermack, K. a., F. Mussett, and H. W. Rigney. 
1973. The lower jaw of Morganucodon. Zoolog- 
ical Journal of the Linnean Society, 53: 87-175. 

. 1981. The skull oi Morganucodon. Zoological 

Journal of the Linnean Society, 71: 1—158. 

KUHNE, W. G. 1956. The Liassic therapsid Oligoky- 
phus. London: Trustees of the British Museum 
(Natural History), x -I- 149 pp. 

LUO, Z. 1994. Sister-group relationships of mammals 
and transformations of diagnostic mammalian 
characters, pp. 98-128. In N. C. Frazer and H.- 
D. Sues (eds.). In the Shadow of the Dinosaurs. 
Cambridge, United Kingdom: Cambridge Uni- 
versity Press. X -I- 435 pp. 

Martinez, R. N., C. L. May, and C. A. Forster. 
1996. A new carnivorous cynodont from the Is- 
chigualasto Formation (Late Triassic, Argentina), 
widi comments on eucynodont phylogeny. Jour- 
nal of Vertebrate Paleontology, 16: 271-284. 

Osborn, J., AND A. W. CroMPTON. 1973. The evo- 
lution of mammalian from reptilian dentitions. 
Breviora, 399: 1-18. 

Parrington, F. R. 1946. On the cranial anatomy of 
cynodonts. Proceedings Zoological Society of 
London, 116: 181-197. 

ROMER, A. S. 1970. The Chanares (Argentina) Tri- 
assic reptile fauna. VI. A chiniquodontid cyno- 
dont with an incipient squamosal— dentary jaw ar- 
ticulation. Breviora, 344: 1-18. 

ROMER, A. S., AND A. Lewis. 1973. The Chaiiares 
(Argentina) Triassic reptile fauna. XIX. Postcra- 
nial materials of the cynodonts Probelesodon and 
Probainognathus. Breviora, 407: 1—26. 

SA Teixeira, a. M. 1979. Um novo cinodonte car- 
nivoro {Probelesodon kitchingi) do Triassico do 
Rio Grande do Sul, Brasil. Unpubhshed M.S. 
dissertation. Universidade Federal do Rio 
Grande do Sul, Brazil: Postgraduation in Geo- 
sciences. 71 pp. 

SCHULTZ, C. L. 1986. Osteologia parcial do postcran- 
io de Scaphonyx sulcognathus Azevedo 1982 
(Lepidosauria, Rhynchocephalia, Rhynchosauri- 
dae). Unpublished M.S. dissertation. Universi- 
dade Federal do Rio Grande do Sul, Brazil: Post- 
graduation in Geosciences, 139 pp. 

Shapiro, M. D., and F. A. Jenkins, Jr. 2001. A cy- 
nodont from the Upper Triassic of East Green- 
land: tooth replacement and double-rootedness. 
Bulletin of the Museum of Comparative Zoology, 
156: 49-58. 

Shubin, N. H., a. W. Crompton, H.-D. Sues, and 
P. E. Olsen. 1991. New fossil evidence on the 
sister-group of mammals and early Mesozoic fau- 
nal distribution. Science, 251: 1063-1065. 

Sues, H.-D. 1985. The relationships of the Tritylo- 
dontidae (Synapsida). Zoological Journal of the 
Linnean Society, 85: 205-217. 

Sun, A.-L. 1984. Skull morphology of the tritylodont 
genus Bienotheroides of Sichuang. Scientia Sini- 
ca, B 27: 270-284. 



Abstract. The inner ear structures underwent fun- 
damental changes during the evolution from non- 
mammalian cynodonts ("mammallike reptiles") to 
earlv mammals. The petrosal bone in mammals has 
an enlarged pars cochlearis containing an elongate 
cochlea. The elongation of the bony cochlear canal is 
usually correlated with the development of a ventral 
eminence of the pars cochlearis, known as the pro- 
montorium. Both cochlear canal and promontorium 
are regarded as apomorphies of mammals. In con- 
trast, nonmammalian cynodonts lack the promonto- 
rium. If present at all, the bony cochlear recess is 
small, globular, and poorly differentiated from the 
vestibule in most cynodonts. The trit)'lodontid Yun- 
nanodon has an intermediate condition. Unlike most 
nonmammalian cynodonts, Yiinnanodon has a dis- 
tinctive cochlear canal in an enlarged pars cochlearis, 
but it lacks the promontorium and retains the basi- 
sphenoid wing, a primitive feature of many cyno- 
donts. These characters in tritylodontids suggest that 
a distinctive cochlear canal in an enlarged pars coch- 
learis probably evolved in the common ancestor to 
trit^'lodontids, tritheledontids, and mammahaforms, 
and before the emergence of the petrosal promon- 
torium in mammaliaforms. The promontorium sub- 
sequently formed by the displacement of the neigh- 
boring parasphenoid-basisphenoid complex and ba- 
sioccipital bone by the pars cochlearis in the early 
evolution of mammaliaforms. 


The bony structure surrounding the in- 
ner ear differs between mammals and 
nonmammahan cynodonts. The inner ear 
of cynodonts is enclosed by the prootic 
and the opisthotic, known collectively as 
the periotic bones, as well as by the ex- 
occipital and the basioccipital. The basi- 
sphenoid wing (or parasphenoid ala), 
which is a large coinponent of the paras- 

' Section of \'ertebrate Paleontology, Carnegie Mu- 
seum of Natural History, Pittsburgh, Pennsylvania 

penoid-basisphenoid complex, overlaps 
the prootic and indirectly contributes to 
the cochlear housing. The bony housing 
for the inner ear in cynodonts is formed 
by multiple bones, as documented in great 
detail for Thrinaxodon (Olson, 1944; Es- 
tes, 1961; Fourie, 1974; Rowe et al, 1993), 
Probelesodon, Massetognathus (Quiroga, 
1979), and Probainognathiis (Allin, 1986). 
This mosaic pattern is primitive for cyno- 
donts because it is also present, albeit in a 
slightly different condition, in noncyno- 
dont therapsids (Olson, 1944; Cox, 1962; 
Sigogneau, 1974). 

In contrast, the bony housing of the in- 
ner ear in early mammaliaforms (inodified 
from Rowe [1988] to include Adelobasi- 
leus and Sinoconodon) is formed exclusive- 
ly by the petrosal, which is the single bone 
composed of the fused prootic and opis- 
thotic elements of nonmammalian cyno- 
donts (Kermack et al., 1981; Rowe, 1988; 
Luo et al., 1995). The parasphenoid-basi- 
sphenoid coinplex, the basioccipital, and 
the exoccipital are excluded by an enlarged 
petrosal from the bony housing for the in- 
ner ear, as has been docuinented in a wide 
range of mammaliaforms and early main- 
mals, such as Sinoconodon (Luo et al., 
1995), morganucodontids (Kermack et al., 
1981; Gow, 1985; Graybeal et al, 1989; 
Luo and Ketten, 1991; Crompton and 
Luo, 1993), triconodontids (Kermack, 
1963; Crompton and Luo, 1993; Rougier 
et al., 1996), docodonts (Lillegraven and 
Krusat, 1991), monotremes (Kuhn, 1971; 
Zeller, 1989; Luo and Ketten, 1991; Fox 
and Meng, 1997), multituberculates 
(Miao, 1988; Luo and Ketten, 1991; Lil- 

Bull. Mus. Comp. ZooL, 156(1): 81-97, October, 2001 81 


Bulletin Museum of Comparative Zoology, Vol. 156, No. 

legraven and Hahn, 1993; Meng and Wyss, 
1995; Fox and Meng, 1997; Hurum, 1998), 
and archaic therians (Wible et al., 1995; 
Hu et al., 1997). In mammals, the pars 
cochlearis that contains the cochlear canal 
is much larger than in cynodonts, forming 
a ventrolateral eminence known as the 
promontorium, which is a very conspicu- 
ous external feature in the mammalian 
basicranium (Gow, 1985; Hopson and Bar- 
ghusen, 1986; Rowe, 1988; Luo et al., 

Mammals and nonmammalian cyno- 
donts differ also in features of the inner 
ear. In the noncynodont therapsids, such 
as dicynodonts (Cox, 1962) and gorgon- 
opsids (Sigogneau, 1974), the sacculococh- 
lear cavity (or recess) is not differentiated 
from the vestibular cavity (Olson, 1944; 
Cox, 1962; Sigogneau, 1974). At the an- 
terior end of this sacculocochlear cavity is 
the fenestra vestibuli or oval window, 
which accommodates the stapes that trans- 
mitted sound vibrations from the middle 
to the inner ear. The cochlear part of the 
osseous inner ear is more distinctive from 
the vestibule in primitive cynodonts such 
as Thrinaxodon (Estes, 1961; Fourie, 
1974). In advanced cynodonts (Quiroga, 
1979; Allin and Hopson, 1992), the bony 
cochlear structure is more developed than 
in Thrinaxodon and noncynodont therap- 
sids. The bony cochlea is represented by a 
small and globular cavity but is too short 
to be termed the cochlear canal (except for 
tritylodontids, and perhaps tritheledontids; 
see below). 

In contrast, in Early Jurassic mammalia- 
forms, the cochlear canal is elongate and 
differentiated from the saccular cavity 
(Graybeal et al., 1989; Luo and Ketten, 
1991; Luo et al, 1995). The elongate co- 
chlear canal is a shared derived feature of 
diverse mammalian groups during the Me- 
sozoic, as documented in multitubercula- 
tes (Miao, 1988; Luo and Ketten, 1991; 
Lillegraven and Hahn, 1993; Meng and 
Wyss, 1995; Fox and Meng, 1997; Hurum, 
1998), docodonts (Lillegraven and Krusat, 
1991), and possibly in symmetrodont ther- 

ians (Wible et al, 1995; Hu et al., 1997). 
The elongate bony cochlear canal suggests 
a better-developed cochlear duct, which 
may indicate a better sensitivity to high- 
frequency sound that is very important in 
the hearing function of all extant mam- 
mals, and probably important for at least 
some of the earliest mammals (Rosowski 
and Graybeal, 1991; Rosowski, 1992; Hu- 
rum, 1998). 

Given these differences in the structure 
of the inner ear and its bony housing be- 
tween mamiuals on the one hand and non- 
mammalian cynodonts on the other, the 
ear structures must have undergone exten- 
sive transformation during the early evo- 
lution of mammals after their divergence 
from nonmammalian cynodonts. To eluci- 
date the pattern of this phylogenetic trans- 
formation, it is essential to obtain some de- 
tailed anatomical information on the inner 
ear and the surrounding bones in such de- 
rived cynodonts as tritylodontids and tri- 
theledontids. Some earlier studies of the 
ear region of tritylodontids reported the 
presence of a cochlear canal (Kiihne, 1956; 
Hopson, 1965). Two additional studies 
(Gow, 1986; Sun and Cui, 1987) offered 
observations on the basicranial structures 
surrounding the inner ear. This paper de- 
scribes the inner ear and its bony housing 
in the tritylodontid Yunnanodon, and their 
anatomical relationships as revealed by se- 
rial sections. The new information has im- 
plications for the evolution of the ear re- 
gion through the transition from nonmam- 
malian cynodonts to mammals, given the 
fact that tritylodontids are considered by 
some to be closely related to mammals 
(Kemp, 1983; Rowe, 1988; Wible, 1991; 
Wible and Hopson, 1993; Luo, 1994; Luo 
and Crompton, 1994; but see the alterna- 
tive phylogeny by Crompton [1972], Sues 
[1985a], and Hopson and Barghusen 


Fossil remains of Yunnanodon (Cyno- 
dontia, Tritylodontidae) are from the Up- 
per Red Beds of the Lower Lufeng For- 

mation of Yunnan, China (Sun et al., 
1985), which is considered to be Early Ju- 

Irassic (Sinemurian to Phensbachian) by re- 
cent studies (Luo and Sun, 1993; Luo and 
Wu, 1995). Yunnanodon is the smallest tri- 
tylodontid known from the Lower Lufeng 
Formation (Cui, 1976, 1986; Luo and Wu, 
J 1994), with a skull length ranging from 36 
' to 47 mm. Yunnanodon has only two cusps 
in the lingual row of the upper postcani- 
nes, the main diagnostic character distin- 
guishing this taxon from other tritylodon- 
tids in the Lower Lufeng, all of which have 
three lingual cusps on the upper teeth 
(Cui, 1976; Luo and Wu, 1994). Yunnan- 
odon is comparable to Dinnebitodon from 
ll the Kayenta Formation of Arizona (Sues, 
1985b) in some derived dental characters 
(Luo and Wu, 1995). 

Several skull specimens of Yunnanodon 
(Cui, 1976; Sun and Cui, 1987) were re- 
examined in tliis study (Institute of Verte- 
ij brate Paleontology and Paleoanthropology, 
'I Beijing: IVPP 5071 [holotype]; 7204; 7205; 
7219). A duplicate skull (courtesy of A.-L. 
Sun) was sectioned by using a Croft Grind- 
er to expose its internal structures (Croft, 
1950; Crompton, 1955). Camera lucida 
drawings and photographs were taken of 
each serial (transverse) section exposed by 
grinding, in the place of the original spec- 
imen. The reconstruction of the basicranial 
bones and the inner ear was made from the 
serial sections by using the Slicer Dicer® 
program by Visualogic, Inc., Bellevue, 
Washington. Measurement of the length 
and internal diameters of the cochlear canal 
was based on the original sections (camera 
lucida drawing and photos, both with 
scales). The measurement of the internal 
diameter of the semicircular canal was 
made from the original sections. The di- 
ameter of the arc of the semicircular canal 
{sensu Hurum, 1998) was taken from the 
graphic models of inner ear endocasts as 
rendered by the Slicer Dicer program. 



The petrosal forms the bony housing for 
the entire inner ear in Yunnanodon. The 

Tritylodontid Inner Ear • Luo 83 

petrosal excludes all other bones from the 
immediate bony housing of the inner ear. 
No suture marks the separation of the two 
periotic bones (prootic and opisthotic) in 
Yunnanodon (Figs. 1, 2). This has been 
documented in other tritylodontids (Kii- 
hne, 1956; Hopson, 1964, 1965; Sun, 
1984; Cow, 1986; Sues, 1986). The ossifi- 
cation of the cartilaginous otic capsule of 
the chondrocranium begins in several os- 
sification centers in amniotes (de Beer, 
1937). The absence of the prootic-opis- 
thotic suture implies that the separate em- 
bryonic ossification centers coalesced into 
a single bone in adult tritylodontids, in- 
stead of two separate prootic and opis- 
thotic bones of other therapsids (Olson, 
1944) and extant diapsid reptiles (de Beer, 
1937). The absence of the prootic— opis- 
thotic suture in tritylodontids is a derived 
condition in comparison to many other 
nonmammalian cynodonts, as pointed out 
by many authors (Kemp, 1983; Sun, 1984; 
Hopson and Barghusen, 1986; Sues, 1986; 
Rowe, 1988; Wible, 1991; Luo, 1994). 

The pars cochlearis that encloses the 
bony cochlear canal is large relative to the 
rest of the petrosal, as shown in the serial 
sections and in the broken basicranial 
specimens (e.g., IVPP 5071 as described 
by Sun and Cui [1987]). The promonto- 
rium, defined as the ventral eminence of 
the pars cochlearis in extant mammals 
{sensu Williams et al. [1989] and Luo et 
al. [1995]), is represented by a bulging 
area posterior to the basisphenoid wing 
and anterior to the crista interfenestralis in 
the skull with an incomplete basisphenoid 
wing (Fig. 2). In a more or less intact bas- 
icranium (IVPP 7219), the medial part of 
the pars cochlearis is covered medially by 
the basioccipital, and anteriorly by the ba- 
sisphenoid wing (more details below). The 
ventral (and external) exposure of the pars 
cochlearis in the intact skulls appears to be 
much smaller than its entire size (Fig. 1) 
because much of the pars cochlearis is su- 
perficially covered by the sphenoid com- 
plex and the basioccipital bone. 

In some possibly juvenile skulls in which 

84 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 1. Basicranium and inner ear of Yunnanodon (Tritylodontidae, Cynodontia) in ventral views. (A) Basicranium and its 
position in sl<ull (ventral view of basicranium; the squamosal is not illustrated in the stipple drawing; modified from Luo [1994]; 
skull outline modified from Cui [1976] and Luo and Wu [1994]). (B) Approximate position of the inner ear (purple) to the sur- 

Tritylodontid Inner Ear • Luo 


the petrosal, the basisphenoid, and the ba- 
sioccipital are separated from one another, 
the pars cochlearis can be broadly ex- 
posed. This led Sun and Cui (1987) to sug- 
gest that Yunnanodon developed a pro- 
montoriumlike structure ( = the pars coch- 
learis), an observation that is confirmed 
here. The medial aspect of the pars coch- 
learis has a flat facet that may be exposed 
in some specimens in which the basioccip- 
ital has become detached from the basi- 
cranium. This facet appears to be identical 
to the medial facet of the promontorium 
in Sinoconodon (Crompton and Luo, 1993; 
Luo et al., 1995). The overlap of this flat 
facet by the lateral lappet of the basioccip- 
ital in the intact skulls is very similar to the 
condition in subadult specimens of Sino- 

The fenestra vestibuli (oval window) is 
oval in outline, with a long diameter of 
— 1.8 mm and a short diameter of —1.5 
mm (based on IVPP 7219). The fenestra 
vestibuli is separated by a thin crista in- 
terfenestralis from the perilymphatic fo- 
ramen (round window or foramen cochle- 
ae). The latter is located in the same de- 
pression as (but separated from) the jug- 
ular foramen. The lateral trough of the 
petrosal is bound medially by the pars 
cochlearis and laterally by the lateral 
flange. The large ventral opening of the 
cavum epiptericum is anterior to the lat- 
eral trough. The facial foramen is located 
anterior to the fenestra vestibuli. The lat- 
eral flange is perforated by two vascular 
foramina. The pterygoparoccipital fora- 
men, which probably carried the superior 
ramus of the stapedial artery (Wible and 
Hopson, 1995), is posterior to the lateral 
flange. The anterior paroccipital process of 

the petrosal is bulbous. It supports the 
quadrate in the intact skull. The posterior 
paroccipital process is represented by a 
horizontal ridge with a free-standing lat- 
eral (distal) end. The anterior and poste- 
rior paroccipital processes are separated 
by the stapedial muscle fossa and its as- 
sociated groove. The dorsal aspects of the 
anterior and posterior paroccipital pro- 
cesses are in contact with the squamosal 
(Figs. 1, 2). These petrosal features are 
characteristic of all tritylodontids. The tab- 
ular bone is present on the occiput, cov- 
ering much of the mastoid part of the pe- 
trosal posteriorly. This primitive feature is 
shared by many cynodonts but is absent in 
Morganucodon and more derived inam- 

Sphenoid Complex 

The basisphenoid is an endochondral 
ossification whereas the parasphenoid is an 
intramembranous ossification that under- 
lies the former (Goodrich, 1930; de Beer, 
1937); the two elements are fused early in 
development to form the basisphenoid— 
parasphenoid complex in extant diapsids 
(Goodrich, 1930; de Beer, 1937; Bellairs 
and Kamal, 1981; Rieppel, 1993), and this 
complex reaches posteriorly to border on 
the basioccipital. In one cranial study of 
living diapsids (Oelrich, 1956), the un- 
paired anterior median element of this 
complex is considered to be the paras- 
phenoid in adult diapsids, whereas the 
paired posterolateral parts of this complex 
are identified as the basisphenoid. In living 
mammals, the intramembranous paras- 
phenoid ossification forms an unpaired 
median structure in the rostrum of the ba- 
sisphenoid-parasphenoid complex (Jollie, 

rounding basicranial bones (note that the basisphenoid [green] is superficially overlapping the petrosal [gray] but does not directly 
envelope the cochlea [purple]). (C) Basicranial structure. (D) Inner ear endocast (ventral view). Abbreviations (color code for 
bones): app, anterior paroccipital process; asc, anterior semicircular canal; bo, basioccipital (yellow); bs, basisphenoid (green); 
bsw, basisphenoid wing (=parasphenoid ala [green]); ce, cavum epiptericum; cif, crista interfenestralis; co, bony cochlear canal; 
80, exoccipital (blue); fc, foramen cochleae; fc-jf, the confluent foramen cochleae and jugular foramen; ff, facial foramen (VII); 
fst, fossa for stapedial muscle; fv, fenestra vestibuli; jf, jugular foramen; If, lateral flange of the petrosal; Isc, lateral semicircular 
canal; It, lateral trough; oc, occipital condyle; od, odontoid notch of basioccipital; pe, petrosal (gray); ppp, posterior paroccipital 
process of the petrosal; psc, posterior semicircular canal; sq, squamosal (pink). 

86 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 



Figure 2. Basicranium and inner ear of Yunnanodon (Tritylodontidae, Cynodontia). (A) Basicranium (detailed features in the 
ventrolateral view, with skull tilted, based on IVPP 7219) and its position In skull (skull outline In lateral view, with zygoma 
removed). The dashed line in the stipple drawing represents the posterior limit of the preserved part of the baslsphenoid wing, 
which is most probably incomplete In the illustrated skull. As a result, the pars cochlearis (cochlear housing) Is more exposed 
than in the intact skull. (B) Approximate anatomical relationships of Inner ear (purple) to the surrounding baslcranlal structures 
(basicranium Is tilted ventrolaterally, more so than In Fig. 2A). (C) Baslcranlal structure. (D) Inner ear endocast (lateral view). 
Abbreviations (color code for bones): ar, articular bone; app, anterior parocclpltal process; asc, anterior semicircular canal; bo, 
basiocclpltal (yellow); bs, baslsphenoid (green); bsw, baslsphenoid wing ( = parasphenoid ala [green]); ce, cavum epiptericum; 
CO, bony cochlear canal; dfx, dorsal flex of the baslsphenoid; eo, exoccipital (blue); fc-jf, the confluent foramen cochleae and 
jugular foramen; ff, facial foramen; fst, fossa for stapedial muscle; fv, fenestra vestibuli; If, lateral flange of the petrosal; Isc, 
lateral semicircular canal; It, lateral trough; oc, occipital condyle; od, odontoid notch; pc, pars cochlearis (cochlear housing); pe, 
petrosal (gray); ppp, posterior parocclpltal process of the petrosal; psc, posterior semicircular canal; q, quadrate; sq, squamosal 

Tritylodontid Inner Ear • Luo 87 

1962), whereas the posterolateral parts of 
this complex are formed of endochondral 
ossification (J. R. Wible, personal com- 
munication). The anatomical term "basi- 
sphenoid wing" (Kiihne, 1956; Crompton, 
1964; Luo et al., 1995) is synonymous with 
the "parasphenoid ala" in other studies of 
tritylodontids (Hopson, 1964; Gow, 1986; 
Sues, 1986). Both terms have been applied 
to the paired posterior extensions from the 
main body of the sphenoid complex. The 
basisphenoid wing is used here for cyno- 
donts and mammaliaforms. 

The basisphenoid has a strong dorsal 
flexion posteriorly in Yiinnanodon, as is 
typical of other tritylodontids. The basi- 
sphenoid has a ventral tuberosity. Bifur- 
cating from the ventral tuberosity are the 
crests of the basisphenoid wings (Fig. 1). 
The basisphenoid wing covers the anterior 
and ventral aspects of the pars cochlearis, 
and conceals the anterior part of the pars 
cochlearis from ventral view (Figs. IC, 
3B). The basisphenoid wing does not 
reach the rim of the fenestra vestibuli in 
Yunnanodon (Figs. IB, C). 

The development of the basisphenoid 
dorsal flexion and the basisphenoid wing 
may vary considerably in relation to overall 
skull size among tritylodontids. In the larg- 
er skulls o{ Bienotheriiun (Hopson, 1964), 
Bienotheroides (Sun, 1984), and Kaijen- 
tatherium (Sues, 1986), the crest on the 
basisphenoid wing is hypertrophied and 
accentuates the dorsal flexion of the basi- 
cranium. The hypertrophied basisphenoid 
crest reaches posterolaterally near the fe- 
nestra vestibuli. In Tritylodon, the basi- 
sphenoid wing borders on the fenestra ves- 
tibuli and reaches the facial foramen 
(Gow, 1986). The basisphenoid wing is 
much larger and more pronounced in 
these larger tritylodontids than in smaller 
tritylodontids, such as Yunnanodon (Fig. 
1) and Bocatherium (Clark and Hopson, 
1985). The pars cochlearis is always pre- 
sent in tritylodontids. However, its external 
exposure as the promontorium is a variable 
feature in the basicranium, partly because 
of the allometric effect from the wide 

range of skull sizes in this diverse group. 
The pars cochlearis tends to be better ex- 
posed in small tritylodontids with a weaker 
basisphenoid wing, but the pars cochlearis 
may not have any external exposure at all 
if covered by a hypertrophied basisphe- 
noid wing in large tritylodontids, such as 
Bienotherium and Tritylodon. 


The basioccipital is elongate and plate- 
like. Its anterior part intrudes between the 
two basisphenoid wings (Fig. IB). The lat- 
eral part of the basioccipital forms a lappet 
and overlaps the ventral surface of the pars 
cochlearis extensively (Fig. 3B), as evi- 
denced by a specimen in which the suture 
of two bones is visible. Therefore, in the 
intact specimens of Yunnanodon, the me- 
dial part of the pars cochlearis is concealed 
from ventral view by the basioccipital. The 
overlap of the basioccipital lateral lappet 
on the pars cochlearis may be more exten- 
sive in large tritylodontids than in such 
small taxa as Yunnanodon. The posterior 
part of the basioccipital does not seem to 
border on the jugular foramen, which is 
encircled by the petrosal and the exoccip- 
ital (Fig. 1). An odontoid notch is present 
on the posterior border of the basioccipi- 

Cochlear Housing 

The serial sections of Yiinnanodon show 
that the cochlear canal is entirely envel- 
oped by the pars cochlearis, which is ex- 
ternally covered by the basisphenoid wing 
and by the lateral lappet of the basioccip- 
ital (Fig. 3B). This is consistent with an 
earlier observation on Tritylodon by Gow 
(1986) that the basisphenoid wing (paras- 
phenoid ala) is a superficial part of the 
compound bony structure around the co- 
chlea. However, the serial sections indicate 
that the basisphenoid wing and the pars 
cochlearis are distinctive structures in 
Yunnanodon, but not homologous to each 
other as suggested by Gow (1986); the ba- 
sisphenoid wing does not directly envelop 
the cochlear canal. It should be pointed 


Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Larger tritylodonts Pa 

E. Multituberculates 


A. Probelesodon 

Figure 3. Comparison of tine basicranial structures in cynodonts and mammals (schematic transverse section at the level of 
the posterior part of the cavum epiptericum and/or the anterior part of the cochlea). (A) Probelesodon lewisi (Romer, 1970; 
based on MCZ 3774, serial section 212; Museum of Comparative Zoology, Harvard University; courtesy of A. W. Crompton; the 
basisphenoid and basioccipital are indistinguishable in the section). (B) Yunnanodon brevirostre (composite sl<etches based on 
several sections of a specimen provided by A.-L. Sun) and a generalized large tritylodont (showing the proportional difference 
of the basisphenoid wing and the basioccipital lappet). (C) Sinoconodon rigneyi (based on sections of IVPP 8689; Institute of 
Vertebrate Paleontology and Paleoanthropology, Beijing). (D) Morganucodon watsoni (based on sections of MCZ 20998, Gray- 
beal et al. [1989]). (E) Nemegtbaatar gobiensis (modified from Hurum, 1998, fig. 5). Abbreviations: al, anterior lamina of the 
petrosal; bo, basioccipital; bs, basisphenoid; bsw, basisphenoid wing (parasphenoid ala); ce, cavum epiptericum (for the trigem- 
inal ganglion); co, cochlear canal; ep, epipterygoid (alisphenoid); If, lateral flange of the petrosal; pa, pila antotica; pc, pars 
cochlearis; pr, promontorium (ventral or external eminence formed by the pars cochlearis). 

out that the presence of the promontorium 
in Yunnanodon, as first noted by Sun and 
Cui (1987), is reconfirmed here. However, 
the promontorium on their specimen 
(IVPP 5071) is exposed because of the 
postmortem detachment of the basisphe- 

noid and the basioccipital (see Fig. 2). In 
the intact specimens of Yunnanodon, the 
pars cochlearis is mostly concealed in ven- 
tral view. In one specimen (IVPP 7219), 
the basisphenoid wing seems to cover at 
least one half of the pars cochlearis on one 

Tritylodontid Inner Ear • Luo 89 

side, but its absence on the other side The saccular and utricular recesses are 

leaves much of the pars cochlearis ex- divided in Yiinnanodon, similar to those of 

posed. Bienotheriutn as illustrated by Hopson 

(1965, fig. 12). The junction of the bony 

Inner Ear utricular recess and the ampullae of the 

The cochlear canal is a tubular structure anterior and lateral semicircular canals is 

with a bulge in the middle part (best seen slightly more inflated than the rest of the 

in lateral view; Fig. 2). The canal is ap- vestibule. All three bony semicircular ca- 

proximately 1.9 mm in length as measured nals have somewhat irregular shapes. The 

from the anterior border of the fenestra bony tubes of the semicircular canals 

vestibuli to the apex of the cochlear canal range from 0.4 to 0.5 mm in diameter. The 

{sensu Luo et al. [1995]), or about 3.7 mm anterior semicircular canal has the largest 

if measured from the posterior border of arc with a maximum radius of —1.7 mm. 

the fenestra to the apex (sensu Rosowski The posterior semicircular canal has a 

and Graybeal [1991]). The inner surface of maximum radius of 1.1 mm. The posterior 

the bony cochlear canal is simple and de- part of this canal is bent to form an angle, 

void of internal structures. The lateral semicircular canal is the small- 

The canal is slightly constricted anterior est, with a radius to its arc of 0.8 mm. The 
to the fenestra vestibufi (Fig. 2A). From lateral and posterior semicircular canals 
serial sections it appears that the cochlear are located within the petrosal deep to the 
canal is connected to the bony saccular fossa for the stapedial muscle (Fig. 2). The 
cavity by a relatively narrow and short anterior semicircular canal is located with- 
channel. This channel probably housed in the petrosal portion of the side wall for 
the ductus reuniens, the membranous the braincase and dorsal to the bulbous 
structure that connects the saccule to the anterior paroccipital process. On the en- 
basal (proximal) part of the cochlear duct, docranial surface of the braincase, the an- 
On the basis of these bony features, it may terior semicircular canal forms the rim of 
be inferred that the cochlear part was bet- the subarcuate fossa and almost encircles 
ter differentiated from the saccular part of the fossa, 
the membranous labyrinth than in the p-.o-^. looinM 
primitive condition of other cynodonts. LJIoUUbolUN 

The middle portion of the cochlear canal In the primitive condition of noncyno- 

is bulging with a maximum diameter about dont therapsids such as dicynodonts (Fig. 

1.7 mm but its anterior portion tapers to- 4A; Olson, 1944; Cox, 1962) and gorgon- 

wards the apex that is slightly turned in opsids (Olson, 1944; Sigogneau, 1974), the 

dorsolateral direction. The bulging middle sacculocochlear cavity ( = "lagenar recess" 

portion of the cochlea is reminiscent of the of Sigogneau [1974]) is not differentiated 

globular outline of the cochleas in other from the rest of the bony vestibular cavity 

advanced nonmammalian cynodonts (Qui- in the inner ear (Fig. 4A). The fenestra 

roga, 1979). The bone near the anterior vestibuli may be either on the lateral as- 

apex is fractured in serial sections (repre- pect, or on the ventrolateral aspect in the 

sented by dashed line on the endocasts in sacculocochlear cavity. 

Figs. 1 and 2). The floor (fundus) of the A smaH and globular cochlear cavity is 

internal acoustic meatus is fully ossified, as differentiated from the main part of the 

already described in Oligokijphus (Kuhne, saccular recess in the cynodont Thrinaxo- 

1956), Bienothehum (Hopson, 1964), and don. The cochlear cavity is distinctive from 

Tritylodon (Gow, 1986). The cochlear and the fenestra vestibuli in the lateral view 

vestibular branches of the vestibulococh- (Fig. 4B) and from the vestibule in the 

lear cranial nerve (VIII) had separate fo- medial view (not illustrated; Fourie, 1974; 

ramina to the inner ear. Rowe et al, 1993; E. F. Allin, personal 


Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 4. Comparison of inner ear endocasts of cynodonts and primlfive mammals. (A) A generalized inner ear endocast of a 
dicynodont (modified from Olson [1944] and Cox [1962]). (B) A generalized inner ear endocast of Thhnaxodon (a composite 
figure based on descriptions and figures of Olson [1944], Fourie [1974], and illustrations by courtesy of Dr. Allin). (C) Probele- 
sodon (modified and reversed from Quiroga [1979]). (D) Probainognathus (modified from Allin [1986]). (E) Yunnanodon (baaed 

Tritylodontid Inner Ear •Luo 91 

communication). However, the cochlear 
cavity does not extend anteriorly beyond 
the fenestra vestibuli. The bony housing of 
the inner ear has contributions from sev- 
eral bones: the prootic, the opisthotic, the 
exoccipital, and the supraoccipital, none of 
which are fused with one another. The co- 
chlear component of the prootic bone is 
very small (Fourie, 1974). The basisphe- 
noid wing contributes to the cochlear 
housing extensively. On the external sur- 
face of the basicranium, the basisphenoid 
wing reaches and participates in the rim of 
the fenestra vestibuli, concealing the pro- 
otic froin the ventral view. 

The more advanced Probelesodon and 
Probainognathus have a globular cochlear 
cavity that is larger in proportion to the 
whole inner ear (Figs. 4C, D; Quiroga, 
1979; Allin, 1986). The cochlear structure 
also extends anterior to the fenestra ves- 
tibuli (Figs. 4C, D). A larger size and a 
more anterior orientation of the cochlear 
structure are both derived characteristics, 
compared to that o{ Thrinaxodon, and very 
different from the poorly differentiated 
sacculocochlear recess in dicynodonts and 

Probelesodon and Probainognathus are 
more primitive than Yiinnanodon in that 
their cochlear cavity is iTiuch smaller (Fig. 
4), as is the pars cochlearis of the prootic. 
The pars cochlearis does not form an ex- 
ternal promontorium. The cochlear cavity 
does not extend anteromedially, dorsal to 
the basisphenoid wing. The basisphenoid 
wing bears a prominent crest and reaches 
near the fenestra vestibuli. The rim of the 
fenestra vestibuli is formed by an elevated 
and thickened ring (Lucas and Luo, 1993; 
Luo, 1994). The prootic and the opisthotic 
are separate bones. All these characters 

are unlike those of tritylodontids and 

The tritylodontid Yunnanodon (Fig. 4E) 
is more derived than other nonmammalian 
cynodonts including Probelesodon, Pro- 
bainognathus, and Massetognathus (Qui- 
roga, 1979; Allin, 1986) in possessing a 
longer and larger cochlear canal. The co- 
chlear canal extends far beyond the ante- 
rior rim of the fenestra vestibuli. The prox- 
imal part of the canal is distinctive from 
the saccular region. The cast of the co- 
chlear canal of Yunnanodon is similar to 
those of Sinoconodon (Luo et al., 1995) 
and Morganucodon (Graybeal et al., 1989; 
Luo and Ketten, 1991). The cochlear canal 
is even longer in Yunnanodon than in Sin- 
oconodon in absolute size. Its proportion 
to the overall length of the skull is about 
the same as in the latter. The only feature 
of the cochlea that may be interpreted as 
a primitive character is the bulging middle 
part that somewhat resembles the more 
globular cochleas in Probainognathus (Al- 
lin, 1986) and Massetognathus (Quiroga, 

The cochlear canal does not extend the 
entire length of the pars cochlearis in Yun- 
nanodon, thus resembling those of Sino- 
conodon and Haldanodon, in the propor- 
tion of the cochlear canal to the pars coch- 
learis. Although the cochlea is housed in a 
pars cochlearis with an externally exposed 
promontorium in all known mammali- 
aforms and mammals, the presence of the 
petrosal promontorium is not necessarily 
correlated to a fully elongated cochlea, as 
evidenced by Sinoconodon (Luo et al., 
1995) and the late Jurassic docodont Hal- 
danodon (Lillegraven and Krusat, 1991). 
In both mammals the cochlea is short rel- 
ative to the size of the pars cochlearis. 

on reconstruction from serial sections from a specimen courtesy of A.-L. Sun). (F) Sinoconodon (based on sections of IVPP 
8689). (G) Morganucodon (modified from Graybeal et al. [1989] and Luo and Ketten [1991]). Figures not to the same scale. 
Abbreviations; asc, anterior semicircular canal; co, cochlear canal; cr, globular cochlear recess (undifferentiated from the vesti- 
bule); fc-jf, confluent foramen cochleae and jugular foramen; fv, fenestra vestibuli (oval window); Isc, lateral semicircular canal; 
psc, posterior semicircular canal; sc, undifferentiated osseous sacculocochlear structure (essentially a part of the vestibule). 

92 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Yunnanodon, Sinoconodon, and Haldano- 
don share this primitive condition, in con- 
trast to the more derived Morganiicodon, 
triconodontids, and multituberculates, in 
which the cochlear canal extends the full 
length of the pars cochlearis (and the pro- 

The overlap of the basioccipital on the 
flat medial facet of the pars cochlearis is 
present in both Yunnanodon and Sinocon- 
odon (Figs. 3B, C). However, unlike the 
condition in Sinoconodon and other early 
mammaliaforms in which the basisphenoid 
wing is lost and the pars cochlearis is ven- 
trally exposed (Fig. 3C), the pars coch- 
learis of Yunnanodon is not well exposed, 
because of coverage by the well-developed 
basisphenoid wing (Fig. 3B). It should be 
noted that the presence of a pars coch- 
learis may be a shared derived condition 
of mammals, tritylodontids, tritheledon- 
tids, and possibly probainognathids. As re- 
ported by Crompton (1994), the trithele- 
dontid Pachygenelus, a taxon closely relat- 
ed to mammals, has a small cochlea con- 
tained within the pars cochlearis that is 
also concealed by the basisphenoid and 
the basioccipital — very similar to the con- 
figuration in Yunnanodon as described in 
this paper. A juvenile skull of a probain- 
ognathid cynodont (Bonaparte and 
Crompton, 1994) also has a proinonto- 
riumlike structure in the ear region. Al- 
though this structure is not known in the 
adult specimens of Probainognathus, Pro- 
bainognathus possibly has a small pars 
cochlearis that is covered by the basisphe- 
noid and basioccipital and not exposed as 
the promontorium in the fully grown 

Tritylodontids and tritheledontids are 
both considered to belong to the mam- 
maliamorphs (modified from Rowe [1988] 
to include tritheledontids; see also Wible 
[1991], Wible and Hopson [1993], Luo 
[1994], and Luo and Crompton [1994]). 
The characteristics of the inner ear and its 
bony housing in these two groups, as de- 
scribed by Crompton (1994) and here, of- 
fer fresh insight into the pattern of early 

evolution of the anatomical structures of 
the basicranium and the inner ear (Fig. 5). 

The development of an elongate cochle- 
ar canal and change in the pattern of os- 
sification of the basicranial bones are cor- 
related (Fig. 5). It is hypothesized that in 
advanced cynodonts, such as Probainog- 
nathus (Allin, 1986), Probelesodon, and 
Massetognathus (Quiroga, 1979), an ex- 
pansion of the cochlear recess occurs with- 
in the prootic. This development is related 
to a greater reduction of the basisphenoid 
wing in these derived cynodonts (Fig. 5: 
node B) than in Thrinaxodon and other 
basal cynodonts. 

In mammaliamorphs (Fig. 5, node C, 
including tritheledontids), embryonic os- 
sifications of the otic capsule must have 
fused into a single petrosal bone in the ful- 
ly grown adults, instead of two separate os- 
sifications (prootic and opisthotic). This 
change may have made the bony housing 
for the entire inner ear more rigid; as a 
result the inner ear may be better insulat- 
ed and less susceptible to interference. 
Related to the change in ossification, a 
short but distinctive cochlear canal is de- 
veloped within a neomorphic pars coch- 
learis, as described for Yunnanodon here, 
and for Pachygenelus by Crompton (1994). 
In Sinoconodon and more derived mam- 
mals (Fig. 5, node D), the basisphenoid 
wing is lost so that the pars cochlearis is 
exposed on the ventral surface of the bas- 
icranium. However, the ventral coverage 
of the pars cochlearis by the basioccipital, 
a primitive condition, is retained at least in 
part in Sinoconodon. 

In Morganiicodon and more derived 
mammals except docodonts (Fig. 5, node 
D), the cochlear canal is more elongate 
than those of Sinoconodon and Yunnano- 
don, and extends the full length of the pars 
cochlearis. Related to the elongation of the 
cochlear canal, the external surface of the 
promontorium is also more inflated. The 
basioccipital is shifted medially (Fig. 3). 
Correspondingly, the flat medial facet on 
the promontorium as seen in Sinoconodon 
and Yunnanodon is lost. 




^ Placentals 























^ Monotremes 


— Sinoconodon 









Tritylodontid Inner Ear • Luo 



Inner Ear 

canal coil 

duct coil 




Bony Housing 

inflated promontorium 
& bo reduction 

- promontorium 
& loss of bsw 

periotic fusion 

bsw reduction 

prootic housing 

Figure 5. Stepwise transformation of the bony housing of inner ear and the cochlear structures from nonmammalian cynodonts 
to mammals. Phylogeny is based on Kemp (1983), Rowe (1988), Wible (1991), Wible and Hopson (1993), Luo (1994), Luo and 
Crompton (1994), and Hu et al. (1997). Definitions of mammaliamorphs and mammaliaforms are modified from Rowe (1988). 
Character evolution from nodes A through E is explained in text. The membranous labyrinth of the cochlea (cochlear duct) is 
coiled (node F) in all living mammals (Zeller, 1989; Luo and Ketten, 1991; Fox and Meng, 1997), although in monotremes the 
bony labyrinth of the cochlea (cochlear canal) lacks the corresponding coil. Multituberculates (node G) have a straight or slightly 
curved cochlear canal without coil (Luo and Ketten, 1991; Meng and Wyss, 1995; Hurum, 1998), as in the symmetrodont 
Zhangheotherium (node H; see Hu et al. [1997]). Only the living marsupials and placentals (node I) have a fully coiled bony 
labyrinth (cochlear canal) in correlation with the coiled membranous labyrinth (cochlear duct; Zeller, 1 989; Luo and Ketten, 1 991 ; 
Fox and Meng, 1997). Either the coiled membranous cochlear duct in living monotremes must be considered as convergent to 
those of living therians, or the uncoiled cochleas of multituberculates and Zhangheotherium (Hu et al., 1997) must be regarded 
as an atavistic reversal to those of mammaliaforms. The coiled cochlear structures within the pars cochlearis are homoplasic 
among main lineages of the mammalian crown group. Abbreviations: bo, basioccipital; bsw, basisphenoid wing. 

Mammals are most specialized among 
living vertebrates in their hearing adapta- 
tion, much of which is attributable to their 
derived inner and middle ear structures. 
The pars cochlearis containing a cochlea is 
one of the most complex character systems 
of the mammalian skull, and is crucial for 
more sensitive hearing, especially for high 
frequency sound. The assembly of such a 
complex character system udth significant 
functional adaptation occurred in several 
incremental steps during the morphologic 

evolution of nonmammalian cynodonts 
and early mammals (Fig. 5). 

The development and the emergence of 
the pars cochlearis in the basicranium are 
correlated with the transformation of the 
cochlear canal. It is hypothesized that, 
through the transition from nonmammali- 
an cynodonts to early mammaliaforms, the 
enlarged pars cochlearis with a cochlear 
canal had preceded the development of 
the petrosal promontorium. The promon- 
torium is developed by the emergence of 

94 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

the pars cochlearis on the basicranium to 
displace the neighboring sphenoid com- 
plex and basioccipital bone. Inflation of 
the external promontorium of the pars 
cochlearis is associated with the elongation 
of the cochlear canal. The seemingly dras- 
tic transformation of the ear region in the 
basicranium is achieved in a stepwise 
transformation though the phylogenetic 
transition from nonmammalian cynodonts 
to mammals. 

After an elongate cochlea developed 
within the enlarged pars cochlearis of the 
petrosal in mammaliaforms, further evo- 
lution of the cochlear coiling in the crown 
group of mammals shows a considerable 
degree of homoplasy (Fig. 5). The living 
monotremes have a coiled cochlear duct 
(membranous labyrinth) but without the 
corresponding coil of the bony cochlear 
canal (bony labyrinth). Multituberculates 
have a straight or slightly curved bony co- 
chlear canal without coil (Luo and Ketten, 
1991; Meng and Wyss, 1995; Hurum, 
1998), as in the symmetrodont Zhangh- 
eotherium (Hu et al., 1997). Only the liv- 
ing marsupials and placentals have a fully 
coiled membranous labyrinth (cochlear 
duct) that is intricately associated with the 
coiled bony labyrinth (cochlear canal; Zell- 
er, 1989; Luo and Ketten, 1991; Fox and 
Meng, 1997). Either the coiling of the 
membranous cochlear duct in living mono- 
tremes must be considered as convergent 
to those of li\dng therians, or the uncoiled 
cochleas of multituberculates and Zhangh- 
eotherium (Hu et al., 1997) must be re- 
garded as an atavistic reversal to those of 
mammaliaforms (Fig. 5). 


Professor A. W. Crompton has been an 
inspiration for this work. I thank Profes- 
sors A.-L. Sun and Z.-M. Dong, X.-C. Wu, 
and Mr. G. Cui for generously providing 
the tritylodontid specimens and their casts. 
For access to comparative materials, I 
thank Professors A. W. Crompton and F. 
A. Jenkins, Jr., and Mr. C. R. Schaff (Har- 
vard University); Professor J. A. Hopson 

(University of Chicago); and Professor A.- 
L. Sun (Institute of Vertebrate Paleontol- 
ogy and Paleoanthropology). Professor 
Crompton provided the Croft grinder. 
Professor Sun helped with the serial sec- 
tions. Mr. J. A. Georgi assisted in comput- 
erized reconstruction and Mr. M. A. Klin- 
gler assisted in illustrations. During this 
study, I benefited from discussion with 
Drs. E. F. Allin, A. W. Crompton, J. A. 
Hopson, T Rowe, and J. R. Wible. Dr. Al- 
lin graciously provided his unpublished 
drawings of Thrinaxodon for this study. 
The manuscript benefited from the critical 
and editorial reviews by Drs. Allin, Hop- 
son, Jenkins, Shapiro and Wible, and an 
anonymous reviewer. This research was 
supported by a National Science Founda- 
tion CAREER Award (DEB 9527892), the 
National Geographic Society, and the Net- 
ting and O'Neil Funds of Carnegie Mu- 
seum of Natural History. 


Allin, E. F. 1986. The auditory apparatus of ad- 
vanced mammal-like reptiles and early mammals, 
pp. 283-294. In N. Hotton, III, P. D. MacLean. 
J. J. Roth, and E. C. Roth (eds.). The Ecolog\ 
and Biology of Mammal-like Reptiles. Washing- 
ton, D.C.: Smithsonian Institution Press, x + 326 

Allin, E. R, and J. A. Hopson. 1992. Evolution of 
the auditory system in Synapsida ("mammal-like 
reptiles" and primitive mammals) as seen in the 
fossil record, pp. 587-614. In D. B. Webster, R. 
R. Fay, and A. N. Popper (eds.). The Evolution- 
ar\' Biology of Hearing. New York: Springer- Ver- 
lag. li 4- 859 pp. 

Bellairs, a. d'A., and a. M. Kamal. 1981. The 
chondrocranium and the development of the 
skull in recent reptiles, pp. 1—263. In C. Gans 
and T. S. Parsons (eds.). The Biology of Reptiles. 
Vol. 11. London: Academic Press, xi -I- 475 pp. 

Bonaparte, J. F, and A. W. Crompton. 1994. A 
juvenile probainognathid cynodont skull from 
the Ischigualasto Formation and the origin of 
mammals. Revista del Museo Argentino de Cien- 
cias Naturales, Paleontologia, 5: 1-12. 

Clark, J. M., and J. A. Hopson. 1985. Distinctive 
mammal-like reptile from Mexico and its bear- 
ings on the phylogeny of Tritylodontidae. Nature, 
315: 398-400. 

Cox, C. B. 1962. A natural cast of die inner ear of a 
dicynodont. American Museum Novitates, 2116: 

Tritylodontid Inner Ear • Luo 95 

Croft, W. N. 1950. A parallel grinding instrument 
for the investigation of fossils by serial sections. 
Journal of Paleontolog>', 24: 693-698. 

CromptoN, a. W. 1955. Techniques for the study of 
Permo-Triassic Fossils of South Africa. South Af- 
rican Museums Association Bulletin, 6: 57-60. 

. 1964. On the skull of Oligokyphiis. Bulletin 

of the British Museum (Natural History'), Geol- 
ogy, 9: 70-82. 

. 1972. Postcanine occlusion in c\aiodonts and 

tritylodonts. Bulletin of the British Museum 
(Natural History), Geology, 21: 27-71. 

. 1994. Masticatory function in non-mamma- 
lian cynodonts and early mammals, pp. 55—75. In 
J. J. Thomason (ed.). Functional Morphology in 
Vertebrate Paleontology. Cambridge, United 
Kingdom: Cambridge University Press, xi + 277 

CROMPTON, A. W., AND Z. Luo. 1993. The relation- 
ships of Liassic mammals Sinoconodon, Morgan- 
iicodon oehleri and Dinnetherium, pp. 30^4. In 
F. S. Szalay, M. J. Novacek, and M. C. McKenna, 
(eds.), Mammtil Phylogeny: Mesozoic Differen- 
tiation, Multituberculates, Early Therians, and 
Marsupials. New York: Springer- Verlag. x + 249 

Cui, G. 1976. Yunnania, a new tritylodontid from Lu- 
feng, Yunnan. Vertebrata PalAsiatica, 25: 1-7 (in 

. 1986. Yunnonodon, a replacement name for 

Yunnania Cui, 1976. Vertebrata PalAsiatica, 24: 
9 (in Chinese). 

DE Beer, G. R. 1937. The development of the ver- 
tebrate skull. O.xford, United Kingdom: The 
Clarendon Press, xxii + 552 pp. 

ESTES, R. 1961. Cranial anatomy of the cynodont 
reptile Thrinaxodon liorhiniis. Bulletin of the 
Museum of Comparative Zoology, Harvard Uni- 
versity, 125: 165-180. 

FOURIE, S. 1974. The cranial morpholog)' of Tltri- 
noxodon liorhiniis Seeley. Annals of the South 
African Museum, 56: 337^00. 

Fox, R. C, AND J. Meng. 1997. An X- radiographic 
and SEM study of the osseous inner ear of mul- 
tituberculates and monotremes (Mammalia): im- 
plications for mammalian phylogeny and evolu- 
tion of hearing. Zoological Journal of the Lin- 
nean Society (London), 121: 249-191. 

Goodrich, E. S. 1930. Studies of the Structure and 
Development of Vertebrates. London: Macmil- 
lan. xxxiv -I- 837 pp. 

Gow, C. E. 1985. Apomoi-phies of the Mammalia. 
Soudi African Journal of Science, 81: 558-560. 

. 1986. The side wall of the braincase in cy- 
nodont therapsids and a note on the homology 
of the mammalian promontorium. South Africa 
Journal of Zoology, 21: 136-148. 

W. CROMPTON. 1989. Inner ear structure in Mor- 
ganucodon, an early Jurassic mammal. Zoological 

Journal of the Linnean Society (London), 96: 

HOPSON, J. A. 1964. The braincase of the advanced 
mammal-like reptile Bienotheriiim. Postilla, 87: 

. 1965. Tritylodontid therapsids from Yunnan 

and the cranial morphologv' of Bienotheriiim. 
Ph.D. dissertation. Chicago: The University of 
Chicago. 295 pp. 

analysis of therapsid relationships, pp. 8.3-106. In 
N. Hotton, 111, P D. MacLean, J. J. Roth,, and 
E. C. Roth (eds.). The Ecology and Biology of 
Mammal-like Reptiles. Washington, D.C.: Smith- 
sonian Institution Press, x + 326 pp. 

Hu, Y, Y. Wang, Z. Luo, and C. Li. 1997. A new 
symmetrodont mammal from China and its im- 
plications for mammalian evolution. Nature, 
390: 137-142. 

HURUM, J. H. 1998. The inner ear of two Late Cre- 
taceous multituberculate mammals, and its im- 
plications for multituberculate hearing. Journal 
of Mammalian Evolution, 5: 65-93. 

Jollie, M. 1962. Chordate Morphology. New York: 
Reinhold Publishing Co. 478 pp. 

Kemp, T S. 1983. The interrelationships of mammals. 
Zoological Journal of the Linnean Society (Lon- 
don), 77: 353-384. 

Kermack, K. a. 1963. The cranial structure of the 
triconodontids. Philosophical Transactions of the 
Royal Society of London, B, 246: 83-103. 

Kermack, K. A., F. Musset, and H. W Rigney. 
1981. The skull of Morganitcodon. Zoological 
Journal of the Linnean Society (London), 71: 1- 

KUHN, H.-J. 1971. Die Entwicklung und Morpholo- 
gic des Schadels von Tachtjglossus aciileatus. Ab- 
handlungen der Senckenbergischen Naturfor- 
schenden Gesellschaft, Frankfurt am Main, 528: 

KtJHNE, W G. 1956. The Liassic Therapsid Oligo- 
kyphiis. London: British Museum (Natural His- 
tory). X + 149 pp. 

LiLLEGRAVEN, J. A., AND G. Hahn. 1993. Evolution- 
ary' analysis of the middle and inner ear of Late 
Jurassic multituberculates. Journal of Mammali- 
an Evolution, 1: 47-74. 

LiLLEGRAVEN, J. A., AND G. Krusat. 1991. Cranio- 
mandibular anatomy of Haldanodon exspectatiis 
(Docodonta; Mammalia) from the Late Jurassic 
of Portugal and its implications to the evolution 
of mammahan characters. Contributions to Ge- 
ology, University of Wyoming, 28: 9-138. 

Lucas, S. G., and Z. Luo. 1993. Adelobasileus from 
the Upper Triassic of West Texas: the earliest 
mammal. Journal of Vertebrate Paleontologx', 13: 

Luo, Z. 1994. The sister taxon relationships of mam- 
mals and transformations of the diagnostic mam- 
malian characters, pp. 98-128. In N. C. Eraser 
and H.-D. Sues (eds.). In the Shadow of the Di- 


Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

nosaurs — Early Mesozoic Tetrapods. Cambridge, 
United Kingdom: Cambridge University Press, x 
+ 435 pp. 
LUO, Z., AND A. W. Crompton. 1994. Transforma- 
tions of the quadrate (incus) through the transi- 
tion from non-mammalian cynodonts to mam- 
mals. Journal of Vertebrate Paleontology, 14: 

Luo, Z., A. W. Crompton, and S. G. Lucas. 1995. 
Evolutionary origins of the mammalian promon- 
torium and cochlea. Journal of Vertebrate Pale- 
ontology, 15: 113-121. 

Luo, Z., AND D. R. Ketten. 1991. CT scanning and 
computerized reconstiaictions of the inner ear of 
multituberculate mammals. Journal of Vertebrate 
Paleontology, 11: 220-228. 

Luo, Z., AND A.-L. Sun. 1993. Oligokyphus (Cyno- 
dontia: Tritylodontidae) from the Lower Lufeng 
Formation (Lower Jurassic) of Yunnan, China. 
Journal of Vertebrate Paleontology, 13: 477^82. 

Luo, Z., AND X.-C. Wu. 1994. The small vertebrate 
fauna of the Lower Lufeng Formation, Yunnan, 
pp. 251-270. In N. C. Eraser and H.-D. Sues 
(eds). In the Shadow of the Dinosaurs — Early 
Mesozoic Tetrapods. Cambridge, United King- 
dom: Cambridge University Press, x + 435 pp. 

. 1995. Correlation of vertebrate assemblage 

of the Lower Lufeng Eormation, Yunnan, China, 
pp. 83-88. In A. Sun and Y. Wang (eds.), Sixtli 
Symposium on Mesozoic Terrestrial Ecosystems 
and Biotas, Short Papers. Beijing: China Ocean 
Press, vi -I- 250 pp. 

Meng, J., AND A. R. Wyss. 1995. Monotreme affin- 
ities and low-frequency hearing suggested by 
multituberculate ear. Nature, 377: 141-144. 

MlAO, D. 1988. Skull moi"phology of Lanibdopsalis 
bulla (Mammalia, Multituberculata) and its phy- 
logenetic implications to mammalian evolution. 
Contributions to Geology, University of Wyo- 
ming, Special Paper, 4: 1—104. 

Oelrich, T M. 1956. The anatomy of the head of 
Ctenosaura pectinata (Iguanidae). Miscellaneous 
publications. Museum of Zoology, University of 
Michigan, 94: 1-122. 

Olson, E. C. 1944. Origin of mammals based upon 
cranial morphology of therapsid suborders. Spe- 
cial Papers of the Geological Society of America, 
55: 1-122. 

QUIROGA, J. C. 1979. The inner ear of two cynodonts 
(Reptilia— Therapsida) and some comments on 
the evolution of the inner ear from pelycosaurs 
to mammals. Gegenbaurs morphologisches Jahr- 
buch, 125: 178-190. 

RiEPPEL, O. 1993. Patterns of diversity in the reptil- 
ian skull, pp. 344-390. 7;i J. Hanken and B. K. 
Hall (eds.). The Skull. Vol. 2. Chicago: The Uni- 
versity of Chicago Press, xiii -I- 566 pp. 

ROMER, A. S. 1970. The Chaiiares (Argentina) Tri- 
assic reptile fauna. VI. A chiniquodontid cyno- 
dont with an incipient squamosal— dentary artic- 
ulation, Breviora, 344: 1-18. 

ROSOWSKI, J. A. 1992. Hearing in transitional mam- 
mals: predictions from the middle-ear anatomy 
and hearing capabilities of extant mammals, pp. 
615-631. In D. B. Webster, R. R. Fay, and A. N. 
Popper (eds.). The Evolutionary Biology of 
Hearing. New York: Springer- Verlag. li + 859 pp. 

RosowSKi, J. A., AND A. Graybeal. 1991. What did 
Morganucodon hear? Zoological Journal of the 
Linnean Society (London), 101: 131-168. 

1996. Basicranial &r\a.tomy o{ Priacodonfniitaen- 
sis (Triconodontidae, Mammalia) from the Late 
Jurassic of Colorado, and a reappraisal of mam- 
maliaform interrelationships. American Museum 
Novitates, 3183: 1-28. 

ROWE, T 1988. Definition, diagnosis and origin of 
Mammalia. Journal of Vertebrate Paleontology, 
8: 241-264. 

RowE, T, W Carlson, and W Bottorff. 1993. 
Thrinaxodon — digital atlas of the skull. Austin, 
Texas: University of Texas Press (CD-ROM). 

Sigogneau, D. 1974. The inner ear of Gorgonops 
(Reptilia, Therapsida, Gorgonopsia). Annals of 
the South African Museum, 64: 53-69. 

Sues, H.-D. 1985a. The relationships of the Tritylo- 
dontidae (Synapsida). Zoological Journal of the 
Linnean Society (London), 85: 205—217. 

. 1985b. Dinnebitodon amarali, a new tritylo- 

dontid (Synapsida) from the Lower Jurassic of 
western North America. Journal of Paleontology, 
60: 758-762. 

1986. The skidl and dentition of two trity- 

lodontid synapsids from tlie Lower Jurassic of 
Western North America. Bulletin of the Museum 
of Comparative Zoology, Harvard University; 
151: 217-268. 

Sun, A.-L. 1984. Skull moi-phology of the tritylodont 
genus Bienotheroides of Sichuan. Scientia Sinica, 
Series B, 27: 270-284. 

Sun, A.-L., and G. Cul 1987. Otic region in trity- 
lodont Yitnnanodon. Vertebrata PalAsiatica, 25: 

Sun, A.-L., G. GUI, Y. Li, and X.-C. Wu. 1985. A 
verified list of Lufeng Saurischian Fauna. Ver- 
tebrata PalAsiatica, 22:1-12. 

Wible, J. R. 1991. Origin of Mammalia: the craniod- 
ental evidence reexamined. Journal of Vertebrate 
Paleontology, 11: 1-28. 

WlBLE, J. R., AND J. A. HOPSON. 1993. Basicranial 
evidence for early mammal phylogeny, pp. 45- 
62. In F. S. Szalay, M. J. Novacek, and M. C. 
McKenna (eds.). Mammal Phylogeny: Mesozoic 
Differentiation, Multituberculates, Early Theri- 
ans, and Marsupials. New York: Springer- Verlag. 
X 4- 249 pp. 

. 1995. Homologies of the prootic canal in 

mammals and non-mammalian cynodonts. Jour- 
nal of Vertebrate Paleontology, 15: 331-356. 

WlBLE, J. R., G. W. Rougier, m"'j. Novacek, M. C. 
McKenna, and D. Dashzeveg. 1995. A mam- 
malian petrosal from the Early Cretaceous of 

Tritylodontid Inner Ear • Luo 


Mongolia: implications for the evolution of the 
ear region and mammaliamorph interrelation- 
ships. American Museum Novitates, 3149: 1-19. 
Williams, P. L., R. Williams, M. Dyson, and L. 
H. Bannister. 1989. Grays Anatomy, 37th edi- 
tion. New York: Churchill Livingstone. 1,598 pp. 

Zeller, U. 1989. Die Enhvicklung und Morphologie 
des Schadels von Ornithorhijnchus anatinus 
(Mammalia: Prototheria: Monotremata). Abhan- 
dlungen der Senckenbergischen Naturforschen- 
den Gesellschaft, Frankfurt am Main, 545: 1— 



Abstract. Batodon tenuis Marsh, 1892, a rare, mi- 
nute, eutherian mammal, is a member of several Late 
Cretaceous (Lancian North American Land Mammal 
Age), North American local faunas. A hitherto un- 
described fragment of maxilla from the Hell Creek 
Formation, Garfield Count>', Montana, documents 
tlie association of M- and M^ Analysis of the small 
available sample of upper and lower molars in func- 
tional orientation strengthens the basis for their as- 
sociation, provides a functional explanation for the 
enlargement of the talonid of Mj, and suggests only 
one species is represented. Currently, B. tenuis, with 
an estimated body mass of approximately 5 g, is the 
smallest known eutherian mammal from the Creta- 
ceous. The phylogenetic position of B. tenuis in the 
poorlv documented. Cretaceous radiation of euthe- 
rians is still unclear. 


In a search of University of California 
Museum of Paleontology (UCMP) collec- 
tions to find uncataloged and fragmentary 
teeth of Late Cretaceous marsupials for 
enamel microstructure research (Wood et 
al., 1999), an important new specimen re- 
ferable to Batodon tenuis came to light 
(Wood and Clemens, 1990). Batodon ten- 
uis Marsh 1892 is a tiny, very rare. Late 
Cretaceous eutherian mammal previously 
known from isolated teeth and only three 
jaw fragments containing more than one 
tooth (see Lillegraven [1969], Clemens 
[1973], Archibald [1982], Storer [1991], 
and Lofgren [1995]). The type specimen, 
USNM 2139, is a dentary fragment with 

' Department of Biology, Providence College, 
Providence, Rhode Island 02918. 

- Museum of Paleontology, University of Cahfomia, 
Berkeley, California 94720. 

Pi^i in place. The new specimen, UCMP 
136091, is a fragment of a right maxillaiy 
with almost undainaged M-^^ in place (Fig. 
1). This specimen is only the second max- 
illary fragment of B. tenuis containing 
more than one tooth to be discovered and 
contains the first record of M^ for the spe- 
cies and genus. The systematic affinities of 
Batodon are unclear, and, as discussed be- 
low, we choose to regard this rare genus 
as incertae sedis within the Placentalia. 

Butler (1961, 1972a), Fox (1975), and 
Crompton and Kielan-Jaworowska (1978), 
among others, have advocated a functional 
approach to description and phylogenetic 
analysis of tribosphenic mammals. A great 
deal of this and later work (including this 
paper) are due to the influence of Profes- 
sor A. W Cromptons foundational work 
on the origin and function of tribosphenic 
molar teeth (see Crompton [1971] as a 
landmark example; also Crompton and 
Hiiemae [1969] and Crompton and Sita- 
Lumsden [1970]). 

Our first goal in this study was to deter- 
mine whether functional correspondence 
of the new specimen (UCMP 136091) 
with the M2_3 present in UA 3721 (the only 
known, associated last two lower molars 
referred to B. tennis [Lillegraven, 1969]) 
reasonably substantiates allocation of both 
upper and lower molars to the same spe- 
cies. Furthermore, we wished to deter- 
inine if the molars preserved in the new 
specimen show close size and morphologic 
similarity to all other previously referred 

Bull. Mus. Comp. Zool., 156(1): 99-118, October, 2001 


100 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


Figure 1. Stereophotograph of UCMP 136091, Batodon tenuis, maxillary fragment with M^ 

specimens, so that greater confidence 
could be expressed in their allocation to a 
single species. This paper perhaps may 
serve as a case study in the utility of func- 
tional association of isolated Mesozoic 
postcanine elements, an easy and yet of- 
ten-neglected method. Finally, we com- 
pare the estimated body mass of B. tenuis 
to those of other small eutherians and an- 
alyze hypotheses of its phylogenetic rela- 



Here we follow McKenna and Bell 
(1997) in formally recognizing Placentalia 
as the appropriate name for the clade that 
frequently is dubbed Eutheria. Informally, 
reflecting vernacular use, the terms euthe- 
rian and placentalian are recognized as 

Provenance and History 

The new maxillaiy fragment, UCMP 
136091, described here comes from con- 
centrates obtained in 1974 by screen wash- 
ing sedimentary rocks of the Hell Creek 
Formation at the Flat Creek Locality 5 (V- 
73087) in Garfield County, Montana. Ar- 

chibald (1982: 166) described four isolated 
teeth of B. tenuis from this locality. The 
new specimen was discovered after he 
coinpleted his monographic study. 

Marsh (1892) established B. tenuis on a 
fragment of a dentary (USNM 2139) con- 
taining Po^ from the Lance Formation, Ni- 
obrara County, Wyoming. The type locality 
is UCMP locality no. V-5003 (Clemens, 
1973), also referred to as Mammal locality 
no. 1 (Lull, 1915). Additional specimens of 
B. tenuis have been reported from Lancian 
North American Land Mammal Age 
(NALMA) local faunas of the Lance For- 
mation, Wyoming (Clemens, 1973); the 
Scollard Formation, Alberta (Lillegraven, 
1969); the Hell Creek Formation, Mon- 
tana (Archibald, 1982); and the French- 
man Formation, Saskatchewan (Storer, 

A boundary between the Lancian NAL- 
MA and the older, still poorly character- 
ized "Edmontonian" NALMA, has not 
been defined. In part this reflects the pres- 
ence of marine units (e.g., the Beaipaw 
Shale) separating the terrestrial deposits 
producing Lancian and "Edmontonian" lo- 
cal faunas (see Lillegraven and McKenna, 
1986). Recently Hicks et al. (1999) esti- 

Functional Molar Association in Batodon • Wood and Clemens 101 

mated the duration of deposition of the 
Hell Creek Formation in North Dakota, 
which has yielded Lancian local faunas, as 
encompassing approximately the last 1.7 
million years (ca. 65.5-67.2 million years 
before the present) of the Cretaceous. 
This can be taken as a minimum duration 
of the Lancian NALMA and probably in- 
cludes the ages of all known occurrences 
of B. tenuis. 

Until recently, the genus Batodon had 
not been reported in older, Judithian 
NALMA local faunas of the North Amer- 
ican Western Interior. Some of these are 
known from large samples obtained by 
screen washing (e.g., Montellano, 1992). 
In an abstract, Carrano et al. (1997) re- 
cently noted the discovery of Batodon sp. 
at an unspecified locality in the type area 
of the Judith River Formation. When fully 
documented, this record would greatly ex- 
tend the range of the genus over approx- 
imately the last 13 milHon years of the 

Batodon tenuis was not represented in 
the large sample of the early Paleocene 
(Puercan NALMA) Hells Hollow local 
fauna (Archibald, 1982). Lofgren (1995) 
recovered four isolated teeth referable to 
B. tenuis from channel fillings in the Hell 
Creek Formation, McCone County, Mon- 
tana, which he interpreted to have been 
deposited during the Puercan. Because of 
reworking, these channel fillings contain a 
mixture of fossils of Puercan and Lancian 
vertebrates. Sloan and Van Valen (1965) 
did not report the occurrence of B. tenuis 
in their collections from the Bug Creek 
Anthills locality in McCone County. The 
UCMP collections made at this locality 
(Bug Creek Anthills, V-87038, 87074, 
87151) include four isolated molars refer- 
able to this species. The time of deposition 
of the channel filling at Bug Creek Anthills 
and the composition of its vertebrate fauna 
have been disputed. The strongly support- 
ed hypothesis that the Bug Creek Anthills 
channel filling was deposited in the Puer- 
can and contains a mixture of latest Cre- 
taceous and early Paleocene vertebrate 

fossils has not been falsified (for discussion 
of this problem see Lofgren et al. [1990] 
and Lofgren [1995]). In the following anal- 
ysis we include data on the molars from 
Bug Creek Anthills locality and the chan- 
nel fillings investigated by Lofgren (1995) 
but do not interpret them as documenting 
an extension of the range of Batodon into 
the Puercan. At least in the North Amer- 
ican Western Interior, the genus Batodon 
appears to have become extinct at the end 
of the Cretaceous. 

Functional Orientation 

To obsei-ve their functional orientation 
teeth are rotated around their anterior- 
posterior axes until the line of sight is par- 
allel with the direction of movement of the 
lower teeth as they were drawn into cen- 
tric occlusion with the upper teeth. In this 
orientation the paracone and metacone 
just mask views of their respective conules. 
All polished wear facets on upper and low- 
er teeth are produced by simultaneous, 
parallel shear in this direction ("phase 
one," see Kay and Hiiemae [1974]). When 
the proper degree of rotation has been at- 
tained, the shear facets disappear from 
view because they are aligned parallel with 
the line of sight. This orientation best il- 
lustrates the relative lengths of shearing 
blades and other functional elements of 
the crown. In this orientation, upper and 
lower teeth can be drawn separately on 
tracing paper, or as computer-based im- 
ages, and then superimposed to show pre- 
cise occlusal relationships of all their cor- 
responding parts. Wood et al. (1979) used 
this technique to support association of 
rare, isolated upper and lower teeth of the 
Paleocene "primate," Torrejonia. Howev- 
er, in general, this technique has not en- 
joyed wide application in the study of tri- 
bosphenic dentitions. 

In this paper our figures are labeled as 
"functional" when specimens are illustrat- 
ed in the functionally rotated view, and 
"crown" in the orientation that has tradi- 
tionally been labeled as "occlusal view" 

102 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

(but see Crompton and Kielan-Jaworows- 
ka [1978] for additional terms). 



P 2004.565 





American Museum of Nat- 
ural History 

North American Land 
Mammal Age 

Specimen numbers given 
in this format are from the 
Saskatchewan Museum of 
Natural Histoiy 
University of Alberta 
University of California 
Museum of Paleontology 
Natural History Museum, 
Smithsonian Institution 
Locality designations given 
in this format are from the 
UCMP locality catalog 

Description of New Specimen 

When discovered much of the new max- 
illary fragment (UCMP 136091) was ob- 
scured by loosely adhering sand grains. 
The parastylar region of M- was missing. 
During cleaning the M- separated from 
the M^ allowing the adjacent sides of the 
molars to be illustrated fully. Subsequent- 
ly, the maxillary fragments were rejoined 
and the molars exactly restored to their 
original positions. Figure 1 is a stereopho- 
tograph of the cleaned and restored spec- 
imen in crown view. Figure 2 is made up 
of line drawings of the two molars in tra- 
ditional crown, lingual oblique, labial, an- 
terior, functional, and posterior views. 

The crown view of M^ of UCMP 136091 
(Fig. 2 A) does not clearly illustrate some 
postmortem damage to the tooth. At high- 
er magnifications of the anterior view (Fig. 
2E) the enamel is obviously broken and 
lacking from the extremely narrow area of 
attachment of a parastylar lobe. The an- 
terior root beneath the paracone is also 
freshly broken. The parastylar lobe is bro- 

ken away from all other known upper mo- 
lars except UA 4081, P 2004.565, and the 
M^ of UCMP 136091. On M^ of UA 4081 
(Fig. 3), the parastylar lobe is a prominent 
but very thin and delicate structure; one 
may surmise that it survived only because 
it was protected by the stout metastylar 
blade of M^. 

Superimposed crown views (not illus- 
trated) of the M2 of UCMP 136091 and of 
UA 4081 are remarkably coincident, es- 
pecially in the areas of the protocones, 
paracones, metacones, and conules. Dif- 
ferences are in a slightly inore robust labial 
edge of the metastylar area of UCMP 
136091 and its more robust precingulum. 
The postcingulum of UA 4081 is slightly 
more robust than that of UCMP 136091 
and is absolutely larger in its labial exten- 
sion beneath the metaconule. Although 
not the case on the molars of UCMP 
136091, on some molars of B. tennis the 
pre- and postcingula meet on the lingual 
slope of the protocone (Lillegraven, 1969). 
Some apical wear is apparent on the rims 
of the pre- and postcingula of M-s of B. 
tenuis. Evidently they did not produce a 
shear. The cingula appear to have served 
as stops for the lower teeth as they came 
to their limits in centric occlusion. Perhaps 
selective pressures would be less for pre- 
cision in shape of these structures than 
would have been the case for the func- 
tional shearing blades higher on the crown 
(but see Polly, 1998a). 

The specimen UCMP 102909 lacks 
clearly developed internal wings of the 
conules. On UCMP 117649, one of the 
specimens described by Archibald (1982), 
the metaconule has a distinct internal 
wing; a weaker but still distinct internal 
wing is present on the paraconule. The 
stages of wear of UCMP 136091 and UA 
4081 are approximately equivalent. The 
conules of both have distinct internal 
wings that are emphasized by wear, which 
produced chevrons of dentine within the 
enamel lining of the trigon basin. 

In labial and lingual oblique views of 
Batodon molars, the paracones and meta- 


Functional Molar Association in Batodon • Wood and Clemens 103 










Figure 2. Line drawings of M^ ^ UCMP 136091, Batodon tenuis. (A) Crown views; arrows to M=' indicate projecting metacone 
and metaconule; arrowhead to M^ indicates damaged part of parastylar area. Posterior is to left and anterior is to right. (B) 
Ungual oblique views; additional large arrow on M^ indicates heavily worn groove between nnetacone and paracone. (C) Labial 
views; arrowhead indicates damaged parastylar area of M=^. (D) Functionally rotated views. (E) Anterior views; arrowhead indi- 
cates damaged parastylar region of M=. (F) Posterior views. Abbreviations: me, metacone; pa, paracone; pr, protocone; ac, 
anterior protocone cingulum (precingulum); pc, posterior protocone cingulum. 

cones are closely conjoined from their ba- 
ses to perhaps one half or two thirds of 
the distance to their apices. Such mor- 
phology might suggest that little or no 
shear occurred directly between the para- 
cone and inetacone. However, in worn 
specimens, it is evident that the hypoconid 

of the lower molar scraped a strong groove 
on the lingual surfaces of those cusps be- 
fore passing between the internal conule 
wings and into the deeply excavated trigon 
basin. A comparably large hypoconid is not 
evident in Cenozoic genera such as Pa- 
laeoryctes or Didelphodus, but is approxi- 

104 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 







Figure 3. Line drawings of M'-^, from cast of UA 4081, Batodon tenuis, from the Scollard Formation, Alberta, Canada (see 
Lillegraven [1969]). (A) Crown view. (B) Lingual oblique view of M^ only. (C) Labial view. (D) Functional view, orientation as in 
Figure 2D. (E) Posterior view of M== only. 

mated in some Lancian species of Cimo- 
lestes. Lipotyphlans, many condylarths, 
and their descendants often emphasize 
shearing between the hypoconid and the 
blades of the centrocrista that Hnk the 
metacone and paracone, which usually are 
well separated to their bases. 

Superimposed labial and lingual oblique 

views of M2 of UCMP 136091 and M^ of 
UA 4081 demonstrate that the former has 
a slightly less bowed ectoflexus and a more 
robust and absolutely larger metacone and 
metastylar blade. Slightly greater wear on 
the metastylar blade of UA 4081 might 
overemphasize the difference, but we sus- 
pect that the difference is real, yet within 

Functional Molar Association in Batodon • Wood and Clemens 105 

the limits of individual variation. A small cone. The pre- and postcingula are prom- 
cuspule or expansion of the external edge inent, but the postcingulum is relatively 
of the central stvlar shelf is present on small in comparison to those of other mo- 
both specimens. On M-s of B. tenuis the lars. The cingula are more widely separat- 
paracone is taller than the metacone. ed by the strong lingual slope of the pro- 
Using the cervical limit of enamel as a tocone than on M-. 
base line, in anterior and posterior views The metacone is reduced in size, but 
of M- the protocones of all specimens are still quite prominent; it forms a distinct 
almost as high as the paracones, and about projection in the lingual oblique view (Fig. 
the same height or higher than the meta- 2B). Metacone and paracone are closely 
cones. The profile of the protocone is es- conjoined at their bases, as on M-, and 
pecially pointed and triangular. As Lille- wear emphasizes the greater angle at 
graven (1969: 82) noted, because of the which they diverge. A wear facet extends 
exaggerated protocone, the stylar shelf of down the conjoined lingual slopes of the 
Batodon appears relatively narrow, but in metacone and paracone showing that the 
absolute width its proportions are compa- hypoconid of Mj sheared down into the 
rable to the stylar shelf of Ciniolestes cer- trigon basin past the distinct internal wings 
beroides, for exainple. of the conules. The internal surface on the 
In anterior view (Fig. 2E), the paracone paraconule wall is worn flat. Resembling 
is quite steep-sided and apically almost the metacone, the metaconule projects 
rectangular in outline. In contrast, the pos- posteriorly. The posterior inclination of the 
terior profile of the metacone sweeps la- metacone and metaconule accoinmodated 
bially into the large metastylar blade. The strong hypoconid shear within the trigon. 
posterior metaconule wing is clearly a 
strong enechelon shearing blade contact- 
ing the preprotocristid of the lower molar Storer (1991) described the mammals of 
after it passed the metastylar blade. The the Lancian Gryde local fauna, Saskatch- 
anterior wing of the paraconule also forms e\van. The sample included an upper right 
a strong shearing blade for the postproto- molar, P 2004.565 (Fig. 4), that has been 
cristid, but the primary blade on the an- identified tentatively as an M^ of B. tenuis. 
terior slope of the paracone, the prepara- Storer (1991) noted that the specimen has 
crista, is less distinct, as often is the case a more anteriorally directed preparacrista 
in "proteutherians" (see Crompton and than the M- of UA 4081 (Fig. 3). Also, P 
Kielan-Jaworowska [1978]). 2004.565 has a relatively smaller parastylar 

lobe. These moi^phologic differences sup- 
port Storer s tentative identification, which 

The specimen UCMP 136091 is the first is accepted here, 
speciinen of B. tenuis to preserve M^ in 

association with other identifiable molars. Lower Dentition 

The M^ of the new specimen has a robust The type specimen of B. tenuis, USNM 
parastylar lobe with a deep groove to re- 2139, is an anterior dentaiy fragment con- 
ceive the protoconid of M3. In anterior taining P2_4. The specimen AMNH 58777 
view, the paracone-parastylar shearing preserves P^, a major part of P3, and P4 and 
blade is more distinct than on M^, and the Mj.^ (illustrated in Clemens, 1973). Com- 
shearing blade of the anterior paraconule parable small size and morphologic simi- 
wing is also strong. The protocone is con- larity of P4 to that of USNM 2139 are the 
siderably less triangular in anterior view, basis for reference of AMNH 58777 to B. 
but is as triangular as that of M- in pos- tenuis. Likewise small size and close mor- 
terior view. The protocone is taller than phologic similarity of the molars of AMNH 
the metacone but not as tall as the para- 58777 and UA 3721 (Fig. 5) are the bases 

106 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 










Figure 4. Line drawings of ?M\ P 2004.565, Batodon tenuis, from the Gryde local fauna, Frenchman Formation, Saskatchewan, 
Canada (see Storer [1991]). Note that in all views the parastylar area is intact. (A) Crown view. (B) Lingual oblique view. (C) 
Labial view. (D) Functional view. (E) Anterior view. (F) Posterior view. 

Functional Molar Association in Batodon • Wood and Clemens 107 


Mp pad med M-a 







Figure 5. Line drawings of M2_3, from cast of UA 3721, Batodon tenuis, from the Scollard Formation, Alberta, Canada (see 
Lillegraven [1969]). (A) Crown view. (B) Lingual view. (C) Labial view. (D) Anterior view of M^ and posterior view of M3. (E) 
Functionally rotated orientation. Abbreviations: pad, paraconid; med, metaconid; prd, protoconid; hyd, hypoconid; hycid, hypo- 
conulid; entd, entoconid. 

for reference of the latter to this species. 
Detailed descriptions of the morphology 
of these specimens can be found in Lille- 
graven (1969) and Clemens (1973). 

Functional Relationships of Upper and 
Lower Molars 

Figure 6 illustrates correspondingly 
numbered functional shear blades on up- 
per and lower molars (after Crompton and 
Hiiemae [1969]). Figures 7 and S are com- 
posite (same-scale) drawings of the M-^^ of 
UCMP 136091 and the M^ and fragment 

of M^ of UA 4081, each shown in occlusion 
with M,^3 of UA 3721 (Fig. 5). All the 
teeth are illustrated in functional orienta- 
tion. The occlusal fit of UCMP 136091 
with UA 3721 (Fig. 7) is, overall, some- 
what better than for UA 3721 and the up- 
per molars of UA 4081 (Fig. 8). The M^-^ 
of UCMP 136091 occlude very well with 
M2_3 of UA 3721 (Fig. 7), especially in the 
areas between the hypoconid and para- 
cone— metacone embrasure. With these 
functional units of opposing second and 
third molars in place, some other function- 

lOS Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

4a 3o 

2a -^ 

Figure 6. Line drawings of M= ^ UCMP 136091, and UA 3721, Batodon tenuis, to indicate corresponding functional shear 
blades (see Crompton and Hiiemae [1969] and Crompton and Kielan-Jaworowsl<a [1978]). (A) M^^^, UCMP 136091 in crown 
view. Abbreviations: la, preparacrista; lb, preparaconule crista (anterior wing); 2a, postmetacrista; 2b, postmetaconule crista 
(posterior wing); 3a, postparacrista; 3b, postmetaconule crista (posterior wing); 4a, premetacrista; 4b, premetaconule crista 
(anterior wing); 5, preprotocrista; 6, postprotocrista. (B) UA 3721 in functional view. Abbreviations: siiear blades: 1, postproto- 
cristid; 2, preprotocristid; 3, cristid obliqua; 4, posthypocristid; 5, postmetacristid and posterior wall of metaconid; 6, pre-entoconid 

al units are not quite in perfect occlusion. 
The slight gap between metacone and hy- 
poconulid of the second molar could be 
due to slightly greater wear on the upper 
tooth or simply to intraspecific variation in 
size. However, the protocones of M-"^^^ are 
slightly too extended lingually to fit com- 
fortably into the talonid basins of Mo_3. 
This most probably indicates individual 
variability in size. The function of the po- 
stcingulum of M^ as an embrasure stop is 

clearly reflected in the narrowed and dor- 
soventrally lowered configuration of the 
paraconid of M3. 

Lillegraven (1969: 84) and subsequent 
students of B. tenuis have commented on 
the posterior projection of the hypoconu- 
lid and, therefoi^e, the "extended talonid" 
of M3. The large, posteriorly projecting 
metacone and metaconule of M^ are func- 
tionally related to the extended talonid of 
M3. Shearing facets on the anterior face of 

Functional Molar Association in Batodon • Wood and Clemens 109 

UCMP 136091 

UA 3721 


Figure 7. Functional views of IVl^-^ (UCMP 136091) and M2_3 (UA 3721, reversed), Batodon tenuis. (A) Separated views of 
molars. (B) Upper and lower molars shown in centric occlusion. Corresponding strong shear blades 3 and 4 on the third molar, 
as well as a functional postcingulum and extended hypoconulid, indicate that the upper and lower dentitions represent the same 

the metacone (facet 4a, Fig. 6) and inter- 
nal metaconule wing (facet 4b, Fig. 6) oc- 
cluded against the posterior side of the hy- 
poconid and labial side of the enlarged hy- 

poconulid. The remainder of the hypocon- 
ulid is covered by the postcingulum of the 
upper molar, but it is not clear whether 
crushing occurred between thein at full 

110 Bulletin Museum of Comparative Zoologtj, Vol. 156, No. 1 




Figure 8. Functional views of fragmentary M' and M^ (UA 4081, reversed) and Mj 
of molars. (B) Upper and lower molars shiown in centric occlusion. 

(UA 3721 , reversed). (A) Separated views 

occlusion. Although not unique among eu- 
therians, enlargement of the hypoconulid 
and posterior projection of metacone and 
metaconule are not seen in many palaeo- 
ryctids (Cimolestes and Procerbenis ex- 
cepted) or in undoubted lipotyphlans in 
which there is a trend to reduce the entire 
size of the last molar. This morphology 
could be either a plesiomorphic or an au- 
tapomorphic condition of B. tenuis. In Ba- 
todontoides (see Bloch et al. [1998]), M, is 

smaller than Mo, as is the case in the larger 
geolabidids (Lillegraven et al., 1981). Par- 
antjctoides (Fox, 1979, 1984) has an ex- 
tended Mj talonid, and Fox (1984: 15) 
considered an M3 hypoconulid "strongly 
developed, projecting upward in finger- 
like fashion" to be a primitive character 
state for eutherian mammals. 

Although the overall fit is not as good as 
for UCMP 136091, M^ of UA 4081 oc- 
cludes reasonably well with M^ of UA 3721 

Functional Molar Association in Batodon • Wood and Clemens 111 

Table 1. Measurements (mm) of molars referred to Batodon tenuis* 








Upper molars 



UCMP 136091 




Lacks parastyle 



UCMP 136091 




Parastyle present 

UA 4081 




Parastyle present 

UA 4081 



Excluding parastyle 



UCMP 117649 




Lacks parastyle 



UCMP 102909 




Lacks parastyle 



UCMP 133080 


1.00 (b) 




Lacks parastyle 

Lower molars 



UCMP 117651 





UA 3721 





UA 3721 







UCMP 117652 





UCMP 92590 







UCMP 100638 







UCMP 98188 







UCMP 133764 







UCMP 132174 






AM 58777 


1.25 (a) 

0.80 (a) 

0.65 (a) 



AM 58777 


1.30 (a) 

0.75 (a) 

0.70 (a) 



UCMP 117650 






UCMP 133081 






AM 58777 


1.25 (a) 

0.50 (a) 



USNM 2139 


1.20 (a) 

* All measurements taken by C.B.W. w^th the exception of those taken by W.A.C. (a) and Donald Lofgren 


(Fig. 8). In fact, for occlusion of the pro- 
tocone into the talonid basin, UA 4081 has 
a better size and fit. The iTiain discrepancy 
is in the area of the paracone-cristid obli- 
qua, or (if adjusted there) between pro- 
toconid and parastylar area. The differenc- 
es are not great and may be as expected 
with an attempt to occlude the upper and 
lower dentitions of different individuals 
from the saine species. 

Reference of Specimens to One Species, 
Batodon tenuis 

One of the purported diagnostic char- 
acters of B. tenuis is its diminutive size. 
The fossils referred to this species are the 
smallest eutherian teeth found in Lancian 
local faunas of the North American West- 
em Interior. Geographically these sites ex- 
tend froin central Alberta in the north to 
central eastern Wyoming in the south. 

Making allowances for individual and pos- 
sible latitudinal variation, are the patterns 
and ranges of variation in dimensions what 
might be expected for a primitive euthe- 
rian species? 

Measurements and locality data for the 
available sample of B. tenuis are given in 
Table 1. In order to avoid introduction of 
variation through use of different instru- 
ments, coefficients of variation (CVs) were 
calculated only for specimens iTieasured by 
C.B.W. In general CV values are within 
the ranges found in other early tribos- 
phenic eutherian species known from 
much larger sainples (see Polly [1998b] for 
a recent discussion of CVs in smaller 
mammals). Of course, the range of coef- 
ficients for B. tennis might be the product 
of small sample size and uncertainty in dis- 
tinguishing between isolated specimens 
such as those representing M^ and M^. Co- 

112 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Table 2. Statistical summary for dimensions (mm) of teeth referred to Batodon tenuis.* 








M,, all specimens 

M2, excluding UCMP 132174 



















Width, trigonid 












Width, trigonid 






Width, talonid 












Width, trigonid 






Width, talonid 






* OR, Observed range; SD, Standard deviation; CV, coefficient of variation. 

efficients for widths are generally lower, 
and it is this dimension that usually is least 
affected by position in the tooth row. One 
glaring exception is in width of trigonid of 
M,. However, when measurements of 
UCMP 132174 are removed from the cal- 
culation, the discrepancy is much less ev- 
ident (see Table 2). Some reassessment of 
the identity of UCMP 132174, an isolated 
molar, may be indicated. It is evident by 
inspection that some of the dimensions on 
specimens from the Scollard Formation, 
Alberta, are on the small end of the size 
range. However, this is not consistently 
true for all dimensions. With these excep- 
tions, at present, variation in dental di- 
inensions does not deinand recognition of 

Table 3. Measurements (mm) of anterior and 
posterior margins of M^ of Batodon tenuis ex- 
cluding THE STi'LAR AREA. 







UCMP 136091 



UA 4081 




UCMP 117649 




UCMP 102909 




UCMP 133080 



Statistical summary of dimensions of margins for M-'* 







Anterior margin 
Posterior margin 






* OR, Observed range; SD, standard deviation; CV, 
coefficient of variation. 

more than one species of Batodon in the 
Lancian local faunas sainpled to date. 

Interestingly, ineasurements of the an- 
terior and posterior margins (protocone to 
paracone, protocone to inetacone in crown 
view, excluding the stylar areas) of M- of 
B. tenuis (Table 3) are much more uniform 
than standard measurements for length on 
the tribosphenic crown. Measureinents of 
these nonstandard dimensions for the new 
specimen (UCMP 136091) and for UA 
4081 are, in fact, identical. Mention was 
made above of the coincidence of proto- 
cone and conule outline in superimposed 
drawings, and of the possibility that this 
area could be under greater selective pres- 
sure for uniformity than would other parts 
of the teeth. Such a possibility would be 
purely hypothetical at present, of course, 
and is based on an inadequate sample, but 
extended comparative studies of better- 
known genera would be a worthwhile test 
of this idea. Although Polly (1998a) has re- 
ported that developmental factors may 
correlate better than functional factors in 
tooth crown measurements of viverravid 
carnivores, new data from extant Sorex 
may demonstrate different patterns of var- 
iability (D. Polly, personal communica- 

In smnmary, reference of the smallest 
tribosphenic eutherian teeth from Lancian 
localities in the North American Western 
Interior to a single species, B. tenuis, is 
supported by several lines of evidence. 

Functional Molar Association in Batodon • Wood and Clemens 113 

\Mien merged into a single, albeit small 
sample, observed ranges and coefficients 
of variation of taxonomically significant 
tooth dimensions are what would be ex- 
pected to characterize a single species 
(Polly, 1998b). Sizes and configurations of 
cusps, basins, and shearing crests of upper 
and lower teeth document a common oc- 
clusal pattern. Possibly diagnostic special- 
ization of occluding elements of the pos- 
terior parts of M^ and M., add support to 
the proposed association of upper and low- 
er molars. 



Analyses of patterns of occlusion using 
the functional orientation have strength- 
ened the basis for association of the upper 
and lower dentitions of B. tenuis. Its upper 
dentition is now known from P^ through 
M3 (P^ illustrated by Lillegraven [1969] 
but lost before publication; Storer [1991] 
offers a possible P\ P 2004.30); the lower 
is knowTi from P^ through M3. Although 
more anterior teeth are unknown or rep- 
resented only by fragments, it is clear that 
the canine was large (USNM 2139) and 
that the dental formula was I?, CI, P4, 

Body Mass 

By Cenozoic standards. Cretaceous 
marsupials, eutherians, and their closely 
related sister groups were animals of small 
to very small body mass. Lillegraven et al. 
(1987) argued that small body masses, in 
the context of metabolic and reproductive 
constraints, played a significant role in the 
origin and diversification of tlie earliest 
marsupials and eutherians. Because most 
Cretaceous and early Cenozoic mammals 
were known from only isolated teeth, 
many early estiinates of body mass were, 
at best, educated guesses. Then, based on 
data from living species, Gingerich and 
Smith (1984) pioneered the development 
of a method to use the area of the crown 
of M, to estimate the body mass of extinct 

eutherians. Recently, Bloch et al. (1998) 
presented a version of this technique mod- 
ified specifically to estimate the body mas- 
ses of lipotyphlans and proteutherians. In 
their sui-vey of the modem lipotyphlans, 
Bloch et al. (1998) divided the group into 
seven classes of body mass. The three 
smallest classes and the percentage of 
modem species included in each were 1— 
3 g (3%), 3-7 g (26%), and 7-20 g (35%). 
Bloch et al. (1998) demonstrated that, in 
comparison to the range of variation of 
body masses of extant lipotyphlans, late 
Paleocene (Clarkforkian) and early Eocene 
(Wasatchian) purported lipotvphlans oc- 
cupied the lower end of the range of var- 
iation. The body mass of the extinct spe- 
cies, Batodontoides vanhouteni, from the 
Wasatchian of Wyoining, was estiinated to 
have been approximately 1.3 g, and, thus, 
was the smallest, nonvolant, Cenozoic eu- 
therian yet knov\ai. 

Bloch et al. (1998) did not extend their 
study to include Cretaceous eutherians. In 
Table 4 we present measureinents of lower 
first molars and estimates of body mass of 
the smallest eutherians or probable euthe- 
rians in several local faunas ranging from 
the Aptian-Albian (ca. 110 million years 
before the present) through the Lancian. 
The estimated body mass (8.05-8.53 g) of 
Montanalestes keebleri of Early Creta- 
ceous (Aptian-Albian) age, die most prim- 
itive pui-ported eutherian (Cifelli, 1999) 
from North America, falls sfightly above 
the arbitrary boundary between the sec- 
ond and third modern lipotyphlan body 
mass classes. Similarly, Prokennalestes mi- 
nor, the smallest known Asian eutherian, 
also of Early Cretaceous age, has an esti- 
mated body mass of 6.76 g, falling slightly 
below this boundary. The oldest (Aquilan— 
Judithian) and most primitive North 
American genus of unquestioned eutheri- 
an affinity is Faranijctoides, which is 
knov\Ti from two species, P. nialeficus (Fox, 
1984) and P. sternbergi (Fox, 1979). Esti- 
mates of body mass of these species range 
between 9 and 16 g, that is, within the 

114 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Table 4. Estimates of body mass of species of Batodon, Paranyctoides, Prokennalestes, and Mon- 



usurements nf M 








Estimated mass 

Notes, references 

Batodon tenuis 





AMNH 58777, Clemens, 1973 

Paranyctoides stembergi 





UA 14822, C.B.W. measurements 

Paranyctoides nialeficus 





UA 16168, Fo.x, 1984 





UA 17170, Fox, 1984 





UA 16171, Fox, 1984 





UA 16175, Fox, 1984 





UA 16181, Fox, 1984 

Prokennalestes minor 





Kielan-Jaworowska and Dashzeveg, 

Montanalestes keebleri 





Left Mj, R. Cifelli, personal com- 





Right Ml, R. Cifelli, personal 


third class of body mass recognized by 
Bloch et al. (1998). 

Isolated lower first and second molars of 
B. tenuis cannot be distinguished with cer- 
tainty, and we have not used them for es- 
timates of body mass. Only one Mj of B. 
tenuis has been found in place in a dentary 
(AMNH 58777). The body mass of this in- 
dividual is estimated as having been 5.39 
g. On the basis of the estimates derived 
from the formula provided by Bloch et al. 
(1998), B. tenuis is the smallest known 
Cretaceous eutherian, but it still falls with- 
in the second class of body masses of mod- 
ern lipotyphlans. Batodon tenuis was not 
as minute as the Eocene Batodontoides 
vanhouteni or the modern Suncus etrus- 


A variety of taphonomic and collecting 
biases limit the chances of recovery of the 
remains of very small maiumals, and the 
available sample of Cretaceous eutherians 
is small in absolute number of specimens 
and biogeographically patchy. The possi- 
bility that even smaller Cretaceous euthe- 
rians will be discovered cannot be exclud- 
ed. However, it is of interest that the 
smallest currently known Cretaceous eu- 
therians have estimated body inasses in 
the range of 3-20 g. This is the range of 

body masses that includes 61% of modem 
lipotyphlans (Bloch et al., 1998) and small 
luembers of other, more distantly related 
eutherian orders. 

Systematic Affinities 

Currently, a lively debate is swirling 
around the questions of the time of origin 
of crown-group Mammalia in general and 
eutherian orders in particular (see Gib- 
bons [1998]). Several molecular phyloge- 
neticists (for example see Springer [1997] 
and Kumar and Hedges [1998]) have re- 
ported data from molecular clock esti- 
mates, suggesting that modern eutherian 
clades may have begun to diverge as long 
ago as the Early Cretaceous, despite the 
lengthy gap in the fossil record that such 
dates would imply (but see Nessov et al., 
1998). Studies by Foote et al. (1999) and 
Alroy (1999) are examples of quantitative 
arguments from the fossil record that cast 
doubt on the molecular data. Novacek 
(1999) has addressed this question from a 
phylogenetic slant. These authors argue 
that the major ordinal level clades of the 
crown-group Placentalia did not differen- 
tiate until after the extinction of the non- 
avian dinosaurs inarking the end of the 

Functional Molar Association in Batodon • Wood and Clemens 115 

Debate over the times of origin of the spheres demonstrate that the tribosphenic 
major clades of eutherians is part of an ex- dentition was but one outcome in early ex- 
tensive revision of our understanding of periments in the evolution of more com- 
the pattern of the early evolution of the plex triangularly symmetrical teeth (see 
Mammalia. For example, eutherians and Kielan-Jaworowska et al. [1998] and Bon- 
marsupials were long thought to be char- aparte [1996]). 

acterized by the synapomorphy of a tri- Against this background of rapidly ex- 

bosphenic dentition. Discoveries of many panding knowledge of the complexities of 

mammals with tribosphenic dentitions, early mammalian evolution, interpreta- 

which cannot be confidently referred to tions of the phylogenetic affinities of B. 

the crown-groups Placentalia and Marsu- tenuis play a role in discussions of the be- 

pialia, such as Montanalestes (Cifelli, ginning of the radiation of the eutherian 

1999), show that this type of dentition is a crown group. Is Batodon a member of a 

sviiapomoiphy of a more inclusive group, lineage within the crown-group Placenta- 

A recently discovered Middle Jurassic lia, thus favoring the hypothesis that the 

mammal from Madagascar, Ambondro radiation of modern eutherian orders be- 

(Flynn et al., 1999), suggests a more an- gan in the Late Cretaceous, or is it a mem- 

cient origin of the tribosphenic dentition ber of a lineage not involved in their an- 

than previously expected. cestry? Lillegraven (1969) and several later 

Other recent discoveries reveal greater workers referred Batodon to the Palaeo- 

complexity in the evolutionary radiation of ryctidae, which was classified in the order 

mammals with a reversed triangular sym- Proteutheria, an admittedly paraphyletic 

metry of their cheek teeth, the holotheri- taxon with unclear phylogenetic affinities 

ans, during the Jurassic and Cretaceous. A (see Butler, 1972b). Others opted for an 

newly discovered Early Cretaceous Austra- even less specific reference of the Palaeo- 

lian mammal, Ausktribosphenos, exliibits a ryctidae placing it in the order Insectivora, 

surprising combination of dental and man- incertae sedis (Clemens, 1973) or the in- 

dibular characteristics. Rich et al. (1997, fraclass Eutheria, incertae sedis (Archi- 

1999) maintain that the dentition of Aws/c- bald, 1982; Lofgren, 1995). In contrast, 

tribosphenos is not only fully tribosphenic Novacek (1976) and Lillegraven et al. 

but also exhibits eutherian characteristics (1981) suggested that Batodon is an early 

closer to those of the Erinaceomorpha lipotyphlan, probably allied to or even an- 

than any other group. This dental mor- cestral to the Geolabididae. Fox (1984: 19) 

phology is combined with very plesiom- doubted these suggestions of lipotyphlan 

Orphic mandibular structures such as a affinity. McKenna and Bell (1997) classi- 

Meckelian groove and postdentaiy bone fied Batodon and Batodontoides as mem- 

sulci or facets (Rich et al., 1999). Alter- bers of Geolabididae, within Lipotyphla. 
native interpretations suggest that Ausktri- Bloch et al. (1998) reviewed the system- 

bosphenos is a representative of an endem- atic assignments of Batodon and Batodon- 

ic Australian radiation of holotherians toides. They reported a necessarily limited, 

(Kielan-Jaworowska et al., 1998; Rich et computer-assisted cladistic analysis of den- 

al, 1998). An even older, isolated tribos- tal characters that indicated that Batodon 

pheniclike upper molar has been reported is a basal member of a monophyletic Geo- 

from Late Jurassic deposits in China labididae, with Ciniolestes, Palaeoryctes, 

(Wang et al., 1998). The authors maintain and Asioryctes as successive outgroups. Al- 

that it is probably the missing upper molar though a helpful beginning, this outcome 

of the pseudotribosphenic genus Shuoth- may be variable depending on which and 

erium (Chow and Rich, 1982). These and how many characters are selected, and by 

other holotherians recently discovered in inclusion of other taxa such as other mem- 

both the northern and southern hemi- bers of Palaeoryctidae and/or perhaps a se- 

116 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

ries of additional taxa between Palaeoryc- 
tes and Asioryctes. Differences between 
Botodon and Batodontoides tliat would 
bear reexamination in the more extended 
study would be M3 morphology, M3 size 
relative to Mj ,, talonid cusps in all three 
lower molars, and presence or absence of 
upper molar conules. 

MacPhee and Novacek (1993) summa- 
rized the issues concerning relationships 
between proteutherians, palaeoryctids, 
and lipotyphlans. An important part of the 
problem, perhaps presently insurmount- 
able, is the lack of comparative cranial ma- 
terial for most of the earlier taxa. Known 
material, such as that for Asioryctes (Kie- 
lan-Jaworowska, 1981), is mostly devoid of 
apomorphic characters needed for such an 
analysis. Palaeoryctids (see Thewissen and 
Gingerich [1989]) and leptictids (see No- 
vacek [1986]), skulls of which are known, 
have a few characters that can be parsi- 
moniously inteipreted as synapomoiphies 
of the Lipotyphla. Given the limited 
amount of information on its dentition and 
the lack of cranial data, we conclude that 
currently B. tenuis is best classified as 
Placentalia, incertae sedis. 


This work would not have occurred 
without the encouragement and support of 
Professor A. W. Crompton. We sincerely 
thank Mr. Al Coleman, who took the ster- 
eophotograph reproduced in Figure 1, and 
L. Laszlo Meszoly for the drawings (Fig- 
ures 2-8, based on sketches made by 
C.B.W. through a camera lucida attached 
to a Wild-Heerbrug M3 stereo micro- 
scope) that also illustrate this paper. 
Thanks are also due to Dr. John Storer for 
facilitating reillustration of P 2004.565. Fi- 
nancial support for this study came, in 
part, from the University of Galifornia Mu- 
seum of Paleontology and a series of grants 
(most recently EAR 9505841) from the 
National Science Foundation. Providence 
College provided additional support by 
means of sabbatical leave and grants from 
its Committee to Aid Faculty Research. 

Finally, thanks go to P. David Polly, Ri- 
chard Cifelli, an anonymous reviewer, and 
several other colleagues who provided 
data, helpful discussions of aspects of the 
study, or reviews of drafts of the manu- 


Alroy, J. 1999. The fossil record of North American 
mammals: evidence for a Paleocene evolutionary 
radiation. Systematic Biology, 48: 107-118. 

Archibald, J. D. 1982. A study of Mammalia and 
geology across the Cretaceous— Tertiary boundary 
in Garfield County, Montana. University of Cal- 
ifornia Publications in Geological Sciences, 122: 

Bloch, J. 1., K. D. Rose, and P. D. Gingerich. 
1998. New species of Batodontoides ( Lipotyi^hla, 
Geolabididae) from the early Eocene of Wyo- 
ming: smallest known mammal? Journal of Mam- 
malogy, 79: 804-827. 

Bonaparte, J. F. 1996. Cretaceous tetrapods of Ar- 
gentina. Miinchner Geowissenschaftliche Abhan- 
dlungen, 30: 73-130. 

Butler, P. M. 1961. Relationships between upper 
and lower molar patterns, pp. 117-126. In G. 
Vandebroek (ed.), International Colloquium on 
the Evolution of Lower and Non-specialized 
Mammals. Part I. Letteren en Schone Kunsten 
van Belgie. Brussels, Belgium: Koninklijke 
Vlaamse Academic voor Wetenschappen. 320 pp. 

. 1972a. Some functional aspects of molar evo- 
lution. Evolution, 26: 474-^83. 

1972b. The problem of insectivore classifi- 

cation, pp. 253-265. In K. A. Joysey and T. S. 
Kemp (eds.). Studies in Vertebrate Evolution. 
New York: Winchester Press. 284 pp. 

Carrano, M. T, R. W Blob, J. J. Flynn, R. R. 
Rogers, and C. A. FORSTER. 1997. The mam- 
malian fauna of the Judith River Formation type 
area (Campanian, Central Montana) revisited. 
Journal of Vertebrate Paleontology, 17(3): 36A. 

Chow, M., and T H. V. Rich. 1982. Shuotherium 
dongi, n. gen. and sp., a therian with pseudo- 
tribosphenic molars from the Jurassic of Sichuan, 
China. Australian Mammology, 5: 127-142. 

Cifelli, R. L. 1999. Tribosphenic mammal from the 
North American Early Cretaceous. Nature, 401: 

Clemens, W A., Jr. 1973. Fossil mammals of the 
type Lance Formation, Wyoming — part III. Eu- 
theria and summary. University of California 
Publications in Geological Sciences, 94: 1—102. 

Crompton, A. W 1971. The origin of the tribos- 
phenic molar, pp. 65—87. //! D. M. Kermack and 
K. A. Kermack (eds.). Early Mammals. Zoologi- 
cal Journal of tlie Linnaean Society, 50(Suppl. 1): 
xiv + 1-203. 

Functional Molar Association in Batodon • Wood and Clemens 117 

Crompton, a. W., and K. Hiiemae. 1969. How 
mammalian teeth work. Discovery, 5: 23—34. 

Crompton, A. W., and Z. Kiel,\n-J.\\voro\\'ska. 
jj 1978. Molar structure and occlusion in Creta- 

ceous therian mammals, pp. 249—287. In P. M. 
Butler and K. A. Joysey (eds.). Development, 
Function, and Evolution of Teeth. New York: Ac- 
ademic Press, v -I- 523 pp. 

Crompton, A. W., and A. Sita-Lumsden. 1970. 
Functional significance of the therian molar pat- 
tern. Nature, 227: 197-199. 

Flynn, J. J., J. M. Parrish, B. Rakatosamimanana, 
W. F Simpson, and A. R. Wyss. 1999. A Middle 
Jurassic mammal from Madagascar. Nature, 401: 

FooTE, M., J. P. Hunter, C. M. Janis, and J. J. 
Sepkoski, Jr. 1999. Evolutionary and preserva- 
tional constraints on origins of biologic groups: 
divergence times of eutherian mammals. Sci- 
ence, 283: 1310-1314. 

Fox, R. C. 1975. Molar structure and function in the 
earlv Cretaceous mammal Pappotherium: evolu- 
tionar\' implications for Mesozoic Theria. Cana- 
dian Journal of Earth Sciences, 12: 412^42. 

. 1979. Mammals from the Upper Cretaceous 

Oldman Formation, Alberta. III. Eutheria. Ca- 
nadian Journal of Earth Sciences, 16: 114—125. 

. 1984. Paranijctoides maleficus (new species), 

an early eutherian mammal from the Cretaceous 
of Alberta. Special Publication of the Carnegie 
Museum of Natural History, 9: 9-20. 

Gibbons, A. 1998. Genes put mammals in age of di- 
nosaurs. Science, 280: 675—676. 

Gingerich, p. D., and B. H. Smith. 1984. AUome- 
tric scaling in the dentition of primates and in- 
sectivores, pp. 257-272. In W. L. Jungers (ed.). 
Size and Scaling in Primate Biology. New York: 
Plenum Publishing Corporation, xiv -I- 491 pp. 

Hicks, J. R, K. R. Johnson, L. Tauxe, D. Clark, 
and J. D. Obradovich. 1999. Geochronology of 
the Hell Creek Formation of southwestern 
North Dakota: a multidisciplinary approach us- 
ing biostratigraphy, isotopic dating, geochemistry, 
and magnetostratigraphy. Geological Society of 
America, Abstracts with Programs, 31(7): A-71. 

Kay, R. F, and K. M. Hiiemae. 1974. Jaw movement 
and tooth use in recent and fossil primates. Jour- 
nal of Physical Anthropology, 40: 227-256. 

Kielan-Jaw'OROW'SKA, Z. 1981. Evolution of therian 
mammals in the Late Cretaceous of Asia. Part 
IV. Skull structure of Kennelestes and Asioryctes. 
Palaeontologica Polonica, 42: 25—78. 

Kielan-Jaworowska, Z., R. L. Cifelli,, and Z. 
LUO. 1998. Alleged Cretaceous placental from 
down under. Lethaia, 31: 267-268. 

Kielax-Jaworowska, Z., and D. Dashzeveg. 1989. 
Eutherian mammals from the Early Cretaceous 
of Mongolia. Zoologica Scripta, 18: 347-355. 

Kumar, S., and S. B. Hedges. 1998. A molecular 
timescale for vertebrate evolution. Nature, 392: 

Lillegra\t:n, J. A. 1969. Latest Cretaceous mam- 
mals of upper part of Edmonton Formation of 
Alberta, Canada, and review of marsupial-pla- 
cental dichotomy in mammalian evolution. Uni- 
versity of Kansas Paleontological Contributions, 
Article 50(Vertebrata 12): 1-122. 

Lillegra\t;n, J. A., and M. C. McKenna. 1986. 
Fossil mammals from the "Mesaverde" Forma- 
tion (Late Cretaceous, Judithian) of the Bighorn 
and Wind River Basins, Wyoming, witli defini- 
tions of late Cretaceous North American Land- 
Mammal Ages. American Museum Novitates, 
2840: 1-68. 

KRISHTALK.A. 1981. Evolutionary' relationships of 
Middle Eocene and younger species of Cente- 
todon (Mammalia, Insectivora, Geolabididae) 
wdth a description of the dentition of Ankylodon. 
University of Wyoming Publications, 45: 1-97. 

AND J. L. PatTON. 1987. The origin of the eu- 
therian mammals. Biological Journal of the Lin- 
nean Society, 32: 281-336. 

LOFGREN, D. 1995. The Bug Creek problem and the 
Cretaceous-Tertiary transition at McGuire 
Creek, Montana. University of California Publi- 
cations, Geological Sciences, 140: 1-185. 

Lofgren, D., C. Hotton, and a. Runkel. 1990. 
Reworking of Cretaceous dinosaurs into Paleo- 
cene channel deposits, upper Hell Creek For- 
mation, Montana. Geology, 18: 874—877. 

Lull, R. S. 1915. The mammals and homed dino- 
saurs of the Lance Formation, Niobrara County, 
Wyoming. American Journal of Science, Series 4, 
40: 319-348. 

MacPhee, R. D. E., and M. J. Novagek. 1993. Def- 
inition and relationships of Lipotyphla, pp. 13- 
30. In F S. Szalav, M. C. McKenna, and M. T. 
Novacek (eds.). Mammal Phylogeny: Placentals. 
New York: Springe r-Verlag. v + 321 pp. 

Marsh, O. C. 1892. Discovery of Cretaceous Mam- 
malia, part HI. American Journal of Science, Se- 
ries 3, 48: 249-262. 

McKenna, M. C, and S. K. Bell. 1997. Classifi- 
cation of Mammals Above the Species Level. 
New York: Columbia University Press, xii + 631 

MONTELLANO, M. 1992. Mammalian fauna of the Ju- 
dith River Formation (Late Cretaceous, Judithi- 
an), northcentral Montana. Universit\' of Califor- 
nia Publications, Geological Sciences, 136: I- 

Nessov, L. a., J. D. Archibald, and Z. Kielan- 
Jaworowska. 1998. Ungulate-like mammals 
from the Late Cretaceous of Uzbekistan and a 
phylogenetic analysis of Ungulatomorpha. Bul- 
letin of the Carnegie Museum of Natural His- 
tory, 34: 40-88. 

Novacek, M. J. 1976. Insectivora and Proteutheria 
of the later Eocene (Uintan) of San Diego Coun- 
ty, California. Los Angeles County Natural His- 

118 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

tory Museum, Contributions to Science, 283: 1- 

. 1986. The skull of leptictid insectivorans and 

the higher-level classification of eutherian mam- 
mals. Bulletin of the American Museum of Nat- 
ural Histoiy, 183: 1-112. 

. 1999. 100 million years of land vertebrate 

evolution: the Cretaceous-Early Tertiary transi- 
tion. Annals of the Missouri Botanical Gardens, 
86: 230-258. 

Polly, P. D. 1998a. Variabihty, selection, and con- 
straints: development and evolution in viverravid 
(Camivora, Mammalia) molar morphology. Pa- 
leobiology, 24: 409-429. 

. 1998b. Variability in mammalian dentitions; 

size-related bias in the coefficient of variation. 
Biological Journal of the Linnean Society, 64: 

Rich, T H., T. F. Flannery, and P. Vickers-Rich. 
1998. Alleged Cretaceous placental from down 
under: reply. Lethaia, 38: 346-348. 

Rich, T H., P. Vickers-Rich, A. Constantine, T 
F. Flannery, L. Kool, and N. van Klaveren. 
1997. A tribosphenic mammal from the Meso- 
zoic of Australia. Science, 278: 1438-1442. 

. 1999. Early Cretaceous mammals from Flat 

Rocks, Victoria, Australia. Records of the Queen 
Victoria Museum, 106: 1-35. 

Sloan, R. E., and L. Van Valen. 1965. Cretaceous 
mammals from Montana. Science, 232: 220-227. 

Springer, M. S. 1997. Molecular clocks and the tim- 
ing of the placental and marsupial radiations in 
relation to the Cretaceous— Tertiary boundary'. 
Journal of Mammalian Evolution, 4: 285-302. 

Storer, J. E. 1991. The mammals of the Gryde Lo- 
cal Fauna, Frenchman Formation (Maastrich-| 
tian: Lancian), Saskatchewan. Journal of Verte- 
brate Paleontology, 11: 350-369. 

Skull and endocranial cast of Eoryctes melaniis, 
a new palaeoryctid (Mammalia: Insectivora) from 
the early Eocene of western North America. 
Journal of Vertebrate Paleontology, 9: 459^70. 

Wang, Y., W. a. Clemens, Y. Hu, and C. Li. 1998. 
A probable pseudo-tribosphenic upper molar 
from the late Jurassic of China and the early ra- 
diation of the Holotheria. Journal of Vertebrate 
Paleontology, 18: 777-787. 

Wood, C. B., G. C. Conroy, and S. G. Lucas. 1979. 
New discoveries of fossil primates from the type 
Torrejonian (Middle Paleocene) of New Mexico. 
Folia Primatologica, 32: 1-7. 

Wood, C. B., and W A. Clemens. 1990. An undes- 
cribed upper M2-3 of Batodon tenuis: a func- 
tional assessment. Journal of Vertebrate Paleon- 
tology, 10(3): 50A. 

Wood, C. B., E. R. Dumont, and A. W Crompton. 
1999. New studies of enamel microstructure in 
Mesozoic mammals: a review of enamel prisms 
as a mammalian synapomorphy. Journal of Mam- 
malian Evolution, 6: 177-214. 



Abstract. The developmental and reproductive 
strategies of marsupial mammals differ from those of 
placental mammals. In marsupials, most maternal nu- 
tritional support of the developing young is through 
lactation. The young are bom at an extremely altricial 
state and undergo luost development while attached 
to the teat. In order to achieve functional indepen- 
dence at an altricial state, the marsupial embryo 
accelerates the development of certain bones of the 
facial region, most cranial musculature, and a few ad- 
ditional stmctures. At the same time, relative to pla- 
centals, marsupials delay significantly the develop- 
ment of central nervous system structures, in partic- 
'ilar the forebrain. In this paper I present preliminary 
results concerning the origins of these heterochronies 
in ontogeny and phylogeny. In ontogeny, heteroch- 
ronies are initiated in marsupials by shifting the tim- 
ing of neural crest differentiation and migration rel- 
ative to eutherians and other amuiotes. Further, early 
fore- and midbrain differentiation is delayed relative 
:o the hindbrain. Preliminary data from nonmam- 
nalian amniotes and monotremes is discussed to as- 
sess phylogenetic origins. Comparisons with non- 
iiammalian amniotes suggest that the pattern ob- 
iei"ved in marsupials is derived, and that observed in 
olacentals is primitive. Preliminary data on mono- 
remes suggest that the monotreme condition is 
iomewhat intermediate beKveen the two therian taxa. 
Finally, the implications of these results for contro- 
. ersies regarding the evolution of mammalian repro- 
luction are discussed. 


Mammalian reproduction is character- 
zed by distinctive adaptations for maternal 
lutrient provision to the young. In therian 
mammals this provisioning occurs via two 
routes. First, during a period of intrauter- 
ne embryonic development exchange of 
lutrient material between the mother and 
s oung occurs through a placenta. Although 
:his adaptation is most often identified 

' Department of Biology, Duke University, Dur- 
ham, North Carolina 27710. 

with mainmals, intrauterine development 
and the existence of a placenta is not un- 
common among vertebrates (e.g., squa- 
mate reptiles; Shine, 1985). Second, and 
virtually unique to mammals, nutrition is 
provided postnatally to the young through 
specialized mammary glands. The origin 
and evolution of these distinctive traits has 
been a topic of much discussion. This dis- 
cussion has been enriched in part because 
the two clades of living therian inam- 
mals — inarsupials and placentals — possess 
quite different strategies of reproduction, 
with differential emphasis on these two 
processes of maternal investment. (The 
terms marsupial-placental and metatheri- 
an— eutherian are each to some degree un- 
satisfactory to distinguish the two clades; 
however, they are used informally and in- 
terchangeably throughout the text. In par- 
ticular, the characters discussed in this pa- 
per are only accessible in extant taxa, and 
inferences cannot be extended to mein- 
bers of any clade known only in the fossil 

Marsupials are considered lactational 
specialists, where a relatively short intra- 
uterine period of maternal— fetal inter- 
change is followed by an extended period 
of lactation (e.g., Renfree, 1983, 1993, 
1995). In contrast, eutherians are charac- 
terized by relatively longer periods of in- 
trauterine development, with extensive fe- 
tal— inaternal interchange, and variable re- 
liance on lactation. Because the period of 
organogenesis is so short in marsupials, the 
neonates show minimal development of 
most systems and are always highly altri- 
cial. Eutherian neonates exliibit a range of 
developiTient from altricial to highly pre- 

Bull. Mus. Comp. ZooL, 156(1): 119-135, October, 2001 119 

120 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

cocial; however, even the most altricial eu- 
therian is far more developed than the 
most precocial marsupial. 

Because the reproductive mode and rel- 
ative state of the neonate in marsupials 
and placentals is so different, a rich liter- 
ature exists that contrasts these strategies 
and speculates on the evolutionary signif- 
icance of the observed patterns. Three ba- 
sic inteipretations have been made. In the 
first intei-pretation the marsupial condition 
is seen as a primitive condition, in which 
significant constraints prevent long periods 
of intrauterine development (e.g., Lille- 
graven, 1975; Lillegraven et al., 1987). Hy- 
pothesized constraints have included an 
inability to develop an efficient maternal- 
fetal exchange system, physical constraints 
on embryo size due to the configuration of 
the reproductive tract, or an inability to 
develop immunologic protection of the fe- 
tus. At times explicitly, but always implic- 
itly, the marsupial condition is seen as a 
primitive and less efficient mode of repro- 
duction, as evidenced by the competitive 
difficulties marsupials face in the presence 
of eutherians. 

A second view is that the marsupial 
mode has evolved in response to a number 
of specific selective pressures, with partic- 
ular adaptive advantages (e.g., Hayssen et 
al., 1985; Kirsch, 1977a,b; Parker, 1977). It 
is argued that because little initial mater- 
nal investinent occurs during intrauterine 
development, the female can reduce or 
abandon her litter in response to harsh or 
uncertain conditions with ininimal loss of 
lifetime reproductive effort. This view 
clearly assumes that the marsupial condi- 
tion is derived, and that if a competitive 
inferiority exists, it is due to other factors, 
such as the consequences of marsupials 
having evolved in greater isolation than 
have eutherians. 

A third view is that marsupial and eu- 
therian modes of reproduction are simply 
part of a single continuum (Tyndale-Biscoe 
and Renfree, 1987). Marsupials and pla- 
centals each provide nutrition to the de- 
veloping young through both placental and 

lactational exchange. The hypothesis has 
been made that the primitive therian con- 
dition was characterized by an altricial ne- 
onate as a consequence of small body size 
in the earliest therian. During evolution, 
marsupials and placentals simply empha- 
sized different ends of the maternal in- 
vestment continuum. Eutherians took the 
strategy of greater and greater maternal in- 
vestment through placentation, and be- 
cause of longer periods of intrauterine de- 
velopment, neonates are less altricial. 
Highly precocial young are thought to be 
correlated with the evolution of large body 
size. In contrast, marsupials have relied on 
the strategy of more investment via lacta- 
tion and have reduced the period of intra- 
uterine development, resulting in a more 
altricial neonate. The differences between 
the two are simply the results of two dif- 
ferent, but not necessarily inferior or su- 
perior, strategies of provisioning the 
young. Both marsupials and placentals are 
assumed to have diverged from a primitive 
condition that was somewhat intermediate. 
This debate has proved difficult to re- 
solve, in part because mammalian repro- 
duction is so distinctive, and in part be- 
cause reproductive modes are difficult to 
reconstruct in the fossil organisms. In this 
paper I will focus on the evolution of 
mammalian development. I argue that eu- 
therian and marsupial reproductive strat- 
egies are reflected in distinct developmen- 
tal patterns; therefore, information on the 
evolution of development can provide new 
data for phylogenetic analysis of the evo- 
lution of reproduction. I first review pre- 
viously published work on organogenesis 
in marsupials and placental mammals. I 
then introduce new comparative work that 
extends this previously published work, to 
address questions about the origins of 
these differences in development and in 


It has long been recognized that, rela- 
tive to eutherians, marsupials accelerate 

Evolution of Mammalian Development • Smith 


the development of certain structures such 
as the tongue, the bones around the oral 
apparatus and the bones and muscles of 
the forehmb (e.g., Hill and Hill, 1955; Lee 
and Cockburn, 1985; Khma, 1987; Maier, 
1987, 1993; Tyndale-Biscoe and Renfree, 
1987, and references therein; Hughes and 
Hall, 1988; Nelson, 1988; Filan, 1991; 
Clark and Smith, 1993; Gemmell and Sel- 
wood, 1994). This advancement is inter- 
preted as an adaptive response to the func- 
tional requirements placed on the neonate 
by the marsupial life histoiy The extreme- 
ly altricial neonate must independently 
travel to, identify, and enter the pouch or 
teat region, and recognize and attach to 
the teat. The neonate must have sufficient 
functional maturity to suckle and process 
food while it completes its development. 
However, by and large, no broad-based, 
detailed comparisons of development have 
been made to identify the specific heter- 
ochronies that characterize marsupials. In 
a series of studies I presented such an 
analysis for major craniofacial stmctures 
(Smith, 1996, 1997; Nunn and Smith, 

In these studies, relatively complete de- 
velopmental series of six placental and four 
inarsupial mammals were examined. The 
placentals include the laboratoiy mouse, 
Mus musculus (Rodentia); the doinestic 
cat, Felis domestica (Camivora); the do- 
mestic pig, Siis scrofa (Artiodactyla); the 
pangolin, Manis javanica (Pholidota); the 
tarsier, Tarshis spectrum (Primates); and 
the tree shrew, Tupaia javanica (Scanden- 
tia). The four marsupials are the gray 
short-tailed opossum, Monodelphis domes- 
tica (Didelphidae); the tammar wallaby, 
Macropiis eiigenii (Macropodidae); the 
eastern quoll, a species of marsupial "cat," 
Dasyunis viverrinus (Dasyuridae); and a 
bandicoot, Perameles nasuta (Perameli- 
dae). Care was taken to chose taxa that 
represent the phylogenetic breadth of 
their clades. Figure 1 illustrates the phy- 
logenetic relations ainong these taxa. Most 
specimens were part of the Hubrecht 
Comparative Embryology collection in 

2 12 

2 12 2 


2 12 

2 2 11 

2 2 11 

2 2 11 

2 2 20 

2 2 12 

2 2 12 











Figure 1. Phylogeny of the taxa used in the comparative 
studies. Phylogeny for marsupials taken from Sanchez-Villagra 
(1 999) and Springer et al. (1 998); that for placentals taken from 
Novacek (1990). The numbers on each line illustrate results 
from the event-pair analysis (Smith, 1997) and represent the 
character states for the following pairs of events (in order): 
parietal-telencephalon, dentary-telencephalon, alisphenoid- 
exoccipital, jugal-craniofacial muscles. Character state 
means the first element in the pair occurs before the second; 
character state 1 means the first and second elements appear 
at same time (in the sample available); character state 2 
means the first element in the pair occurs after the second. 
These event pairs represent a variety of phylogenetic patterns. 
The pair parietal-telencephalon has a uniform pattern in ther- 
ians — the parietal always ossifies after the telencephalon 
evaginates. The pair dentary-telencephalon separates mar- 
supials and placentals. In marsupials the dentary ossifies at 
the same time or before the telencephalon appears; in euthe- 
rians the dentary always ossifies after appearance of the tel- 
encephalon. The alisphenoid also separates the two clades 
with the exoccipital preceding the alisphenoid in all marsupials, 
and the alisphenoid and exoccipital ossifying at the same time 
in all eutherians (with the exception of Mus). Finally, the pair 
jugal-craniofacial muscles represents a case in which no phy- 
logenetic pattern of the developmental relation of these two 
elements is apparent. See Smith (1997) for further discussion 
of these events. 

The Netherlands, although others were 
obtained from a number of sources. For 
each taxon at least 10 stages were available 
for the period between the first and last 
developmental events examined. All spec- 
imens examined were serially sectioned 
embiyos, originally embedded in paraffin 
and stained with common histologic stains. 
Details on the taxa, sources, and stages ex- 
amined can be found in Smith (1997). 
Each specimen was exainined to deter- 

122 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

mine the state of 28 elements of the cra- 
nial skeletal, muscular, and central nervous 
systems (CNSs), which serve as landmarks 
for the most critical stages in the differ- 
entiation of craniofacial structures. The 
kinds of events examined are briefly sum- 
marized here; further details are provided 
in Smith (1997). The initial ossification 
center of 12 bones of the dermal and en- 
dochondral skeletons was documented. 
Other conditions of the cranial skeleton in- 
cluded, for example, the first contact be- 
tween the membrane bones over the cra- 
nial roof, the differentiation of cartilage in 
the cranial base, closure of the secondaiy 
palate, and the development of a joint cap- 
sule at the dentary-squamosal joint (see 
Clark and Smith [1993, and references 
therein] for more detail on development 
of the cranial skeleton). 

Three stages, ranging from the first fu- 
sion in myoblasts to the age at which all 
craniofacial muscles were distinguishable, 
were used as measures of muscle devel- 
opment (for more detail on the assessment 
of muscle development see Smith [1994]). 
Finally, six events were examined that in- 
dicate maturation of the CNS and cranial 
sense organs. These included, for example, 
the evagination of the telencephalic vesi- 
cles, the filling of the lens vesicle by pri- 
mary lens cells, and the appearance of at 
least four distinct layers in the cortex. 

Two different approaches were taken to 
analyze the comparative data (see Smith 
[1997] and Nunn and Smith [1998] for de- 
tails on methods). The first method con- 
structs a matrix for each taxon in which the 
timing of each event is compared to the 
timing of every other event. This creates a 
series of pair- wise comparisons, where the 
timing of event A is compared to event B, 
C, D, and so on (forming pairs A-B, A-C, 
A-D, as well as B-C, B-D, and so on). The 
data set studied here included 28 events, 
which produced 378 event pairs. Each pair 
was assigned one of three character states, 
reflecting the relative timing of the two 
events. The three states were character 
state 0, when event A (the first event in 

the pair) occurred before event B (the sec- 
ond event in the pair); character state 1, 
when A occurred in the same stage as B; 
and character state 2, when A occurred af- 
ter B. The character state for each pair of 
events was then mapped on a phylogeny, 
to determine whether any group of taxa 
(e.g., marsupials or placentals, or subsets 
within a major clade) had a unique char- 
acter state distribution (Fig. 1; see Smith 

The second approach is quantitative. 
Each event in the sequence was given a 
rank order number between 1 and 28 (be- 
cause there were 28 events), with events 
occurring at the same time ranked as a tie. 
An analysis of variance ( ANOVA) was then 
performed to determine which events had 
a significantly different rank between mar- 
supials and placentals. In addition, meth- 
ods were also developed to correct for 
phylogenetic nonindependence in the as- 
sessment of significance (see Nunn and 
Smith [1998]). 

Although the two analytical methods are 
quite different, they provide congruent re- 
sults and allow a determination of which 
shifts in relative timing — heterochrony — 
characterize craniofacial organogenesis in 
marsupial and placental mammals. In the 
ANOVA the following 11 events had sig- 
nificantly different ranks in the two clades: 
the evagination of the telencephalon; con- 
tact between the olfactory bulb and the ol- 
factory epithelium; layering in the cortex; 
the differentiation of the thalamus and hy- 
pothalamus; filling of the lens vesicle by 
primary lens cells; the initial ossification of 
the dentary, maxillary, premaxillaiy, and 
exoccipital bones; the closure of the sec- 
ondary palate; and the meeting of the der- 
mal bones over the cranial roof. The initial 
ossification of the dentary, maxillary, pre- 
maxillary, and exoccipital bones and the 
closure of the secondary palate occurred 
early in marsupials relative to placentals 
(i.e., they had a significantly lower rank); 
the other events were late in marsupials 
when compared to placentals. This same 
set of characters exliibited shifts in the 

Evolution of Mammalian Development • Stnith 


Eutherian cranial development 

m. Mi 



Metathehan cranial development 






Figure 2. A summary of the relative timing of development of craniofacial features in metattierians and eutherians. The upper 
set of boxes (light stippling) for each clade represents events in central nervous system (CNS) development; the lower set of 
boxes (dark stippling) represents events in the development of skeletal-muscular systems. The arrow at the bottom represents 
time. The letter B represents the approximate time of birth in each group. For comparison, the two taxa were scaled and aligned 
relative to the timing of the events of the skeletal-muscular system; the CNS in each clade was plotted relative to the scaled 
skeletal-muscular system events. Key to numbers: 1, evagination of telencephalon; 2, pigment in retina; 3, connection between 
the olfactory epithelium and olfactory bulb, layering present in the cortex, thalamus and hypothalamus present, primary lens cells 
fill lens vesicle; 4, tongue muscle cells begin fusion and cartilage present in basicranium; 5, ossification of premaxillary, maxillary, 
and dentary bones; 6, closure of secondary palate, all major components of craniofacial muscle present, muscle maturation, 
differentiation of tooth buds, appearance of ear ossicles, first ossification in most membrane bones; 7, ossification of most 
endochondral bones, meeting of membrane bones over cranial roof, and differentiation of mandibular joint cartilage and capsule. 

event-pair character analysis. In the event- 
pair analysis, 58 of the 378 event-pairs had 
character states that distinguished marsu- 
pial and placental mammals. Fifty-seven of 
these 58 event-pairs contained either one 
or two events that were found to differ in 
the ANOVA. 

These specific results reveal that cranio- 
facial development in marsupials and pla- 
cental is distinguished by major shifts in 
the relative timing of the differentiation of 
the somatic structures of the head relative 
to the differentiation of the CNS (Fig. 2). 
These heterochronies have two major 
components. First, in eutherians the onset 
of morphogenesis of the CNS begins long 
before the appearance of any cranial skel- 
etal or inuscular tissues. In inarsupials cra- 
nial skeletal and muscular tissues begin 

develop inent early relative to CNS differ- 
entiation. Second, in eutherians the events 
of CNS development examined are com- 
pleted before most somatic stiiictures be- 
gin differentiation, whereas in marsupials 
morphogenesis of these same elements ex- 
tends long into the period of cranial skel- 
etal development. More broadly, relative 
to eutherians, marsupial development can 
be characterized by two steep heteroch- 
ronies, or shifts in developmental timing: 
cranial musculoskeletal tissues are highly 
advanced in onset and rate of develop inent 
relative to the tissues of the CNS, and in 
the body as a whole, the rostral portion is 
highly advanced relative to the caudal por- 
tion (Smith, 1996, 1997; Nunn and Smith, 

The early development of musculoskel- 

124 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

etal tissues is almost certainly a result of 
the necessity for the marsupial neonate to 
possess certain functional abilities relative- 
ly early in its normal developmental peri- 
od. The fact of this early adaptation has 
long been recognized. However, the ad- 
aptations of the marsupial neonate do not 
simply involve the advancement of the 
forelimbs and a few structures around the 
oral apparatus and birth at an altricial 
state. Instead, the developmental trajec- 
tory of all cranial tissues seems to be shift- 
ed. This whole-scale shift is best inter- 
preted as the interaction of the adaptive 
innovations with constraints imposed by 
two developmental processes (Smith, 

The first process involves CNS devel- 
opment and the constraints arising from 
the sensitivity of nervous tissue to ener- 
getic fluctuations during differentiation. 
The absence of nutrients during organo- 
genesis of the brain can lead to long-last- 
ing neural deficiencies (e.g., Dobbing, 
1972; Winick et al, 1972; Cheek, 1975; 
Dodge et al, 1975; Winick, 1976, 1979; 
Shoemaker and Bloom, 1977; Hetzel and 
Smitli, 1981; Dhopeshwarkar, 1983; Hersch- 
kowitz, 1989). Because of these critical re- 
quirements, Sacher and Staffeldt (1974) 
proposed that neurogenesis is the rate-lim- 
iting step in mammalian development. 
The second set of processes involves the 
development of mesenchymal-derived tis- 
sues and the importance of initial cell con- 
densation size during skeletogenesis (Grii- 
neberg, 1963; Atchley and Hall, 1991; 
Hall, 1991; Hall and Miyake, 1992, 1995; 
Dunlop and Hall, 1995; Miyake et al, 
1996, 1997). This work shows that in gen- 
eral, condensation must be adequate be- 
fore cartilage will differentiate and bone 
formation will begin. Therefore, it is likely 
that the processes of skeletogenesis re- 
quire the developing embryo to allocate a 
sufficient number of cells, and presumably 
energy, to the skeletal system at the veiy 
earliest stage of differentiation. 

In eutherian mammals the onset of neu- 
rogenesis and the initial period of growth 

of the CNS begin early, when little com- 
petition exists from other tissues, and 
growth and differentiation continue 
throughout the extended embryonic and 
fetal periods when nutrition is relatively 
constant. However, metatherians face the 
competing demands of the adaptations 
that allow function of the systems most 
critical to independent suivival of the al- 
tricial neonate, the extremely short period 
from primitive streak to birth (which av- 
erages 6 days for marsupials as a group and 
is less than 3 days in dasyurids; Tyndale- 
Biscoe and Renfree, 1987), the necessity 
for sufficient allocation to these systems 
for moiphogenesis, and the rate-limiting 
nature of neurogenesis. Marsupials appar- 
ently avoid the constraints arising from 
these competing demands by shifting the 
bulk of neural differentiation to the ex- 
tended postnatal period and devoting em- 
bryonic resources to tissues that must be 
functional at birth (see Smith [1997] for 
more discussion of this hypothesis). 


The discussion above focused on events 
that occur during organogenesis, after the 
basic systems have appeared. These data 
do not address when these heterochronies 
originate in development. At least two 
competing hypotheses have been devel- 
oped. First, these shifts possibly represent 
patterns of acceleration and deceleration 
of moi"phogenesis of cranial structures 
once the major elements of the embiyo 
have differentiated. Therefore, these 
changes would represent relatively minor 
terminal shifts in development, and sug- 
gest that there is significant conservation 
of the basic body plan. Alternatively, it may 
be that these shifts occur early in devel- 
opment and represent major changes in 
patterning of the tissues of the head, and 
indeed the embryo as a whole. Of partic- 
ular interest is the fact that the bones and 
connective tissues of the facial region, 
which are greatly accelerated relative to 
the CNS in marsupials, are in fact derived 

Evolution of Mammalian Development • Smith 


Figure 3. Photographs of a 10.5-day-gestation embryo Monodelphis domestica (approximately six somites); (A) is a dorsal view 
and (B) is an anterior-dorsal view of same specimen. Neural crest migration occurs early relative to neural tube differentiation 
in marsupials. Although no closure of the neural tube has occurred, streams of neural crest have migrated into the first arch 
region, are migrating into the second arch region, and appear to be about to migrate into posterior regions. Further, at this time 
the hindbrain is fairly well differentiated, with recognizable rhombomeres, yet little or no development of midbrain or forebrain 
regions has occurred. This is quite different from the pattern seen in eutherians. Key: C, cervical region; O, otic sulcus (region 
of rhombomeres 5 and 6); PC, preotic sulcus (between rhombomeres 2 and 3); FB, forebrain region; 1 , the first stream of neural 
crest, which appears to populate the first arch and frontonasal region; 2, the second stream of neural crest, which appears to 
provide cells to the second arch; 3, the third stream, which appears to go to the third through sixth branchial arches. 

from a neural tissue — the neural crest (re- 
viewed in Le Douarin [1982], Noden 
[1983, 1987, 1991], Hall [1987], and Hall 
and Horstadius [1988]). The relative tim- 
ing of neural crest differentiation serves as 
the earliest "decision point" in embryonic 
allocation to neural or to mesenchymal tis- 
sues. Do these differences originate with 
shifts in the relative timing or pattern of 
neural crest migration? A positive answer 
would support the hypothesis that this is a 
fundamental change in development and 
tliat early development is potentially plas- 

Neural crest migration has been studied 
extensively in a number of nonmammalian 
vertebrates, particularly in the quail-chick 
system (e.g., Le Douarin, 1982; Noden, 
1983, 1987, 1991; Hall and Horstadius, 
1988). The studies of mammals thus far 
have indicated essential similarity \\dth 
other vertebrates although a few important 
differences exist (see, for example, Nichols 
[1981, 1986, 1987], Serbedzija et al. 
[1992], Morriss-Kay et al. [1993], Tan and 
Morriss-Kay [1985, 1986], Trainor and 
Tam [1995], and Peterson et al. [1996]). 
One difference is that in the mammals 

studied neural crest migration begins rel- 
atively early when the anterior part of the 
neural tube is still open, whereas in other 
vertebrates migration is typically after neu- 
ral tube closure (e.g., Le Douarin, 1982; 
Hall and Horstadius, 1988; Hanken et al., 
1997). In both mice and rats neural crest 
appears to begin migration at the five- to 
six-somite stage (8 or 9 days; Nichols, 
1981; Morriss-Kay et al, 1993). Other 
than the unpublished studies of Hill and 
Watson (1958), no studies of neural crest 
migration have been conducted in any 

Preliminary results from a study of neu- 
ral crest inigration in inarsupials suggest 
that the shift in the differentiation of the 
CNS and somatic tissues is initiated by a 
shift in the relative timing of neural crest 
differentiation relative to neural tube dif- 
ferentiation. In a five- to six-somite em- 
bryo (approximately 10 days gestation) of 
M. domestica (Fig. 3), significant neural 
crest migration has already occurred; how- 
ever, no folding has taken place in the neu- 
ral plate. In marsupials substantial migra- 
tion of neural crest into the first arch and 
future frontonasal regions has occurred 

126 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 4. Whole-mount staining of embryos with antibody to Distal-less protein. Methods follow Hanken et al. (1992). (A) 
Monodelphis after 10.5 days of gestation (10-12 somites); (B) Mus at 10- to 12-somite stage; (C) Monodelphis after 10 days, 
20 hours of gestation (16 somites); (D) Mus at 14- to 16-somite stage. (A) and (B) represent early stages in neural crest 
accumulation in branchial arches. Note that although the anterior neural tube is open in tjoth embryos, the fore- and midbrain 
are much more robust (both in thickness of neuroepithelium and length of region) in the mouse, but that much less accumulation 
of neural crest has occurred in the frontonasal and first arch regions in this animal. In (C) and (D) the same pattern continues — 
in Monodelphis the neural tube is still open anteriorly, whereas in the mouse the neural tube is complete and regional differ- 
entiation is beginning. However, massive accumulations of neural crest occur in the facial region, particularly in the maxillary 
process, in Monodelphis relative to the mouse. Key: arrows represent approximate forebrain-midbrain and midbrain-hindbrain 
junctions; o, otic vesicle; mx, maxillary process; md, mandibular process. 

before any closure of the tube occurs at 
any point along its length. Further, al- 
though the hindbrain is well differentiated 
at this stage, with evidence of all rhom- 
bomeres, little proliferation or differenti- 
ation occurs in either the fore- or midbrain 
regions. In eutherians significant prolifer- 
ation of tissues in the fore- and midbrain 
regions occurs before neural crest migra- 
tion begins, and rhombomere subdivision 
occurs after this period. These preliminaiy 
studies do indicate that the neural crest 
arises from the same rhombomeric seg- 
ments that have been reported in other 
vertebrates (e.g., Noden, 1991). However, 
because little differentiation of fore- and 
midbrain regions occurs at this stage, it is 

difficult to assess the contribution of these 
regions to the neural crest in marsupials. 
Migration of the neural crest was fur- 
ther examined using an antibody to Distal- 
less proteins. These proteins, produced by 
genes of the mammalian Dlx family, bind 
to a number of cell types, including mi- 
grating neural crest (Robinson and Ma- 
hon, 1994; Panganiban et al., 1995; Han- 
ken et al., 1997). This antibody is not an 
exclusive marker of neural crest, but it 
does stain migrating neural crest and al- 
lows comparison of neural crest migration 
in marsupials and eutherians relative to 
other tissues (Fig. 4). Apparently, relative 
to neural tube development, significantly 
more neural crest occurs in the branchial 

Evolution of Mammalian Development • Smith 127 

arch region in Monodelphis than in Mus. 
For example, in both the 10- to 12-somite 
and 16-somite stages in Mus (Figs. 4B, D) 
the neural tube is considerably advanced 
relative to Monodelphis (Figs. 4A, C) yet 
the relative size of the first and second 
arches is much smaller in Mus. In partic- 
ular, little or no accumulation of neural 
crest is apparent in the maxillary region in 
Mus. Analysis of the preliminary data sug- 
gests a number of features of the pattern 
of neural crest migration in marsupials. In 
marsupials neural crest seems to migrate 
at a tiine that is earlier, relative to neural 
tube closure, than in eutherians, or other 
amniotes (and indeed, apparently other 
vertebrates). In addition, the neural tube 
seems to differentiate neural crest cells in 
larger populations relative to allocation to 
neural structures in inarsupials. Finally, 
differentiation of the hindbrain, the region 
that supplies much of the neural crest to 
the branchial arches, seems to be ad- 
vanced relative to the forebrain in marsu- 
pials, so that the major delay in CNS de- 
velopment is concentrated primarily in 
forebrain structures. 

Analysis of these data on early devel- 
opment suggests that the differences be- 
tween marsupials and placentals in the rel- 
ative maturation of neural and somatic tis- 
sues occur during the early events in tissue 
differentiation. Therefore, the heteroch- 
ronies are not shifts in the relative rates of 
growth or differentiation of terminal struc- 
tures. These results are consistent with the 
hypothesis that early development is fairly 
plastic and may be modified to meet spe- 
cific demands at a distinctive stage in de- 
velopment (see Raff [1996]). 


Thus far I have considered two clades — 
eutherians and metatherians. Although I 
have implied that the marsupial condition 
is derived, I have not yet provided the ev- 
idence. Below, the patterns observed in 
marsupials and placentals first will be com- 
pared with those of nonmammalian am- 

niotes. I will then discuss preliminary data 
on early development in monotremes. 

Early Development of Amniotes 

Marsupial development, relative to that 
of eutherians, was shown above to be char- 
acterized by at least three major sets of 
heterochronies. First, a relative delay oc- 
curs in differentiation of the CNS and in 
particular in the forebrain region. Second, 
the differentiations of the branchial arch 
and facial regions are advanced. These 
shifts seem to be effected in part by shift- 
ing forward the relative timing of neural 
crest differentiation and migration. Third, 
not discussed in detail above, is the exis- 
tence in marsupials of an extreme rostral- 
caudal gradient of development. Although 
to some degree a rostral-caudal gradient 
exists in eutherians, so that at a given stage 
the forelimb is advanced relative to the 
hind limb, this gradient is extreme in mar- 
supials (Fig. 5). For example, this gradient 
is reflected by the relative differentiation 
of the somites, where in marsupials pos- 
terior seginents develop relatively late. 
However, the most striking expression of 
this gradient is the relative development of 
the fore- and hind limb buds. In inarsu- 
pials the foreliinb bud is massive at a time 
when the hind limb bud is not yet present. 

These three features may be defined as 
three character complexes (each of which 
contains a multitude of individual charac- 
ters) that inay be examined in a broader 
phylogenetic context. In Figure 6 early 
embryos of a chicken (Gallus) and snap- 
ping turtle {Chelijdra), are compared with 
those o{ Monodelphis and Mus. Eutherians 
share with the nonmammalian amniotes 
the advancement of the neural tube, the 
relatively small branchial arches, and the 
relative similarity of the rate of fore- and 
hind limb development. In each, the mar- 
supial condition is quite distinct, and inust 
be intei-preted as derived. Therefore, in 
this context, placentals possess what must 
be taken as the priinitive amniote condi- 

128 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 5. Three stages of development in Monodelphis do- 
mestica: (A) approximately 11 days of gestation; (B) approxi- 
mately 12 days of gestation; (C) approximately 13 days of ges- 
tation. Note that in each specimen the anterior part of the body 
is greatly advanced relative to the posterior in size and relative 
degree of differentiation. In particular, the forelimb is highly 
advanced relative to the hind limb at all stages. 

Comparisons Across Mammalia 

To assess the condition at the node 
Mammaha, information on monotremes, 
the third major clade of extant mammals, 
is needed. Clear possession by mono- 
tremes of the derived elements of marsu- 

pial development would be parsimoniously 
interpreted as a shared derived resem- 
blance. On the other hand, resemblance of 
monotremes to the eutherian condition 
(which is shared with nonmammalian am- 
niotes) would further highlight the derived 
and specialized nature of marsupial devel- 
opment and reproduction. 

Few monotreme embryos are available 
for study. Most are in the Hill Collection, 
part of the Hubrecht Comparative Embry- 
ology Laboratoiy. Preliminary evaluation 
of some of this material indicates that 
inonotremes exliibit a mosaic of marsupi- 
allike and placentallike developmental 
characters. First, monotreines share with 
marsupials and nonmammalian amniotes 
many primitive characteristics of the ear- 
liest embiyo. For example, all develop as 
a flat blastodisc on a large yolk, in a man- 
ner that is quite distinct froin that of eu- 
therians (Hughes, 1993). In addition to 
these shared primitive characters of early 
development in marsupials and mono- 
tremes, monotremes and marsupials share 
some derived characters. For example, 
early in development in monotremes the 
branchial arches apparently are accelerat- 
ed relative to the neural tube. 

Somewhat later in development, mono- 
tremes seeiu to resemble eutherians more 
closely. Figure 7 shows sections of embry- 
onic Monodelphis and Mtis, and a pre- 
hatching Ornithorhynchus (platypus) em- 
bryo. In order to define an equivalent 
landmark, they are matched for the same 
relative stage of development of the eye. 
In Miis, the telencephalon is differentiated 
as distinct hemispheres and cell prolifera- 
tion is well underway in both the telen- 
cephalon and diencephalon. However, the 
cells that will form the cartilages, bones, 
and muscles of the face show little or no 
evidence of condensation or differentia- 
tion. In contrast, in Monodelphis the tel- 
encephalon has just begun evagination 
(not shown in this section) but little or no 
proliferation of cells has occurred in either 
the telencephalon or diencephalon. Yet, at 
this stage cartilage is fully differentiated 

Evolution of Mammalian Development • Smith 


Figure 6. Embryos of (A) Monodelphis; (B) Mus; (C) Gallus; and (D) Chelydra. Note that in (B) through (D) the forelimb bud 
(FL) and hind limb bud (HL) are approximately the same size; in (A) the forelimb bud is massive, whereas the hind limb is not 
yet at the bud stage. Further note that in (B) through (D) the telencephalon (T), as well as the other regions of the brain are 
recognizable as distinct swellings; no such divisions yet exist in (A). Finally note that the branchial arches and frontonasal region 
(N) are massive in (A), and relatively small in the other taxa. 

and present in the nasal and basicranial re- 
gions; bone is present in the dentary, pre- 
maxilla, and maxilla; and the tongue mus- 
culature has differentiated (see Smith 
[1994, 1997]). The Omithorhynchus em- 
bryo is intermediate between these con- 
ditions, although it is more similar to the 
eutherian than inetatherian condition. The 
major subdivisions are present in the neu- 
ral tube and proliferation of the neuroep- 
ithelium is well underway in both the tel- 
encephalon and diencephalon, yet like eu- 
therians no cartilage, bone, or muscle is 
present. Therefore, monotremes do not 
exhibit the same degree of advancement 

of cranial musculoskeletal tissues as mar- 

Until more monotreme material is ob- 
tained and analyzed, the issue of the con- 
dition at the node Mammalia is obscure. 
Monotremes apparently share many prim- 
itive characters with marsupials, as well as 
some derived features of early develop- 
ment. However, indication also exists that 
aspects of the developmental trajectory of 
monotremes resemble that of eutherians 
(and nonmammalian amniotes). Under- 
standing the mosaic of patterns is essential 
to our efforts to reconstruct the phyloge- 
netic relations of mainmals and model the 

130 BtiUetin Museum of Comparative Zoologij, Vol. 156, No. 1 

^^99 ( Ji^^^JftlA 

■'  '9 


t ': 1'° "^'^SH 

^ VK,OtiMnBQ||HW 



origins of mammalian developmental ad- 


Developmental Plasticity and Conservation 

The processes that distinguish marsupial 
and placental maminals begin at the ear- 
liest point in the differentiation of tissues 
of the craniofacial region. They involve 
fundamental shifts in early patterning 
events, and comprise changes in a complex 
series of events. These changes may be 
traced back to the appearance of the neu- 
ral plate, where at this stage large numbers 
of cells differentiate into migratory neural 
crest cells, rather than neural tissues. 
Within the neural tube the hindbrain dif- 
ferentiates early and the midbrain and 
forebrain are delayed. In addition, a local- 
ized acceleration of somitic differentiation 
occurs in the cervical and upper thoracic 
regions and a marked delay occurs in cau- 
dal somites. The distinction is not a simple 
shifting foi"ward in time, or speeding up 
the rate of development of a few features, 
nor is it due to the establishment of a sim- 
ple anterior— posterior gradient of acceler- 
ation along the body axis. The changes in- 
volve multiple advancements and delays of 
sets of cells, tissues, and organs, within and 
between regions. 

Developmental differences between mar- 

Figure 7. Sections through the head of (A) Monodelphis; (B) 
Ornithorhynchus; (C) Mus. Specimens were chosen for ap- 
proximate match in the relative development of the eye. In 
Monodelphis the neural tube is at an early stage with no sig- 
nificant proliferation of the neural epithelium (although the tel- 
encephalon has evaginated — not visible in this section). How- 
ever, at this time the maxillary, dentary, and premaxillary 
bones have begun ossification; cartilage is present in the ba- 
sisphenoid and basiocclpital regions; and muscle has differ- 
entiated in the tongue. In Mus the telencephalon is evaginated, 
and significant proliferation of neural epithelium has occurred 
in all regions of the brain, but no cartilage, bone, or muscle 
have begun differentiation. Ornithorhynchus resembles Mus: 
no bone, muscle, or cartilage are present, yet the neural epi- 
thelium has started proliferation. However, unlike Mus, con- 
densations for bones, muscles, and cartilages apparently have 
been initiated. Key: C, connective tissue in basicranium; TEL, 
telencephalon; arrow, ossification in the maxillary bone; T, 

Evolution of Mammalian Development* Smidi 


supials and placentals thus are not late 
changes or terminal additions to a conser- 
vative mammalian developmental pro- 
gram. Little evidence exists that develop- 
ment in the two groups of therians can be 
characterized simply as two ends of a con- 
tinuuin. The developmental trajectory in 
marsupials is highly modified from very 
early stages in order to produce a specific 
adaptive configuration of the neonate. This 
suggests that development, even at its ear- 
hest stages, is highly plastic. 

The observation of significant early plas- 
ticity, even in animals in which the adults 
are quite siinilar, is important for under- 
standing the ways in which development 
and evolution interact. If it is common for 
early development to be shifted in funda- 
mental ways, with little change in adult 
structure, then the degree to which devel- 
opmental processes impose constraints on 
the generation of form may have been 
overestimated. These issues, on relative 
consen'ation or plasticity of development, 
and therefore the possible severity of de- 
velopmental constraints, can only be re- 
solved by studies that are both broad phy- 
logenetically and detailed developmentally 
(e.g., Richardson et al., 1977; Hall, 1984; 
Wray and Raff, 1991; Hanken et al., 1992, 
1997; Swalla et al., 1993; Richardson, 
1995; Olsson and Hanken, 1996; Raff, 
1996; Lowe and Wray, 1997; Smith, 1997). 

Evolution of Mammalian Development 

Comparative patterns of development in 
marsupials and placentals can be mapped 
in relation to two outgroups. When mono- 
tremes are added to the comparison, veiy 
preliminary observations suggest that the 
primitive condition for mammals is some- 
what intermediate, but probably is char- 
acterized by an altricial neonate and an 
early shift in the relative development of 
branchial arches relative to neural tissue. 
However, when mammals are compared 
with nonmaminalian amniotes, the inar- 
supial condition appears highly derived, 
and the pattern seen in eutherians closely 

resembles the outgroup, or primitive con- 

This set of observations implies one of 
three scenarios. First, it is possible that the 
primitive mammalian developmental con- 
dition was monotreine- or marsupiallike, 
and the resemblance of placentals to other 
amniotes is an evolutionary reversion. 
However, it must be emphasized that 
these are not simple shifts in growth rates 
of terminal structures, but instead changes 
in the early patterning of multiple systems. 
Because of the complexity of the charac- 
ters, this is not a parsimonious hypothesis. 
Further, if this scenario is true, then sev- 
eral vital questions remain unanswered. 
Are the shared characters of marsupials 
and monotremes reflective of the initial 
adaptations of mammalian reproduction? 
If so, why did development change so rad- 
ically in response to the initial mammalian 
reproductive pattern? Did this primitive 
reproductive pattern place the constraints 
on embryo resource allocation hypothe- 
sized above or are other factors in opera- 
tion? Finally, why did eutherian develop- 
ment revert to what appears to be a prim- 
itive amniote pattern? 

Second, it is possible that marsupials 
and monotremes share derived characters 
relative to eutherians + nonmammalian 
amniotes. This pattern would provide sup- 
port for the recently revived Marsupionta, 
a hypothesized monophyletic group con- 
sisting of marsupial and monotreme mam- 
mals (e.g., Gregoiy, 1947). The existence 
of the Marsupionta has been supported re- 
cently by molecular data (i.e., Janke et al., 
1996, 1997; Penny and Hasegawa, 1997; 
Kirsch and Mayer, 1998); however, virtu- 
ally all morphologic and paleontological 
evidence firmly supports the Theria (e.g., 
Crompton, 1980; Rowe, 1988; Jenkins, 
1990; Hopson and Rougier, 1993; Wible 
and Hopson, 1993; Zeller, 1993; Kirsch 
and Mayer, 1998). Finally, it is possible 
that the derived conditions that marsupials 
and monotremes share are independently 
derived. As I have argued above, this is a 
complex series of characters, and thus 

132 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

such convergence would involve complex 
series of convergent changes. 

The resolution of the relative merit of 
these scenarios requires more research on 
development in a variety of animals, in 
particular monotremes. Focus on two is- 
sues — the degree to which marsupials and 
monotremes share derived characters, and 
the complexity of the developmental pro- 
cesses involved — ^will help to assess the 
likelihood of reversion or convergence in 
the lands of characters and processes dis- 
cussed here. 

To conclude, the patterns of develop- 
ment described here allow a number of 
generalizations about the evolution of 
mammalian reproductive strategies. First, 
marsupials are not simply primitive with 
regard to eutherians, at least in the context 
of amniotes as a whole. Early develop inent 
in marsupials is derived when compared 
with other amniotes. Second, early devel- 
opment is quite distinct in marsupials, so 
that the entire developmental trajectory in 
marsupial and placental mammals differs. 
These two sets of observations seem to re- 
fute the hypothesis that marsupials and 
placentals are merely two ends of a con- 
tinuum. The developmental data support 
the hypotheses that marsupials and pla- 
centals have followed two distinct paths, 
each derived in its own right. I hope that 
further detailed information will provide 
infonnation on the most likely state at the 
node Mammalia and may allow the specif- 
ic functional correlates of monotreme, 
marsupial, and placental developmental 
patterns to be assessed. Analysis of these 
data may help resolve phylogenetic issues 
as well as further efforts to reconstruct the 
evolution of mammalian reproduction. 


I thank Drs. J. Hanken, W. M. Kier, and 
J. A. W. Kirsch for comments; G. Pangan- 
iban for the antibody to dll; Alex van Niev- 
elt for comments and technical support; 
the curators at the Hubrecht Comparative 
Embryology Collection at the National 
Laboratory of Developmental Biology, 

Utrecht, the Netherlands, and the Cornell 
University Comparative Embryology Col- 
lection, Ithaca, New York, for allowing me 
access to specimens in their care; and Na- 
tional Science Foundations grants IBN 
9407616 and 9816985 for support. Finally, 
I wish to express particular gratitude to Dr. 
A. W. Crompton for his influence on me 
during my career, and his numerous con- 
tributions to our understanding of verte- 
brate, and particularly mammalian, evolu- 
tion, function, and morphology. 


Atchley, W. R., and B. K. Hall. 1991. A model for 
development and evolution of complex morpho- 
logical structures. Biological Reviews, 66: 101- 

Cheek, D. B. 1975. Fetal and Postnatal Cellular 
Growth. New York: John Wiley and Sons, xii + 
538 pp. 

Clark, C. T, and K. K. Smith. 1993. Cranial oste- 
ogenesis in Monodelphis domestica (Didelphi- 
dae) and Macropiis eugenii (Macropodidae). 
Journal of Morphology, 215: 119-149. 

Crompton, A. W. 1980. Biology of the earliest mam- 
mals, pp. 1-12. In K. Schmidt-Nielsen, L. Bolis, 
and C. R. Taylor (eds.). Comparative Physiology: 
Primitive Mammals. Cambridge, United King- 
dom: Cambridge University Press, .xiv -I- 196 pp. 

DhoPESHWARKAR, G. a. 1983. Nutrition and Brain 
Development. New York: Plenum Press, viii + 
326 pp. 

DOBBING, J. 1972. Vulnerable periods in brain de- 
velopment, pp. 9-20. In K. Elliot and J. Knight 
(eds.). Lipids, Malnutrition and the Developing 
Brain: Ciba Foundation Symposium. Amster- 
dam: Elsevier, viii -I- 326 pp. 

Dodge, P. R., A. L. Prensky, and R. D. Feigin. 
1975. Nutrition and the Developing Nervous 
System. St. Louis, Missouri: C. V. Mosby Com- 
pany, xvi + 538 pp. 

DUNLOP, L.-L., AND B. K. Hall. 1995. Relationships 
between cellular condensation, preosteoblast for- 
mation and epithelial-mesenchymal interactions 
in initiation of osteogenesis. International Journal 
of Developmental Biology, 39: 357—371. 

FiLAN, S. L. 1991. Development of the middle ear 
region in Monodelphis domestica (Marsupialia, 
Didelphidae): marsupial solutions to early birth. 
Journal of Zoology, London, 225: 577—588. 

Gemmell, R. T, and L. Selwood. 1994. Structunil 
development in the newborn marsupial, the 
stripe-faced dunnart, Sminthopsis macroura. 
Acta Anatomica, 149: 1-12. 

Gregory, W. K. 1947. The monotremes and the pa- 
limpsest theory. Bulletin of the American Mu- 
seum of Natural History, 88: 1-52. 

Evolution of Mammalian Development • Smith 133 

GruneBERG, H. 1963. The Pathology of Develop- 
ment. A Study of Inherited Skeletal Disorders in 
Animals. New York: John Wiley and Sons, xiv + 
309 pp. 

Hall, B. K. 1984. Developmental processes under- 
lying heterochrony as an evolutionary mecha- 
nism. Canadian Journal of Zoology, 62: 1-7. 

. 1987. Tissue interactions in the development 

and evolution of the vertebrate head, pp. 215- 
259. In P. F. A. Maderson (ed.). Developmental 
and Evolutionary Aspects of die Neural Crest. 
New York: Wiley-Interscience. xiv + 394 pp. 

1991. Cellular interactions during cartilage 

and bone development. Journal of Craniofacial 
Genetics and Developmental Biology, 11: 238- 

Hall, B. K., and S. Horstadius. 1988. The Neural 
Crest. London: Oxford University Press, viii -I- 
303 pp. 

Hall, B. K., and T. Miyake. 1992. The membranous 
skeleton: the role of cell condensations in ver- 
tebrate skeletogenesis. Anatomy and Embryolo- 
gy, 186: 107-124. 

. 1995. Divide, accumulate, differentiate: cell 

condensation in skeletal development revisited. 
International Journal of Developmental Biology, 
39: 881-893. 

Hanken, J., D. H. Jennings, and L. Olsson. 1997. 
Mechanistic basis of life-history evolution in an- 
uran amphibians: direct development. American 
Zoologist, 37: 160-171. 

Hanken, J., M. W. Klymkowsky, C. H. S. D. W. 
Summers, and N. Ingebrigtsen. 1992. Cranial 
ontogeny in the direct-developing frog, Eletith- 
erodactyhis coqui (Anura: Leptodactyhdae) ana- 
lyzed using whole-mount immunohistochemistry. 
Journal of Morphology, 211: 95-118. 

Hayssen, v., R. C. Lacy, and P. J. Parker. 1985. 
Metatherian reproduction: transitional or tran- 
scending? American Naturalist, 126: 617—632. 

HersCHKOWITZ, N. 1989. Brain development and 
nutrition, pp. 297-304. 7h P. Evrard and A. Min- 
kowski (eds.). Developmental Neurobiology. 
New York: Vevey/Raven Press, xix -I- 315 pp. 

Hetzel, B. S., and R. M. Smith. 1981. Fetal Brain 
Disorders. Amsterdam: Elsevier/North Holland, 
xiii + 489 pp. 

Hill, J. P, and W. C. O. Hill. 1955. The growth 
stages of the pouch young of the native cat (Das- 
ijiinis viverrinus) together with observations on 
the anatomy of the newborn young. Transactions 
of the Zoological Society, London, 28: 349-453. 

Hill, J. P., and K. M. Watson. 1958. The early de- 
velopment of the brain in marsupials; prelimi- 
nary communication. Journal of Anatomy, 92: 

Hopson, J. A., AND G. W Rougier. 1993. Braincase 
structure in the oldest known skull of a therian 
mammal: implications for mammalian systemat- 
ics and cranial evolution. American Journal of 
Science, 293: 268-299. 

Hughes, R. L. 1993. Monotreme development with 
particular reference to the extraembryonic mem- 
branes. Journal of Experimental Zoology, 266: 

Hughes, R. L., and L. S. Hall. 1988. Structural 
adaptations of the newborn marsupial, pp. 8-27. 
In C. H. Tyndale-Biscoe and P. A. Janssens 
(eds.). The Developing Marsupial. Models for 
Biomedical Research. Berlin: Springer vii + 245 

A. Vhaeseler, and S. Paabo. 1996. The mito- 
chondrial genome of a monotreme — the platy- 
pus. Journal of Molecular Evolution, 42: 15.3— 

jANKE, A., X. Xu, AND U. Arnason. 1997. The com- 
plete mitochondrial genome of the wallaroo {Ma- 
cropus robiistus) and the phylogenetic relation- 
ship among Monotremata, Marsupialia and Eu- 
theria. Proceedings of the National Academy of 
Sciences, 94: 1276-1281. 

Jenkins, F. A., Jr. 1990. Monotremes and the biology 
of Mesozoic mammals. Netherlands Journal of 
Zoology, 40: 5-31. 

Kirsch, J. A. W 1977a. Biological aspects of the mar- 
supial-placental dichotomy: a reply to Lillegrav- 
en. Evolution, 31: 898-900. 

. 1977b. The six-percent solution: second 

thoughts on the adaptedness of the Marsupialia. 
American Scientist, 65: 276-288. 

Kirsch, J. A. W, and G. C. Mayer. 1998. The platy- 
pus is not a rodent: DNA hybridisation, amniote 
phylogeny, and the palimpsest theory. Philosoph- 
ical Transactions of the Royal Society, Series B, 
353: 1221-1237. 

Klima, M. 1987. Early development of the shoulder 
girdle and sternum in marsupials (Mammalia: 
Metatheria). Advances in Anatomy, Embryology, 
and Cell Biology, 109: 1-91. 

Le Douarin, N. 1982. The Neural Crest. Cam- 
bridge, United Kingdom: Cambridge University 
Press, xi + 259 pp. 

Lee, a. K., and A. CocKBURN. 1985. Evolutionary 
Ecology of Marsupials. Cambridge, United King- 
dom: Cambridge University Press, viii + 274 pp. 

LiLLEGRAVEN, J. A. 1975. Biological considerations 
of the marsupial-placental dichotomy. Evolution, 
29: 707-722. 

AND J. L. PatTON. 1987. The origin of eutherian 
mammals. Biological Journal of the Linnean So- 
ciety, 32: 281-336. 

Lowe, C. J., and G. A. Wray. 1997. Radical alter- 
ations in the roles of homeobox genes during 
echinoderm evolution. Nature, 389: 718-721. 

Maier, W 1987. The ontogenetic development of the 
orbitotemporal region in the skull of Monodel- 
phis domestica (Didelphidae, Marsupialia), and 
the problem of the mammalian alisphenoid, pp. 
71-90. In H.-J. Kuhn and U. Zeller (eds.), Mor- 

134 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

phogenesis of the Mammalian Skull. Hamburg, 
Germany: Verlag Paul Parey. 144 pp. 

-. 1993. Cranial moiphology of the therian 

common ancestor, as suggested by the adapta- 
tions of neonatal marsupials, pp. 165—181. In F. 
S. Szalay, M. J. Novacek, and M. C. McKenna 
(eds.). Mammal Phylogeny — Mesozoic Differ- 
entiation, Multituberculates, Monotremes, Early 
Therians and Marsupials. New York: Springer, x 
+ 249 pp. 

MiYAKE, T, A. M. Cameron, and B. K. Hall. 1996. 
Stage-specific onset of condensation and matrix 
deposition for Meckel's and other first arch car- 
tilages in inbred C57B1/6 mice. Journal of Cra- 
niofacial Genetics and Developmental Biology, 
16: 32-47. 

. 1997. Stage-specific expression patterns of 

tilkaline phosphatase during development of the 
first arch skeleton in inbred C57BL/6 mouse em- 
bryos. Journal of Anatomy, 190: 239—260. 

Morriss-Kay, G., E. Ruberte, and Y. Fukiishl 
1993. Mammalian neural crest and neural crest 
derivatives. Annals of Anatomy, 175: 501-507. 

Nelson, J. E. 1988. Growth of the brain, pp. 86- 
100. In C. H. Tyndale-Biscoe and P. A. Janssens 
(eds.). The Developing Marsupial. Berlin: 
Springer, vii -I- 245 pp. 

Nichols, D. H. 1981. Neural crest formation in the 
head of the mouse embryo as observed using a 
new histological technique. Journal of Embiyol- 
ogy and Experimental Morphology, 64: 105-120. 

. 1986. Formation and distribution of neural 

crest mesenchyme to the first phaiyngeal arch 
region of the mouse embryo. American Journal 
of Anatomy, 176: 221-231. 

. 1987. Ultrastructure of neural crest forma- 
tion in the midbrain/rostral hindbrain and peri- 
otic hindbrain regions of the mouse embiyo. 
American Journal of Anatomy, 179: 143-154. 

NODEN, D. M. 1983. The role of the neural crest in 
patterning of avian cranial skeletal, connective, 
and muscle tissues. Developmental Biolog)-, 96: 

. 1987. Interactions between cephalic neural 

crest and mesodermal populations, pp. 89—119. 
In P. F. A. Maderson (ed.). Developmental and 
Evolutionary Aspects of the Neural Crest. New 
York: Wiley-Interscience. xiv + 394 pp. 

. 1991. Vertebrate craniofacial development: 

the relation between ontogenetic process and 
moqihological outcome. Brain, Behavior and 
Evolution, 38: 190-225. 

Novacek, M. J. 1990. Moqahology, paleontology and 
the higher clades of mammals, pp. 507-543. In 
H. H. Genoways (ed.). Current Mammalogy. Vol. 
2. New York: Plenum, x-viii + 577 pp. 

NUNN, C. L., AND K. K. Smith. 1998. Statistical anal- 
yses of developmental sequences: the craniofa- 
cial region of marsupial and placental mammals. 
American Naturalist, 152: 82-101. 

OlssoN, L., and J. Hanken. 1996. Cranial neural- 

crest migration and chondrogenic fate in the Ori- 
ental fire-bellied toad Bomhina orientalis: defin- 
ing the ancestral pattern of head development in 
anuran amphibians. Journal of Morphology, 229: 

Panganiban, G., a. Sebring, L. Nagy, and S. B. 
Carroll. 1995. The development of crustacean 
limbs and the evolution of arthropods. Science, 
270: 1363-1366. 

Parker, P. 1977. An ecological comparison of mar- 
supial and placental patterns of reproduction, pp. 
269-285. In B. Stonehouse and D. Gilmore 
(eds.). The Biology of Marsupials. London: Mac- 
millian Press, Ltd. vii -I- 486 pp. 

Penny, D., and M. Hasegawa. 1997. The platypus 
put in its place. Nature, 387: 549-550. 

Peterson, P. E., T N. Blankenship, D. B. Wilson, 
AND A. G. Hendrickx. 1996. Analysis of hind- 
brain neural crest migration in the long-tailed 
monkey {Macaca fascicularis). Anatomy and Em- 
bryology, 194: 235-246. 

Raff, R. A. 1996. The Shape of Life: Genes, Devel- 
opment and the Evolution of Animal Form. Chi- 
cago: University of Chicago Press, xxiii -I- 520 pp. 

Renfree, M. B. 1983. Marsupial reproduction: the 
choice between placentation and lactation, pp. 
1-29. In C. A. Finn (ed.), O.xford Reviews of Re- 
productive Biology. Vol. 5. Oxford, United King- 
dom: Oxford University Press. 

. 1993. Ontogeny, genetic control, and phylog- 
eny of female reproduction in monotreme and 
therian mammals, pp. 4-20. In F S. Szalay, M. 
J. Novacek, and M. C. McKenna (eds.). Mammal 
Phylogeny — Mesozoic Differentiation, Multitu- 
berculates, Monotremes, Early Therians and 
Marsupials. New York: Springer, x + 249 pp. 

. 1995. Monotreme and marsupial reproduc- 
tion. Reproduction, Fertility and Development, 
7: 1003-1020. 

Richardson, M. K. 1995. Heterochrony and the 
phylotypic period. Developmental Biology, 172: 

Richardson, M. K., J. Hanken, M. L. Goonerat- 
NE, C. PiEAU, A. Raynaud, L. Selwood, and 
G. M. Wright. 1997. There is no highly con- 
served embryonic stage in the vertebrates: im- 
plications for current theories of evolution and 
development. Anatomy and Embiyology, 196: 

Robinson, G. W, and K. A. Mahon. 1994. Differ- 
ential and overlapping e\-pression domains of 
Dl.x-2 and Dl.x-3 suggest distinct roles for Distal- 
less homeobox genes in craniofacial develop- 
ment. Mechanisms of Development, 48: 199- 

ROW'E, T 1988. Definition, diagnosis and the origin 
of Mammalia. Journal of Vertebrate Paleontolo- 
gy, 8: 241-264. 

Sacher, G. a., and E. F. Staffeldt. 1974. Relation 
of gestation time to brain weight for placental 
mammals: implications for the theory of verte- 

Evolution of Mammalian Development • Smith 135 

brate growth. American Naturalist, 108: 593- 

SAnchez-Villagra, M. 1999. A Study of Cranial 
Evolution in Marsupials with Description of New 
Fossil Forms. Ph.D. thesis. Durham, North Car- 
olina: Duke University. 538 pp. 

Serbedzija, G. N., M. Bronner-Fraser, and S. E. 
FraseR. 1992. Vital dye analysis of cranial neural 
crest cell migration in the mouse embryo. De- 
velopment, 116: 297-307. 

Shine, R. 1985. The evolution of viviparity in rep- 
tiles: an ecological analysis, pp. 605-694. In C. 
Cans and F. Billett (eds.). Biology of the Reptilia. 
Vol. 15, Development B. New York: John Wiley 
and Sons, x + 731 pp. 

Shoemaker, W. J., and F. E. Bloom. 1977. Effect 
of undernutrition on brain morphology, pp. 147- 
192. In R. J. Wurtman and J. J. Wurtman (eds.). 
Nutrition and the Brain. Vol. 2. New York: Raven 
Press, vi + 313 pp. 

Smith, K. K. 1994. The development of craniofacial 
musculature in Monodelphis domestico (Didel- 
phidae, Marsupialia). Journal of Morphology, 
222: 149-173. 

. 1996. Integration of craniofacial structures 

during development in mammals. American Zo- 
ologist, 36: 70-79. 

. 1997. Comparative patterns of craniofacial 

crest in the rat embryo. Cell Tissue Research, 
240: 40.3-416. 

-. 1986. Analysis of cranial neural crest cell mi- 

development in eutherian and metatherian mam- 
mals. Evolution, 51: 1663-1678. 
Springer, M. S., M. Wasserman, J. R. Kavanagh, 

C. Kraje\\'SKI. 1998. The origin of the Austra- 
lasian marsupial fauna and the phylogenetic af- 
finities of the enigmatic monito del monte and 
marsupial mole. Proceedings of the Royal Soci- 
ety of London, Series B, 265: 2381-2386. 

SWALLA, B. J., K. W. Makabe, N. Satoh, and W. R. 
Jeffery. 1993. Novel genes expressed differen- 
tially in ascidians with alternate modes of devel- 
opment. Development, 119: 307-318. 

Tan, S. S., and G. M. MORRISS-Kay. 1985. The de- 
velopment and distribution of the cranial neural 

gration and early fates in postimplantation rat 
chimeras. Journal of Embryology and Experi- 
mental Morphology, 98: 21-58. 

Trainor, p. a., and p. p. L. Tam. 1995. Cranial par- 
axial mesoderm and neural crest cells of the 
mouse embryo: co-distribution in the craniofacial 
mesenchyme but distinct segregation in branchi- 
al arches. Development, 121: 2569-2582. 

Tyndale-Biscoe, H., and M. Renfree. 1987. Re- 
productive Physiology of Marsupials. Cambridge, 
United Kingdom: Cambridge University Press, 
xiv + 476 pp. 

Wible, J. R., AND J. A. HOPSON. 1993. Basicranial 
evidence for early mammal phylogeny, pp. 45- 
62. In F. S. Szalay, M. J. Novacek, and M. C. 
McKenna (eds.). Mammal Phylogeny — Mesozoic 
Differentiation, Multituberculates, Monotremes, 
Early Therians and Marsupials. New York: 
Springer, x + 249 pp. 

WiNICK, M. 1976. Malnutrition and Brain Develop- 
ment. New York: Oxford University Press, xv + 
169 pp. 

. 1979. Nutrition, Pre- and Postnatal Devel- 
opment. New York: Plenum Press, xx + 496 pp. 

W^INICK, M., P. Rosso, and J. A. Brasel. 1972. Mal- 
nutrition and cellular growth in the brain: exis- 
tence of critical periods, pp. 200-206. 7;! K. El- 
liot and J. Knight (eds.). Lipids, Malnutrition and 
the Developing Brain: Ciba Foundation Sympo- 
sium. Amsterdam: Elsevier, viii + 326 pp. 

Wray, G. a., and R. a. Raff. 1991. The evolution 
of developmental strategy in marine inverte- 
brates. Trends in Ecology and Evolution, 6: 45- 

ZelleR, U. 1993. Ontogenetic evidence for cranial 
homologies in monotremes and therians, with 
special reference to OrnithoHujnchns, pp. 95- 
107. In F. S. Szalay, M. J. Novacek, and M. C. 
McKenna (eds.). Mammal Phylogeny — Mesozoic 
Differentiation, Multituberculates, Monotremes, 
Early Therians and Marsupials. New York: 
Springer, x + 249 pp. 



Abstract. A sample of 20 Late Triassic theropod 
footprints from Greenland preserves evidence of ped- 
al integument. Skin impressions range from dimples, 
valleys, peaks, and ridges, to parallel striations. These 
features were created by tlie scale-covered digital 
pads as the skin— sediment interface was broken. 
Therefore, sldn impressions document aspects of 
both the direction and timing of sldn motion, allowing 
foot movements during the stance phase of locomo- 
tion to be inferred. Sldn impressions represent a pre- 
viously unrecognized source of functional data for re- 
constructing theropod locomotion. 


Dinosaur footprints vary widely in the 
amount of detail they preserve. The ma- 
jority of tracks show diffusely contoured 
iinprints of each digit, although some are 
more clearly defined by digital pad and 
claw impressions. Very few preserve the 
finest level of detail — evidence of integu- 
mentary structures (Lockley, 1989). Many 
factors are responsible for the relative rar- 
ity of tracks bearing traces of minute fea- 
tures. Almost all of these factors relate to 
scale, which can range over two orders of 
magnitude from the entire foot to an in- 
dividual epidermal tubercle. For example, 
larger features permanently deform sub- 
strates of widely differing properties, 
whereas smaller features require much 
more stringent conditions to leave their 
mark (Allen, 1997; Currie et al, 1991). 
Imprints of gross structures are also more 
likely to be preserved and discovered as 
undeqDrints or overprints (Langston, 1986; 
Lockley, 1989). In contrast, fine details are 

' Department of Ecology and Evolutionary Biology, 
Box G, Brown University, Providence, Rhode Island 

only visible if the "true" track survives and 
is directly exposed. Finally, large imprints 
are more resistant to erosion than are 
small ones, both before burial and after 

Reports of skin impressions in fossil di- 
nosaur tracks are uncommon in the ich- 
nological literature. Classic works include 
only brief references to "papillae", "tuber- 
cles", and "pits" (Hitchcock, 1858; Lull, 
1953), and one illustration (Hitchcock, 
1858, plate X). "Striations", "striae", and 
"furrows" are also described in passing 
(Baird, 1957; Hitchcock, 1858; Woodhams 
and Hines, 1989). The most thoroughly 
documented pedal skin imprints have 
been attributed to ornithischians (Currie 
et al., 1991). In this study I report on Late 
Triassic theropod tracks from Greenland 
that preserve evidence of integumentary 
detail. I analyze sldn impressions as re- 
cords of foot movement and discuss their 
contribution to reconstructing theropod 


In 1989, extensive horizons containing 
dinosaur tracks were discovered in the 
0rsted Dal Member of the uppermost 
Fleming Fjord Fonnation of Jameson 
Land, East Greenland (Gatesy et al., 1999; 
Jenkins et al., 1994). These cyclically bed- 
ded siliciclastic and carbonate-bearing 
strata were deposited in an extensive rift 
lake system of Norian— Rhaetic age (Clein- 
mensen et al., 1998). Herein, I report on 
20 tracks with skin impressions that were 
collected froin four localities (see map in 
Jenkins et al., 1994): eight prints at Tait 

Bull. Mus. Comp. Zool., 156(1): 137-149, October, 2001 137 

138 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Bjerg (L layer), two at Macknight Bjerg (S 
layer), one at Sydkronen (SS layer), and 
nine at Wood Bjerg (C layer). These spec- 
imens are, with one exception, isolated 
prints. Individual trackways were either 
too poorly exposed or impossible to iden- 
tify because of a high density of similarly 
sized tracks. Specimens will be housed at 
the Geological Museum at the University 
of Copenhagen; herein, temporary identi- 
fication numbers are used. 

All tracks were exposed by natural 
weathering; excavation and mechanical 
preparation have proven largely unsuc- 
cessful. Original material was collected 
and analyzed for this study, but silicone 
(Silastic) or alginate molds were made of 
one half of the tracks and some were cast 
in plaster. Impressions on tracks, molds, 
and casts were studied using a Wild M7-S 
binocular dissecting microscope and fiber- 
optic illuminators. An axis drav^i along the 
length of the iinprint of digit III was used 
as a reference to measure the relative me- 
diolateral orientation of features within a 
track. The distribution of skin imprints 
within and among specimens was tabulat- 
ed by dividing each track into 22 subre- 
gions (see below. Fig. 3, and Table 1). Un- 
fortunately, it is typically extremely diffi- 
cult to discern whether a region lacks skin 
impressions because of a true absence 
rather than incomplete exposure or local- 
ized damage. Therefore, distribution fre- 
quencies represent raw percentages that 
were not adjusted to coinpensate for dif- 
ferential preseivation. 

All figures show impressions as if they 
were made by right feet; left footprints 
were reversed to ease coinparison among 

tracks. Illustrations of skin impressions 
(Fig. 1) were made from camera lucida 
drawings. Stipple density was varied to 
represent depth, rather than shadow. 
Track outlines and regions of skin impres- 
sion were traced over digitized video im- 
ages. Stereo images of skin iinpressions 
(Fig. 2) were created by image processing 
digitized video images in Adobe Photo- 
shop 2.5. Three-dimensional vectors were 
modelled and rendered in Studio 8.5 soft- 
ware froin Alias|Wavefront using data inea- 
sured from striations with a protractor and 



Skin impressions are only found in rel- 
atively shallow tracks (maximum depth 4— 
21 miTi). All preserve imprints of digital 
pads and claws, but lack evidence of the 
manus or hallux (Gatesy et al., 1999; Jen- 
kins et al., 1994). Tracks in the sample 
range from 15 to 23 cin in length (Table 
1; estimated in nearly complete speci- 
mens), with a mean of 18.8 ± 2.7 cm (N 
= 19). Such prints are referable to the 
ichnogenus Grallator {Anchisaiiripii.s; 
Hitchcock, 1858; Lull, 1904; Olsen and 
Galton, 1984), which has been attributed 
to small to medium-sized theropod dino- 
saurs (e.g., Farlow and Lockley, 1993; Ol- 
sen et al, 1998). 

Identification of Skin Impressions 

Tracks with skin impressions were rec- 
ognized in the field by their distinctive re- 
ticulate texture (Figs. 1, 2). The floor, and 
less frequently the walls, of digital pad im- 

Figure 1. Examples of skin impressions in Late Triassic theropod prints from Greenland. Reticulate patterns are found on ttie 
floor and walls of digital pad impressions; micromorphology varies from hexagonally arranged dimples (a, L.04) to bumps, ridges, 
and valleys (b, L.OO). Striations are found on the borders of depressions. Entry striations (c, L.01) were formed by scales plowing 
down and forward in early stance, whereas exit striations (d, L.OO) were created as digital pads were withdrawn in late stance. 
All tracks are drawn as those made by right feet. Shaded areas designate preservation of skin impressions. Illustrations of skin 
impressions show depth, not shadow; deeper areas are darkly stippled relative to lighter, elevated areas. Textures in a-d are 
drawn from a perspective perpendicular to the impressed sediment, rather than the horizontal bedding plane. 

Scale bar: 5 cm for track outlines and 2 mm for skin impression details. 

Theropod Skin Impressions • Gatesy 139 



^■^ X 



^•4 n fi^..'^ 


• I'S^J^^' ' ;:.'.■'■"'■:■•'■'•.':- ' j^S^' 





■.■■\,.. /  ;" .■.-.•■■.■.•-•.■.•.■;.;:.. -.-^y 

140 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

pressions bear structures forming densely 
packed, loosely hexagonal arrays. The spe- 
cific microtopography of these arrays is 
variable within and among tracks. In some, 
concave dimples are separated by raised 
laminae (Figs, la, 2a). This morphology 
may represent a relatively accurate mold 
of convex reticulate scales (Lucas and Stet- 
tenheim, 1972) covering the digital pads. 
In others, arrays are formed from convex 
peaks of sediment. Between these two ex- 
tremes are less regular patterns of peaks, 
ridges, dimples, and troughs (Figs, lb, 2b). 
The size of components making up all ar- 
rays is relatively consistent, both within 
and among tracks. Dimples are approxi- 
mately 1 mm in diameter and peaks are 
spaced approximately 1 mm apart. These 
characteristic reticulate patterns are not 
found on sediment outside of dinosaur 

Along the borders of digital pad depres- 
sions, and more rarely within them, are re- 
gions bearing parallel grooves. Grooves are 
spaced 0.5—1.5 mm apart. In some cases 
these furrowed areas abut reticulate ar- 
rays, but many are isolated patches. 
Grooved sediment is found within para- 
bolic tracts (Figs. Ic, 2c, 3) posterior to 
pad imprints of digits II and IV. These are 
likely entry striations that formed as the 
scale-covered digital pads plowed through 
the sediment before settling into their 
deepest position. Similar grooves are lo- 
cated along the medial and lateral walls of 
the impression of digit III (Figs. Id, 2d, 
3). Such grooves are interpreted as exit 
striations, which were created by scales 
scraping the sediment as the digital pads 
of digit III were lifted from the substrate. 
In two specimens, claw scrapes and stria- 
tions sweep posteriorly; these were formed 
when the foot slipped backward before 

breaking ground contact (Thulborn and 
Wade, 1984, 1989). 

Distribution of Skin Impressions 

Reticulate arrays are found within all 
nine digital pad depressions (two for II, 
three for III, and four for IV), as well as 
in the area around the base of the toes. 
However, skin impressions are unequally 
distributed within and among prints; some 
regions preserve skin imprints much more 
frequently than others (Fig. 3; Table 1). 
Arrays produced by the proximal pad of 
digit II, proximal pad of digit III, and 
proximal-middle pad of digit IV are pres- 
ent in 60%, 65%, and 70% of the tracks 
sampled, respectively. Impressions tend to 
decrease in frequency distally, with the ex- 
ception of the proximal pad of digit IV. 
Overall, reticulate imprints of skin are rel- 
atively common for each digit (II, 50%; 
III, 38%; IV, 51%) as well as for the toe 
base area (50%). 

Striations are much less widespread 
(Fig. 3). Entry striations are lacking from 
digit III, but present at low frequencies 
(5-25%) posterior to pad imprints of digits 
II and IV. The impression of digit III can 
bear exit striations medially, laterally, or 
both; medial grooves produced by the 
withdrawal of the distal pad are most com- 
mon (20%). Evidence of backward slip- 
ping is relatively rare. 

The uneven distribution of skin impres- 
sions within tracks likely stems from an in- 
teraction among several influences. Each 
portion of every pad depression exliibits 
sldn texture in at least one of the 20 tracks, 
showing that the entire plantar surface was 
capable of leaving reticulate arrays. How- 
ever, the timing or magnitude of pressure 
could have differentially altered the sedi- 
ment's cohesiveness and adhesion, affect- 

Figure 2. Stereophotographs of skin impressions showing variety of reticulate (a, L.04 and b, L.OO) and striated (c, L.01 and 
d, L.OO) textures. In all cases thie lighting is from the upper left. 

Scale bars: 5 mm. 

Theropod Skin Impressions • Gatesy 141 




'-'*"''- •" 


7i^  ■rrTSi'" 

142 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Table 1. Distribution of skin impressions in 20 theropod footprints. Presence of reticulate 

texture (r) and entry or exit striations (s) is indicated for the nine toe pads and the toe base. 

Entry and exit tracts not bearing striations are designated by t. 


Side (em) 


Di0t II 

Digit III 

Digit III 

Digit III 

Digit IV 

Digit IV 

Digit IV 

Digit IV 



L 21 






R t 






R 23 














R 21 












R 16 









R inc. 






R 20 







L? 17 






L 22 










L 21 

















L 18 









R 19 




R 19 











R 21 



R t 




L 18 











R 17 












R 16 







R 16 





L 23 










R 15 







R 15 








Percent of sample 











40 5 









Figure 3. Relative frequency of sl<in impressions within 
prints. Numbers show the percentages of the track sample (N 
= 20) having reticulate arrays or striations in each of 22 sub- 
regions. Shaded areas posterior to the impressions of digits II 
and IV represent entry tracts, many of which bear striations 
(see Table 1). Shaded areas medial and lateral to the impres- 
sion of digit III are exit tracts. 

ing the likelihood of skin leaving a mark 
(Currie et al., 1991). Intraprint depth var- 
iation may also play a role. The distal por- 
tions of each toe often left the deepest, 
narrowest depressions. Such contour 
might make skin imprints from distal pads 
less likely to be freed froin overlying ma- 
trix by natural erosion. Field identification 
and collection could impose an additional 
bias, because a reticulate texture was more 
easily recognized than entry or exit stria- 
tions. An analysis of variation within a sin- 
gle trackway could elucidate the relative 
contribution of these factors, but such data 
are as yet unavailable from Greenland. 

Formation Dynamics 

With every step, walking theropods ap- 
plied forces against the ground. On a re- 
ceptive substrate, sediment conformed to 
the plantar surface of the penetrating foot. 
However, skin impressions were only cre- 
ated as the integuinent moved off the sub- 
strate. If the skin— sediment interface di- 

Theropod Skin Impressions • Gatesy 143 

Figure 4. Simplified depiction of skin impression formation. 
As a section of a digital pad is driven vertically into thie ground 
(a to b), thie receptive substrate molds itself to the reticulate 
scales, creating a skin-sediment interface (b). If the pad is 
withdrawn at a steep angle, a relatively accurate representa- 
tion of the integument is exposed (c). In contrast, if the pad 
plows through the sediment, scales on the skin tangential to 
the direction of motion will create striations (d). Therefore, skin 
impressions can act as three-dimensional records of skin 

vided cleanly, each subregion of the track 
would reflect the integumentary stinicture 
it last apposed. This perspective, empha- 
sizing the dynamic nature of skin impres- 

sion formation, has both spatial and tem- 
poral connotations. 

Motion of the skin relative to the sedi- 
ment during separation strongly influences 
sldn impression morphology (Fig. 4). This 
relationship is easily visualized by two ex- 
amples, both starting with a patch of skin 
in contact with a receptive substrate (Fig. 
4b). If the patch is lifted normal to the 
skin-sediment interface, it will have a 
good chance of leaving behind a relatively 
accurate mold of its integumentary surface 
(Fig. 4c). Clear impressions of a reticulate 
scale pattern are indicative of sldn with- 
drawn relatively steeply up and away from 
the sediment. In contrast, if the patch is 
dragged through the sediment, its scales 
will leave behind a series of parallel stria- 
tions (Fig. 4d). The last scales to contact 
the substrate will plow furrows along the 
path of skin movement. Thus, the dynam- 
ics of separation allow skin impressions to 
act as three-dimensional records of integ- 
umentary motion. 

Time is also represented, because all 
skin imprints in a track are not formed si- 
multaneously. The moving toes generate 
skin impressions sequentially as localized 
subregions of the pedal integument vacate 
their underlying area of substrate. For the 
remainder of the discussion, I focus on the 
breaking of the sldn-sediment interface 
and the formation of skin impressions in 
specific portions of the track. Based on 
such evidence, I analyze aspects of thero- 
pod foot movement in three periods with- 
in the stance phase. 

Early Stance: Entry Tracts and Entry 

The stance phase of the stride cycle be- 
gins with ground contact. In early stance 
("touch-down" of Thulbom and Wade, 
1989), the area of skin-sediment contact 
increased as the digital pads penetrated 
the substrate. However, Triassic theropod 
feet did not follow a simple vertical path. 
The convex plantar surface of most digital 
pads entered the substrate obliquely, 
forming a teardrop-shaped impression. 

144 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Such parabolic entry tracts, primarily pro- one pad. Spreading (abduction) of digits II | 

duced by digits II and IV, are preserved in and IV can be inferred from one print 

70% of the prints. In 50% of the sample, (S.Ol), but is not clearly supported by an- 

entry tracts bear striations that were other (L.Ol). Again, trackways bearing 

formed as scales plowed through the sed- multiple prints with skin impressions could 

iment. Specifically, striations are furrows test these hypotheses, but await discovery 

made by the last scales to vacate the sub- in the 0rsted Dal. 

strate. Such scales are located on those Digit III must have followed a down 
portions of a pad's surface tangential to the and foi"ward trajectoiy similar to digits II 
direction of penetration (Figs. 4d, 5b). and IV, but clear entry tracts are only pre- 
Thus, entry striations are three-dimension- served in a single specimen (C.A). One 
al motion vectors, which verify that the possible explanation is overprinting, 
digital pads moved down and forward im- caused by motion nearly parallel to the 
mediately after ground contact. lo'^g ^^s of digit III in early stance. The 
An oblique penetration is expected if distal pad of digit III may have produced 
theropods walked using the same general an entry tract that was quickly obliterated 
mechanism employed by living tetrapods by the advancing middle pad, which in 
(e.g., Alexander 1977). A stance limb ini- turn had its entry tract erased by the prox- 
tially pushes both down and forward imal pad. A similar mechanism could ex- 
against the substrate, causing the animal to plain the variation in entry tracts and stri- 
decelerate. Under such a loading regime, ations made by digits II and IV. Motion of 
the direction and distance a theropods the foot parallel to the long axis of digit III 
digital pads move after contact depend on would minimize overprinting in the diver- 
substrate consistency. Entry striations re- gent side digits. For example, two speci- 
veal that feet penetrated at relatively shal- mens with entry striation angles of — 1° 
low slopes (Fig. 6). Of the 12 specimens (C.B) and —2° (C.I) preserve entry tracts 
measured, the average slope was only 16°. from all six pads of digit II and IV. In con- 
Digital pad movements also vary in their trast, a more oblique trajectory would tend 
mediolateral direction with respect to the to obliterate all but the most proximal en- 
long axis of the footprint. Striation orien- try tract of a digit oriented parallel to foot 
tations vary from —39° (intorted) to +15° movement. A specimen (C.A) showing en- 
(extorted) with respect to the long axis of try striations made by distal pads of digit 
digit III. Entry striations produced by the IV angled medially shows entry tracts from 
proximal pad of digit II are intorted (mean the middle and distal pads of digit III, but 
— 19°) in all four specimens showing this very little signs of entry from digit II. In 
detail (Fig. 6). If the impression of digit this case, the foot was likely laterally ro- 
III lies parallel to the theropods direction tated (toed-out) at impact, possibly during 
of travel, digit lis proximal pad would a sharp turn, 
have moved forward and laterally in early 

stance. Alternatively, if the entire foot was ^id Stance: Pad Deformation 
medially rotated (toed-in) at the time of As the limb force increased and became 

contact (e.g., Padian and Olsen, 1989; more vertically oriented, each digital pad 

Thulborn and Wade, 1989), the proximal ceased moving foiAvard. Loading would 

pad of digit II may have slid primarily for- have caused a theropods pads to deform 

wards. The sequence of pad touchdown and settle into their mid stance ("weight- 

and relative movement among pads could bearing" of Thulborn and Wade, 1989) po- 

potentially be recorded in prints with a sitions (Figs. 5a— c). Although the amount 

complete complement of entiy striations. of deformation that took place is difficult 

Unfortunately, only tliree prints have striae to quantify, the presence of at least some 

documenting the entry path of more than vertical compression and horizontal expan- 

Theropod Skin Impressions • Gatesy 145 


Figure 5. Effects of pad deformation and recoil on sl<in im- 
pressions. A digital pad entering the substrate obliquely in ear- 
ly stance (a to b) creates an entry tract witfi entry striations (*). 
In mid stance the limb drives the pad down vertically (b to c), 
flattening the pad and obliterating the entry tract as new skin 
is forced against the substrate (small arrows). In late stance 
the load is reduced, allowing the viscoelastic pad to return to 
its unflattened shape before liftoff (d). Peripheral skin is peeled 
off at a relatively steep angle (small arrows), leaving reticulate 
arrays. In most cases the pad can exit its depression without 
contacting the walls. Pad deformation has been exaggerated 
for clarity. 

sion can be inferred. The viscoelastic paw 
pads of living mammals have been found 
to reduce impact forces (Alexander et al., 
1986); avian digital pads also deform in 
early stance (Gatesy, personal observa- 

In Triassic theropods, the profile of dig- 
ital pad depressions in Greenlandic tracks 
is direct evidence of pad flattening. Some 
prints have almost horizontal floors, de- 
spite having entry tracts with a fluted, U- 
shaped profile. Such impressions are best 

explained by pad deformation, rather than 
toes with a flat plantar surface (Lockley 
and Hunt, 1994: 38). As a pad compressed, 
new skin around the periphery would con- 
tact the substrate. This increase in depres- 
sion diameter could partially or even com- 
pletely obliterate entry tracts and striations 
made earlier in stance (Figs. 5b, c). In 
tracks retaining such traces of entry, the 
pads must have plowed forward far 
enough to escape mid-stance overprinting. 
Thus, pad deformation may explain the 
relative size of entry tracts from digits II 
and IV and their absence in some prints. 
Movements of the proximal pad of digit 
III, for which an entry tract has never 
been found, are difficult to infer. One pos- 
sibility is that this pad made an entry tract 
that it subsequently oveiprinted during 
mid-stance deformation. Alternatively, the 
proximal pad of digit III may not have 
contacted the substrate in early stance. 
The pad could have descended and de- 
formed only after forward motion of the 
foot was complete. The toe base region, 
which also never shows signs of oblique 
entry, likely delayed contact as well. Thus, 
mid stance was a period during which skin 
contact was maximized, leading to the de- 
stioiction of earlier sldn impressions rather 
than the creation of new ones. 

Late Stance: Reticulate Arrays, Exit 
Striations, and Backward Scrapes 

During the second half of stance, the 
limb's force against the ground decreased 
and was directed down and backwards. In 
late stance ("Idck-off of Thulbom and 
Wade, 1989), the foot was withdrawn, 
thereby breaking sldn— sediment interfaces 
and creating all skin impressions other 
than entry striations. Three main types of 
imprints were left behind: reticulate ar- 
rays, exit striations, and backward slips. 

Reticulate arrays were exposed on the 
walls and floors of the digital pad depres- 
sions. In the formation of most tracks in 
the sample, regions of skin moved up and 
away from the indented sediment at an an- 
gle steep enough to prevent striations. 

146 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


Figure 6. Summary stereo renderings of the tliree-dimensional orientation of entry and exit striations for 17 pads from nine 
different specimens. Vectors represent the direction of the striations only, not their length, and have been placed around a 
diagrammatic track outline adjacent to their appropriate pads. Entry striations are roughly parallel with the long axis of the track 
and slope downwards at a low angle, indicating that the pads of digits II and IV mostly slid forward after contact. In contrast, 
exit striations created by the middle and distal pads of digit III have a very small forward component; most motion was upward 
upon toe withdrawal. 

This scenario is reasonable for the central 
regions of the floor of each pad impres- 
sion, but what about the peripheral areas 
and walls? Could skin have moved normal 
to all surfaces simultaneously? One possi- 
ble explanation is pad recoil. If a pad was 
deformed in mid stance, a decreased load 

would allow pad soft tissues to rebound, 
thereby vacating the walls and outer por- 
tions of its depressions floor first (Fig. 5d). 
Skin abutting the center of the floor of the 
pad depression would rise slightly later. 
Such a mechanism could explain the ab- 
sence of exit striations for most pads. Re- 

Theropod Skin Impressions • Gatesy 147 

coil would reduce a pad's transverse di- bom, 1982), or interpreted trackways as 

ameter enough to allow a clean withdrawal evidence of limb posture (e.g., Lockley 

\\dthout wall contact. and Hunt, 1995; Padian and Olsen, 1989). 

Not all pads avoided scraping the walls A wealth of locomotor information has 

of their depression upon withdrawal. In been gleaned from the two-dimensional 

some relatively deep tracks, pads of digit position of prints with respect to one an- 

III intersected with sediment, creating other (stride length, step angle, trackway 

striation-bearing exit tracts. The three-di- width, and toed-in versus toed-out), but 

mensional orientation of exit striations is little emphasis has been given to the mor- 

based on a small sample (Fig. 6), but some phology of single prints. One exception is 

patterns can be discerned. The distal and Thulbom and Wade (1984, 1989), who 

middle pads of digit III left relatively ver- were able to explain much of the variation 

tical striations upon withdrawal at the very in their large sample by relating specific 

end of stance (Figs. Id, 2d, 6). When mea- features of a track to events during the 

sured in the horizontal plane, skin move- stance phase. Walking dynamics have also 

ment was primarily transverse with respect been inferred from subsurface sediment 

to the long axis of the print. These trajec- deformation (Avanzini, 1998). Most re- 

tories are consistent with digit III being cently, deep tracks have been shown to 

lifted up and out of its depression rather preserve the three-dimensional foot move- 

than being dragged forward (Fig. 6). How- ments of Late Triassic theropods from 

ever, neither medial nor lateral exit stria- Greenland (Gatesy et al., 1999). One con- 

tions predominate; both are present in two cern with these data is the possible effects 

specimens. In most cases the third toe of sinking on locomotion. Did theropods 

seems to have been removed without sig- walk differently on soft and firm sub- 

nificant transverse deviation. strates? 

At least two prints show evidence of Data from skin impressions may help 

backward movement of toes before liftoff, answer this question, because they are 

Such slippage occurred when the down found in relatively shallow tracks. Evi- 

and backward limb force overcame friction dence of pad movement during locomo- 

between sldn and sediment. In one case tion on a firm substrate can be compared 

(L.06) slipping produced an elongate claw to foot trajectories preserved in deep 

scrape ("retro-scratches" of Thulborn and prints. The effects of sinking on stance 

Wade, 1989) as well as scale-induced stri- phase movements, if any, can then be as- 

ations. These marks converge posteriorly, sessed in order to discern locomotor func- 

indicating adduction of digits II-IV upon tion under a variety of substrate condi- 

flexion of the metatarsophalangeal joints tions. Results from this analysis of sldn im- 

(Baird, 1957; Gatesy et al., 1999). " pressions of 20 Greenlandic tracks must be 

considered preliminaiy, but they contrib- 

Previous Work and Future Directions ^^^ to ^ more detailed documentation of 

Study of vertebrate tracks has increased foot movements in basal theropods. Pedal 

dramatically in the past two decades (e.g., function may have been very similar to 

Farlow and Chapman, 1997; Gillette and that seen in living ground-dwelling birds, 

Lockley 1989; Lockley 1991, 1997, 1998; but this hypothesis requires testing and 

Thulborn, 1990). Along with this revival verification. 

has been a heightened awareness of the Clearly, much more work is needed, 

contribution footprints can make to un- Trackways with skin impressions would be 

derstanding dinosaur locomotion. Many particularly informative about the effects 

workers have applied Alexanders (1976) of speed, turning, and preservational vari- 

method to calculate speed (e.g., Farlow, ation. A broader suivey of theropod tracks 

1981; Russell and Belland, 1976; Thul- with skin impressions from other locafities 

148 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

and ages is sorely needed. Analyses of fos- 
sils will also benefit from a better under- 
standing of track formation in living ani- 
mals, particularly birds. Surprisingly, a 
complete description of avian foot motion 
during terrestrial locomotion is unavailable 
(for preliminary accounts see Cracraft, 
1971; Gatesy, 1999; Gatesy et al., 1999). 
Techniques such as high-speed film or vid- 
eo should allow features such as tracts and 
striations to be related directly to skin mo- 
tion. Simple manipulations may elucidate 
variation in reticulate patterns and permit 
a more precise interpretation of reticulate 
arrays in fossil tracks. Extant forms also al- 
low the magnitude, direction, and position 
of the limb's force against the substrate to 
be measured (e.g., Roberts, 2001) and re- 
lated to foot movement. Finally, a more 
sophisticated analysis of substrate proper- 
ties and their effects on track formation 
dynamics (e.g., Allen, 1997) is needed for 
dinosaurs. In time, a combination of such 
approaches may delineate both general 
features of theropod locomotion as well as 
more specific details of its evolutionary 


Footprints were collected as part of a 
joint Harvard University and University of 
Copenhagen expedition, with support 
from the National Science Foundation, the 
Carlsberg Foundation, and the Putnam 
Expeditionary Fund of Harvard University. 
I particularly thank F. A. Jenkins, Jr., N. 
H. Shubin, W. W. Amaral, K. M. Middle- 
ton, and other members of the field crew 
for their help, and J. O. Farlow for com- 
ments and advice. Sldn impressions in Fig- 
ure 1 were drawn by Peggy Price. This pa- 
per is dedicated to my advisor. Fuzz 
Crompton, who taught me that variation is 
often information, not just noise to be av- 
eraged out. 


Alexander, R. McN. 1976. Estimates of speeds of 

dinosaurs. Nature, 261: 129-130. 
. 1977. Mechanics and scaling of terrestrial lo- 

comotion, pp. 93-110. In T. J. Pedley (ed.). Scale 
Effects in Animal Locomotion. London: Aca- 
demic Press. XX + 545 pp. 

Alexander, R. McN., M. B. Bennett, and R. F. 
Ker. 1986. Mechanical properties and function 
of the paw pads of some mammals. Journal of 
Zoology, London, 209: 405-419. 

Allen, J. R. L. 1997. Subfossil mammalian tracks 
(Flandrian) in the Severn Estuary, S. W. Britain: 
mechanics of formation, preservation and distri- 
bution. Philosophical Transactions of the Royal 
Society of London B, Biological Sciences, 352: 

AVANZINI, M. 1998. Anatomy of a footprint: biotur- 
bation as a key to understanding dinosaur walk 
dynamics. Ichnos, 6: 129—139. 

Baird, D. 1957. Triassic reptile footprint faunules 
from Milford, New Jersey. Bulletin of the Mu- 
seum of Comparative Zoology, 117: 449—520. 

Clemmensen, L. B., D. V. Kent, and F. A. Jenkins, 
Jr. 1998. A Late Triassic lake system in East 
Greenland: facies, depositional cycles and pa- 
laeoclimate. Palaeogeography, Palaeoclimatology, 
Palaeoecology, 140: 135-159. 

Cracraft, J. 1971. The functional morphology of the 
hind limb of the domestic pigeon, Columbo livia. 
Bulletin of the American Museum of Natural 
History, 144: 171-268. 

Currie, P. J., G. C. Nadon, and M. G. Lockley. 
1991. Dinosaur footprints with sldn impressions 
from the Cretaceous of Alberta and Colorado. 
Canadian Journal of Earth Sciences, 28: 102- 

Farlow, J. O. 1981. Estimates of dinosaur speeds 
from a new trackway site in Texas. Nature, 294: 

Farlow, J. O., and R. E. Chapman. 1997. The sci- 
entific study of dinosaur footprints, pp. 519-553. 
In J. O. Farlow and M. K. Brett-Surman (eds.). 
The Complete Dinosaur. Bloomington, Indiana: 
Indiana University Press, xi -I- 752 pp. 

Farlow, J. O., and M. G. Lockley. 1993. An os- 
teometric approach to the identification of the 
makers of early Mesozoic tridactyl dinosaur foot- 
prints, pp. 123-131. In S. G. Lucas and M. Mo- 
rales (eds.). The Nonmarine Triassic. New Mex- 
ico Museum of Natural History and Science Bul- 
letin, 3. 478 pp. 

Gatesy, S. M. 1999. Guineafowl hind limb function 
1: cineradiographic analysis and speed effects. 
Journal of Morphology, 240: 127-142. 

Gatesy, S. M., K. M. Middleton, F. A. Jenkins, 
Jr., and N. H. Shubin. 1999. Three-dimensional 
preservation of foot movements in Triassic the- 
ropod dinosaurs. Nature, 399: 141-144. 

Gillette, D. D., and M. G. Lockley. 1989. Di- 
nosaur Tracks and Traces. Cambridge, United 
Kingdom: Cambridge University Press, xvii -I- 
454 pp. 

Hitchcock, E. 1858. Ichnology of New England: A 
Report on the Sandstone of the Connecticut Val- 

Theropod Skin Impressions • Gatesy 149 

ley. Especially its Fossil Footmarks, Made to the 
Government of the Commonwealth of Massa- 
chusetts. Boston, Massachusetts: W. White, xii + 
220 pp. 

Jenkins, F. A., Jr., N. H. Shubin, W. W. Amaral, S. 
M. Gatesy, C. R. Schaff, L. B. Clemmensen, 
W. R. Downs, A. R. Damdson, N. Bonde, and 
F. OSBAECK. 1994. Late Triassic continental ver- 
tebrates and depositional environments of the 
Fleming Fjord Formation, Jameson Land, East 
Greenland. Meddelelser om Gr0nland, Geosci- 
ence, 32: 1-25. 

LanGSTON, W. 1986. Stacked dinosaur tracks from 
the Lower Cretaceous of Texas — a caution for 
ichnologists, p. 18. In D. D. Gillette (ed.). First 
International S^inposium on Dinosaur Tracks 
and Traces, Abstracts with Program. Albuquer- 
que, New Mexico: New Mexico Museum of Nat- 
ural History. 31 pp. 

LOCKLEY, M. G. 1989. Summary and prospectus, pp. 
441^47. In D. D. Gillette and M. G. Lockley 
(eds.). Dinosaur Tracks and Traces. Cambridge, 
United Kingdom: Cambridge University Press, 
xvii -I- 454 pp. 

. 1991. Tracking Dinosaurs: A New Look at 

our Ancient World. Cambridge, United King- 
dom: Cambridge University Press, xii -I- 238 pp. 

. 1997. The paleoecological and paleoenviron- 

mental utility of dinosaur tracks, pp. 554—578. In 
J. O. Farlow and M. K. Brett-Surman (eds.). The 
Complete Dinosaur. Bloomington, Indiana: In- 
diana University Press, xi + 752 pp. 

. 1998. The vertebrate track record. Nature, 

of North America. Memoirs of the Boston Soci- 
et>' of Natural History, 1904: 461-557. 

1953. Triassic life of the Connecticut V'alley. 

396: 429-432. 

LoCKLEY, M. G., AND A. P. HUNT. 1994. Fossil foot- 
prints of the Dinosaur Ridge area. Publication of 
the Friends of Dinosaur Ridge and the Univer- 
sity of Colorado at Denver Dinosaur Trackers 
Research Group, 2: i-53. 

. 1995. Ceratopsid tracks and associated ich- 

nofauna from the Laramie Formation (Upper 
Cretaceous: Maastrichtian) of Colorado. Journal 
of Vertebrate Paleontology, 15: 592-614. 

Lucas, A. M., and P. R. Stettenheim. 1972. Avian 
Anatomy: Integument. 2 vols. Agriculture Hand- 
book 362. Washington, D.C.: U.S. Department 
of Agriculture. 750 pp. 

Lull, R. S. 1904. Fossil footprints of the Jura-Trias 

Connecticut State Geological and Natural His- 
tory Survey, Bulletin, 81: 1-336. 

Olsen, p. E., and p. M. Galton. 1984. A review of 
the reptile and amphibian assemblages from the 
Stormberg of southern Africa, with special em- 
phasis on the footprints and the age of the 
Stormberg. Palaeontologia Africana, 25: 87-110. 

Olsen, P. E., J. B. Smith, and N. G. McDonald. 
1998. Type material of the type species of the 
classic theropod footprint genera Eubrontes, An- 
chisauripus, and GroUator (Early Jurassic, Hart- 
ford and Deerfield Basins, Connecticut and Mas- 
sachusetts, U.S.A.). Journal of Vertebrate Pale- 
ontology, 18: 586-601. 

Padian, K., and p. E. Olsen. 1989. Ratite footprints 
and the stance and gait of Mesozoic theropods, 
pp. 231-241. In D. D. Gillette and M. G. Lock- 
ley (eds.). Dinosaur Tracks and Traces. Cam- 
bridge, United Kingdom: Cambridge University 
Press, xvii + 454 pp. 

Roberts, T. J. 2001. Muscle force and stress during 
running in dogs and wild turkeys. Bulletin of the 
Museum of Comparative Zoology, 156: 283-295. 

Russell, D. A., and P. Belland. 1976. Running 
dinosaurs. Nature, 264: 486. 

Thulborn, R. a. 1982. Speeds and gaits of dino- 
saurs. Palaeogeography, Palaeoclimatology, Pa- 
laeoecology, 38: 227-256. 

. 1990. Dinosaur Tracks. London: Chapman 

and Hall, x-vii + 410 pp. 

Thulborn, R. A., and M. Wade. 1984. Dinosaur 
trackways in the Winton Formation (mid-Creta- 
ceous) of Queensland. Memoirs of the Queens- 
land Museum, 21: 413-517. 

. 1989. A footprint as a history of movement, 

pp. 51-56. In D. D. Gillette and M. G. Lockley 
(eds.). Dinosaur Tracks and Traces. Cambridge, 
United Kingdom: Cambridge University Press, 
xvii -I- 454 pp. 

WOODHAMS, K. E., AND J. S. HiNES. 1989. Dinosaur 
footprints from the Lower Cretaceous of East 
Sussex, England, pp. 301-307. In D. D. Gillette 
and M. G. Lockley (eds.). Dinosaur Tracks and 
Traces. Cambridge, United Kingdom: Cam- 
bridge University Press, x-vii -I- 454 pp. 



Abstract. A diminutive eudimorphodontid ptero- 
saur, from the Late Triassic Fleming Fjord Formation 
of East Greenland, possesses relatively short wings, 
short ulnae and tibiae, and long metatarsals. The new 
species, smaller than any known individual of Eudi- 
morphodon, is unique among known pterosaurs in 
having pro.ximal limb segments (humerus, ulna, fe- 
mur, tilaia) of nearly equal length. Although the 
Greenlandic pterosaur is probably a juvenile, as in- 
dicated primarily by the lack of synostosis of axial and 
limb girdle components, the appendicular propor- 
tions of the specimen are too different from those in 
other knowai pterosaurian taxa to be accounted for 
solely by immaturit)'. The bicondylar fourth metacar- 
pophalangeal joint, in which the dorsal condyle has a 
larger radius of curvature and a more extensive artic- 
ular surface than the ventral condyle, appears to be 
intermediate between a primitive unicondylar joint 
and the asymmetric trochlea common among ptero- 
saurs. This spectrum of joint configurations repre- 
sents increasing mechanical stability, consonant with 
the interpretation that the mechanism evolved among 
basal pterosaurs to accommodate wing folding during 
the upstroke in flapping flight. 


The earliest well-documented records 
of pterosaurs are from Late Triassic (No- 
rian) deposits in Italy. Eiidimorphodon 
ranzii, first described from a single, nearly 
coinplete skeleton froin the Zorzino lime- 
stones (Middle to Upper Norian) near 

' Department of Organismic and Evolutionary Bi- 
ologv, and Museum of Comparative Zoology, Harvard 
Universit)', Cambridge, Massachusetts 02138. 

- Department of Organismal Biology and Anatomy, 
University of Chicago, Chicago, Illinois 60637. 

^ Department of Ecolog)- and Exolutionar)' Biology, 
Brown University, Providence, Rhode Island 02912. 

^ Department of Integrative Biology, and Museum 
of Paleontology, University of California, Berkeley, 
Cahfomia 94720. 

Bergamo (Zambelli, 1973), is now known 
from four additional specimens, including 
juveniles (Wild, 1978, 1994). Another 
specimen, designated as a different species 
(E. rosenfeldi), derives from the lower part 
of the Dolomia di Fomi (Middle Norian) 
in Udine Province (Dalla Vecchia, 1995). 
Other contemporaneous taxa from Berga- 
mo Province include Peteinosaunis zam- 
bellii (represented by two specimens; 
Wild, 1978), also from the Zorzino lime- 
stones, and Preondactijlus huffarinii 
(known from a single specimen; Wild, 
1984; Dalla Vecchia, 1998) from the Do- 
lomia di Fomi. A compacted assemblage 
of pterosaur bones, interpreted as a gastric 
pellet, was referred to P. hujfarinii by Dal- 
la Vecchia et al. (1989) principally on the 
basis of estimated limb length ratios. The 
specimen derives from a fossiliferous, 
Middle Norian zone in the Dolomia di 
Fomi (Roghi et al., 1995), 150-200 m low- 
er in the section that yielded tlie type of 
P. bujfarinii (Dalla Vecchia et al, 1989). 

Some pterosaurs of the Late Triassic al- 
ready had attained moderate size. Padian 
(1980) described a partial wing skeleton 
from a pterosaur with a wingspan of 1.5 m 
that he recognized was neither Eiidimor- 
phodon nor Peteinosaunis. Wild (1984), in 
his description of the type of Preondactij- 
lus bujfarinii, referred the wing skeleton 
to this taxon on the basis of phalangeal 
proportions, although the type of P. biif- 
farinii is smaller, with a wingspan estimat- 
ed at 45 cm (Wellnhofer, 1991) or "a little 
less than 50 cm" (Dalla Vecchia, 1998: 

Bull. Mus. Comp. Zool., 156(1): 151-170, October, 2001 151 

152 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

358). Adult Eiidimorphoclon ranzii had a 
wingspan of 1 m (Wild, 1978). 

Here we describe a tiny pterosaur, 
which we interpret as a young individual 
of a new species of Eiidirnorphodon, with 
an estimated 24-cm wingspan, from the 
Late Triassic Fleming Fjord Formation, 
Jameson Land, East Greenland (Jenkins et 
al, 1994). 

The following abbreviations of institu- 
tional names are used: MCSNB, Museo 
Civico di Scienze Naturali, Bergamo; 
MCZ, Museum of Comparative Zoology, 
Harvard University, Cambridge, Massa- 
chusetts; MFSN, Museo Friulano di Storia 
Naturale, Udine; and MGUH, Geological 
Museum, University of Copenhagen. 


Class Reptilia Laurent!, 1768 

Subclass Archosauria Cope, 1869 

Order Pterosauria Kaup, 1834 

Family Eudimorphodontidae Wellnhofer, 

Genus Eudimorphodon ZambeWi, 1973 

Eudimorphodon cromptonellus new 

Holotype. MGUH VP 3393 (MCZ field 
no. 13/91G; Fig. 1). 

Etymology. The specific epithet coin- 
bines a patronym honoring A. W. Cromp- 
ton for his contributions to our under- 
standing of the evolution of vertebrate 
structure and function, with Latin -elliis, 
diminutive in reference to the small size 
of the type. 

Diagnosis. A very small pterosaur that 
shares with Eudimorphodon ranzii (as di- 
agnosed by Wild, 1978: 179, and later 
emended by Wild, 1994: 97-102) a het- 
erodont dentition composed of uni-, tri-, 
and quinticuspid teeth (vinknown in any 
other pterosaurian taxon); additionally, 
some teeth are quadricuspid. Eudimor- 
phodon rosenfeldi (Dalla Vecchia, 1995) 
also has quinticuspid teeth but the denti- 
tion is incompletely known. The tooth 

count is estimated at 11 or 12 postpre- 
maxillary teeth, which is 14 or 15 fewer 
than the type of E. ranzii and three or four 
fewer than the number reconstructed for 
the Milano juvenile specimen of E. ranzii 
(Wild, 1978, figs. 25, 27). The tibia is rel- 
atively shorter than in any known ptero- 
saur (104% of the femur); the ulna is rel- 
atively shorter than in most other ptero- 
saurs (111% of the humerus; some speci- 
mens of Canipylognathoides zitteli have 
coinparable humero-ulnar proportions, see 
Table 2). As in the Milano juvenile, no ev- 
idence is found of the two enlarged, fang- 
like maxillary teeth situated beneath the 
ascending process that are characteristic of 
adult E. ranzii. Differs from the juvenile 
specimen (MCSNB 8950) referred to E. 
ranzii by Wild (1994) in that the metatar- 
sals are approximately 25% longer, where- 
as all other limb bone lengths are substan- 
tially shorter. The new taxon differs from 
E. rosenfeldi (Dalla Vecchia, 1995) in hav- 
ing a huinerus that is shorter than the fe- 


Horizon and Locality. Lower part of the 
Carlsberg Fjord beds in the 0rsted Dal 
Member of the Fleming Fjord Formation, 
Scoresby Land Group, Jameson Land, East 
Greenland. On the southern flank of Mack- 
night Bjerg (Geodsetisk Instituut, Gr0nland 
series 1:250,000, Carlsberg Fjord Quadran- 
gle), a localized bone assemblage was dis- 
covered in 1989 by William W. Amaral at 
71°22.277'N, 22°33.341'W (the Macknight 
Bjerg quarry; the latitude and longitude co- 
ordinates given here, taken in 1995 by av- 
eraging multiple global positioning system 
readings to reduce anomalies due to drift, 
differ slightly from those reported by Jen- 
kins et al. (1994) based on a single reading 
from a hovering helicopter). Excavated in 
1991 and 1992, this locality presented a 
rich taphocoenosis of vertebrate remains, 
predominantly the plagiosaurine Gerro- 
thorax cf. pulcherrimus but also the capi- 
tosaur Cyclotosaurus cf posthumus (Jen- 
kins et al., 1994). In addition to the ptero- 
saur, the only otlier associated skeletal ma- 
terial of a terrestrial tetrapod found at this 

Triassic Pterosaur From Greenland • Jenkins et al. 153 

Figure 1 . Stereophotographs of Eudimorphodon cromptonellus (MGUH VP 3393), new species, preserved in two matrix blocks 
(A, see Fig. 3 for details; B, see Fig. 2). 

154 Bulletin Museuin of Comparative Zoology, Vol. 156, No. 1 

site was that of a Plepidosauromoiph rep- 
resented by a partial postcranial skeleton 
and lower jaws. 

Age. Late Triassic (PNorian-Rhaetian) 
(Jenkins et al., 1994; Clemmensen et al., 

Material. A fairly complete but largely 
disarticulated and partly crushed skeleton. 
Identifiable cranial bones include both 
mandibles and maxillae, as well as a nasal, 
lacrimal, jugal, quadrate, and squamosal. 
Other cranial elements are too damaged to 
offer a basis for useful description. Post- 
cranial remains include numerous ceivical 
and dorsal vertebrae (most neural arches 
are disassociated from centra), several cau- 
dal vertebrae, and rib fragments. Appen- 
dicular elements include the right scapula; 
a partial coracoid; the right humerus, ra- 
dius, ulna, fourth metacarpal, and wing 
phalanges; both femora, a tibia, and a fib- 
ula; metatarsals; and numerous pedal pha- 

Comments. The specimen was discov- 
ered in the process of splitting coarsely 
bedded matrix in the Macknight Bjerg 
quarry; parts of the skeleton are thus pre- 
served on part and counterpart blocks 
(Figs. 1—3). Postmortem tissue maceration 
resulted in disarticulation of most of the 
bones, but transport was minimal and 
some natural associations are preserved 
(skull, cervical vertebrae, right manus, 
right hind limb). 



Maxilla. The right maxilla (Figs. 2, 4), 
largely complete except for some damage 
to its rostral and caudal ends, is 13.5 mm 
in length. A posteriorly recurved ascending 
process is preseived, which in Eudimor- 
phodon ranzii separates the antorbital fe- 
nestra from the external naris (Wild, 1978, 
fig. 1). The maxilla bears 11 teeth with an 
apparent diastema between the third and 
fourth. The diastema, which is situated an- 
teroventral to the ascending process of the 
maxilla approximately in the locus of the 

enlarged fanglike teeth of adult E. ranzii 
(Wild, 1978, fig. 25b), shows no evidence 
of alveoli. The last seven teeth are poste- 
rior to the ascending process of the max- 
illa. The left maxilla (in medial aspect, Fig. 
2), partly overlain and obscured by the 
right maxilla, is fractured and deformed 
but is complete posteriorly where it lies in 
contact with the jugal. Nine teeth are pre- 
sent, but most are incompletely preserved; 
a gap (Pdiastema) between the fourth (in 
the process of eruption) and fifth is suffi- 
cient to have accommodated three tooth 

Other Cranial Bones. Most cranial 
bones are obscured by postmortem col- 
lapse, crushing, and disarticulation of the 
skull. However, a few can be identified, 
but offer little detail that warrants further 
description beyond that illustrated (Figs. 
2, 3). Crushed bone superorostral to the 
anterior ends of the maxillae represents 
part of the right nasal, and possibly the 
posterior part of the premaxilla. No evi- 
dence is found of premaxillaiy teeth. The 
slender right lacrimal lies behind the as- 
cending process of the maxilla. The jugal 
is represented by a postorbital process. 
Both the right squamosal and quadrate lie 
separate from the skull. The squamosal has 
a deep notch that represents the superior 
border of the inferior temporal fenestra. 
The quadrate bears a large, bulbous artic- 
ular condyle. 

Dentition. The teeth are buccolingually 
narrow and vary in mesiodistal length from 
0.42 to 1 mm. The relatively simple, uni- 
cuspid mesial teeth, exemplified by the 
most mesial tooth presei'ved in the right 
maxilla (Fig. 4) and left mandible (Fig. 2), 
are the smallest (0.42, 0.48 mm in length, 
respectively); the mesial crest that de- 
scends from the apical cusp is slightly 
more convex than the distal crest. The re- 
mainder of the dentition comprises mul- 
ticusped teeth sti*ucturally similar to those 
oi EtidimorpJiodon ranzii (Wild, 1978) but 
unlike that in any other known pterosaur. 
The enamel is smooth and without surfi- 
cial grooves; Wild (1978, fig. 28) regarded 

Triassic Pterosaur From Greenland 'Jenkins et al. 155 


CO .9 Q) 

.2 o § 
3) OJ o 

^ - (D 

< -^ o 

X . 

^ C (0 
CO (0 > 

LL Q. 2 

-^ g o 

^ o ^ 

O U5 -^ 

J3 >s C 

X Q. O 

i: TO . 

?^ s 

t c . _ 
O j£ P 

§^ d 

O . CD 
-n C\J 

CD -C Q. 

^ Q.5 

QJ CO . . 

E "o > 

- CD *- 

-5 iS Q. 

C Q) ^ 

CD C o 

g (0 ra 


TO y <D 

;z ^ _ 

ro.'^ iS 

— Cl U) 

«■ O t3 

-^- ti^" 
^ O CD 

-- i 


is S ^ 


OJ o •= 
.2, -> 

CO CD ^ 


-Q 3 _ 



CD lT gj 
CO it- > 


"5 O 2 
CO — CD 

J2 O 

<"' t: <^ 

CO c c 

^ = =3 

Q) O C 
C _ CD 
^ c " 

^ r O 
" 2.£ 

> ^ . 

i2 "5 ,n 

5 O) SI 

I -r 

o .y ° 

O f« CD 
r- CD -*-- 

y- CD =J 
O Q. O 

UJ CD "= 

H- .c t: 

O *" CD 

^ 6 °- 

O ^ (D 
^ TD £ 

(D c CU 

£•- £ 

o S SS 

Q. E 0). 

^ CD 
C\i L.- ^ 

=> .y 5 

CD iS c 

C 3 CD 

^ o C 

O y 13 

^ CD . 

- _ Q. 

i: CD 3 

i5 5 « 

3 C C 

-Q . 3 

= CD . 

.- ^ ^ 

3 ^.5 

p CD -9 

5 C *- 

to J t 


- E ^ 

c= CO 

1 x^- 

CO __, 

-O CO g 

> 5 E 
"c ro CD 

co" oj 5- 
.^ E w 


"- Q. CO 

5"CD CJ 
J3 3 CO 

<" £ ■- 
■C <D CO 

(D E 3 

> -n 

. Q. 

s ^ 

o . . 



■C M- 


^ CO 


O- (0 

cr Q. 

156 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 







































CO .5 


Triassic Pterosaur From Greenland 'Jenkins et al. 157 

the textural difference in enamel among possibly conceals a second distal accessory 
specimens of E. ranzii as ontogenetic cuspule at the base of the crown. The 
(adults possess grooves that run apicobas- fourth tooth, in the process of eruption, 
ally; juveniles lack them). Accessory cus- displays three cusps, but the basal mesial 
pules are developed along the mesial and and distal cuspules, if present, would be 
distal crests descending from the primary, obscured. The fifth tooth has two acces- 
or apical, cusp; comparable to the vari- sory cuspules mesially, but only one dis- 
ability illustrated by Wild (1978, fig. 7) for tally. Of the six most distal teeth, all appear 
the Milano juvenile specimen of E. ranzii, to be quinticuspid, except for the relatively 
teeth may bear a single accessory cuspule small eighth tooth (mesiodistal length, 
(along the distal crest), or two, three, or 0.45 mm), which appears to be tricuspid, 
four cuspules. Mesial teeth (ranging in The 11 teeth of the right maxilla (Fig. 
mesiodistal length from 0.54 to 0.78 mm) 4) are the best preserved of the entire den- 
tend to be tricuspid, with the accessory tition. A diastema between the third and 
cuspules situated at the mesial and distal fourth teeth is evidence that the maxillary 
base of the crown. Teeth in the distal part tooth count could have been 12 or more, 
of the row (ranging in mesiodistal length The most mesial tooth is unicuspid and 
from 0.83 to 1.08 mm) tend to be quinti- relatively small (mesiodistal length, 0.42 
cuspid. Accessory cuspules, particularly mm); the second also appears to be uni- 
the basal ones, tend to be oriented in pal- cuspid, but the third is clearly tricuspid 
mate fashion, splaying from the central (respective lengths, 0.54, 0.72 mm). The 
(apical) axis; in adult E. ranzii, cuspules remaining eight teeth (4th-llth) vary in 
either parallel the central axis or converge mesiodistal lengths from 0.8 to 1 mm, with 
slightly (Wild, 1978, fig. 8). One tricuspid the exception of the most distal tooth (0.6 
tooth illustrated by Wild (1978, fig. 7) of a mm). The fourth and very probably the 
juvenile E. ranzii (the Milano specimen) fifth are quinticuspid. However, the sixth 
exliibits a similar splaying of accessory cus- and seventh are quadricuspid, with a sin- 
pules, gle accessory cuspule on the distal crest of 

Eleven teeth are present in the left the sixth and mesial crest of the seventh, 

mandible (nine shown in Fig. 3; the two The 9th is tricvispid, the 10th quinticuspid, 

most distal in Fig. 2). The most mesial, and the most distal a small tricuspid. As in 

unicuspid tooth is followed by a bicuspid the smaller, Milano juvenile specimen of 

(with a minute accessory cuspule on the Eudimorphodon ranzii (Wild, 1978, figs, 

distal crest). Most of the remaining left 25, 27), no evidence is found of the two 

mandibular teeth, insofar as preserved, ap- enlarged, fanglike maxillary teeth situated 

pear to be tricuspid, with the exception of beneath the ascending process that are 

the penultimate, which is quinticuspid. characteristic of adult £. ranzii. Few of the 

The 11 teeth of the right mandible, bet- 10 left maxillary teeth preserve any details 
ter preserved than those of the left, all ex- of the crowns; the fifth, sixth, and seventh 
hibit three or more accessory cuspules; the are certainly quinticuspid. 
mesial dentition appears to be unrepre- A definitive tooth count cannot be as- 
sented because there are no uni- or bicus- certained because of postmortem damage; 
pid teeth. The most mesial tooth bears two neither the premaxillary teeth, nor the 
accessory cuspules mesially, and at least fanglike, mesialmost teeth of the lower jaw 
one distally (the basal part of the crown, known in presumably ontogenetically old- 
where a second distal cuspule would be ex s^eciVLxens o{ Eudimorphodon ranzii dLve 
positioned, is obscured by the next over- preserved. Nonetheless, the tooth counts 
lying tooth). The second tooth is tricuspid, in both maxillae and both mandibles are 
The third tooth is at least quadricuspid; sufficiently comparable to estimate 11 or 
the obliquity of its position in the alveolus 12 postpremaxillary teeth, two to three 

158 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

1 mm 

Figure 4. Lateral view of the right maxilla of Eudimorphodon cromptonellus (MGUH VP 3393), new species. 

fewer than the number reconstructed for 
the Milano juvenile (Wild, 1978, figs. 25, 


Postcranial Skeleton 

Vertebrae. Approximately six cervical, 
eight dorsal, one Psacral, and one (possibly 
two) caudal vertebrae are represented. 
However, postmortem disturbance consid- 
erably obscures vertebral details. Although 
some vertebrae are more or less associated 
in a series, inost are disarticulated from 
one another. Furthermore, in most cases 
the neural arches and centra themselves 
are disarticulated; the only complete ver- 
tebra is an elongate midcaudal (Fig. 3; 
centrum length 5.7 min; overall length, 7.6 
mm). Reconstruction is further hampered 
by the overlap of bones that have collapsed 
one on the other. Of the eight isolated cen- 
tra that appear undistorted and are suffi- 
ciently exposed for measurement, all are 
of uniform length (1.7—1.75 mm). The 
most complete neural arch (Fig. 3), which 
is isolated, is 1.6 mm in length (including 

the small pre- and postzygapophyses), 
lacks transverse processes, has a neural ca- 
nal width of 1.25 inin, and has a spinous 
process height of 0.8 inm. The arch's lack 
of transverse processes and relatively re- 
stricted neural canal, considered with its 
proximity to the hind foot and a midcaudal 
vertebra, are evidence that the element is 
derived from the proximal caudal series. 
Two neural arches (also separated from 
their respective centra; Fig. 2) are inter- 
preted as representing dorsal vertebrae by 
virtue of their association with elements of 
the shoulder girdle and the presence of 
transverse processes (approximately 1 mm 
in length) that project horizontally from 
the jvmction of the laminae and pedicles. 
The breadth of the neural canal is 1.5 mm 
in the smaller, and 2.25 mm in the larger 
specimen; likewise, the respective distanc- 
es between the distal ends of the trans- 
verse processes are 4.1 and 4.75 mm. 
However, these ineasurements are only 
approxiinate because of slight postmortem 

Triassic Pterosaur From Greenland 'Jenkins et al. 159 

displacement between the left and right 
halves of the arches. 

Shoulder Girdle. A dissociated scapula 
and partial coracoid are the only shoulder 
elements preserved. The evidence of a 
sternum is equivocal. A comminuted, 
sheetlike expanse of bone associated with 
the humeral head (?st. Fig. 3) may be ster- 
nal, and at one end is a process that re- 
sembles a cristospine. The coracoid (pre- 
sented in medial aspect. Fig. 2) is 6.5 mm 
long as preserved, but the sternal end has 
disintegrated beneath vertebral remains. 
The acrocoracoid process is partly \asible. 
The scapula (in dorsomedial aspect. Fig. 
2) is 12.6 mm in length as presei-ved; deg- 
radation of the caudal end of the blade 
raises the likelihood of somewhat greater 
length (at least 13 mm). 

Forelimb. The right humerus, 18.2 mm 
long, has a slightly sigmoidal, dorsoventral 
curvature; the proximal end as a whole is 
reflected dorsally, and the distal end is re- 
flected ventrally. The deltopectoral crest is 
subtriangular, comparable to that in juve- 
nfle Eudimorphodon ranzii but unlike the 
shape in adults, which is quadrangular 
(Wfld, 1978, fig. 29). The humeral head is 
oriented dorsally, and possesses the typi- 
cally pterosaurian sellar shape (Padian, 
1983). The diaphysis at midshaft, which is 
slightly flattened, is 1.4 mm in width. The 
distal end, 3.3 mm in width, lacks cortical 
bone; radial and ulnar condyles are not 

The right radius and ulna (Fig. 3; esti- 
mated lengths, 19.5 and 20.1 mm, respec- 
tively) lie approximately parallel to each 
other. The proximal shaft of the ulna is 
crushed; the proximal shaft of the radius 
is broken, with the fragmented ends over- 
lapping. Much of the distal ulnar and ra- 
dial shafts lie beneath the humerus, pos- 
terior skull, and other bones. The proximal 
end of the ulna overlaps that of the radius; 
neither is well presei-ved in this region. 
The distal ulna, preserved in lateral view, 
exliibits a bicondylar, topically pterosaurian 
shape, and the distal radius has the char- 

acteristic ventral process that broadens the 
articulation with the proximal cai-pals. 

Metacarpals and manual phalanges lie 
beneath the distal ends of the radius and 
ulna and the adjacent jaw. The right fourth 
or wing metacarpal (8.4 mm length) is pre- 
sented in medial view ( 4, Fig. 3). 
The well-ossified distal articular surface is 
bicondylar (Fig. 5A), unlike the trochlear 
form that is conventional among ptero- 
saurs (Fig. 5B). However, as in other 
pterosaurs, the dorsal (=extensor side) 
condyle has a radius of curvature greater 
than that of the ventral (=flexor side) con- 
dyle. Metacaqjal I (5.6 mm length) lies 
parallel to IV (Fig. 3). In most pterosaurs 
these tv\^o bones are nearly equal in length 
(I is slightly shorter than IV), whereas in 
this specimen metacai'pal I is only 67% of 
IV, comparable to the ratio that can be es- 
timated for the adult holotype of Eudi- 
morphodon ranzii (MCSNB 2888; Wild, 
1978, fig. 17). Metacaipals I and IV are 
separated by a phalanx (2.7 mm) and two 
incomplete elements (3.8 and 3.9 mm) 
that are probably also phalanges (Fig. 3). 
The third metacarpal (8.3 mm), which lies 
beneath the adjacent jaw, overlies another 
metacarpal (here interpreted as a right 
metacarpal II; 7.4 mm length; 2, Fig. 
3) that became fully exposed when meta- 
carpal IV was removed. Associated with 
these bones is a small, rounded, flat bone 
that may be a distal carpal. No manual 
claws are evident. 

Parts of a proximal (first) wing phalanx 
are associated with the distal end of a wing 
metacarpal on one block (Fig. 3) and the 
posterior end of the skull on the counter- 
part block (Fig. 2). The bone was broken 
when the matrix containing the entire 
specimen was first cleaved during quarry- 
ing. Mid-diaphyseal diameter is 1.1 mm, 
but the shaft broadens at both ends; the 
shaft closest to the occiput (presumably 
the distal end) has a diameter of about 2 
mm. As preserved, the restored length of 
the bone is 12.2 mm. However, cross-sec- 
tional diameters of the two broken ends 
differ (1.56, 1.05 mm versus 1.1, 0.9 mm). 

160 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


Figure 5. The distal end of riglit metacarpal IV of (A) Eudimorphodon cromptonellus, new species, and (B) Rhamphorhynchus 
sp. (after Wellnfiofer, 1975a, fig. 13c) in distal (above) and medial, or ulnar, views. Not to scale. 

and thus a diaphyseal section appears to 
be missing. The other proximal wing pha- 
lanx overlies the humeral head and is com- 
plete (18 mm length) although broken at 
midshaft and crushed distally (Fig. 3). 

Other wing phalanges are identified on 
the basis of the dimensions of their artic- 
ular ends. A second wing phalanx (20.5 
mm length) lies behind the skull (Fig. 2); 
the breadth of the proximal and distal ends 
are respectively 1.5 and 1.2 mm. The prox- 
imal half of the other second wing phalanx 
protrudes from beneath the metatarsals 
(Fig. 3); its proximal end (1.35 mm 
breadth) contacts the distal end of a prox- 

imal wing phalanx. The proximal two 
thirds of a third wing phalanx (Fig. 2) is 
tentatively identified on the basis of the 
breadth of its proximal end (1.15 mm). 
Several fragmentary bones may represent 
fourth wing phalanges, but are too incom- 
plete to assess. 

Hind Limb. Both femora are preserved 
in close association (Fig. 3). The complete 
right femur (19.7 mm length) is presented 
in dorsolateral view, with only the distal 
end partially obstructed by overlying bone. 
The left femur, represented by the diaph- 
ysis and distal end (fragment length, 15.75 
mm), was removed for histologic exami- 

Triassic Pterosaur From Greenland •Jenkins et al. 161 

nation (see below), as was another frag- 
ment (length, 4.5 mm), possibly repre- 
senting the proximal end. Some distortion 
of the femora is apparent, but nonetheless 
botli preserve indications of the cuivature 
(i.e., dorsal deflection of the proximal end, 
and lateral deflection of the distal end) 
that is characteristic of pterosaurs and di- 
nosaurs (Padian, 1986). The proximal end 
expands into a distinct head that is slightly 
upturned and inflected medially; a plane 
normal to the broadly convex articular sur- 
face would intersect the longitudinal axis 
of the shaft at about 45°. The distal end of 
the femur bears two contiguous condyles 
separated only by a slight sulcus; the larger 
medial condyle is more hemispheroidal in 
comparison to the ovoid, less convex lat- 
eral condyle. 

A tibia is preserved, and a fibula is ten- 
tatively identified (Fig. 3). Unlike the con- 
dition in adult Eudirnorphodon ranzii 
(Wild, 1978: 214) and most pterosaurs, no 
tibiofibular synostosis is apparent; the 
bones have been completely separated 
postmortem. The right tibia, contiguous 
with the distal end of the right femur, is 
complete, although its proximal end lies 
beneath vertebrae. The length of the ex- 
posed bone is 19 mm; inasmuch as a prox- 
imal expansion is evident, the extent of the 
obscured part is probably no more than 1- 
2 mm, giving an estimated length of 20.5 
mm. The narrowest diameter of the cylin- 
drical shaft is 1.1 mm. Distally the tibia is 
expanded (2.3 mm width) to support an 
astragalar facet that is set transversely to 
the shaft. Identification of the fibula, usu- 
ally established on the basis of tibial asso- 
ciation, cannot be made unambiguously in 
the present specimen. The presumed fib- 
ula (fi. Fig. 3) is incomplete; as preserved, 
the bone is 13.8 mm in length (and thus 
longer than the closely associated metatar- 
sals). The Pproximal end diameter is 1.3 
mm. A displaced fragment near its Pdistal 
end, if fibular, would indicate an overall 
fibular length of about 15 mm. 

No tarsal bones can be identified, with 
the possible exception of a medial distal 

tarsal associated with the phalanges of 
Pdigit V (Fig. 3). However, four complete 
but disarticulated metatarsals are pre- 
served among the hind limb bones (Fig. 
3). Midshaft diameters vary from 0.4 to 0.5 
mm. The distal ends, adjacent to pedal 
phalanges, bear articular surfaces that are 
relatively flat, as in Dimorphodon wein- 
trauhi (Clark et al, 1998). The lengths of 
the metatarsals (12, 11.35, 11.25, and 10.5 
mm) would correspond to the relative pro- 
portions of the second, third, first, and 
fourth metatarsals in the juvenile speci- 
men of Eudirnorphodon ranzii (MCSNB 
8950 B) described by Wild (1994) in which 
II > III > I > IV. In pterosaurs generally, 
either metatarsal II or III is the longest 
(Wellnhofer, 1978). The series of four 
metatarsals in the Greenlandic specimen 
may represent bones from different feet, 
and the identifications here are suggested 
on the basis of relative lengths. Specimen 
MCSNB 8950 B is in every comparable 
feature a larger individual than the Green- 
landic form except in metatarsal lengths, 
which are (I) 8.1 mm, (II) 8.85 mm, (III) 
8.6 mm, and (IV) 7.4 mm (Wild, 1994, fig. 

At least 15 disarticulated pedal phalan- 
ges, including a claw, are preserved in the 
region of the metatarsals. Inasmuch as 
there are primitively only 12 nonungual 
phalanges in a pterosaur foot (excepting 
pterodactyloids, which have only 10), the 
probability that elements of both feet are 
comingled is increased still further. 


The Greenlandic pterosaur is smaller 
than any known individual of Eiidimor- 
phodon or any other Triassic pterosaur 
(Table 1); only some smafl (ostensibly ju- 
venile) specimens of Pterodactijlus spp. 
are of comparable size (Wellnhofer, 1970). 
However, a precise, quantitative compari- 
son of size is limited by disarticulation, and 
especially the lack of any reliable indica- 
tion of axial length. Although the speci- 
mens of Eudimorphodon with which the 
Greenlandic form may be compared are 

162 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 






























-^ . re 



'^] ^. ^ 0; dj OJ (D 


O 2 nv 



-- CQ 




■^"5 /^ CO ^ 



3 s 

••£*i '~ 

= -2'-;- 

p- " rt — " 


*- a o Ci^r" ^ 


lO LO CO CD lO 05 CO 

00 00 l^ CO CO CD '^ t- 05 I- 00 t- t-^ CO 


lO o] oi lo CM 




r- ID OS 

^ CO (M 




—I O 

00 CO 
oq CO 



^ 02 O 

-H CO ^ 



o] 00 

Ol CM 

^ ^ (Li IJ 1; QJ 


ol CO CO CO CO CO ^H Ol 







































1 — 1 


1 — 1 






























. — 1 






- — I 


1 — 1 


. — 1 

1 — 1 




1— H 


EDS ^ 

^ Ol CO ^ 

X X X X 

p 73 cd c^ Co 
rt bJj b£ M b/j 





Co Co Co 

V3 crt on c/: 



S> > 

AC 1; 

ctf cS cS 

<W -k^ ■4-' -^^ -1_ 

cj rs :« CS cS ^ 

.— -M -tJ -M *J 

f^ CU Qj Qj ^ F^H 

p s s ;s s w 






* -E 

Triassic Pterosaur From Greenland 'Jenkins et al. 163 

articulated, none is complete and the 
lengths of some limb elements can only be 
estimated. Nonetheless, the size range 
represented by the available sample of Eu- 
dimorphodon may be approximated by 
summing the lengths of the humerus, ulna, 
metacarpal IV, wing phalanx 1, femur and 
tibia (S h + . . . t, Table 1). By this index, 
MGUH VP 3393 is one-third the size of 
the type and largest specimen of E. ranzii, 
and two-thirds the size of two juvenile 
specimens (Fig. 6). 

Taxonomic Assignment 

The distinctive dentition in MGUH VP 
3393 offers secure evidence on which to 
assign the specimen to the genus Eiidi- 
moiyhodon, but allocation to a known spe- 
cies is problematic. Conspecificity with E. 
ranzii appears improbable on the basis of 
two proportional disparities. First, the 
lengths of the metatarsals in the juvenile 
E. ranzii (MCSNB 8950 B) described by 
Wild (1994) from Ponte Giurino are on 
the order of 25% shorter than those in 
MGUH VP 3393 (Table 1), although al- 
most all other long bone lengths indicate 
that MCSNB 8950 is larger (Fig. 6). Al- 
though the Ponte Giurino Eiidimoij)hodon 
lacks a skull. Wild (1994: 112-115) ex- 
pressed confidence in referring the speci- 
men to E. ranzii based primarily on the 
close correspondence of limb bone lengths 
and proportions to those of the Milano ju- 
venile. Second, femoral lengths in the 
Greenlandic specimen and the two juve- 
nile but larger E. ranzii are essentially the 
same (Table 1), with the improbable im- 
plication for conspecificity that all long 
bones, except the femur, increased in 
length during early development. 

Differences in limb proportions obviate 
the possibility that the Greenlandic ptero- 
saur might be an immature conspecific of 
Eudimorphodon rosenfeldi. Eudimorpho- 
don rosenfeldi, decribed by Dalla Vecchia 
(1995) on the basis of a single, smaller 
specimen than the type of E. ranzii, was 
differentiated in part from E. ranzii by the 
comparable lengths of the tibia and ulna. 

and the greater length of the tibia relative 
to the humerus; these proportions are 
more or less shared by the Greenlandic 
specimen (Table 1). However, unlike the 
condition in E. rosenfeldi, the humerus is 
shorter than the femur in the Greenlandic 
specimen, a primitive condition known 
among pterosaurs only in Preondactijlus 
buffarinii (Dalla Vecchia, 1998, table 5). 
Furthermore, the hypothesis that the 
Greenlandic specimen is an immature E. 
rosenfeldi entails seemingly improbable al- 
lometric reversals in comparison to those 
known in congeners. In E. ranzii, the hu- 
meral/femoral index decreases from juve- 
nile to adult stages (from about 135% to 
115%; Table 2), but the same index in- 
creases when the Greenlandic specimen 
(92%) and E. rosenfeldi (109%) are com- 
pared. A comparable reversal is seen in the 
ulnar/femoral index (Table 2). Finally, the 
femoral/tibial index in E. ranzii increases 
slightly from juvenile to adult stages, 
whereas an approximately 30% decrease 
occurs in the same index between MGUH 
VP 3393 and E. rosenfeldi (Table 2). 

Assessment of Ontogenetic Stage 

Various criteria have been employed to 
differentiate immature from adult ptero- 
saurs: relative body size, degree of ossifi- 
cation, and osteometric ratios (Wellnhofer, 
1970, 1975a-c); synostosis, epiphyseal os- 
sification, and bone histology (Bennett, 
1993); and a combination of morphomet- 
ric and histologic features (Padian et al., 

Insofar as these criteria may be applied, 
the Greenlandic pterosaur would appear 
to be neither a hatchling nor an adult, and 
is most probably a juvenile. The lack of 
intracranial fusion, as well as the lack of 
synostosis between vertebral arches and 
centra, and between scapula and coracoid, 
are indicative of a preadult stage. The di- 
minutive size of the individual is sugges- 
tive of immaturity, but not conclusive. The 
limb proportions may be interpreted as 
ontogenetically immature, or phylogeneti- 
cally primitive, or both. In the series of 

164 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

5 cm 

Triassic Pterosaur From Greenland 'Jenkins et al. 165 

Pterodactylus spp. studied by Wellnhofer images; evidence of calcified cartilage is 
(1970), juveniles generally possess niore preserved in a few areas. Although no 
comparable humeral/radial and femoral/ trace is found of a subchondral bone plate 
tibial lengths than do adults. Although the or transphyseal canals, a delineation is vis- 
limb proportions of the Greenlandic ible between the diaphyseal and epiphy- 
pterosaur represent the juvenile end of seal— inetaphyseal regions. Histology thus 
this allometric spectrum, these propor- provides primarily negative evidence with 
tions are also intermediate between vari- which to assess the ontogenetic stage of 
ous basal pterosaurs and nonpterosaurians the Greenlandic pterosaur. 
(Table 2). 

Histologic sections of the the left femur Appendicular Proportions 

yielded less than definitive results because Certain appendicular proportions of the 

of extensive diagenetic alteration and re- tiny Greenlandic form lie within the range 

sultant artifacts. As in pterosaurs and known among other pterosaurs (Table 2). 

birds, the diaphyseal cortex is thin (about The humerus is slightly shorter (92%) than 

15— 20% of shaft diameter). The wide med- the femur, a condition unusual among 

ullary cavity, infilled with calcite, shows no pterosaurs; comparable proportions are 

trace of trabecular projections that have known only in Preondactijlus biijfarinii 

been identified in chelonian, crocodilian, (Dalla Vecchia, 1998, table 5) and in Pter- 

and dinosaurian embiyos (Homer et al., odactylus antiqiius, P. suevicus, and P. mi- 

2001). The bone cortex is mostly parallel cronijx (Wellnhofer, 1970; see Bennett, 

fibered, with only indistinct indications of 1996, for a taxonomic reallocation of P. 

localized lamellar deposition; the osteocy- suevicus to the genus Cycnorhamphus). In 

tic spaces are variably distributed. No fea- both juvenile and adult Eiidimorphodon 

tures are present that might be expected ranzii as well as in other pterosaurs, these 

in an embryo or rapidly growing neonate, proportions are reversed. In other appen- 

nor are indicators present of growth stasis dicular proportions the Greenlandic ptero- 

and maturity. The parallel-fibered matrix saur is distinctive (Table 2). Relative to the 

differs from the fibrolamellar architecture humerus or femur, the ulna is shorter than 

knowii in later pterosaurs, which has been in any known pterosaur. Similarly, relative 

interpreted as indicative of rapid growth to the femur, the tibia is shorter than in 

(Bennett, 1993). Vascular canals are nearly any known pterosaur. Among pterosaurs, 

all longitudinally oriented; the canals ap- the Greenlandic pterosaur is distinctive in 

pear to be primary, but no evidence is pre- having the relative lengths of brachium to 

sent of primary osteonal development. The antebrachium, and femur to cms, both 

bone is less vascularized than that of ein- nearly equal. 

bryonic and hatchling dinosaurs, and is The length of the metatarsus also ap- 

more comparable to that in hatchling alii- pears to be unusual. Metatarsals in the 

gators (see Homer et al., 2001). Longitu- Greenlandic pterosaur vary from 132 to 

dinal sections through the epiphysis reveal 156% of the lengths represented in a ju- 

endosteo-endochondral trabeculae, most venile eudimoiphodontid (MCSNB 8950 

of which are diagenetically altered to ghost B), which is particularly notable because 


Figure 6. Limb proportions of the adult, type specimen of Eudimorphodon ranzii (MCSNB 2888), top, compared with a juvenile 
E. ranzii (MCSNB 8950), middle, and E. cromptonellus, new species (MGUH VP 3393), bottom. The limbs are positioned in the 
same transverse plane to permit graphic illustration of the relative lengths of the long bones; no postural or kinematic represen- 
tation is intended. Data on E. ranzii are from Wild (1978, 1994). Appendicular bone lengths are estimated for the metatarsus of 
MCSNB 2888; the wing phalanges distal to the break in the proximal wing phalanx of MCSNB 2888; and the first and fourth 
wing phalanges of MGUH VP 3393. 

166 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Table 2. Limb segment proportions. Data from (a) Ewer, 1965; (b) Sereno and Arcucci, 1994; (c) 

PADIAN, personal OBSERVATION; (d) DALLA VECCHIA, 1998; (e) WILD, 1978; (f) WILD, 1994; (g) DALLA 
VECCHIA, 1995; (h) UNWIN, 1988, and PADIAN, personal OBSERVATION (l) WELLNHOFER, 1978, AND Pa- 

DIAN, personal OBSERVATION. •' 

















Euparkeria capensis 







Marasuch us lilloensis 








Scleromoclihis taijlori 







Eudiinorphodon cromptonelhis. 







new species, MGUH VP 3393 

Preondactijhis biiffariuii. 







MFSN 1770 


Peteinosaunis zambeUU, 







MCSNB 3359 


Eudimorplwdon ranzii. 







juvenile, Milano specimen 


Eiidimoiyhodon ranzii. 







juvenile, MCSNB 8950 


Eudimorplwdon ranzii, 







adult, tyi^e, MCSNB 2888 


Eudi moi~phodon rosenfcldi 







MFSN 1797 


Dinior})hodon macroni/x 








Donjgnathus banthensis 







Campylognathoides liasicus 







Campijlognathoides zitteli 








every other limb dimension (wdth the ex- 
ception of femoral length) indicates that 
MCSNB 8950 is a larger animal (Table 1; 
Fig. 6). Similarly, even the shortest meta- 
tarsal of the Greenlandic pterosaur (ca. 11 
mm) is longer than the longest metatarsals 
of Pterodactijlus specimens that, in other 
limb dimensions, appear to be otherwise 
comparably sized or even larger (P. ele- 
gans, no. 49, 7.8 mm; P. elegans, no. 12, 
9.5 iTim; P. inicronyx; no. 42, 8.8 mm; 
Wellnhofer, 1970, fig. 19). In most ptero- 
saurs, the average length of metatarsals I— 
IV is less than that of metacarpal IV. Ex- 
ceptions are Peteinosaunis zamheUii 
(Wild, 1978, pi. 14) and Campijlognathoi- 
des zitteli (Pheninger, 1895: 216-217; 
Wellnhofer, 1978: 38), in which average 

metatarsal lengths are greater (105%, 
129%, respectively). The Greenlandic 
form is even inore exceptional, having 
metatarsals that average 141% of metacar- 
pal IV length. 

Wild (1978) interpreted Peteinosaunis 
zamhellii as the most primitive pterosaur 
then known on the basis of the relative 
shortness of the wing compared to the 
length of the hind limb. Wild (1984: 54) 
later cited various limb segment ratios in 
support of his obseiA/ation that Preondac- 
tijhis hiiffarinii is unique among ptero- 
saurs for "... an unproportionally long 
hind-limb," and a wing that is "extraordi- 
narily short." More recently, a reanalysis of 
Pr. hujfarinii led Dalla Vecchia (1998: 365) 
to conclude that Pe. zanibellii, Pr hiijfar- 

Triassic Pterosaur From Greenland 'Jenkins et ol. 167 

inii, and Dimorphodon macronyx are extended to promote maximum thrust, and 
"probably the most primitive of all known during upstroke the effective wingspan is 
pterosaurs." Direct comparison of these shortened. Nonetheless, the analogy is 
taxa with the Greenlandic pterosaur is lim- limited by anatomical differences. Birds 
ited because the fourth wing phalanx in possess a multiaxial carpometacaipal joint 
MGUH VP 3393 is unknown, and the complex with numerous degrees of free- 
lengths of various phalanges can only be dom and movement possibilities (Vazquez, 
estimated. Nonetheless, a restricted com- 1992). Reduction of aerofoil drag of the 
parison may be made on the basis of distal wing during upstroke in birds may 
summed lengths of the propodial, meso- be further promoted by a rotation of the 
podial, and metapodial elements (S hu- feathers that opens slots between them 
merus + ulna + metacarpal IV/X femur + (the valve function of Norberg, 1985), or 
tibia + average metatarsal length). The by a closure of the imbricating fan of 
proportionate lengths of the proximal fore- feathers. The fiber-stiffened distal patagi- 
limb to proximal hind limb, thus defined, um of pterosaurs (i.e., distal to the meta- 
are 90% in the Greenlandic pterosaur, carpophalangeal joint; Padian and Rayner, 
95% in Pr. hiijfarinii, and 100% in Pe. 1993; see also Unwin et al., 1993) has no 
zauibellii. In D. macronyx the proportion comparable intrinsic mechanisms for 
is 104%, in Campylognathoides liasicus changes in shape, and thus shortening ef- 
124%, in juvenile Eudiinoiyhodon ranzii fective wingspan would appear to be the 
and Dorygnathus banthensis 130%, but in only alternative mechanism for reducing 
Campylognathoides zitteli, with its unusu- aerofoil drag during upstroke. The joint 
ally short hind limb, 94%. that permits the largest range of excursion 

is the metacarpophalangeal joint. Howev- 

The Pterosaurian Wingbeat Cycle ^^ th^ pterosaurian metacaipophalangeal 

Most reconstructions of pterosaurs in joint is a uniaxial joint, with one degree of 

flight depict the wings fully extended in a freedom; the asymmetrical form of this 

position that is suggestive of soaring, al- joint, which entails a screwlike motion 

though general agreement exists that all (Wellnhofer, 1978), provides for radial de- 

except the largest pterosaurs were capable viation of the wing finger upon wing ex- 

of sustained, flapping flight. The only at- tension, and ulnar deviation and flexion 

tempt to illustrate the excursion of an en- upon wing folding. 

tire wing is by Wellnhofer (1991: 153), but The bicondylar form of the metacai-po- 

his diagram of a pterodactyloid only de- phalangeal joint in Eudimorphodon 

picts the downstroke, and shows major cromptoneUiis appears to be intermediate 

flexion at the wrist and little at the meta- between a primitive, unicondylar joint and 

carpophalangeal joint. Subsequently, Pa- the trochlear form of later pterosaurs. This 

dian and Rayner (1993: 143, fig. 13C) sug- spectrum of joint configurations represents 

gested that the pterosaurian "... metacar- increasing mechanical stability, consonant 

pophalangeal joint is ideally adapted for with the interpretation that the joint was 

sweeping the wingtip during the upstroke actively employed in flapping flight. As in 

in a movement analogous to that in birds, other pterosaurs, the joint in Eudimoiyho- 

With their long, thin wings the pterosaur don cromptoneUiis is stiiicturally asym- 

wingbeat would have appeared very simi- metric: the dorsal condyle has a larger ra- 

lar to that of long-winged birds such as dius of curvature, and a more extensive ar- 

gulls or albatrosses." ticular surface, than the ventral condyle 

As a first-order approximation, an avian (Fig. 5). However, the joint surfaces at 

model of a pterosaurian wingbeat cycle their extensor end are evenly aligned, 

fulfills fundamental aerodynamic require- Thus, during the downstroke, the extend- 

ments: upon downstroke the wing is fully ed distal wing (supported by the four pha- 

168 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 7. Reconstruction of metacarpophalangeal (mp) joint function in Eudimorphodon cromptonellus, new species, during 
flapping flight. The wing, in dorsolateral view, is shown in an extended position during early downstroke and in a flexed position 
in early upstroke. The inset above depicts mp joint relations, disarticulated for illustrative purposes, in the extended (E) and 
flexed (F) positions. The axis of the mp joint in an extended position (e) is normal to the metacarpal shaft because of the 
symmetry of the condyles in this region; the proximal phalanx is thus aligned with metacarpal IV. The mp joint axis in a flexed 
position (f) is oblique to the metacarpal shaft because of condylar asymmetry, resulting in a slight rotation and flexion of the 
distal wing as the mp joint undergoes ulnar deviation. The excursion of the mp joint during the upstroke of flapping flight is 
consonant with the distribution of fibers as interpreted by Wild (1994: 109) for the congener E. ranz// (MCSNB 8950). Elongate, 
membrane stiffening fibers occur in the distal wing, whereas short fibers in the proximal wing allow flexibility. 

Triassic Pterosaur From Greenland 'Jenkins et al. 169 

langes) is aligned with the proximal wing 
(Fig. 7) because the joint's axis at the ex- 
tensor end of the metacarpophalangeal 
joint is perpendicular to the plane of the 
entire wing (Fig. 7, top inset). Were the 
metacai-pophalangeal joint to be flexed 
during upstroke, the shift in the joint s axis 
(Fig. 7, top inset), which results from con- 
dylar asymmetry, engenders a rotation and 
flexion of the distal wing out of the plane 
of the proximal wing. Thus, during up- 
stroke, the distal wing would be directed 
ventrally and somewhat everted (i.e., the 
ventral surface turned to face slightly lat- 
erally). On this interpretation, the trafling 
position of the distal wing relative to the 
proximal wing during the upstroke would 
be comparable to that seen in birds, but 
the rotation and flexion would seem 
uniquely pterosaurian. 


Field work, supported by grants from 
the National Science Foundation, the 
Carlsberg Foundation, and the Putnam 
Expeditionary Fund of the Museum of 
Comparative Zoology, was initiated on the 
basis of geological studies by L. B. Clem- 
mensen (University of Copenhagen). We 
thank W. W Amaral for preparation of the 
specimen; A. H. Coleman and P. Chan- 
doha for photography; L. L. Meszoly for 
drawing Figures 2 and 3; and K. Brown- 
Wing for rendering Figures 5, 6, and 7. 
For histologic examination of the femur, 
we are grateful to M. Goodwin (University 
of California Museum of Paleontology) 
and E. Lamm (Museum of the Rockies) 
for technical assistance, A. Paulsen for 
photography, and J. R. Horner and A. de 
Ricqles for consultation. Finally, we thank 
S. C. Bennett, F. M. Dalla Vecchia, K. M. 
Middleton, P. Wellnhofer, and R. Wild for 
authoritative, helpful reviews. 


Bennett, S. C. 199.3. The ontogeny of Pteranodon 
and other pterosaurs. Paleobiology, 19: 92-106. 

. 1996. On the taxonomic status of Cijcnor- 

hamphus and Gallodactylus (Pterosauria: Ptero- 

dactyloidea). Journal of Paleontology, 70: 3.35- 
Clark, J. M., J. A. Hopson, R. Hernandez R., D. 
E. Fastovsky, and M. Montellano. 1998. 
Foot posture in a primitive pterosaur. Nature, 
391: 886-889. 
Clemmensen, L. B.. D. V. Kent, and F. A. Jenkins, 
Jr. 1998. A Late Triassic lake system in East 
Greenland: facies, depositional cycles and pa- 
laeoclimate. Palaeogeography, Palaeoclimatology, 
Palaeoecology, 140: 135-159. 

Dalla Vecchia, F. M. 1995. A new pterosaur (Rep- 
tilia, Pterosauria) from the Norian (Late Triassic) 
of Friuli (northeastern Italy). PreUminary note. 
Gortania — Atti del Museo Friulano di Storia Na- 
turale, 16: 59-66. 

. 1998. New observations on the osteology and 

ta.xonomic status of Preondactijlus buffarinii 
Wild, 1984 (Reptilia, Pterosauria). Bollettino del- 
la Societa Paleontologica ItaUana, 36: 355-366. 

Dalla Vecchia, F. M., G. Muscio, and R. Wild. 
1989. Pterosaur remains in a gastric pellet from 
the Upper Triassic (Norian) of Rio Seazza Valley 
(Udine, Italy). Gortania — Atti del Museo Friu- 
lano di Storia Naturale, 10: 121-131. 

Ewer, R. F. 1965. The anatomy of the thecodont 
reptile Euparkeria capensis Broom. Philosophi- 
cal Transactions of the Royal Society of London 
B, 248: 379-435. 

Horner, J. R., K. Padian, and A. de Ricqles. 
2001. Comparative osteohistology of some em- 
bryonic and perinatal archosaurs: developmental 
and behavioral implications for dinosaurs. Paleo- 
biology, 27: 39-58. 

Jenkins, F A., Jr., N. H. Shubin, W W. Amaral, S. 
M. Gatesy, C. R. Schaff, L. B. Clemmensen, 
W. R. Downs, A. R. Davidson, N. Bonde, and 
F. OSBaeCK. 1994. Late Triassic continental ver- 
tebrates and depositional environments of the 
Fleming Fjord Formation, Jameson Land, East 
Greenland. Meddelelser om Gr0nland, Geosci- 
ence, 32: 1-25. 

NORBERG, R. A. 1985. Function of vane asymmetry 
and shaft curvature in bird flight feathers; infer- 
ences on flight ability oi Archaeoptenjx, pp. 303- 
318. In M. K. Hecht, J. H. Ostrom, G. Viohl, 
and P. Wellnhofer (eds.). The Beginnings of 
Birds. Eichstatt, Germany: Freunde des Jura- 
Museums. 382 pp. 

Padian, K. 1980. Note of a new specimen of ptero- 
saur (Reptilia: Pterosauria) from the Norian (Up- 
per Triassic) of Endenna, Italy. Rivista del Museo 
Civico di Scienze Naturali "E'. Caffi," 2: 119-127. 

. 1983. A functional analysis of flying and 

walking in pterosaurs. Paleobiology, 9: 218-239. 

. 1986. On the type material of Coelophysis 

Cope (Saurischia: Theropoda) and a new speci- 
men from the Petrified Forest of Arizona (Late 
Triassic: Chinle Formation), pp. 45-60. In K. Pa- 
dian (ed.). The Beginning of the Age of Dino- 
saurs: Faunal Change across the Triassic-Jurassic 

170 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Boundary. New York: Cambridge University 
Press, xii + 378 pp. 

Padian, K., and J. M. V. Rayner. 1993. The wings 
of pterosaurs. American Journal of Science, A, 
293: 91-166. 

Padian, K., A. J. de Ricqles, and J. R. Horner. 
1995. Bone histology determines identification of 
a new fossil taxon of pterosaur (Reptilia: Archo- 
sauria). Comptes Rendus de I'Academie des Sci- 
ences, Sciences de la Terre et des Planetes, serie 
Ila, 320: 77-84. 

Plieninger, F. 1895. Campijlognathus Zitteli. Ein 
neuer Flugsaurier aus dem Oberen Lias Schwa- 
bens. Palaeontographica, 41: 193-222. 

1995. Contribution to the conodont biostratig- 
raphy of the Dolomia di Fomi (Upper Triassic, 
Camia, NE Italy). Memorie di Scienze Geolo- 
gische Universita di Padova, 47: 125-133. 

Sereno, p. C, and a. B. Arcucci. 1994. Dinosau- 
rian precursors from the Middle Triassic of Ar- 
gentina: Marasuchiis lilloensis, gen. nov. Journal 
of Vertebrate Paleontology, 14: 53-73. 

Unwin, D. M. 1988. New remains of the pterosaur 
Diiiwrphodon (Pterosauria: Rhamphorhynchoi- 
dea) and the terrestrial ability of early pterosaurs. 
Modern Geology, 13: 57-68. 

Unwin, D. M., D. M. Martill, and N. N. Bak- 
HURINA. 1993. The structure of the wing mem- 
brane in pterosaurs. Journal of Verebrate Pale- 
ontology, 13(3, Suppl.): 61A. 

Vazquez, R. J. 1992. Functional osteology of the avi- 
an \\'rist and the evolution of flapping flight. Jour- 
nal of Morphology, 211: 259-268. 

WellnhoFER, p. 1970. Die Pterodactyloidea (Pter- 
osauria) der Oberjura-Plattenkalke Siiddeutsch- 

lands. Abhandlungen bayerische Akademie der 
Wissenschaften, mathematisch-naturvvdssenschaf- 
tliche Klasse, 141: 1-133. 

. 1975a. Die Rhamphorhynchoidea (Pterosau- 
ria) der Oberjura-Plattenkalke Siiddeutschlands. 
Teil I. Allgemeine Skelettmorphologie. Palaeon- 
tographica A, 148: 1-33. 

. 1975b. Die Rhamphorhynchoidea (Pterosau- 
ria) der Oberjura-Plattenkalke Siiddeutschlands. 
Teil II. Systematische Beschreibung. Palaeonto- 
graphica A, 148: 132-186. 

. 1975c. Die Rhamphorhynchoidea (Pterosau- 
ria) der Oberjura-Plattenkalke Siiddeutschlands. 
Teil III. Palokologie und Stammesgeschichte. Pa- 
laeontographica A, 149: 1—30. 

. 1978. Pterosauria, pp. 1-81. In P Wellnhofer 

(ed.), Handbuch der Palaoherpetologie, Teil 19. 
Stuttgart, Germany: G. Fischer. 81 pp. 

1991. The Illustrated Encyclopedia of Ptero- 

saurs. London: Salamander Books. 192 pp. 

Wild, R. 1978. Die Flugsaurier (Reptilia, Pterosau- 
ria) aus der Oberen Trias von Cene bei Bergamo, 
Italien. Bollettino della Societa Paleontologica It- 
aliana, 17: 176-256. 

. 1984. A new pterosaur (Reptilia, Pterosauria) 

from the Upper Triassic (Norian) of Friuli, Italy. 
Gortania — Atti del Museo Friulano di Storia Na- 
turale, 5: 45-62. 

1994. A juvenile specimen of Eiidimoiyho- 

don ranzii Zambelli (Reptilia, Pterosauria) from 
the Upper Triassic (Norian) of Bergamo. Rivista 
Museo civico di Scienze Naturali "E. Caffi," 16: 
Zambelli, R. 1973. Eudimorphodon ranzii gen. nov., 
sp. nov, uno pterosauro Triassico. Istituto Lom- 
bardo Rendiconti Scienze, B 107: 27—32. 



Abstract. New fossils from the Duncannon Mem- 
ber of the Catsldll Formation provide material for 
hypotheses about the evolution of fin development 
and function in extinct sarcopterygians. The rhizo- 
dontid affinity of these specimens is supported by the 
pattern of overlap between the clavicle and cleith- 
rum, the robustness of the pectoral girdle, the pres- 
ence of unjointed and elongate lepidotrichia in the 
pectoral appendage, and the presence of multiple lat- 
eral line canals. The small bodv size and weakly os- 
sified endochondral skeletons indicate that these in- 
dividuals are immature. The pectoral fin includes 
both dermal and endochondral elements; massive, 
unjointed dermal rays that compose the bulk of the 
appendage envelop the endochondral bones. The em- 
phasis on both dermal and endochondral elements in 
rhizodontids is an unexpected intermediate condition 
between ray-finned and lobe-finned designs that 
could not be predicted from current models of der- 
mal fin development. The large size of rhizodontids 
and, perhaps, aspects of their locomotor and feeding 
strategies, may have necessitated fins with large sur- 
face areas. Expansion of the surface area of the fins 
was accomplished by elaboration of the dermal ra- 
dials. The evolution of an extensive endochondral 
skeleton in rhizodontids may relate to the role of the 
endochondral skeleton in the control of movements 
and shape of this expanded fin. 


Rhizodontid sarcopterygians are large, 
predatory fish that have a variety of unique 
features of the skull, pectoral girdle, fins, 
and lateral line systems (Andrews, 1985). 
Many of these features, such as the un- 
jointed and elongate lepidotrichia in the 
pectoral fins, suggest that rhizodontids 

* Corresponding Author 

' Department of Organismal Biolog)' and Anatomy, 
University of Chicago, 1027 East 57th Street, Chi- 
cago, Illinois 60637. 

- Department of Vertebrate Biolog>', Academy of 
Natural Sciences of Philadelphia, 1900 Benjamin 
Franklin Parkway, Philadelphia, Pennsylvania 19103. 

may have been specialized for modes of 
locomotion and predation unseen in other 
sarcopterygian taxa. Represented by eight 
or nine genera froin the Devonian and 
Carboniferous, these fish typically attain 
lengths greater than 3 m. Because of the 
many similarities between the endochon- 
dral skeletons of rhizodontid fins and tet- 
rapod limbs, these taxa have figured prom- 
inently in studies of the origin of tetrapod 
limbs (e.g., Gregory and Raven, 1941). 

Remains of rhizodontid sarcopterygians 
are major components of Late Devonian 
and Early Carboniferous nonmarine fossil 
assemblages (Andrews, 1985; Young et al., 
1992). Scales, isolated vertebrae, and teeth 
are the most common rhizodontid ele- 
ments encountered; these are usually as- 
signed to the genera Rhizodus and Strep- 
sodus. However, more complete material 
is rare. The most complete specimens are 
known from the Dinantian of Foulden, 
United Kingdom {Strepsodiis, several par- 
tially articulated specimens [Andrews, 
1985]), the Lower Carboniferous of Aus- 
tralia (Barameda, an articulated partial 
skeleton preserved as a natural mold 
[Long, 1989]), the Frasnian of Antarctica 
(Notorhizodon, isolated cranial and girdle 
elements [Young et al, 1992]), and the Fa- 
mennian of New South Wales, Australia 
(Gooloogongia, articulated individuals pre- 
served as natural molds [Johanson and 
Alilberg, 1998]) and of North America 
(Saiiriptenis, isolated pectoral fins and 
scales [Hall, 1843; Daeschler and Shubin, 

A number of features, particularly in the 
paired fins and associated girdles, distin- 

Bull. Mus. Comp. ZooL, 156(1): 171-187, October, 



172 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

guish rhizodontids as a clade. Rhizodontid 
pectoral girdles are unique in both their 
massive relative size as well as the nature 
of the internal articulations between con- 
stituent bones. The ventral laminae of the 
clavicles and cleithra are broad, orna- 
mented plates that extend cranially, en- 
closing the caudal portion of the gular re- 
gion (Andrews, 1985, fig. 10b). In porole- 
piforiTis and osteolepids (for example see 
Eusthenopteron [Andrews and Westoll, 
1970, fig. 1]), the dorsal lamina of the 
cleithrum is overlapped medially by the 
clavicle, whereas the ventral lamina of the 
clavicle is overlapped medially by the 
cleithrum. In rhizodontids the pattern of 
overlap is reversed: the dorsal lamina of 
the cleithrum is overlapped laterally by the 
clavicle, and the ventral lamina of the clav- 
icle is overlapped laterally by the cleith- 
rum. The function of this massive shoulder 
girdle, with its unusual pattern of overlap 
between constituent elements, is un- 

Although rhizodontid pectoral girdles 
are readily distinguished from those of 
other sarcopteiygians, the pectoral fins of 
rhizodontids share a number of similarities 
with the limbs of tetrapods. The best- 
known rhizodontid fins are isolated ap- 
pendages that have been assigned to the 
genus Sauriptenis. Sauriptents possesses a 
limblike arrangement of endochondral 
bones, including distal radials that have 
been compared to tetrapod digits (Gregoiy 
and Raven, 1941; Daeschler and Shubin, 
1998). Despite differences in the size and 
shape of corresponding elements in Sau- 
riptenis and tetrapods, the pattern of en- 
dochondral bones is extremely similar. The 
phylogenetic significance of these similar- 
ities is muddled by the fact that recent cla- 
distic studies do not support a sister-group 
relationship between tetrapods and rhizo- 
dontids. Missing data are probleinatic in 
the cladistic analysis of the higher level re- 
lationships of rhizodontids because few 
overlapping characters occur between taxa 
such as Sauriptenis (consisting of isolated 
fins and scales) and Gooloogongia (con- 

sisting of a relatively complete skull and 
body, but with poorly preserved fins). 

Here we describe immature rhizodontid 
specimens, which we refer to the genus 
Sauriptenis, from the Famennian Catsldll 
Formation of North Ainerica. The relative 
completeness of the skeletons provides an 
opportunity to examine the morphology of 
rhizodontids in greater detail and to assess 
hypotheses on the ontogeny of the paired 
fin skeleton in nontetrapod sarcopterygi- 


The new rhizodontid material was re- 
covered from the Late Devonian Red Hill 
locality in Clinton County, Pennsylvania 
(Fig. 1). Red Hill is a road-cut exposure of 
the Duncannon Member of the Catsldll 
Formation. During the Late Devonian, 
the Catskill Delta extended from the foot- 
hills of the Acadian highlands within the 
Old Red Continent (Euramerica) to the 
epicontinental Catskill Sea that lay to the 
west (Woodrow, 1985). The fossiliferous 
horizons at Red Hill have been interpreted 
to represent channel margin and overbank 
deposits of a wide river flowing across a 
low-gradient floodplain under a subtropi- 
cal climate (Woodrow et al., 1995). The 2- 
m-thick fossiliferous zone at Red Hill has 
produced the most abundant and well-pre- 
served vertebrate fossils discovered to date 
from the Catskill Formation. 

The rhizodontids were recovered from 
a grayish-red, poorly bedded sandy silt- 
stone unit within the fossiliferous zone. 
Their degree of articulation is exceptional, 
even for Red Hill, and suggests that the 
specimens were quickly buried, did not 
undergo significant postmortem transport, 
and were not subsequently reworked. The 
two specimens on which this study is 
based are preserved on opposite sides of 
the same block of matrix. Remarkably, 
these specimens are the only rhizodontids 
known from Red Hill. 

Red Hill has yielded a diverse assem- 
blage of freshwater vertebrates, terrestrial 
plants, and invertebrates. Among the ver- 

Immature Rhizodontids from North America • Davis et al. 


Figure 1 . Location of Red Hill site, Clinton County, Pennsylvania, USA. Shaded areas represent Devonian age sediments. 
Abbreviations; 80, Interstate 80; 120, Pennsylvania State Route 120. 

tebrates from the fossiliferous zone are 
two early tetrapod taxa, at least three taxa 
of osteolepiform sarcopteiygians, an early 
actinopterygian, groenlandaspidid and 
phyllolepid placoderms, gyracanthid acan- 
thodians, and chondrichthyans. This ver- 
tebrate fauna is associated with progyni- 
nosperm and lycopsid plants as well as pa- 
lynomorphs. In addition, trigonotarbid ar- 
thropods are preserved both as body fossils 
and traces. 


Sarcopterygii Romer, 1955 

Rhizodontida Andrews and Westoll, 1970 

cf. Sauripterus Hall, 1843 

Diagnosis. Rhizodontid affinities are 
supported by the following features: 
cleithruin overlaps dorsal lamina of clavi- 
cle medially and ventral lamina of clavicle 
laterally (reversed from the condition seen 

174 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

in osteolepiforms); laterally expanded ven- 
tral lamina of clavicle and cleithrum; ros- 
trally expanded ventral lamina of cleith- 
rum relative to the condition seen in os- 
teolepiforms; scapulocoracoid sits dorsal to 
the junction of dorsal and ventral laminae 
of cleithrum; dermal skeleton of the paired 
fins coinposed of elongate, unjointed lep- 
idotrichia; no postaxial process on the ul- 
nare; accessoiy lateral lines. Assigned to 
the genus Sauriptenis based on the follow- 
ing: radius anteroposteriorly broader than 
ulna relative to the condition seen in Bar- 
ameda (Long, 1989) and Strepsodus (An- 
drews, 1985); ventral laminae of cleithra 
meet at midline. 

These specimens also have unique char- 
acters seen in no other known rhizodontid: 
no entepicondyle on the humerus; no dis- 
tal radials in the pectoral fin; no articula- 
tions between fin endochondral elements; 
disproportionally large pectoral fins rela- 
tive to body size (when compared to all 
rhizodontids and osteolepids for which this 
can be measured); elongate interclavicle 
(craniocaudal length greater than that of 
ventral laminae of clavicle). Many of these 
characters can be interpreted as being re- 
lated to the ontogenetic stage of ossifica- 
tion in the fin endoskeleton. Therefore, 
these specimens are referred to the genus 
Saiiriptenis as immature individuals. 

Horizon and Locality. The Duncannon 
Member of the Catskill Formation. USA, 
Pennsylvania, Clinton County, Red Hill 
(coordinates 41°20.645'N, 77°40.800'W). 

Age. Late Devonian; late Famennian 
stage (Fa2c substage); Riigospora flexu- 
osa-Grandispora cornuta palynomorph 
zone (Traverse, in press). 

Material. ' All specimens are housed in 
the Academy of Natural Sciences of Phil- 
adelphia (ANSP): ANSP 20980, a well- 
preserved skeleton lacking most cranial el- 
ements and median fins, preserved as part 
and counterpart (Figs. 2, 3); ANSP 20981, 
a smaller individual consisting of a left 
dentary, paired gulars, pectoral girdle and 
fins, preserved on the reverse side of the 
ANSP 20980 counterpart block (Fig. 4). 


Operculogular Series. Lateral gulars, a 
left operculum, and portions of the sub- 
mandibular series are preserved on ANSP 
20981. The lateral gulars are preserved in 
internal view, with the left gular, which is 
relatively more complete, overlapping the 
right. Both sides lack the rostral margin. It 
is difficult to determine whether a median 
gular was present. The ornamentation on 
the internal surface consists of shallow 
grooves that radiate caudally from the ros- 
trolateral margin. Medial to this margin, 
two small teeth are presei"ved in cross sec- 
tion (Fig. 4, f. vom). These teeth may have 
been derived from the left derinopalatine 
or vomer, both of which are not preserved. 
A recessed edge along the lateral margin 
of the left gular marks the contact for the 
submandibulars. The submandibular se- 
ries lies medial to the labial margin of the 
dentary and lateral to the left gular. The 
bone in this area appears very thin and 
broken, making it impossible to identify 

A left opercular is preseived in ANSP 
20981 (Fig. 4). The external surface is 
sparsely ornamented with a series of shal- 
low, parallel grooves that extend dorsoven- 
trally across the bone. The margins of the 
opercular lack this ornamentation and ap- 
pear relatively smooth. The opercular is 
narrow rostrocaudally and deep dorsoven- 
trally This shape is similar to that of Goo- 
loogongia (Johanson and Ahlberg, 1998, 
fig. 2g), but unlike the rounded operculars 
of Barameda (Long, 1989, fig. 5b) and 
Strepsodus (Andrews, 1985, fig. lb). In po- 
rolepiforms and osteolepids the opercular 
extends caudally to contact the cranial 
margin of the dorsal lamina of the cleith- 
rum. The proportions of the opercular in 
ANSP 20981 make it unlikely that it abut- 
ted against the dorsal lamina of the cleith- 
rum. The resulting gap was likely filled by 
soft tissue, as in Latinieria (Jarvik, 1980). 
A preopercular lies rostral to the opercu- 
lar. The poor preservation of the sutures 

Immature Rhizodontids from North America • Davis et al. 175 

10 mm 

Figure 2. ANSP 20980 in part. (A) Photograph of specimen. (B) Labeled drawing: light shading represents fossil or fossil 
impression, dark shading represents matrix. Boxed area refers to Figure 5. 

Abbreviations for Figs. 2-7: ano, anocleithrum; bo, internal boss of scale; cen, centra; dm, cleithrum; dm. dl, dorsal lamina of 
cleithrum; dm. vl, ventral lamina of cleithrum; civ. dl, dorsal lamina of clavicle; civ. vl, ventral lamina of clavicle; d1, d8, digits 1, 
8; de, dentary; epi, epicaudal lobe; f. par, parasymphysial fang; f. vom, vomerine fang; gul, gular; H, humerus; hyp, hypocaudal 
lobe; iciv, interdavide; i, intermedium; j. lep, jointed lepidotrichia; Ibw, lateral body wall; m. ex, median extrascapular; mil, main 
lateral line; op, opercular; pcf, pectoral fin; pif, pelvic fin; pop, preopercular; pot, posttemporal; R, radius; subm, submandibular 
series; sll, secondary lateral line; su, supradeithrum; U, ulna; u, ulnare; unj. lep, unjointed lepidotrichia. 

176 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 3. ANSP 20980 in part. (A) Photograph of specimen. (B) Labeled drawing. Boxed area refers to Figure 6. 

Immature Rhizodontids from North America • Davis et al. 177 

5 mm 

Figure 4. ANSP 20981 in visceral view. (A) Photograph of specimen. (B) Labeled drawing. 

178 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

makes it impossible to identify the rostral 
and dorsal margins of this bone. 

Dentanj. On ANSP 20981, a left den- 
tary is preserved lateral to the left gular 
and submandibulars (Fig. 4). The distal 
portion of the dentary is obscured. A series 
of mesiodistally oriented striations orna- 
ment the labial surface of the dentary. No 
sutures are \dsible that would indicate the 
presence of infradentary bones, nor is it 
possible to identify coronoids or coronoid 
teeth. Total dentary length is estimated to 
be approximately 20 mm. 

Four incomplete teeth are preserved on 
the dentary. These teeth are oval in cross 
section and are buccolingually com- 
pressed. The largest and most mesial of 
the four is a parasymphysial fang (Fig. 4, 
f. par). Rhizodontid teeth are generally sig- 
moid in shape and possess lingual stria- 
tions (see Andrews, 1985; Long, 1989). 
The extreme small size of the teeth and 
state of preservation make it difficult to 
determine whether ANSP 20981 possesses 
these tooth characters. The dentaiy of 
ANSP 20981 is similar to that ascribed to 
PStrepsodus anculonaniensis (RSMGY 
1980.40.36 [Andrews 1985, figs. 9b, c]) in 
the overall proportion of the dentary and 
the position of the parasymphysial fang. 

Extrascapular Series. An incomplete 
median extrascapular is the only cranial el- 
ement that can be identified with confi- 
dence on ANSP 20980. The bone is rough- 
ly trapezoidal in shape, narrowing crani- 
omedially. A craniolaterally directed occip- 
ital commissural canal crosses the center 
of the median extrascapular (Fig. 5). In 
Gooloogpngia and Bararneda, only a small 
area of contact is present between the me- 
dian extrascapular and the postparietals 
(Johanson and Ahlberg, 1998, figs. 2d, f). 
It is not possible to determine the nature 
of this contact on ANSP 20980 because 
the rostral margin of the median extra- 
scapular is incomplete and postparietals 
are not preserved. 

Pectoral Girdle. The articulation be- 
tween the dorsal laminae of the clavicle 
and cleithrum is preserved on both left 

and right sides of ANSP 20980 and on the 
right side of ANSP 20981. The rostral and 
caudal margins of the dorsal lamina of the 
cleithrum are subparallel to each other 
(Fig. 5). This condition is also seen in the 
specimen of Sauripterus described by 
Daeschler and Shubin (1998) and in Goo- 
loogongia (Z. Johanson, personal commu- 
nication). In other rhizodontids the dorsal 
lamina tends to narrow at midlength be- 
fore expanding to meet the ventral lamina. 

Rhizodontid cleithra possess a de- 
pressed flange that extends along the cau- 
dal margin of the dorsal lamina; this flange 
is absent on ANSP 20980 and ANSP 
20981. The right cleithrrun associated with 
the fin of Sauriptenis described by Daes- 
chler and Shubin (1998) also lacks this 
flange, as does that of Gooloogongia (Z. 
Johanson, personal communication). 

The cleithra are split internally between 
part and counterpart, leaving cancellous 
bone exposed in many places. However, 
the dorsal lamina of the right cleithrum is 
completely preserved in internal view. The 
external surface is exposed at the intersec- 
tion of the dorsal and ventral laminae. 
Both the internal and external surfaces are 
ornamented with subparallel ridges that 
extend dorsoventrally (Fig. 5). 

On ANSP 20980 the ventral laminae of 
the right clavicle and cleithrum overlap 
those of the left as a result of postmortem 
inturning and compression of the entire 
girdle (Fig. 5). In ANSP 20981, the re- 
verse is true with the left ventral laminae 
of both the cleithrum and clavicle overlap- 
ping the right. The lateromedially directed 
ventral contact between clavicle and 
cleithnuTi is preserved in ANSP 20981 
(Fig. 4). 

An interclavicle lies in association with 
the ventral laminae of the clavicles and 
cleithra in both ANSP 20980 and ANSP 
20981. The interclavicles have been dis- 
placed dorsocaudally from the position 
they were likely to have occupied in life 
(Figs. 4, 5). In both specimens the rostro- 
caudal length of the interclavicle appears 
to be greater than that of the ventral cla- 

Immature Rhizodontids from North America • Davis et al. 179 


5 mm 

Figure 5. Shoulder girdle of ANSP 20980 in ventral view. (A) Photograph of specimen. (B) Labeled drawing. 

180 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

vicular lamina. This condition suggests that cleithrum is a rectangular bone that con- 
the interclavicle may have been incorpo- tacts the posttemporal rostrally and the 
rated into part of the contact between ven- rostrodorsal margin of the cleithrum cau- 
tral cleithral laminae. The elongation of dally. The posttemporal is somewhat 
the interclavicle caudally is a unique con- smaller and more triangular in shape than 
dition that is not seen in other rhizodon- the supracleithrum. The posttemporal 
tids or sarcopterygians. An interclavicle is contacts the median extrascapular rostro- 
not preserved with either known specimen laterally and the supracleithrum caudally. 
of San riptenis (Hall, 1843; Daeschler and The anocleithrum is subdermal, a condi- 
Shubin, 1998), making it impossible to de- tion shared with porolepiformes, actinistia, 
termine whether the elongation of the in- and dipnoi (Alilberg, 1989). In osteolepi- 
terclavicle is a characteristic of ANSP formes (e.g., Eusthenopteron [Jarvik, 
20980 and ANSP 20981 only or o£ Saurip- 1980]) the anocleithrum contacts the su- 
teni.s in general. ANSP 20981 is also un- pracleithrum rostrally and the cleithmm 
usual in possessing cleithra that meet at caudally, preventing contact between the 
the ventral midline, lending further sup- supracleithrum and cleithrum. 
port to the possibility that the interclavicle Paired Fins. In ANSP 20980, the lead- 
was fused to the symphysis of the cleithra. ing edge of the pectoral fin forms a gentle 
Contact between the ventral cleithra lam- arc, with the greatest degree of curvature 
inae is also preserved in the adult Saiirip- at midlength (Fig. 6). The trailing edge of 
terus specimen. the fin is composed of poorly presei"ved 

On ANSP 20980, a poorly preserved lepidotrichia that appear jointed. The bulk 

right scapulocoracoid is present medial to of the fin is supported by lepidotrichia that 

the caudal margin of the broken cleith- are unjointed for most of their length. The 

rum. The scapulocoracoid is preserved as lepidotrichia are arranged in two layers, 

a small section of diaphanous bone, similar one forming the dorsal surface of the fin 

in texture to the endochondral elements of and one forming the ventral surface. All 

the pectoral fin. The lack of any diagnostic endochondral elements, except for the hu- 

morphology on the scapulocoracoid is merus, thus are enveloped dorsally and 

most likely due to the ontogenetic stage of ventrally by unjointed lepidotrichia. These 

the specimen, as the endochondral ele- layers span all but the most proximal re- 

ments of the pectoral fin are also weakly gion of the fin, with preaxial lepidotrichia 

ossified (see below). The scapulocoracoid extending more proximally than do those 

appears to lie dorsal to the curvature on the postaxial edge. 

formed by the intersection of the dorsal The endochondral elements of the fin 

and ventral cleithral laminae. This relative- are weakly ossified; the central region of 

ly dorsal position of the scapulocoracoid is each element consists of relatively dense 

seen in other rhizodontids such as Strep- bone matrix, whereas the cortical regions 

sochis (Andrews, 1985) and Sauriptems are thin and translucent. The five elements 

(personal observations). In osteolepids, the that are present correspond to the humer- 

scapulocoracoid assumes a more ventral us, radius, ulna, intermedium, and ulnare 

position than in rhizodontids and rests in of derived sarcopterygians. The humerus 

the curvature formed by the intersection of ANSP 20980 lacks postaxial processes, 

of the cleithral laminae. and thus differs from those of all other rhi- 

All bones of the supercleithral series are zodontids. The radius is anteroposteriorly 

preserved in articulation on the left side of broader than the ulna. The overall pro- 

ANSP 20980. Corresponding elements are portions of the radius of ANSP 20980, rel- 

present on the right side, but are disartic- ative to the ulna and humerus, are very 

ulated and partially obscured by the dorsal similar to those of Sauripterus (Daeschler 

lamina of the right cleithrum. The supra- and Shubin, 1998). Other rhizodontids. 

Immature Rhizodontids from North America • Dovis et al. 181 

5 mm 

Figure 6. Left pectoral fin of ANSP 20980 showing endoskeletal and dermal elements. (A) Photograph of specimen. (B) Labeled 

182 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

5 mm 

Figure 7. Comparison of the pectoral fins of (A) ANSP 20980, (B) Sauripterus, and (C) Barameda. Barameda modified from 
Long (1989). 

such as Barameda, possess a radius that is 
narrower than the ulna (Fig. 7). Further- 
more, a relatively narrow radius is also 
seen in osteolepids, including derived taxa 
such as Panderichthijs (Vorobyeva and 
Schultze, 1991). Like Barameda and Sau- 
riptenis, the ulnare also lacks postaxial 
processes. In addition, the ulnare and in- 
termedium terminate at the same distal 
level, a feature shared with Saiiriptenis, 
Barameda, and basal tetrapods. 

The richly branched series of preaxial 

radials that distinguish the fins of Bara- 
meda and Sauripterus are not obseived in 
the pectoral fins of ANSP 20980. Further- 
more, there are no articulations between 
any of the endochondral elements, despite 
the fact that all bones appear to be pre- 
served in situ. The distinct separations be- 
tween adjacent endochondral bones may 
reflect the immaturity of the specimens, as 
all bones are also weakly ossified. The lack 
of distal radials may also be attributed to 
ontogeny given that they are the most dis- 

Immature Rhizodontids from North America • Davis et al. 183 

tal elements of the fin. If ossification of the the preservation or ontogenetic stage of 

endoskeleton proceeded proximally to dis- the specimen cannot be determined. Rhi- 

tally, then distal radials may not have os- zodontids may have exliibited a high de- 

sified by this stage. ANSP 20981 repre- gree of variation in the timing and degree 

sents a smaller individual and, indeed, of endochondral ossification. For example, 

there is no evidence of endochondral os- Foulden rhizodontids Strepsodus and Rln- 

sification in either pectoral fin. The entire zodus show variation in the degree of os- 

fin is composed of long, unjointed lepi- sification of the axial skeleton that is in- 

dotrichia. The absence of endochondral el- dependent of body size, 
ements may be due to lack of preservation. Unpaired Fins. Vertebrae and the pelvic 

but it suggests that ANSP 20981 may be girdle are not preserved and presumably 

ontogenetically younger than ANSP were unossified. The central lobe of the 

20980. Both specimens indicate that ossi- caudal fin contains a linear series of in- 

fication of the fin dermal skeleton preced- completely developed centra (Fig. 3B). 

ed that of the fin endoskeleton. These ossifications cover the dorsal and 

The pectoral fins of ANSP 20980 are ventral surfaces of small impressions in the 

disproportionally large for a sarcopterygi- central axis of the fin. Although lepidotri- 

an. Ratios of fin length to body length for chia are present for both the epaxial and 

ANSP 20980 can be compared to the type hypaxial lobes of the caudal fin, they are 

of PStrepsodus anculonamensis (RSMGY incomplete distally Dorsal and anal fins 

1980.40.36 [Andrews, 1985]) and other os- are not preserved. 

teolepids. The proximodistal length of the Body Scales. Body scales are split be- 

pectoral fin of Strepsodus is approximately tween part and countei-part leaving most 

15% of the total body length (50 mm fin/ scales exposed in internal view. Like other 

345 mm estimated total length). This ratio rhizodontids, cosmine is absent. Scales are 

is not unusual for a sarcopterygian: Eusth- thin, roughly circular in shape, and have a 

enopteron (14.9%, based on Jarviks 1980 series of bony ridges that radiate from an 

reconstruction), Osteolepis (15.3% [Jarvik, unornamented central plateau (Fig. 6). 

1980]), and Panderichthys (16.5%, based Concentric rings of bone connect these 

on Vorobyeva and Schultzes 1991 recon- ridges, giving the surface a woven appear- 

struction) all share similar proportions, ance. There are none of the breaks or gaps 

ANSP 20980 has a pectoral fin that is near- in these ridges that have been hypothe- 

ly one fourth of the animal's body length sized to be the growth lines described for 

(62 mm fin/254 mm estimated total length, other species such as Strepsodus (An- 

or 24.4%). When corrected for body drews, 1985). Internally, scales possess a 

length, the fin surface area of the new rhi- fusiform boss with the rounded caudal end 

zodontid is 2.4 times that of Strepsodus. underlying the growth center. The sugges- 

The type of PStrepsodus anculonamensis is tion has been made that these bosses are 

also considered to be an immature speci- either points of scale attachment to un- 

men, yet its fin proportions are similar to derlying tissues or structural supports for 

those of adult rhizodontids, as well as oth- the nonoverlapping part of the scale (An- 

er sarcopteiygians. The proportionally en- drews, 1985). 

larged fin of ANSP 20980 may, in part. The thin scales that cover the ventral 
correlate to its ontogenetic stage, but may and dorsal surfaces of the pectoral fin con- 
also be a unique characteristic of either form to the shape of the underlying lepi- 
the genus Sauriptenis, or of this specimen, dotrichia without any sign of fracture. The 

A series of small, unjointed lepidotrichia most distal fin scales extend approximately 
are the only pelvic fin elements that are 15 mm beyond the distal edge of the in- 
preserved (Fig. 3). Whether the lack of en- termedium (approximately three fifths of 
dochondral elements in this area is due to total fin length from base to tip). This pat- 

184 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

tern of overlap contrasts with the speci- 
men of PStrepsodus anculonamensis de- 
scribed by Andrews (1985: 73) where 
scales extend "almost" to the fin margin. 

Lateral Lines. Rhizodontids are highly 
derived in that they have multiple lateral 
line canals. On ANSP 20980, sensory 
pores of the main lateral line canal extend 
along the dorsal body wall from the cleith- 
rum to the caudal fin (Figs. 2, 3). These 
oval-shaped pores possess no discernible 
internal morphology, and are flanked dor- 
sally and ventrally by elongate ridges of el- 
evated bone. Faint traces of accessory lat- 
eral line canals can be seen along the scale 
rows that lie dorsal and ventral to the main 
lateral line. Although no sensory pores can 
be identified, accessory lines can be in- 
ferred from the presence of a furrow that 
runs craniocaudally along the center of the 
scale rows adjacent to the main line. In- 
deed, this furrow corresponds in appear- 
ance to the accessory lines described for 
Strepsodus (Andrews, 1985, fig. 5c). The 
presence of multiple lateral lines may be a 
synapomorphy of Rhizodontida, as all rhi- 
zodontids for which articulated scale rows 
can be identified (Sauriptenis, Gooloogon- 
gia, and Strepsodus) possess accessory lat- 
eral lines. The only other evidence of sen- 
soiy structures on ANSP 20980 is the oc- 
cipital commissural canal on the median 


The small size and incomplete ossifica- 
tion of the skeletons suggest that ANSP 
20980 and ANSP 20981 are immature in- 
dividuals. Indeed these features account 
for most of the differences between the 
fins of ANSP 20980 and ANSP 20981 and 
those of other specimens of Sauripterus. 
These differences are most profound in 
the endochondral portion of the fin skel- 
eton, where these elements are poorly os- 
sified, lack articulations between corre- 
sponding bones, and lack any processes. In 
ANSP 20980 the cortex of each endochon- 
dral element is less ossified than the med- 
ullary regions. Furthermore, the lack of ar- 

ticulations between endochondral ele- 
ments differs greatly from other rhizodon- 
tids; larger specimens of Sauripterus and 
Barameda possess well-defined articula- 
tions between all endochondral bones. 
Likewise, the absence of both distal radials 
and a postaxial process on the humerus 
may be ontogenetic features that relate to 
the degree of ossification of the specimen. 
All rhizodontids, and virtually all sarcop- 
terygians, have postaxial processes on the 
humerus. In addition, all rhizodontids pos- 
sess numerous preaxial radials, many of 
which are richly branched (Fig. 7). 

The degree of ossification of the axial 
skeleton could also be used to assess the 
ontogenetic stage of the new specimens. 
In ANSP 20980, no hemal or neural arch- 
es are observed in any portion of the axial 
skeleton. Although trunk centra are absent 
in the caudal fin, there are the impressions 
of three segmented units, whose dorsal 
and ventral surfaces are ossified. We inter- 
pret these ossifications to represent par- 
tially ossified ring centra. Ring centra are 
known from the Foulden material, from 
Baranneda, and are associated with the 
type of Sauripterus. However, axial ossifi- 
cation does not always correlate to body 
size. Small, presumed immature, individ- 
uals of Rhizodus and Strepsodus possess 
partially ossified neural and hemal arches 
caudal to the first dorsal fin. Similarly, the 
large rhizodontids described from Foulden 
do not possess ossified ring centra (An- 
drews, 1985). The complete lack of neural 
arches, hemal arches, and caudal fin sup- 
ports, and the partially ossified caudal ring 
centra are suggestive, although not defin- 
itive, evidence of an immature condition. 


The discovery of immature Sauripterus 
provides new material for hypotheses 
about the evolution of fin development 
and function in extinct sarcopterygians. 
Three stages of growth are currently 
known for fin development in Sauriptenis. 
In the earliest stage, represented by ANSP 
20981, unjointed lepidotrichia are promi- 

Immature Rhizodontids from North America • Davis et al. 185 

iient and no endochondral ossifications are velopment of the dermal skeleton seems to 
present in the fin. In a later stage, such as occur in rhizodontids, even in immature 
that seen in ANSP 20980, the endochon- forms. The dermal skeleton, both within 
dral skeleton is weakly ossified and lacks the fin and across the entire pectoral girdle, 
the distal preaxial radials seen in adults, is well ossified, whereas endochondral ele- 
Adults, such as those of Sauriptenis, con- ments are weakly developed, 
tain fins with pronounced lepidotrichia It would seem paradoxical that the ex- 
and endochondral radials. The endochon- pansion of the endochondral radials in rhi- 
dral skeleton does not play a role in sup- zodontids is correlated with the origin of 
port and locomotion until later stages of large and unjointed dermal rays. After all, 
growth, after the animal reaches a body the endochondral elements would not in- 
size of at least 25 cm. By the time the an- teract with the substrate: the main surfaces 
imal is an adult, dermal and endochondral of the fin would be entirely defined by the 
fin supports are both greatly expanded and lepidotrichia. Why expand endochondral 
ossified. Therefore, the ossification and skeletal elements that do not seem to play 
elaboration of endochondral fin supports a direct role in support and locomotion? 
may correlate with the functional demands The answer to this question may lie in the 
placed on these large predatory fish. fact that in adult Sauripterus the endo- 
This emphasis on both dermal and en- chondral bones provide surfaces for the at- 
dochondral elements in rliizodontids is an tachment of muscles. The humerus and ra- 
unexpected intermediate condition be- dius, in particular, contain crests and pro- 
tween ray-finned and lobe-finned designs cesses for muscles that presumably would 
and would not be predicted from current have played a role in motions at the shoul- 
models of dermal fin development. Thor- der and elbow. The enhancement of the 
ogood (1991) proposed tliat the differences endochondral skeleton in rhizodontids 
between ray-finned and lobe-finned de- niay be correlated to their increased role 
signs are due to a heterochronic shift dur- ^s control elements for the dermal fin skel- 
ing development. The relative amounts of eton. The large size of rhizodonts and, per- 
dermal skeleton and endoskeleton in the fin ^^pg aspects of their locomotor and feed- 
are hypothesized to relate to the timing of j^g strategies, may have called for fins with 
the shift between an apical ectodermal j^rge surface areas. The enlarged surface 
ridge and apical ectodermal fold during ^rea was established by die expansion of 
early fin bud development (Fig. 8). The hy- ^^g dermal fin supports, whereas the con- 
pothesis suggests that ray-finned fishes, ^j-qJ of movements was provided by elab- 
such as teleosts, have an early shift from oration of the endochondral radials. 
ridge to fold, leading to a greater propor- 
tion of the skeleton being of dermal origin. ACKNOWLEDGMENTS 
A developmentally later shift from ridge to 

fold, or the extreme case of no shift what- This paper is dedicated to Fuzz Cromp- 
soever, would result in an appendage that ton in recognition of his patlibreaking syn- 
is primarily or exclusively endochondral in tliesis of experimental morphology and ex- 
design. The new specimens, and the pre- peditionary paleontology. We would fike to 
viously described Sauripterus fins, repre- thank Doug Rowe for many years of dili- 
sent a mosaic between these morphological gent excavation at Red Hill, C. Frederick 
and developmental extremes. Witli an ex- Mullison for his skillful preparation of the 
tensively developed dermal skeleton, and specimens, and Robert Carroll and FUchard 
interleaved endochondral elements, San- Cloutier for their generous access to Eiisth- 
nptenis does not fit in a model based solely enopteron specimens for comparative pur- 
on comparisons between teleosts and de- poses. We also wdsh to thank Zerina Johan- 
rived sarcopterygians. Indeed, a great de- son and Farish A. Jenkins, Jr., for helpful 

186 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 




Figure 8. (A) Transition from apical ectodermal ridge (AER, left) to apical ectodermal fold (AEF, right) during early fin devel- 
opment. Cross section along proximodistal axis. (B) "Clock model" of heterochrony and phenotype. A developmentally early 
transition from AER to AEF would result in an "actinopterygian" phenotype (here represented by the pectoral fin of Amia). A 
developmentally later transition from AER to AEF would result in a "sarcopterygian" phenotype (here represented by Eusthen- 
opteron). A and B modified from Thorogood (1991). Fins modified from Jarvik (1980; Early) and Andrews and Westell (1970; Late). 

discussions. Kalliopi Monoyios provided 
editorial assistance. This research was sup- 
ported by the Academy of Natural Sciences 
of Philadelphia and the National Science 
Foundation (EAR 9628163 to N. H. S.). 


Ahlberg, p. E. 1989. Paired fin skeletons and rela- 
tionships of the fossil group Porolepiformes (Os- 
teichthyes: Sarcopterygii). Zoological Journal of 
the Linnean Society, 96: 119-166. 

Immature Rhizodontids from North America • Davis et al. 


Andrews, S. M. 1985. Rhizodont crossoptetygian 
fish from the Dinantian of Foulden, Berwick- 
shire, Scotland, with a re-evaluation of this 
group. Transactions of the Royal Society of Ed- 
inburgh, 76: 67-95. 

ANDREWS, S. M., AND T. S. Westoll. 1970. The 
postcranial skeleton of rhipidistian fishes exclud- 
ing Eusthenopterou. Transactions of the Royal 
Societ)' of Edinburgh, 68: 391^89. 

Daeschler, E. B., and N. H. Shubin. 1998. Fish 
widi fingers? Nature, 391: 133. 

Gregory, W. K., and H. C. Raven. 1941. Studies of 
the origin and early evolution of paired fins and 
limbs. Annals of the New York Academy of Sci- 
ence, 42: 273-360. 

Hall, J. 1843. Geology of New- York. Part IV. Com- 
prising the Survey of the Fourth Geological Dis- 
trict. Natural History of New York. Vol. 4. Al- 
bany, New York: Carroll and Cook. 683 pp. 

JARVIK, E. 1980. Basic Structure and Evolution of 
Vertebrates. 2 vols. London: Academic Press, xvi 
+ 575 pp. and xiii + 337 pp. 

JOHANSON, Z., AND P. E. Ahlberg. 1998. A complete 
primitive rhizodont from Australia. Nature, 394: 

Long, J. A. 1989. A new rhizodontiform fish from 
the Early Carboniferous of Victoria, Australia, 
wdth remarks on the phylogenetic position of the 
group. Journal of Vertebrate Paleontology, 9: 1- 

RomeR, A. S. 1955. Herpetichthyes, Amphibiioidei, 
Choanichthyes or Sarcopterygii?. Nature, 176: 

ThorogoOD, P. 1991. The development of the tele- 
ost fin and implications for our understanding of 
tetrapod limb evolution, pp. 347-354. In J. R. 

Hinchcliffe, J. M. Hurle, and D. Summerbell 
(eds.). Developmental Patterning of tlie Verte- 
brate Limb. New York: Plenum Press, xi + 452 

Tr^WERSE, A. (in press). Dating the earfiest tetra- 
pods: a Catskill palvaiological problem in Penn- 
sylvania. Courier Forschungsinstitut Sencken- 

scription and systematics of panderichthyid fishes 
with comments on their relationship to tetra- 
pods, pp. 68-109. In H.-P. Schultze and L. Trueb 
(eds.). Origins of the Higher Groups of Tetra- 
pods: Controversy and Consensus. Ithaca, New 
York: Cornell University Press, xii + 724 pp. 

WOODROW, D. L. 1985. Paleogeography, paleocH- 
mate, and sedimentary processes of tlie Late De- 
vonian Catskill Delta, pp. 51-63. In D. L. Wood- 
row and W. D. Sevon (eds.). The Catskill Delta, 
Special Paper 201. Boulder, Colorado: Geologi- 
cal Society of America, vii + 246 pp. 

WOODROW, D. L., R. A. J. ROBINSON, A. R. Prave, 
A. TRAVERSE, E. B. Daeschler, N. D. Rowe, 
AND N. A. Delaney. 1995. Stratigraphic, sedi- 
mentologic, and temporal framework of Red Hill 
(Upper Devonian Catskill Formation) near Hy- 
ner, CUnton County, Pennsylvania: site of the 
oldest amphibian known from North America, 
pp. 1-8 In J. Way (ed.), 1995 Field Trip Guide, 
60th Annual Field Conference of Pennsylvania 
Geologists. Lock Haven, Pennsylvania. 

Young, G. C, J. A. Long, and a. Ritchie. 1992. 
Crossopterygian fishes from the Devonian of 
Antarctica: systematics, relationships and biogeo- 
graphic significance. Records of the Austriilian 
Museum, 14(Suppl): 1-77. 



Abstract. Mysticete whales trap prey in a sieve of 
baleen, the structure of which varies in such param- 
eters as the number of plates, their overall dimen- 
sions, and the number and density of hairlike fringes 
that form on the medial surface, creating filters of 
different mesh size. Many prey items presumably en- 
tangle in the mat of interwoven fringes, necessitating 
that they be freed for oral transport and swallowing. 
Although the tongue is commonly implicated in such 
removal, this has never been studied. Three hypoth- 
eses for prey release are presented: mechanical scrap- 
ing or shaking of baleen by the tongue, shaking of 
the head or lips to dislodge prey with the aid of grav- 
ity, and use of a rapid flushing "backwash" current to 
draw water into the mouth through baleen. Pre\d- 
ously unpublished data on mysticete tongue structure 
and function support all hvpotheses; behavioral ob- 
servations of foraging whales and morphologic find- 
ings such as baleen growth and wearing also aid in 
their evaluation. All three mechanisms are likely to 
be used depending on the type of baleen and size, 
type, and density of prey. 


For as long as whales have been distin- 
guished from fishes, the role of baleen or 
"whalebone" in trapping prey has been 
recognized. Indeed, given this singular, 
highly derived tissue s unique consti-uction 
and arrangement in rows of serial plates, 
baleen's sievelike function — filtering large 
quantities of small prey from high volumes 
of seawater — seems obvious, all the more 
so with cursory examination of mysticete 
diet and foraging behavior. 

Yet for all that has long been known 
about prey capture in the mammalian sub- 
order Mysticeti ("mustached" whales), one 
aspect is only dimly understood: how are 
entangled prey removed from the filter be- 
fore transport and deglutition? The tongue 

' Department of Biology, Hampden-Sydney Col- 
lege, Hampden-Sydney, Virginia 23943. 

is commonly implicated in this function, 
but this is a grossly inadequate and vague 
answer. Because of manifest logistical lim- 
itations a clear, Jonahs-eye view inside a 
live whales closed mouth is unattainable. 
The experimental techniques of cineradi- 
ography and electromyography that have 
greatly elucidated precise mechanisms of 
lingual function in other inammals are, re- 
grettably, inapplicable. Hence the best 
(and indeed perhaps only) way to address 
this question is to rely on speculative in- 
ference based on the functional anatomy 
of the tongue and associated oral struc- 
tures. This paper introduces and com- 
pares, on the basis of moiphological and 
ecological evidence, three hypotheses con- 
cerning removal of trapped prey from 
mysticete baleen. 


Before proceeding to a discussion of 
plausible prey removal mechanisins it is 
necessary to consider briefly the mecha- 
nisms by which prey are captured. Mysti- 
cetes vary dramatically in foraging meth- 
ods. Variation occurs mainly along family 
lines, and is reflected in the baleen itself 
as well as in other oral stiiictures and (con- 
comitantly) in diet and foraging ecology, all 
of which have consequences for prey re- 

Foraging and Diet 

Right whales (Balaenidae) utilize a type 
of suspension feeding similar to that of 
more primitive vertebrates, which pump 
or push water constantly and unidirection- 
ally through the mouth (Sanderson and 
Wassersug, 1993). Balaenid skim feeders 

Bull. Mus. Comp. ZooL, 156(1): 189-203, October, 2001 


190 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

cruise slowly through swarms of minute ring the bottom and sucking in muddy wa- 
zooplankton, primarily copepods, amphi- ter from which prey are winnowed (Klaus 
pods, and euphausiids (Lowry and Frost, et al., 1990; Weitkamp et al., 1992). Ram 
1984; Carroll et al., 1987). Skim feeding is gulping necessitates forward lunges paral- 
inost commonly observed at the oceans lei to or through the surface; the gray 
surface (Watldns and Schevill, 1979; Mayo whale can feed while stationaiy. In both 
and Mai-x, 1990), yet stomach content data cases prey are filtered as the mouth closes 
and scratched, muddy rostra indicate that to expel engulfed water (and, in the gray 
skim feeding occurs at all levels of the wa- whale, sediment). As Pivorunas noted 
ter column, including the bottom. The (1979), baleen does not actually catch 
enormous head, constituting one third of prey; it merely retains prey as water is ex- 
an adult balaenid's 15- to 20-in length, pelled from the oral cavity, 
functions as an immense plankton tow net As expected, the type, size, and abun- 
(as in other continuous filter feeders such dance of preferred prey correlate with for- 
as manta rays and whale and basking aging ecology. For example, the sei whale 
sharks), although the "seine" is not pulled (Balaenoptera horealis) occasionally eats 
along but rather propelled by the whale's schooling fish as do other rorquals, yet 
foiAvard locomotion at leisurely speeds of more commonly feeds on copepods or oth- 
approximately 4 km per hour during for- er zooplankton and hence engages in skim 
aging (Reeves and Leatherwood, 1985; feeding much like that employed by ba- 
Carroll et al., 1987; Lowry, 1993). A con- laenids (Kawamura, 1974). The gray whale 
stant current of prey-laden water enters primarily uses suction to ingest benthic 
anteriorly, passes through baleen "racks" macroinvertebrates (e.g., gammarid ani- 
on either side of the mouth, and exits lat- phipods, inysids, and molluscs; Murison et 
eral to the phaiyngeal orifice at the trailing al., 1984; Nerini, 1984) yet has been ob- 
edge of the lips. Tiny zooplankton in the served feeding on shoals of fish in luid- 
steady stream of incurrent water are water (Sund, 1975). Although well-docu- 
caught in the finely fringed baleen, and mented ecological partitioning exists with- 
field obseiA/ations suggest that right whales in mysticetes, opportunism is the rule. Var- 
{Euhalaena glacialis) graze for hours with iation in type, size, and abundance of prey 
few interruptions for breathing or closing might alter the behavior of prey capture as 
the mouth (Watkins and Schevill, 1979; well as removal of trapped prey in any 
Mayo and Mai'x, 1990), leading to anec- mysticete species, 
dotal speculation that they may spend long 
periods skimming before gathering ^"^^ MOrpnoiogy 

enough prey to swallow. The differences between continuous 
In contrast, other mysticetes are not and intermittent (ram and suction) filter- 
continuous filter feeders but are intermit- feeding inysticetes are manifested not only 
tent filter feeders that ingest discrete in diet but more aptly in key features of 
mouthfuls of water and separate food from oral anatomy. Unlike other mysticetes, ba- 
this water before expelling it. A single laenids possess no throat grooves yet have 
mouthful is engulfed either by ram feed- a large anterior cleft (the subrostral gap) 
ing in rorquals (Balaenopteridae) such as between baleen racks. The high, arched 
fin and humpback whales (Jurasz and Jur- skull accommodates extremely long, nar- 
asz, 1979; Watkins and Schevill, 1979; row baleen plates (Fig. 1), and the huge 
Hain et al., 1982) or by intraoral suction semicircular lower lips, which extend far 
in the gray whale (E.schrichtiiis rohustus, above the mandible (like the arched, deep- 
Eschrichtiidae), whose tongue is rapidly sided lower jaw of flamingos; Milner, 
depressed and retracted to expand the oral 1981), cover the baleen laterally and en- 
cavity and create negative pressure, stir- fold the narrow rostrum when the mouth 

Mysticete Prey Removal • Werth 


Figure 1. Baleen size, shape, and fraying correlate with prey type and size. Right whales (top) skim microplankton with narrow, 
finely fringed baleen >4 m in length. Rorquals (center) gulp schooling shrimp and fish with shorter (< 1 m), wider, coarser baleen. 
The gray whale (bottom) sucks in benthic invertebrates and filters them with short (<40 cm), coarsely fringed baleen. 

is closed. Like the subrostral gap, the or- 
olabial sulcus — a gutterlike groove medial 
to the lip — promotes continuous, unidirec- 
tional flow as filtered water passes to an 
"exliaust port" at each lip's trailing edge, 
lateral to the phaiyngeal opening. The 
tongue is firm, muscular, and extraordi- 
narily large; it is estimated to average 4- 
6% of total body mass (Omura, 1958), so 
that in a standard 50-ton or 50,000-kg 
whale, the tongue weighs 2,000-3,000 kg 
and measures several meters in length. 

Ceivical vertebrae are fused, yet the man- 
dibular symphysis is loose and the lips 
highly mobile, controlled by labial mus- 
culature (Lambertsen et al., 1989). The 
gray whales tongue is also firm and mus- 
cular and this whale has a small subrostral 
gap, although this is far less pronounced 
than the gap found in balaenids (Wolman, 
1985). The scarred, abraded jaws of Es- 
chrichtiiis disclose its benthic suction 
feeding, as do mud plumes trailing behind 
feeding whales and suction-generated pits 

192 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

that scar the ocean floor (Oliver and Slat- filter itself varies substantially in numerous 

tery, 1985; Nelson and Johnson, 1987). ways (Fig. 1), including the number of 

Likewise, rorquals possess singular mor- flexible triangular laminae or plates in each 
phologic features for lunge feeding. Posi- rack, which ranges from 100 in Eschri- 
tive inertial pressure opens the mouth just chtius to 480 in the fin whale (Balaenop- 
as a bag is opened by pulling it through tera physalus), although the bowhead (Ba- 
air; water and prey are passively enveloped laena niysticetus) and most large rorquals 
rather than displaced forward or sucked generally have about 300—350 plates per 
internally (Orton and Brodie, 1987). Key rack (Leatherwood et al., 1983). Plates are 
innovations include the flaccid and de- arranged transversely like teeth of a comb 
formable tongue and oral floor with inter- and suspended from the maxillae at inter- 
muscular fascial cleft; the cavum ventrale, vals of roughly 1 cm. Plates are generally 
which receives engulfed water and the dis- only a few millimeters thick (anteroposte- 
placed baglike tongue (von Schulte, 1916; riorly) yet vary greatly between species in 
Pivorunas, 1979); accordionlike longitudi- other dimensions, notably length, measur- 
nal throat pleats and elastic throat wall ing just 5—25 cm in Eschrichtius yet often 
(Orton and Brodie, 1987); wide-opening exceeding 4 m in Balaena (requiring that 
jaws with locking temporomandibular joint they fold posteriorly as the mouth closes), 
to prevent opening during rapid locomo- However, baleen seldom grows wider than 
tion and frontomandibular stay to store Id- 35 cm in any species, so that the shape of 
netic energy for jaw closure (Lambertsen, the triangle varies markedly. Baleen also 
1983; Lambertsen et al., 1995); unfused differs in such characteristics as latero- 
mandibular symphysis with fibrocartilage medial curvature (Lambertsen et al., 
arms extending to mandibular raiui (Pivo- 1989), flexibility (very stiff in Eschrichtius; 
nanas, 1977); and flat, streamlined rostrum said to be most pliable in the pygmy right 
(Gaskin, 1976). Storro-Patterson (1981) whale, Caperea niarginata; Leatherwood 
speculated that a blue whale (Balaenop- et al., 1983), and less consequential factors 
tera musculus) might engulf 1,000 tons of such as color, although the latter may re- 
water in a single gulp. Pivorunas (1979) late to prey capture, as has been supposed 
gave a more conservative estimate of at for jaw and flipper color (Mitchell, 1970; 
least 60 m^ (approximately 70 tons) of wa- Brodie, 1977). 

ter, still a huge amount of water equal to Certainly most significant from the 

roughly 50% of a blue whale s total body standpoint of prey retention is the varia- 

volume. Lambertsen s recent calculations tion in type and density of baleen fringes, 

from computer modeling (Zackowitz, Although the plates themselves form a sort 

2000) suggest an almost identical engulf- of rudimentary sieve, the true task of fil- 

ment voluine of 15,000 gallons (56.85 m^) tration is accomplished by the many thin, 

in humpbacks. Specialized behaviors of hairlike projections that develop on each 

rorqual lunge feeding, including ingenious plate's medial side, forming a network of 

bubble entrapment devices einitted by the meshed fibers (Tomilin, 1957; Williamson, 

blowholes (Gormley, 1983; Wiirsig, 1988) 1973). All baleen develops as a dermal— 

or lobtailing, flipper slapping, and flick epidermal interaction in which conical 

feeding to concentrate prey (Evans, 1987; dermal papillae extend ventrally from an 

Clapham et al., 1995), are as remarkable underlying basal plate of connective tissue 

and resourceful as the mechanics of en- through an epidermal layer, at which time 

gulfment. the papillae are enveloped in a homy layer 

of keratin to form long bristles called horn 

°^'®®" tubes (Slijper, 1962). These tough, fibrous 

More important than these divergences strands are likewise surrounded and ce- 

in oral morphology and ecology, the baleen mented together by a layer of compacting 

Mysticete Prey Removal • Werth 193 

horn, while a soft cushioning layer of in- 
termediate horn provides a dense cortex 
covering the anterior and posterior faces 
of the plate. Friction abrades the matrix 
medially, wearing away compacting horn 
to reveal the hollow horn tubes that re- 
main as the frayed fibers comprising the 
sieving apparatus, whereas the long side of 
the scalene triangle facing the lip remains 
smooth. Cells in the rubbery, pliant epi- 
thelium of the guins anchor baleen to the 
palate and proliferate to replenish abraded 
gingival tissue, just as all papillae grow at 
a uniform rate to replace worn horn tubes. 
The visible portions of baleen consist ex- 
clusively of dead comified cells, but the 
dermal process remains a living tissue, so 
that baleen is analogous to the part-living 
tissue in an ungulate hoof (Slijper, 1962). 
Baleen's anisotropic nature, with a homo- 
geneous cortical layer surrounding free, 
hollow, cylindrical tubes, affords maximal 
strength with minimal mass (Slijper, 1962). 
Baleen is tough yet elastic, a suitable ma- 
terial to meet the demands of constant 
friction. Sadly, the exceptional physical 
characteristics of this material also ren- 
dered it an extremely valuable commodity 
that fueled the whaling industry. 

The hairlike bristles that fray on the me- 
dial side intertwine to form a fibrous mat. 
The slender, springy plates of the sldm- 
feeding balaenids and sei whale possess 
35—70 fine filainentous fringes per square 
centimeter (Leathei-wood et al., 1983). Mi- 
croscopic examination of bowhead baleen 
reveals several distinct histologic units, in- 
cluding simple (bristle) and compound 
(hair) filaments, according to their gingival 
origin (Haldiman et al., 1981; Haldiman 
and Tarpley, 1993). In contrast, fibers of 
rorqual and gray whale baleen are luuch 
shorter, coarser (thicker in diameter), and 
fewer in overall number and density 
(about 30 fringes per plate; Watson, 1981); 
they are inore like rough scrub-bioish bris- 
tles than fine hairs, and are typically more 
wavy than the fine strands of skim feeders. 
However, these rough bristles are suffi- 
ciently long that they may link with other 

fibers from the same and adjacent plates 
to form an interwoven, loosely braided 
mesh. In a mathematical model correlating 
plate and fringe features, Pivorunas (1976) 
suggested that the angle at which fringes 
develop from the medial surface seems 
more critical for prey retention than fringe 
density, and that shorter fringes alleviate 
drawbacks of having fewer fringes. Where 
plates angle laterally, as in rorquals (Fig. 
2), many fringes are exposed on the medial 
surface by friction, so that the coarse fibers 
in this tangled, brushlike mat need not 
have such siuall diameter, whereas in ba- 
laenids fewer fringes can be exposed on 
the relatively straighter medial side. In any 
event the most fundamental distinction 
between baleen of different species is the 
spacing between fringes, and thus the po- 
rosity of the filter. 

Because baleen is not a rigid material, 
its filter porosity varies according to hy- 
drodynamic factors such as the swimming 
velocity of ram feeders; size and density of 
retained prey; and direction, rate, and 
pressure of water flow (Sanderson and 
Wassersug, 1990). The mysticete filter- 
feeding apparatus acts simply as a sieve — 
no active adjustment of filter porosity ac- 
cording to prey size or density is made be- 
fore engulfment — ^with filter elements (ba- 
leen fringes) spaced inore closely than the 
sizes of items to be captured. However, al- 
though no particles are directly intercept- 
ed by adhering to a stick)' surface, it is pos- 
sible although unlikely that tiny prey 
(smaller than the filter spacing) may be 
trapped by other methods of aerosol filtra- 
tion, such as inertial impaction or gravita- 
tional deposition (Rubenstein and Koehl, 
1977). Although small suspension feeders 
must contend with the constraints inher- 
ent in moving in a dense, viscous fluid, 
mysticete filtration may depend more on 
inertial forces that prevail at high Reynolds 
numbers (Vogel, 1994). 

Baleen whales consume whole organ- 
isms, whereas small filtering organisms of- 
ten feed on detritus (fragiuented organic 
debris; Sanderson and Wassersug, 1993). 

194 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 



gular groove 

Figure 2. Diagrammatic cross sections through the closed 
mouth of a right whale (top), rorqual (middle), and gray whale 
(bottom), showing the different dimensions and relations of oral 
features (baleen, lips, tongue, and mouth floor) that affect prey 
prehension and extrication in these families. 

Yet as with other suspension feeders, mys- 
ticetes feed in abbreviated trophic chains 
and thus reap great energy input, a critical 
factor in their attainment of huge body 
size and (before their decimation by hunt- 
ing) wide distribution. Each whale con- 
sumes gigantic quantities of prey, with es- 
timates ranging from 200-1,000 kg per 

meal and 200,000-600,000 kg annually 
(Gaskin, 1982). Like other strainers mys- 
ticetes are not selective; they locate patchy 
food sources and trap whatever is there. 
Although the filter porosity determines the 
smallest prey retained, fine filters catch 
large prey as well as small and hence may 
be more versatile (Gaskin, 1982). Howev- 
er, dietary studies indicate that coarse- 
fringed mysticetes are less selective (Nem- 
oto, 1959, 1970), ingesting items ranging 
from plankton and fish to hapless seabirds, 
whereas balaenids are specialized feeders 
with more restrictive diets. In fact, it may 
be unnecessarily costly to use a finer filter 
than is needed to trap preferred prey, par- 
ticularly during continuous filtration, be- 
cause this increases pressure drag, slowing 
an animal and preventing capture of large 
or evasive prey. Watkins and Schevill 
(1976) noted that the water level inside 
the mouth of a right whale during surface 
skim feeding is higher than that of the sur- 
rounding sea (although gravity would then 
force water out through baleen). A pho- 
togrammetric study of bowhead baleen 
curvature by Lambertsen et al. (1989) sug- 
gested that hydrodynamic rather than pas- 
sive hydraulic forces may develop, creating 
Bernoulli and Venturi effects within the 
mouth to improve filtering efficiency. 
Werth (1995) devised mathematical and 
physical models of the bowhead mouth to 
test these predictions and confirmed that 
both hydrodynamic effects might reduce 
turbulent flow and avoid creation of an an- 
terior pressure wave so that balaenids 
could capture elusive prey even at slow 
swimming speeds. Foraging in tight for- 
mation may achieve the same effect (Wiir- 
sig, 1988, 1989). 


The process of prey removal from the 
mysticete filter is an intriguing question 
that has not yet been satisfactorily ad- 
dressed, much less resolved. Given the 
fine porosity of the filter (because it in- 
volves baleen strands rather than simple 
laminae) in all species and the small prey 

Mysticete Prey Removal • Werth 195 

size of many species, prey may become 
trapped not only on but actually in this fil- 
ter, necessitating that they be freed before 
they can be swallowed. 

Consider an analogy with a dip net used 
to clean a swimming pool. One way to re- 
move debris that accumulates on the net 
would be to scrape the mesh or einploy 
some other direct mechanical means to 
brush off collected material. Alternatively, 
the net might be shaken vigorously so that 
debris falls off with the aid of gravity. A 
third method relies on hydrodynamic rath- 
er than mechanical forces: by rapidly jerk- 
ing the net backwards, a backwash flow 
would filter through the net and free 
trapped items. Undoubtedly additional 
ways exist to clean the net, yet these are 
the simplest and most obvious methods. 

Just as a clogged dip net must be 
cleaned periodically for effective filtration, 
so too the baleen sieve must be cleared for 
it to continue removing planktonic or nek- 
tonic prey from ingested water. Clearly the 
mysticete filter is more complex than a dip 
net screen, for its pore size is not fixed and 
is likely pressure dependent. Although 
only continuous skimmers appear routine- 
ly to ingest items small enough to be deep- 
ly ensnared in fringes (i.e., copepods 1—5 
mm in length), the rapid, explosive expul- 
sion of water in intermittent filter feeders 
might serve to drive prey further into the 
meshwork of fringes, as a huge volume of 
water exits the mouth at high velocity and 
pressure. Still, the relationship between 
filter element spacing and prey size in 
most intermittent filter feeders — namely 
their coarser fringes and attendant trend 
toward larger prey ( 10- to 50-cm schooling 
fish and squid) — means that much of their 
food accumulates on rather than within 
the sieve during collection (and is unlikely 
to penetrate it during water expulsion). Yet 
euphausiids (10 cm) are a favored prey of 
most rorquals and the gray whale eats 
many small invertebrates (1-15 cm), all of 
which could easily become enmeshed in 
fringes. Clearly, ingestion of any prey 
(large or small) in large quantities would 

mean that many items (not all, yet enough 
to be swallowed) simply fall onto the 
tongue upon water expulsion. Also, most 
macroscopic prey are negatively rheotropic 
(i.e., preferring to swim against a current); 
if still alive they will attempt to swim away 
from the expulsive flow, out of the ensnar- 
ing mesh and into the center of the mouth. 

Baleen, unlike some dip nets, is not 
meant to gather debris, although Eschri- 
cJitius might be expected occasionally to 
collect sediment along with intended prey 
from the benthic substrate. Inorganic ma- 
terial must be removed from the filter so 
as to prevent clogging of the baleen, as 
well as separated froin food so as to pre- 
vent its ingestion, although sand and peb- 
bles have been found in gray whale stom- 
achs (Tomflin, 1954; Pike, 1962). Unfor- 
tunately, although baleen could easily be- 
come fouled with spilled oil (Geraci and 
St. Aubin, 1990; Loughlin, 1994), this ma- 
terial is unlikely to be removed effectively 
by any means, such that not only the toxic 
effects of its ingestion but also the obvia- 
tion of filtration would pose dire conse- 
quences for all mysticetes. 

The combination of coarse brushlike ba- 
leen and large prey commonly ingested by 
rorquals means that their food is far less 
likely to become entangled than in species 
with fine fringes and correspondingly 
smaller prey, particularly sei and right 
whales. The large prey of intermittent fil- 
ter feeders might simply fall onto the 
tongue or swim out of baleen without need 
of any removal mechanisms. Accumula- 
tions of minute prey might also drop in 
this manner, leaving some organisms re- 
maining in the baleen yet creating a suf- 
ficiently large bolus to be swallowed. Re- 
current dislodging of enmeshed items, es- 
pecially tiny prey, may be metabolically 
costly. However, such costs must be bal- 
anced with the need for periodic baleen 
cleaning to present the freest, least 
clogged filter for optimal prey capture and/ 
or to preclude swallowing of too large a 
bolus or too thick a slurry of prey. Do mi- 
croplankton that gather on gill rakers of 

196 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

filter-feeding sharks eventually migrate to lateral — that could abrade, shake, or oth- 
the pharynx in similarly large masses? As erwise disturb or wear the mat of inter- 
with mysticetes, the act of filtration has re- woven fringes and thereby release entan- 
ceived inuch more attention than the topic gled prey. According to this idea, prey ei- 
of how collected prey are processed in- ther fall onto the tongues central furrow 
traorally. However, even if large prey and and are subsequently transported to the 
accumulated small prey drop off baleen rear of the oral cavity for swallowing via 
spontaneously, small prey clearly become lingual elevation and retraction and de- 
enmeshed in the filter, necessitating that pression, or else prey removal and trans- 
they be removed for transport and swal- port steps occur concurrently with such 
lowing (as well as to restore the filters po- lingual movements, perhaps in a cyclic se- 
rosity and efficiency). ries controlled by a central pattern gen- 

By extension from the dip-net analogy, erator. 
I propose three hypotheses to explain the A major drawback of this mechanism is 
most likely mechanisms for removal of that the tongue might push prey more 
prey trapped in baleen fringes. Two are deeply into fringes, furthering entangle- 
mechanical — direct dislodging of items via ment. Other potential disadvantages in- 
lingual scraping and indirect release by elude inefficient clearing of prey from 
vigorous head shaking, whereas one is hy- fringes that do not directly contact the 
drodynamic, relying on a powerful, rapid tongue as well as rapid abrasion and, ulti- 
reversed-flow backwash to flush items, mately, removal of baleen. Baleens occa- 
These options need not be mutually exclu- sional presence in whale feces is offered as 
sive: a species might use all three process- evidence that it regularly wears away, pro- 
es to "cleanse the palate" depending on viding compelling circumstantial support 
prey type and density or other circum- for this hypothesis. Isotopic studies of 
stances. Additionally, the extent to which bowhead baleen confirm that its growth 
individuals might devise unique methods varies with age (Schell and Saupe, 1993), 
of prey removal ought not to be discount- exceeding 50 cm of growth in the first year, 
ed. then decreasing by about 10 cm per year 

Although no discussion of baleen clean- until stabilizing at about 20 cm or less in 

ing has been published previously, super- older animals. Whether the bowhead s an- 

ficial references implicate the tongue in nual addition of this much baleen is suf- 

passing. Indeed the tongue plays a central ficient to offset potential loss from abrasive 

role in two of the baleen cleaning strate- prey removal can only be addressed in the- 

gies presented and analyzed here on the ory. Yet, although abrasion might be hy- 

basis of anatomical and observational evi- pothesized to be reduced in nonsldmming 

dence. Specifically, those mechanisms in- mysticetes that need not release tiny prey 

volving the tongue depend predominantly from fine fringes, Ruud's (1940, 1945) 

on changes in its position rather than studies of fin whale baleen growth provide 

shape, which is supported by preliminary data comparable to those from bowheads, 

study of lingual myology. suggesting a constant level of abrasion not 

correlated with diet and foraging method. 

Prey Removal Via Direct Lingual Scraping ^ more serious fault of this fine of rea- 

or bnaKing soning is that lingual scraping would 

The most common supposition is that abrade baleen's medial surface far more 

the tongue is applied directly to scrape ba- than its lateral surface, resulting in differ- 

leen and free trapped items. Although this ential wear and continually narrowing 

is mainly presumed to involve lingual ele- plates. However, not only does plate shape 

vation and retraction, it might entail any not change with age, but growth is uni- 

motion- — anteroposterior, dorsoventral, or form along the entire base of the plate 

Mysticete Prey Removal • Werth 197 

(Ruud, 1940, 1945), so that such differ- 
ential wear could not be countered by dif- 
ferential growth. Nor could baleen be 
scraped solely from below to wear evenly, 
because baleen angles laterally in all mys- 
ticetes and the tongue contacts only its 
medial surface (Fig. 2). Still, support for 
the lingual scraping hypothesis might 
come from anotlier type of differential ba- 
leen wear seen in gray whales: they seem 
predominantly right handed, with signifi- 
cantly shorter baleen on this side (Kasuya 
and Rice, 1970). This asymmetry has been 
ascribed to friction from benthic suction 
ingestion on the right side (head scarring 
and barnacle placement are likewise asym- 
metrical), yet the wear might be incurred 
not during prey capture but during sub- 
sequent scraping removal of prey and sed- 
iment that is trapped chiefly in the right 

The effect of such scraping on the 
tongue must also be considered, and at 
least in the case of Balaena the dorsum is 
covered by a thick, keratinized stratified 
squamous epithelium with a well-devel- 
oped stratum comeum (Tarpley 1985; 
Haldiman and Tarpley, 1993). A homy, 
cornified corium is similarly present on the 
tongue of Euhalaena (Werth, 1990, 1993), 
yet no data are available for other mysti- 
cetes. Unfortunately, although compara- 
tive mysticete lingual myology would shed 
light on the ability of the tongue to per- 
form the movements necessaiy for the 
scraping motions outlined above, few pub- 
lished data exist. 

The lingual inovements necessary for 
this manner of prey removal involve 
changes in the tongue's position rather 
than its shape. Use of the human tongue 
to remove food particles trapped between 
teeth or on the palate is familiar. Yet, al- 
though the mysticete tongue might deform 
to shorten or curl and thereby contact lo- 
calized regions of baleen, it is likely that 
prey become uniformly distributed 
throughout the filter, so that displacement 
of the entire tongue organ via protraction 
and retraction, elevation and depression. 

and lateral shifting would probably be 
more effective in prey removal than lin- 
gual shape deformation. This view accords 
with myologic findings of balaenid tongues 
(Werth, 1990, 1993), which have extrinsic 
muscles (originating outside the tongue) 
that appear to be greater contributors to 
its body (by mass and cross-sectional area) 
than are intrinsic muscles, which exist 
solely within the tongue. Analysis of un- 
published data (Werth, in preparation) 
froin fresh, frozen, and preserved fetal, ne- 
onate, and adult right and bowhead whales 
suggests that although scattered fibers of 
the inusculus (m.) lingualis proprius, es- 
pecially verticalis (perpendiculares) and 
transversus fibers, are found on the dor- 
sum of the tongue root and tip as they in- 
tergrade with plentiful adipose tissue (pre- 
sumably for nutritional storage or ther- 
moregulation), the m. genioglossus is a 
much larger contributor to the tongue 
body, based on gross examination and cal- 
culation of relative cross-sectional area. 
The m. hyoglossus and m. styloglossus, al- 
though significantly smaller than the m. 
genioglossus, nonetheless are well devel- 
oped in all age classes (see also Lambert- 
sen et al., 1989). Taken together, analysis 
of these data on component muscles (sim- 
ilar to data from an odontocete "great 
whale," the sperm whale; Werth, 1998) 
suggests that although the balaenid tongue 
possesses a limited ability for shape 
change, it is well suited to the elevation, 
retraction, and depression that underlie 
the lingual scraping hyjDothesis. 

Although gray whales also possess a 
large, firm, elevated, muscular tongue, the 
conspicuous flaccidity of the adult rorqual 
tongue would seem to preclude much of 
the activity described here. Pivorunas 
(1979) noted major changes in the balaen- 
opterid tongue as it transformed from a 
solid, muscular structure used in suckling 
to the deformable, flaccid sheet seen in 
adults. From birth until around weaning 
muscle fiber is replaced with adipose and 
elastic connective tissues as the increasing- 
ly saccular organ flattens and spreads lat- 

198 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

erally. Intrinsic muscle fibers are scattered sibly, plates rapping together, although the 

and poorly developed; as in balaenids, the rattle is presumably associated with skim 

tongue may sei-ve as a seasonal store of foraging, not prey release. Although ob- 

adipose tissue (Howell, 1930; Tarpley, servation of lateral head movement has 

1985). The tongue is thought to play a role been limited, any degree and direction of 

in expelling water from the oral cavity in motion (including dorsoventral shaking, 

all mysticetes, yet the elastic recoil of gular which has not been documented) might 

closure may accomplish much of this func- dislodge prey. Unfortunately, fine baleen 

tion in rorquals and obviate the need for fringes might adhere together closely (the 

a muscular tongue in this family. way gill lamellae clump) in air, impeding 

A related yet alternative notion is that proper prey removal. Head shaking un- 
the tongue could simply slap or shake prey derwater might resolve this problem, yet 
free, perhaps by vibrating plates or waving because many prey are neutrally buoyant 
them anteroposteriorly. This could allevi- in seawater gravitational forces would not 
ate baleen abrasion (and to a lesser extent prove effective; prey would more likely 
lingual abrasion, although it would involve float or even swim off fringes in water cur- 
similar muscle actions), as well as the in- rents inside the oral cavity generated by 
ability of the tongue to free prey from head shaking. 

fringes not directly contacting the tongue. As with the lingual scraping hypothesis. 

Any motion that might jostle plates or a serious shortcoming of this plan is that 

knock them together could release prey, it might not release prey adequately, yet 

although this would likely be less efficient because balaenids skim for hours with lit- 

with minute prey trapped in fine, filamen- tie apparent swallowing (at least with 

tons fringes. However, the fact that skim- sparsely distributed prey), a good head 

mers continuously filter might obligate shake would likely dislodge sufficient food 

them to purge their filter continuously, in- to swallow. A major disadvantage of head 

stead of at long intervals as was suggested shaking would seem to be the metabolic 

earlier. This is especially crucial in light of cost of moving the entire head, which, al- 

the fact that the filter is at best inefficient though it need not be particularly rapid 

and at worse useless when clogged with nor vigorous, would certainly require more 

prey or other items. Balaenids' long plates energy than simply moving the tongue. Yet 

might also clean themselves to some extent simpler ways may exist to move the head, 

by rubbing or squeezing together when In one of the few published mentions of 

they fold posteriorly as the mouth closes; prey release, Gaskin (1982) postulated that 

however, this is not feasible for the much the short, lunging inishes of right whales 

shorter plates of the skim feeding sei during bouts of skim feeding could agitate 

whale. and reinove clinging food particles. 

Another solution, at least in balaenids 

Prey Removal Via Head or Lip Shaking ^i^^^e large lower lips abut the baleens 

A second hypothesis, also mechanical in lateral edge (Fig. 2), would be to flap the 
nature, stems from obsei-vations of occa- lips or shake only the lower jaw rather than 
sional head-shaking behavior in southern the entire head. Lambertsen et al. (1989) 
right whales (Payne, in press). Whales described the bowhead's strong labial mus- 
have been seen shaking their heads rapidly culature, particularly the temporalis and 
from side to side above the surface with a deep masseter, which have extensive in- 
sound audible from a great distance. This sertions on the coronoid process. Although 
sound is not unlike the "baleen rattle" of these muscles, like the specialized mysti- 
right whale skim feeding (Watkins and cete temporomandibular articulations and 
Schevill, 1976), produced by water lapping mandibular symphysis, have been impli- 
over partially submerged plates, and, pos- cated in mandibular adduction for balaen- 

Mysticete Prey Removal • Werth 


id feeding (especially to establish the or- 
olabial sulcus and support baleen; Es- 
chricht and Reinhardt, 1866), they could 
play a further role in prey removal. Prelim- 
inary study also discloses the presence of 
small slips of labial musculature arising 
solely from the lower jaw of Balaena 
(Lambertsen et al, 1989; Werth, 1993); 
the extent to which these control the lips 
is unknown. A combination of surface and 
submerged head shaking, forward lunging, 
mandibular rotation, and cheek flapping in 
right whales might allow for mechanical 
removal of trapped prey without direct 
contact between tongue and baleen and 
substantial wear on either. The study of 
Ray and Schevill (1974) of benthic suction 
feeding in a young captive gray whale con- 
firmed that each lip could be moved in- 
dependently and curled away from the ba- 
leen. Mandibular rotation has also been as- 
sociated with enhanced gape and enlarge- 
ment of the oral cavity for improved water 
engulfment in balaenopterids (Lillie, 1915; 
Howell, 1930), yet the ability of rorqual 
lips to knock baleen is doubtful, for al- 
though the lips contact the short baleen, 
they protrude little above the mandible 
and are likely far less mobile than those of 
gray and right whales. 

Prey Removal Via Hydrodynamic Flushing 

A third cleaning strategy entails back- 
washing a small amount of water into the 
mouth to remove items from baleen and 
deposit them on the tongue for transport 
and swallowing. As in the first hypothesis, 
the tongue is directly involved, although it 
would not contact baleen. By rapidly de- 
pressing and/or retracting the tongue, the 
oral space would enlarge, briefly generat- 
ing a suction pressure to draw water 
through baleen into the mouth. Just as a 
dip net can be rinsed by rapidly jerking it 
back to reverse the flow through it, so wa- 
ter might momentarily enter a whale's 
mouth from the sides and thereby release 
captured items, so long as gape was suffi- 
ciently closed to prevent water from en- 
tering ventral to the baleen racks. Rapid 

abduction of the jaws might be coupled 
with lingual depression to create sufficient 
negative pressure to pull water in. The en- 
suing current need not be strong, merely 
sufficient to reverse the water flow and de- 
liver prey into the center of the oral cavity. 
Note that this idea differs from the pre- 
ceding two in that it depends on hydro- 
dynainic rather than mechanical forces, 
with water (rather than a solid object) sup- 
plying the cleaning mechanism. 

Not only would this flushing method re- 
quire substantial lingual (and likely labial) 
mobility, but its efficacy would vary de- 
pending on such mobility as well as other 
factors — namely prey size and type and 
coarseness of baleen strands — that deter- 
mine how likely items are to lodge in fring- 
es. Although rorquals might not possess a 
sufficiently firm and muscular tongue to 
achieve even weak intraoral suction pres- 
sures, their coarser fringes and typically 
larger prey (with the exception of the sei 
whale) ought to ensure that even euphau- 
siids would not become entangled, but 
would simply drop onto the tongue once 
the mouthful of engulfed water was ex- 
pelled. Certainly the suction-feeding gray 
whale could generate sufficient negative 
pressure to flush baleen effectively. In es- 
sence the sole difference between suction- 
generated prey capture and release would 
be gape and, to a minor extent, lip posi- 
tion: although a wide gape would allow for 
prey ingestion, a narrow gape would sim- 
ply result in a stream of cleansing water 
through baleen plates. A foraging gray 
whale could right its body or remain in a 
side-swimming position (although not con- 
tacting the substrate) for this backflushing. 
Just as the mouthful of engulfed water is 
expelled from the mouth through baleen, 
the mouthful of water for prey flushing 
could likewise be expelled by intermit- 
tently filter- feeding mysticetes, either be- 
fore or after deglutition of accumulated 
prey. However, note that although whales 
can handle the osmotic load of swallowed 
seawater, they do not drink seawater (Sli- 
jper, 1962). The potential for increased 

200 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


seawater ingestion from this rinsing flow 
might pose a serious problem for baleen 
cleaning via backwash. 

Although balaenid tongues should also 
prove adequate to generate backwash flow, 
their delicate hairlike baleen fringes might 
preclude prey release. Indeed, any extra 
water flow (in any direction) might only 
serve to ensnare items further. Such hy- 
drodynamic considerations are crucial, for 
recall that this is a nonstatic three-dimen- 
sional filter with porosity dependent on 
flow (Sanderson and Wassersug, 1990). 
The higher the water pressure, the denser 
the filter becomes (i.e., the smaller the 
pores), promoting prey capture and allow- 
ing ever smaller prey to be captured. As 
flow through the filter slows, the compact- 
ed mesh expands, so that it can be rinsed 
much more easily. Hence, a backwash flow 
should not be notably rapid or powerful. 
However, the fact that most swimming 
prey are negatively rheotropic further 
complicates matters; live prey would tend 
to swim upstream and burrow deeper into 
baleen during backwashing. 

Although comparative mysticete tongue 
muscle studies are sorely lacking, myologic 
studies of right whales suggest that die 
tongue is capable of undergoing the move- 
ments needed to generate backwash flow. 
The rapid mouth closure and water expul- 
sion observed in right whales by Mayo and 
Marx (1990) may relate to such prey re- 
moval. Although this behavior (which was 
observed to occur roughly once an hour) 
is described as flushing, no direct evidence 
exists to confirm that it indeed frees 
trapped prey. Limited observation of "nod- 
ding behavior," in which a right whale 
quickly dips its head and jerks it back 
(Gaskin, 1982; Mayo and Marx, 1990), 
might likewise be construed as supporting 
the backwash hypothesis, although this be- 
havior might also support the claim of prey 
removal via head shaking or brief forward 
lunging described previously. 


Baleen cleaning is not so simple as 
might initially be assumed, with many fac- 

tors to be taken into consideration, chief 
among these the relation between the fil- 
tration apparatus and filtered items. Given 
the respective strengths and weaknesses of 
the three prey removal hypotheses, all 
seem equally likely to occur, especially in 
skim feeders. Confirmation of baleen 
wearing and replenishment, along with the 
purported actions of tongue musculature, 
support the claim of baleen cleaning by 
tongue scraping — although the lack of dif- 
ferential wear calls this into serious ques- 
tion — or by gentler rubbing, which would 
minimize abrasive loss of baleen. Limited 
observations of lateral head shaking and 
nodding in Eubalaena provide indirect 
support for the other hypotheses. Moipho- 
logic evidence seems to sustain rather than 
preclude each conjectural means of prey 
removal. It may well be that different spe- 
cies and individuals in different situations 
use all three mechanisms. 


I am greatly indebted to Laurie Sander- 
son, whose critical comments greatly im- 
proved the content and clear expression of 
the ideas presented here. Richard Wasser- 
sug and Jim Mead also provided many use- 
ful insights in their careful reviews of this 
paper. Discussions with Scott Kraus, Tom 
Albert, Larry Barnes, Roger Payne, Tom 
Ford, Butch Rommel, Dan Hillmann, 
Stormy Mayo, and Craig George helped 
me to formulate and develop the hypoth- 
eses of prey removal and their respective 
strengths and weaknesses. The ongoing 
anatomical study of bowhead tongues de- 
scribed in this paper was supported finan- 
cially and logistically by the North Slope 
Borough, Department of Wildlife Man- 
agement, Barrow, Alaska (contract C2189) 
and Alaska Eskimo Whaling Commission, 
which generously permitted me to use 
data from haivested whales and examine 
specimens housed at the Lousiana State 
University School of Veterinary Medicine. 


BroDIE, p. F. 1977. Form, function, and energetics 
in Cetacea: a discussion, pp. 45-58. In R. J. Har- 

Mysticete Prey Removal • Werth 201 

rison (ed.). Functional Anatomy of Marine Mam- 
mals. Vol. 3. New York: Academic Press, x + 428 

Carroll, G. M., J. C. George, L. L. Lowry, and 
K. O. COYLE. 1987. Bowhead whale {Balaena 
inijsticetus) feeding near Point Barrow, Alaska, 
during the 1985 spring migration. Arctic, 40: 

Clapham, p. J., E. Leimkuhler, B. K. Gray, and 
D. K. Mattila. 1995. Do humpback whales ex- 
hibit lateralized behaviour? Animal Behaviour, 
50: 73-82. 

EscHRiCHT, D. R, and J. T. Reinhardt. 1866. On 
die Greenland right whale {Balaena mysticetus 
L.), with special references to its geographical 
distribution and migration in times past and pre- 
sent, and to its external and internal character- 
istics. Ray Society Publications, 40: 3-150. 

Evans, P. G. H. 1987. The Natural History of Whales 
and Dolphins. New York: Facts on File, x-vi + 
343 pp. 

Gaskin, D. E. 1976. The evolution, zoogeography, 
and ecology of Cetacea. Oceanography and Ma- 
rine Biology Annual Review, 14: 247-346. 

. 1982. The Ecology of Whales and Dolphins. 

Portsmouth, New Hampshire: Heinemann Edu- 
cational Books, xii + 459 pp. 

Geracl J. R., AND D. J. St. Aubin (eds.). 1990. Sea 
Mammals and Oil: Confronting the Risks. Toron- 
to: Academic Press, x-vi + 282 pp. 

GORMLEY, G. 1983. Hungry humpbacks forever 
blowing bubbles. Sea Frontiers, 29: 258-265. 

Hain, J. H. W, G. R. Carter, S. D. Kraus, C. A. 
Mayo, and H. E. Winn. 1982. Feeding behavior 
of the humpback whale, Megaptera novaean- 
gliae, in the western North Atlantic. Fishery Bul- 
letin, 80: 259-268. 

Haldiman, J. T, Y. Z. Abdelbakl D. W. Duffield, 
W G. HENK, and R. W Henry. 1981. Deter- 
mination of the gross and microscopic structure 
of the lung, kidney, brain and skin of the bow- 
head whale, Balaena mysticetus (RU 1380), pp. 
305-662. In T. F Albert (ed.). Tissue Structural 
Studies and Other Investigations on the Biology 
of Endangered Whales in the Beaufort Sea. Final 
Report to the Bureau of Land Management from 
the Department of Veterinary Science, Univer- 
sity of Maryland, College Park, Maryland. NTIS 
PB86-153583/AS. iv + 953 pp. 

Haldiman, J. T, and R. J. Tarpley. 1993. Anatomy 
and physiology, pp. 71—156. In J. J. Bums, J. J. 
Montague, and C. J. Cowles (eds.). The Bow- 
head Whale. Lawrence, Kansas: Society for Ma- 
rine Mammalogy, xxxvi + 787 pp. 

Howell, A. B. 1930. Aquatic Mammals. Springfield, 
Illinois: Charles C. Thoinas. xii + 338 pp. 

JURASZ, C. M., AND V. p. JURASZ. 1979. Feeding 
modes of the humpback whale, Megaptera no- 
vaeangliae, in southeast Alaska. Scientific Re- 
ports of the Whales Research Institute, 31: 69- 

Kasuya, T, and D. W Rice. 1970. Note on baleen 
plates and on arrangement of parasitic barnacles 
of gray whale. Scientific Reports of the Whales 
Research Institute, 22: 39^3. 

Kavvamura, a. 1974. Food and feeding ecology in 
the southern sei whale. Scientific Reports of the 
Whales Research Institute, 26: 25-144. 

KLAUS, A. D., J. S. Oliver, and R. G. Kvitek. 1990. 
The effects of gray whale, walrus, and ice goug- 
ing disturbance on benthic communities in the 
Bering Sea and Chukchi Sea, Alaska. National 
Geographic Research, 6: 470—484. 

LambertSEN, R. H. 1983. Internal mechanism of 
rorqual feeding. Journal of Mammalogy, 64: 76— 

A. HiRONS, K. J. Kreiton, and C. Moor. 1989. 
Characterization of tlie functional morphology of 
the mouth of the bowhead whale, Balaena mys- 
ticetus, with special emphasis on feeding and fil- 
tration mechanisms. Report to the Department 
of Wildlife Management, North Slope Borough, 
Barrow, Alaska, xui + 134 pp. 

Frontomandibular stay of Balaenopteridae: a 
mechanism for momentum recapture during 
feeding. Journal of Mammalogy, 76: 877-899. 

Leatherwood, S., R. R. Reeves, and L. Foster. 
1983. The Sierra Club Handbook of Whales and 
Dolphins. San Francisco: Sierra Club Books, xviii 
+ 302 pp. 

LILLIE, D. G. 1915. Cetacea. British Antarctic (Terra 
Nova) Expedition, Zoology, 1: 85-125. 

LOUGHLIN, T. R. (ed.). 1994. Marine Mammals and 
the Exxon Valdez. London: Academic Press, xix 
+ 395 pp. 

Lowry, L. F 1993. Foods and feeding ecology, pp. 
201-238. In J. J. Bums, J. J. Montague, and C. 
J. Cowles (eds.). The Bowhead Whale. 
Lawrence, Kansas: Society for Marine Mammal- 
ogy, xxxvi + 787 pp. 

Lowry, L. F, and K. J. Frost. 1984. Foods and 
feeding of bowhead whales in western and north- 
em Alaska. Scientific Reports of the Whales Re- 
search Institute, 35: 1—16. 

Mayo, C. A., and M. K. Marx. 1990. Surface for- 
aging behaviour of the North Atlantic right 
whale, Eubalaena glacialis, and associated zoo- 
plankton characteristics. Canadian Journal of Zo- 
ology, 68: 2214-2220. 

MiLNER, A. 1981. Flamingos, stilts, and whales. Na- 
ture, 289: 347. 

Mitchell, E. D. 1970. Pigmentation pattern and 
evolution in delphinid cetaceans: an essay in 
adaptive coloration. Canadian Journal of Zoology, 
48: 717-740. 

Murison, L. D., D. J. Murie, K. R. Morin, and J. 
Da Silva Curiel. 1984. Foraging of the gray 
whale along the west coast of Vancouver Island, 
British Columbia, pp. 451^63. In M. L. Jones, 
S. L. Swartz, and S. Leatherwood (eds.). The 

202 Bulletin Museum of Comparative Zoologij, Vol. 156, No. 1 

Gray whale Eschrichtius robustus. Orlando, 
Florida: Academic Press, xxiv + 600 pp. 

Nelson, C. H., and K. R. Johnson. 1987. Whales 
and walruses as tillers of the sea floor. Scientific 
American, 256: 112-117. 

Nemoto, T. 1959. Food of baleen whales with ref- 
erence to whale movements. Scientific Reports 
of the Whales Research Institute, 14: 149-241. 

. 1970. Feeding pattern of baleen whales in 

the ocean, pp. 241-252. In J. H. Steele (ed.). 
Marine Food Chains. Edinburgh: Oliver and 
Boyd, viii + 552 pp. 

Nerini, M. 1984. A review of gray whale feeding 
ecology, pp. 423^50. In M. L. Jones, S. L. 
Swartz, and S. Leatherwood (eds.). The Gray 
whale Eschrichtius robustus. Orlando, Florida: 
Academic Press, xxiv + 600 pp. 

Oli\'ER, J. S., AND p. N. Slattery. 1985. Destruction 
and opportunity on the sea floor: effects of gray 
whale feeding. Ecology, 66: 1965-1975. 

Omura, H. 1958. Nortli Pacific right whale. Scientific 
Reports of the Whales Research Institute, 13: 1- 

Orton, L. S., and p. F. Brodie. 1987. Engulfing 
mechanics of fin whales. Canadian Journal of Zo- 
ology, 65: 2898-2907. 

Payne, R. S. (In press). Behavior of Southern Right 
Whales {EuhaJaena au.stralis). Chicago: Univer- 
sity of Chicago Press. 

Pike, G. C. 1962. Migration and feeding of the gray 
whale {Eschrichtius gibbosus). Journal of the 
Fisheries Research Board of Canada, 19: 815— 

PiVORUNAS, A. 1976. A mathematical consideration 
on tlie function of baleen plates and their fringes. 
Scientific Reports of the Whales Research Insti- 
tute, 28: 37-55. 

. 1977. The fibrocartilage skeleton and related 

structures of the ventral pouch of balaenopterid 
whales. Journal of Morphology, 151: 299-.314. 

. 1979. The feeding mechanisms of baleen 

whales. American Scientist, 67: 432—440. 

Ray, G. C, and W. E. SCHEVaLL. 1974. Feeding of 
a captive gray whale, Eschrichtius robustus. Ma- 
rine Fisheries Review, 36: 31—38. 

Reeves, R. R., and S. Leatherwood. 1985. Bow- 
head whale, pp. 305-344. In S. H. Ridgway and 
R. J. Harrison (eds.). Handbook of Marine Mam- 
mals. Vol. 3: The Sirenians and Baleen Whales. 
San Diego: Academic Press, xviii -I- 362 pp. 

Rubenstein, D. I., and M. a. R. Koehl. 1977. The 
mechanisms of filter feeding: some theoretical 
considerations. American Naturalist, 111: 981- 

RUUD, J. T. 1940. The surface structure of the baleen 
plates and a possible clue to age in whales. Hval- 
radets Skrifter, 23: 1-24. 

. 194.5. Further studies on the structure of the 

baleen plates and their application to age deter- 
mination. Hvalradets Skrifter, 29: 1-69. 

Sanderson, S. L., and R. WassERSUG. 1990. Sus- 

pension-feeding vertebrates. Scientific American, 
262(3): 96-101. 

. 1993. Convergent and alternative designs for 

vertebrate suspension feeding, pp. 37—112. In J. 
Hanken and B. K. Hall (eds.). The Skull. Vol. 3: 
Functional and Evolutionary Mechanisms. Chi- 
cago: University of Chicago Press, x + 460 pp. 

ScheLL, D. M., and S. M. Saupe. 1993. Feeding and 
growth as indicated by stable isotopes, pp. 491— 
509. In J. J. Bums, J. J. Montague, and C. J. 
Cowles (eds.). The Bowhead Whale. Lawrence, 
Kansas: Society for Marine Mammalogy, xxxvi + 
787 pp. 

Slijper, E. J. 1962. Whales. New York: Basic Books. 
475 pp. 

Storro-Patterson, R. 1981. Great gulping blue 
whales. Oceans, 14: 16. 

SUND, P. N. 1975. Evidence for feeding during mi- 
gration and of an early birth of the Califomian 
gray whale (Eschrichtius robustus). Journal of 
Mammalogy, 56: 265. 

TarpLEY, R. J. 1985. Gross and microscopic anatomy 
of the tongue and gastrointestinal tract of the 
bowhead whale (Balaena mijsticetus). Ph.D. dis- 
sertation. College Station, Texas: Texas A&M 
University, xiv +141 pp. 

TOMILIN, A. G. 1954. Adaptive types of the order 
Cetacea. Zoologichesldi Zhurnal Moscow, 33: 

. 1957. Mammals of the USSR and Adjacent 

Countries. Vol. IX: Cetacea. Jerusalem: Israel 
Program for Scientific Translations, xxi + 717 pp. 

VOGEL, S. 1994. Life in Moving Fluids: The Physical 
Biology of Flow, second edition. Princeton: 
Princeton University Press, xiii + 467 pp. 

VON SCHULTE, H. W. 1916. The sei whale (Balaen- 
optera borealis Lesson). Anatomy of a foetus of 
Balaenoptera borealis. Monographs of the Pacific 
Cetacea. Memoirs of the American Museum of 
Natural History (New Series), 1: 389^99. 

Watkins, W. a., and W. E. Sghevill. 1976. Right 
whale feeding and baleen rattle. Journal of Mam- 
malogy, 57: 58-66. 

. 1979. Aerial obsei^vation of feechng behavior 

in four baleen whales: Eubalaena glacialis, Ba- 
laenoptera borealis, Megaptera novaeangliae, and 
Balaenoptera phijsalus. Journal of Mammalogy, 
60: 155-163. 

Watson, L. 1981. Whales of the World. London: 
Hutchinson. 302 pp. 

Weitkamp, L. a., R. a. Wissmar, C. a. Simenstad, 
K. L. Fresh, and J. G. Odell. 1992. Gray 
whale foraging on ghost shrimp, Callianassa cal- 
ifomiensis, in littoral sand flats of Puget Sound, 
USA. Canadian Joumd of Zoology. 70: 227.5-2280. 

Werth, a. J. 1990. Functional anatomy of the right 
whale tongue [abstract]. American Zoologist, 30: 

. 1993. Functional morphology of balaenid 

whale tongues [abstract], p. 112. In G. A. J. Wor- 
thy (ed.), Proceedings of the Tenth Biennial 

Mysticete Prey Removal • Werth 203 

Conference on the Biology of Marine Mammals, 
Galveston, Texas. 130 pp. 

. 1995. Models of hydrodynamic flow in the 

bowhead filter feeding apparatus [abstract]. 
American Zoologist, 35: 59A. 

1998. Functional anatomy of the sperm 

whale tongue [abstract]. American Zoologist, 38: 
Williamson, G. R. 1973. Counting and measuring 
baleen and grooves of whales. Scientific Reports 
of the Whales Research Institute, 25: 279-292. 

WOLMAN, A. A. 1985. Gray whale, pp. 67-90. /;) S. 
H. Ridgway and R. J. Harrison (eds.). Handbook 
of Marine Mammals. Vol. 3: The Sirenians and 
Baleen Whales. San Diego: Academic Press, wiii 
+ 362 pp. 

WURSIG, B. 1988. The behavior of baleen whales. 
Scientific American, 258: 102-107. 

. 1989. Cetaceans. Science, 244: 1550-1557. 

Zackowitz, M. 2000. Doing a whale of a research 
job. National Geographic, 197: 140. 



Abstract. Since the first description of cyclical 
tongue movements in feeding (opossum and cat), 
studies on a range of mammals (fruit bat, tenrec, rab- 
bit, hyrax, macaque, and man) have been completed. 
This review examines those data to determine wheth- 
er a pattern of tongue-jaw linkage in feeding is com- 
mon to all mammals. Major directional changes in 
hyoid and tongue surface movement occur at com- 
parable points in tlie jaw movement cycle regardless 
of craniofacial anatomy and dietetic specialization. In 
all cases, the hyoid and tongue surface move forward 
during the early part of jaw opening, with a return 
movement during later jaw opening and closing. The 
role of this cychcal tongue movement in the transport 
and manipulation of food is common to all mammals 
studied, save that triturated food is moved through 
the palatoglossal-palatopharyngeal arches for bolus 
formation earlier in the tongue cycle in macaque and 
man. New data on swallowing in man suggest that 
Homo has a specialized pattern for liquid bolus for- 
mation but otherwise retains the basic mammalian 


In a letter to Nature, Crompton et al. 
(1975) reported that the hyoid bone moves 
continuously in feeding in the American 
opossum {Didelphis virginiana). This sim- 
ple observation refuted the then conven- 
tional wisdom, based on limited human 
studies, that the hyoid only moves in swal- 
lowing. Attention has since focused on the 
interrelationships between jaw and hyoid 
movement, jaw movement, and hyoid and 
tongue surface movement, and the role of 
these movements in the acquisition, pro- 

' Institute for Sensory Research, and Department 
of Bioengineering and Neuroscience, Syracuse Uni- 
versity, Syracuse, New York 13244-5290. 

- Department of Physical Medicine and Rehabih- 
tation. The Johns Hopkins University School of Med- 
icine, Good Samaritaji Hospital, Baltimore, Maryland 

cessing, and transport of food in mammals. 
In 1978, Hiiemae, Thexton, and Crompton 
posited that cyclical movements of the 
tongue, by facilitating the intraoral man- 
agement of food, including its transport to 
the pharynx for bolus accumulation and 
swallowing, were integral to the feeding 
process. A sufficient body of evidence has 
now accumulated from experimental stud- 
ies (listed in the Appendix) examining 
feeding behavior in representative terres- 
trial maiTunals with very different dietetic 
adaptations to make a synthesis possible. 
This review examines the proposition that 
a pattern of linked jaw and hyoid-tongue 
movement is common to all terrestrial 
mammals witli type I tongues (Doran, 

Although the craniofacial anatomy of 
terrestrial mammals has a common Bau- 
plan, anatomical details vary widely both 
among and within orders (Hiiemae, 2000; 
Turnbull, 1970). However, the basic phys- 
iological functions performed by the teeth, 
jaws, tongue and associated soft tissues do 
not vaiy (Hiiemae, 2000). These functions 
are associated with the acquisition, reduc- 
tion (if needed), transport, and swallowing 
of food for chemical digestion. It follows 
that because all mammals have the same 
Bauplan, subserving the same biological 
functions, it is reasonable to postulate that 
the mechanisms used to fulfill these phys- 
iological functions will be essentially the 
same. If a patterned linkage exists between 
the cyclical moveinents of the jaws and the 
cyclical movements of both the tongue 
surface and the hyoid, that linkage should 
be expressed by a consistent relationship 

Bull. Mus. Comp. Zool., 156(1): 205-217, October, 2001 


206 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

betAveen jaw and tongue— hyoid movement is a basic mammalian pattern; and to brief- 
events in all feeding cycles in all species ly revisit the issue of whether the inam- 
examined. malian pattern could have its antecedents 
Bramble and Wake (1985) developed a in the mechanisms for food transport de- 
theoretical model for a generalized feeding veloped in phylogenetically earlier terres- 
cycle in nonmammalian terrestrial tetra- trial tetrapods. 

pods, largely based on a synthesis of ki- cydcdimcmtai cti mice 

nesiological and electromyographic EXPERIMENTAL STUDIES 

(EMG) data from terrestrial salamanders. Intraoral behavior is particularly difficult 
turtles, and lizards. This model had fea- to study. The only techniques that allow 
tures, such as rate change and direction of the concurrent recording of jaw move- 
jaw and hyolingual motion, similar to those ment, hyoid movement, tongue surface 
reported in mammals (Bramble and Wake movement (provided radiopaque markers 
1985, fig. 13.3). They suggested that there are used), and food position (if treated 
might be homologies between mammalian with barium sulfate) are cinefluorography 
feeding mechanisms, especially intraoral (CFG) or videofluorography (VFG). It 
transport, and those obsei'ved in nonmam- must be emphasized that any CFG or 
malian vertebrates, such that the patterns VFG record is a 2D projection or image 
seen in mammals might have evolved in of 3D events. If a structure, such as the 
earlier terrestrial tetrapods. While ac- hyoid, has little mediolateral movement, 
knowledging the absence of experimental then the 2D image seen in lateral projec- 
data, they extended their analysis to sug- tion probably accurately represents its 
gest that the central pattern generator movement pattern. Data for macaque and 
(CPG) maintaining the rhythmic jaw man confirm that no such confidence can 
movements of feeding identified in mam- be attached to tongue surface movement 
mals (Bremer, 1923; Dellow and Lund, data. Although CFG and VFG have been 
1971), might be present in antecedent used in many studies of feeding in mam- 
forins, and that the output of such a CPG mals representative of omnivores (includ- 
was regulated by sensoiy modulation from ing the Pecora), carnivores, herbivores, ro- 
oral receptors. Smith (1994: 294) argued, dents, and primates, relatively few (see 
for that to be true, neuromotor patterns Hiiemae, 2000; and the Appendix) have 
had to be homologous, that is, despite examined the temporal relationships be- 
changes in peripheral anatomy, the neu- tween jaw and tongue— hyoid movements 
romotor output from the central nervous during feeding sequences. Comparison 
system (CNS) had to be conserved. She and synthesis of those reports to address 
observed that similarities in movement the question of whether a common pattern 
patterns (or, functional behaviors, our exists is complicated by the rate of exper- 
term) had been extended "to hypotheses imental data acquisition, the methods used 
of neuromotor conservatism, to assertions for analysis, the form in which data were 
of evolutionary constraints." Taking a presented, and the need to reconcile the 
broad-brush approach using examples underlying behavioral patterns regardless 
from a broad range of vertebrate taxa to of the terminology used, 
test the hypothesis that terrestrial verte- 

brates have a conservative feeding pro- "'"'^!,'^'^^' ^'vX.?.^,^^^^ 

gram, she found fittle evidence for its sup- HYOID-JAW COMPLEX 
port. Although the proportions of the skull 

The purpose of this review is to examine and lower jaw, as well as the general po- 

what is now known about tongue— jaw sition of the hyoid and the shape of the 

movements in feeding in mammals, in- tongue, differ markedly between mammals 

eluding man; to determine whether there (Hiiemae, 2000), the muscles producing 

Tongue-Jaw Linkages • Hiiemoe and Palmer 207 

jaw, hyoid, or tongue movement are gen- ented pharyngeal surface (Hiiemae, 2000; 
erally homologous. Hyoid position is Hiiemae and Palmer, 1999). If a basic- 
known to be controlled by the interplay of mammalian pattern exists, it follows that 
activity in three groups of muscles. The these morphological changes should not 
anterior suprahyoids (anterior belly of di- affect its expression, although the mechan- 
gastric and geniohyoid) pull the hyoid for- ical outcomes may be different, 
ward and can depress the lower jaw. The Doran (1975) identified two types of 
mylohyoid can also elevate the hyoid, syn- mammalian tongues, both with oral and 
chronously raising the floor of the mouth pharyngeal parts. Type I, found in most 
and so the tongue body. The hyoid is con- mammals (and all those mammals in which 
nected to tlie skull base by the posterior its movement has been studied), can be 
belly of digastric and the stylohyoid (pos- protruded to a maximum of 50% of its 
tenor supraliyoids), which can pull the hy- resting length. Type II tongues are found 
oid backward and upward. The infrahyoids in a few mammals, such as the anteaters, 
connect the hyoid to the sternum (ster- and can be extensively protruded outside 
nohyoid), the scapula (omohyoid), and the the mouth for food gathering. Livingstone 
thyi-oid cartilage (thyrohyoid). The major (1956) argued, first, that the movement of 
infrahyoids (sternohyoid and omohyoid) the tongue depends largely on the move- 
act to pull the hyoid back, down, or both, ment of the hyoid; second, that change of 

The biomechanics of the hyoid complex position, coupled with a change of form 
are poorly understood. Using EMG and results from extrinsic muscle action, and, 
movement data, Crompton et al. (1977) last, that the intrinsic muscles provide for 
demonstrated the mechanism by which a great deal of mobility. It follows that if 
jaw and hyoid movements were produced the tongue base shortens (geniohyoid, my- 
in Didelphis, whether divergent (i.e., the lohyoid), canying the hyoid and the body 
hyoid traveling backward, away from the of the tongue foiAvard, then the effect of 
symphysis, in jaw closing, lengthening the contraction of the genioglossus, which pro- 
tongue base) or convergent (the hyoid tnjdes the tongue, will be augmented, 
traveling forward, toward the symphysis, in Similarly, the action of hyoglossus or sty- 
jaw opening, shortening the tongue base), loglossus, either of which can pull the 
No comparable study has been conducted tongue backward, will be augmented by 
for any other mammal. However, the retraction of the hyoid. However, given the 
changing pattern of shortening and length- pattern of insertion of both genioglossus 
ening in these muscles regulates hyoid po- (medially) and hyoglossus (laterally), con- 
sition and so the length of the tongue base, traction of either will affect the overall 

Nonanthropoid mammals have antero- shape of the tongue body. Nevertheless, as 

posteriorly (AP) elongate and vertically Kier and Smith (1985) and Smith and Kier 

shallow tongues. However, the evolution (1989) have emphasized, the tongue has 

of the anthropoid primates resulted in pro- constant volume, such that a change in any 

gressive change in tongue shape, with con- dimension must be accompanied by 

comitant changes in the anatomy of the changes in the other two. This makes the 

orophaiynx. As described by Thexton and distinction between extrinsic and intrinsic 

Crompton (1998), the shape of the tongue muscles, while anatomically convenient, 

in macaque has changed from that seen in somewhat arbitrary with respect to func- 

Didelphis and most other mammals: it is tion. Schwenk (2001) further advances this 

shorter anteroposteriorly but has greater argument. 

vertical height. In man, the tongue is still To summarize, it follows that move- 
shorter with much greater height, such ments of the lower jaw are linked to hyoid 
that the hyoid is widely separated from its position; movements of the body of the 
oral surface, creating a long, vertically ori- tongue are linked to the length of its base, 

208 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

itself affected by hyoid position and lower terns of activity within cycles and within 

jaw movements; and the working surfaces sequences for those mammals studied 

of the tongue (oral and pharyngeal) have show any commonality across species and 

their shape affected by both of the above, orders. 

but augmented by the activity of the in- 



ISSUES IN SYNTHESIS ^jj j^^ movement cycles have closing 

The rhythmic movements of the jaw in and opening strokes separated by an IP 
feeding, especially chewing, are well de- phase of variable duration (very short in 
scribed. Those movements are now known some carnivores and insectivores, longer in 
to be associated with rhythmic hyoid most omnivores and herbivores with a sub- 
movements. At the same time, the tongue stantial lateromedial or posteroanterior 
not only changes its gross position but also lower molar traverse in occlusion on the 
changes shape depending on the condition working side). For all mammals studied, 
of the food in the mouth and stage in se- the hyoid moves from its most backward 
quence: ingestion with stage I transport; and downward position, at about mini- 
processing; bolus formation with stage II mum gape, to its most forward and up- 
transport; and deglutition (see Hiiemae, ward position during opening, reversing 
2000; Hiiemae and Palmer, 1999). We also direction before maximum gape (Fig. 1). 
now know that the overall jaw cycle time Tongue marker orbits also demonstrate 
varies between species studied, and also this reversal (Fig. 2). 

within each sequence in a given species. To test the hypothesis that all mammals 

such that the time spent in closing (Fast show the same jaw and tongue movement 

Close [FC], Slow Close [SC]/Power Stroke patterns, regardless of cycle duration, or 

[PS]), Intercuspal Phase (IP), and opening phase duration within cycles, and tongue 

(Slow Open [SO] or Ol, 02, Fast Open shape, the available data were analyzed us- 

[FO]) also varies, based on the changing ing distinct movement turnpoints as event 

rate of jaw movement. It follows that if markers (Palmer et al., 1997). The jaw 

jaw, hyoid, and tongue movements are movement event markers were maximum 

linked, that is, their movements are inter- and minimum gape. For the tongue— hy- 

dependent, then there should be some oid, they were maximum forward (MF for 

consistent relationship, regardless of sep- hyoid or TF for tongue); maximum down 

aration in time, between specific events in (MD or TD), maximum back (MB or TB), 

jaw, hyoid, and tongue movement cycles, and maximum up (MU or TU), relative to 

For such linkage to exist, synchrony is not the upper occlusal plane. The available 

essential, rather events should occur in a data were brought to a consistent time 

consistent sequence. scale (normalized) and these turnpoints 

If the hypothesis that there is such link- (hyoid— tongue marker), as reported, were 

age and that same linkage will be found in established relative to maximum and min- 

all mammals is to be exliaustively tested, imum gape and entered into a bar chart as 

then a rigorous comparative analysis using accurately as possible. The results are 

a set of uniform event criteria and tech- shown in Figure 3. 

niques such as interval analysis is required. In every species, TF occurs before max- 

Clearly this is infeasible for the full range imum gape in opening. In all these mam- 

of mammals so far studied given the issues mals, including tenrec and opossum (data 

in data collection and reduction alluded to limited to text descriptions), the tongue 

above. Instead, we are forced to examine marker (anterior tongue marker, ATM, or 

the available qualitative (behavioral) data middle tongue marker, MTM) reaches its 

and attempt to determine whether the pat- most backward position concurrent with or 

Tongue-Jaw Linkages • Hiiemae and Palmer 209 

HYRAX (derived from German and Franks, 1991) 

100 150 200 250 

MACAQUE (derived from Hiiemae et al., 1995) 

100 200 300 400 500 600 
h \ r+ 



RABBIT (derived from Cortopassi and f^uhl, 1990) 

50 100 150 200 
H ^ h-i — h 














Figure 1 . Gape-time (GT) plots for a single cycle of jaw, tongue, and hyoid movement in hyrax, rabbit, and macaque redrawn 
from data in the papers cited (no attempt has been made to represent actual distances traveled in any direction). The pattern 
of upward and fonward movement of the tongue surface in the first part of opening is shown by the extra thicl< lines, the 
synchronous downward movement by thicl< lines. The hyrax and rabbit records (published figures, or text) provide no basis for 
dividing the opening jaw movement into 01 , 02, or FO/03 phases; however, the macaque cycle (a composite) shows the pattern 
of jaw movement when the SO phase has two components. The reversal of ATM and MTM movement at the S02-FO transition 
is clearly shown (see text). 

Abbreviations: ANT. T, anterior tongue; ATM, anterior tongue marker; MID. T, middle tongue; MTM, middle tongue marker; 
POST. T, posterior tongue; PTM, posterior tongue marker. 

210 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 







\^>-^ J> 





^ I Movemen 

Movement in SO (01 , 01 +02) 

Figure 2. The trajectory, shown as loops, of anterior (ATIVI), 
middle (MTM), and posterior (PTM) tongue markers, and the 
hyoid, in opossum, (a) Lapping (a low-amplitude jaw move- 
ment without FO or FC): (b) stage I transport (movement of 
food from an extraoral position or from the front of the oral 
cavity to the molar region); (c) chewing (processing) cycles 
where food has to be repositioned on the occlusal surfaces of 
the postcanines (manipulation); and (d) stage II transport, in 
which triturated food is moved through the palatoglossal arch- 
es for bolus formation and deglutition. Although the tongue and 
hyoid movements in lapping show long elliptical loops, the in- 
troduction of the FO and FC phases when feeding on solid 
food increases their vertical dimension. (FO dashed line, FC 
thin solid line, SC longer dotted line). Regardless of stage in 
sequence, the tongue surface (ATM, MTM) is moving from a 
maximum tongue back (TB) to a maximum tongue fonward 
(TF) position during early opening, reaching its most forward 
position in opening, but before maximum gape. Redrawn from 
Hiiemae and Crompton (1985, fig. 14-12). 

Pteropus giganteus [frultbat, Mid. Tongue] 
Max. Gape MIn. Gape Max. Gape 


Oryctolagus cuniculus [rabbit, Mid. Tongue] 

' J^v ' ^r^v^v r 



Procavia syriacus [hyrax, Mid. Tongue] 


Felis domesticus [cat, Ant. Tongue, Lapping] 


Macaca fascicularis [macaque. Ant. Tongue] 


Homo sapiens [man. Ant. Tongue] 

Figure 3. Tongue turnpoints (TB and TF, with TD and TU 
where available, see text) for chewing cycles in mammals 
where time data is available (no data for opossum are included 
because the gape-time plot data from which Fig. 2 was pre- 
pared were not published). Cycle times have been normalized 
to express the time of events within cycles expressed as the 
percentage of time after initial maximum and before terminal 
maximum gape. The common rhythmic pattern of AP tongue 
movement is shown. 

Vertical bars mark start maximum gape, minimum gape, and 
terminal maximum gape. Forward tongue movement in open- 
ing is shown by the wavy stippling; backward movement is 
shown by the dots. The blank periods (fruit bat, rabbit, ma- 
caque, and man) indicate the time in which the tongue surface 
is rising. 

Tongue-Jaw Linkages • HUemae and Palmer 211 

before minimum gape. The exceptions ap- 
pear to be h>Tax and opossum, where a 
\ery short IP often occurs. However, the 
time compressed data in German and 
Franks (1991, fig. 2) show a short foi-ward- 
backward oscillation before the computed 
minimum gape in stereotypical chewing 
cycles. If TB is taken as the first of these 
backward positions, then the pattern for 
hyrax is consistent with that for the other 
species (as shown in Fig. 3). In some re- 
cords for the opossum (Hiiemae and 
Crompton, 1985), TB occurs before min- 
imum gape. Unfortunately, because the or- 
bits shown in the published figures are ex- 
amples of actual cycles, rather than a sta- 
tistically derived norm based on analysis of 
large numbers of cycles, this result has to 
be taken as strongly supportive, rather 
than confirmatoiy, of a generalized pat- 

It should also be noted that there is lit- 
tle likelihood that the tongue markers 
were in comparable positions in these sep- 
arate experiments, because the published 
data used for each analysis referred to 
mid-tongue or anterior tongue. Further, 
strong evidence exists that although mark- 
ers in the anterior and middle parts of the 
tongue tend to move in synchrony, forward 
movement of the posterior tongue and hy- 
oid may be slightly delayed (Hiiemae et 
al, 1995). This suggests that were the orig- 
inal data to be revisited, a much clearer 
demonstration of a common pattern might 
be obtained. However, we consider it sig- 
nificant that an analytical approach devel- 
oped to describe tongue— hyoid— jaw move- 
ments in macaque and man shows the 
same pattern when applied to other mam- 

The jaw movement cycle is designed to 
assure food reduction in a chewing stroke 
(SC/PS). The remainder of the jaw move- 
ment cycle serves to reposition the lower 
jaw for the next such stroke. Concurrently, 
there is a tongue movement cycle with its 
major activity occurring during opening 
and the FC phase of closing. German and 
Franks (1991) analyzed the temporal re- 

lationships between minimum gape and 
the start of tongue and hyoid protrusion in 
hyrax. They found that the tongue and hy- 
oid tumpoints for the onset of forward 
movement were synchronized to within 
one frame of the computed minimum 
gape, but that no predictable linkage oc- 
curred between tongue-hyoid movement 
events and maximum gape. This study pro- 
vides a convincing demonstration of the 
existence of a possible switch from the jaw 
movement cycle required for food reduc- 
tion to a tongue movement cycle function- 
ing to control intraoral food position and 
food transport in hyrax at minimum gape. 
However, this study cannot be readily ex- 
trapolated to other mammals, especially 
anthropoids, where no minimum gape is 
clearly visually identifiable given a long IP 

Although there is no stereotypical jaw 
movement cycle in macaque (Thexton and 
Hiiemae, 1997), Hiiemae et al. (1995) 
were able to show that TF (ATM and 
MTM), always occurred within 30 milli- 
seconds (usually less) of the last rate 
change in opening, that is, with the initi- 
ation of FO. Furthermore, when 02 was 
present, the amplitude of forward tongue 
movement was greatest. However, gape at 
the 02-F0 transition was always small. If, 
as Thexton and Crompton (1989) argue, 
the Ol and 02 phases of lapping (cat, 
opossum) correspond to SO (Ol ± 02) in 
chewing, then the extensive tongue pro- 
trusion involved in lapping exemplifies the 
anatomical relationship between tongue- 
hyoid movement and gape amplitude, 
leading to the hypothesis that extensive 
tongue protrusion can only occur within a 
relatively narrow range of gape, because 
additional jaw opening (FO) requires hy- 
oid retraction. It also explains the EMG 
data, which show low-level activity in the 
adductors (masseter or medial pterygoid) 
during 02, suggesting an antagonist func- 
tion that resists jaw opening and promotes 
protraction of the hyoid bone and the 
tongue body. 

It must be emphasized that these results 

212 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

must not be construed as implying that and hyoid movement, to functional (food 

tongue and jaw movements are invariably processing) needs cycle to cycle within se- 

time and movement linked. Quite the re- quence. 

verse is true. The interdependence be- To summarize, as the jaw movement cy- 

tween jaw— hyoid— tongue movement dein- cle (close, IP, open) proceeds, there are 

onstrated here applies only to chewing, concomitant inovements of both the hyoid 

lapping, and food transport. Even in chew- and the tongue surface. The hyoid and 

ing sequences, albeit in man, the rhythm tongue move forward and variably upward, 

may be disrupted during the collection starting at minimum gape or within IP, and 

and aggregation of food particles (clear- reverse direction during opening, before 

ance) for bolus formation (Hiiemae et al, maximum gape, or at the small maximum 

1996; Hiiemae and Palmer, 1999; Palmer gape used in lapping (end 02). 
et al., 1997). 

Intrinsic Tongue Movement. Cortopassi MECHANISMS IN FEEDING 
and Muhl (1990) described AP move- Vertical and AP tongue movements dur- 
ments of the rabbit tongue surface as un- ing feeding cycles have a common pattern 
dulating. This could be explained as a across all the adult mammals studied re- 
function of the time delay between the ini- gardless of dietary specialization, stage in 
tiation of an anteriorly or posteriorly di- sequence, and type of food. The question 
rected moveinent between the various arises: how does tliis apparent common 
parts of the tongue. Where comparative pattern subserve the physiological process- 
data are available, such a sequential pat- es required for transmission of swallow- 
tern with the anterior tongue leading able food to the gastrointestinal tract for 
seems to be present. However, there is no chemical digestion? The corollary has to 
doubt that in the fruit bat (de Gueldre and be addressed: how is the linkage between 
de Vree, 1984), cat (Thexton and Mc- forward tongue movement in early jaw 
Garrick, 1988, 1989), and macaque (Hiie- opening correlated with the backward 
mae et al., 1995), there is differential ex- movement of food within and through the 
pansion and contraction of the tongue sur- oral cavity? How does the tongue, acting 
face, as measured by lengthening or short- against the hard palate, and with the 
ening of the Euclidean distance between cheeks (variably developed in inammals), 
tongue inarkers. produce the documented aggregation and 

These studies show that the tongue can distal movement of swallowable food? 
be considered as ha\dng three distinct Feeding sequences are now considered 

components: the anterior tongue (tip to to have four stages: stage I transport 

anterior postcanines), the mid-tongue (re- (movement of food from the anterior oral 

lated to the cheek teeth), and the posterior cavity to the postcanines); processing (food 

tongue (the postfaucial, or pharyngeal, sur- reduction in chewing or by tongue— hard 

face). As might be expected, the amplitude palate compression, or both); stage II 

of possible expansion and contraction is transport (movement of swallowable food 

greatest in the anterior tongue and least in through the palatoglossal— palatopharyn- 

the posterior. However, in chewing cycles geal arches with bolus formation), and, 

(macaque) when little or no anteroposte- last, deglutition. Tongue movements in 

rior food transport is occurring, expan- processing are poorly understood, but 

sion— contraction (measured in lateral pro- clearly involve rotation of the working 

jection) is restricted to the middle seg- (gustatory) surface of the tongue about its 

ment. This may be illustrative not only of long (AP) axis to position food, or maintain 

the 2D representation of 3D events men- food position, in readiness for the next 

tioned above, but also of the tongues ca- chewing cycle (Hiiemae, 2000; Hiiemae 

pacity to respond, independently of jaw and Crompton, 1985). 

Tongue-Jaw Linkages • Hiiemae and Palmer 213 

Tongue and hyoid movement patterns to avoid the risk of aspiration into a respi- 
for stage I and II transport have been de- ratory tract whose aditus (true vocal folds) 
scribed for most of the mammals studied lies well below the oral cavity. We find 
(see the Appendix; Hiiemae and Cromp- (Hiiemae and Palmer, 1999) that in H. sa- 
ton, 1985). In all cases, food, whether liq- piens, boli formed from natural bites (e.g., 
aid or solid, is carried backward through 6-8 g of normal foods) normatively form 
the mouth to the postcanines, or from the in the oropharynx. In short, man is a mam- 
oral cavity to the pharynx, on a backwardly mal. Equally, we argue that for ingested 
traveling tongue surface or by virtue of a liquids, the bolus is formed, contained, 
backwardly traveling tongue-palate con- and organized within the oral cavity, and 
tact. For stage I transport, the only differ- swallowed therefrom. We argue that the 
ences in the mechanism, between mam- pivotal evolutionary change in hominid de- 
nials studied, can be directly correlated velopment has been the development of a 
with the length of the tooth row. The pro- behavioral mechanism for process man- 
cess may take several cycles in macaque agement of liquids, which can flow, in con- 
( German et al., 1989) but can be accom- trast to solids, which, even when triturat- 
plished in a single cycle in man (Hiiemae ed, probably cannot, 
and Palmer, in preparation). Superficially, ^(^^(^, IJSIONq 
the mechanisms of stage II transport show ^UiNULUoiUiMo 

the greatest differences. In the mammals Jaw, hyoid, and tongue movements, dur- 
studied, except macaque and man, aliquots ing the rhythmic cycles of normal feeding 
of swallowable food are moved through in mammals including man, are interde- 
the palatoglossal arches (fauces) during the pendent, that is, there is a consistent event 
upward and backward movement of the order relationship between movements of 
tongue in late FO (03) and FC (Hiiemae the jaw and the hyoid, and the grosser 
and Crompton, 1985). This is described as movements of the tongue. This should oc- 
the squeeze-wedge mechanism. In ma- casion no surprise given the physiological 
caque (Franks et al., 1984, see also Thex- and gross anatomical homologies among 
ton and Crompton, 1998) and man (Hiie- the structural elements involved. That 
mae and Palmer, 1999) the upward and said, the available data cannot support any 
forward movement of the tongue in IP detailed conclusions about the actual be- 
brings the anterior surface of the tongue havior of the tongue surface during feed- 
into contact with the anterior palate, and ing sequences for mammals in general, 
that contact rapidly spreads backward However, if food is to be of metabolic util- 
( middle and posterior tongue) forcing the ity, it has to reach the gastrointestinal tract 
food mass through the fauces. This mech- for chemical digestion. A primary role of 
anism, squeeze-back, depends on forward the tongue— hyoid complex is the move- 
movement of the tongue coupled with a ment of swallowable food into the pharynx 
posteriorly traveling tongue-palate con- for bolus formation and then deglutition, 
tact. Although the mechanisms of stage I and 

All nonanthropoid mammals form the stage II transport are now well document- 
bolus in the oropharynx (piriform fossae- ed (at least for representative species, see 
valleculae). Without prejudice to the air- Hiiemae, 2000; Hiiemae and Crompton, 
way, given an intranarial larynx, the bolus 1985; Hiiemae and Palmer, 1999), the pre- 
is moved into the esophagus from the oro- cise role of the pattern of complex and 
pharynx (see Thexton and Crompton, poorly understood changes in tongue sur- 
1998). It has long been axiomatic that man face position and shape during intraoral 
is different because in Homo sapiens, the food management are not. 
bolus is formed in the oral cavity and pro- The available data do support a generic 
pulsively expelled across the oropharynx, mammalian model that posits that two cy- 

214 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

clical, linked but not interlocked mecha- 
nisms are operative during feeding se- 
quences in mammals. Cyclical jaw move- 
ments position the teeth for food reduc- 
tion in SC, and then complete a cycle to 
reposition the teeth for the next chewing 
stroke. Concomitantly, the hyolingual sys- 
tem is also cycling. Tongue movements, fa- 
cilitated by changes in gross tongue posi- 
tion as a function of hyoid moveinent, and 
implemented by changes in tongue sur- 
face—palate contacts, manipulate and 
transport food within and through the oral 
cavity during the SO-FO and FC phases 
of the jaw inovement cycle. Although the 
order of events is the same for all the 
mammals studied, the temporal linkages 
between them seem tighter in opossum 
and hyrax, where the start of tongue and 
hyoid protrusion is closely associated with 
minimum gape. In macaque and man, 
both of which have more vertically orient- 
ed tongues, the definitive (functional) for- 
ward movement begins during IP and, at 
least in macaque, ends with the last rate 
change in opening. Rate changes do occur 
in opening in man, but are only clearly as- 
sociated with cycles in which stage II 
transport is occurring (work in progress). 

These conclusions are, of necessity, 
based largely on qualitative data. What in- 
ferences can be drawn as to why this pat- 
tern is present and whether it evolved 
from an antecedent premammalian pat- 
tern? For most mammals, solid food has 
to be processed before it can be swal- 
lowed, requiring powered tooth— food- 
tooth contact, that is, the jaws have to sep- 
arate and then close with the food posi- 
tioned between the teeth. The tongue, giv- 
en the variable development of highly 
mobile cheeks in nonanthropoid mam- 
mals, is the primary agent for food manip- 
ulation and positioning. Clearly, food po- 
sitioning must occur in advance of a power 
stroke if such is to be effective and the 
tongue not be traumatized. The amplitude 
of jaw movement must be such as to allow 
large bites of material to be properly po- 
sitioned. It is possible that the rotational 

movements of the tongue surface associ- 
ated with food positioning are a mamma- 
lian adaptation. 

Equally, we argue that, both ontogenet- 
ically and phylogenetically, the fundamen- 
tal functional mammalian behavior is food 
transport. Liquids, requiring no process- 
ing, are simply moved through the mouth, 
into the oropharynx, and swallowed. The 
rhythmic cycle of hyolingual movement is 
the conveyor belt moving material back- 
ward. Tongue contacts with the hard pal- 
ate act as the stop to ensure the unidirec- 
tional (posterior) movement of food. In 
mammalian lapping, the amplitude of jaw 
movement is miniinal, sufficient to pro- 
trude the tongue, allow it to collect an al- 
iquot, and then retract. This low-ampli- 
tude gape (02) is maintained as the 
tongue completes its aliquot collection. 
The jaw cycle is modified when solid food 
requiring processing is introduced. The 
FO phase, with a longer close (FC and 
SC) appears. The underlying hyolingual 
cycle remains the same, its reversal from 
foiAvard to backward movement occurring 
at a gape in the range associated with lap- 

What is significant is that every mammal 
studied adopts, within and between se- 
quences, a cyclical behavior that can, post 
hoc, be correlated with the initial consis- 
tency of the food, and by inference, with 
the effect of its processing. This, given that 
functional behavior can change from cycle 
to cycle, implies continuous sensory feed- 
back to the CNS, which regulates the out- 
put from the CPG(s) producing rhythmic 
jaw and hyolingual movement. 

As Bramble and Wake (1985) observed, 
mammals are by no means unique in hav- 
ing jaws and a hyolingual complex. Their 
model for solid food transport in general- 
ized terrestrial tetrapods looks, as they 
stated, functionally very similar to that 
documented for solid food transport in 
mammals: the hyolingual complex is pro- 
truded as the jaws open and retracted dur- 
ing jaw closure. Does a commonality of 
outcome (food delivered to the digestive 

Tongue-Jaw Linkages • Hiiemae and Palmer 215 

tract) require a commonality of mecha- 
nism? This is the issue at the core of the 
question posed by Smith (1994) when she 
addressed the question: is this consistent 
pattern reflective of a 'conserved neuro- 
motor system'? Two issues must be ad- 
dressed. The first, given phylogenetically 
old osseocartilaginous and soft tissue ele- 
ments, connected by muscle blocks ar- 
ranged to generate movement in specific 
directions, is biomechanical. What muscle 
blocks contract in what way to produce 
what result temporally and spatially? Do 
they do so in a predictable and patterned 
order? If viewed simply as jaw elevators 
and depressors, and hyolingual protractors 
and retractors, then the activity of those 
muscle blocks, as modeled by Bramble 
and Wake (1985), could represent the sub- 
strate on which the more complex mam- 
malian pattern evolved. 

We argue that, given our current knowl- 
edge of the CNS control of rhythmic jaw- 
hyolingual behaviors in mammals and in 
nonmammalian tetrapods, it is premature 
to focus on neuromotor systems per se. 
(Given the foregoing, one might ask "what 
is a neuromotor system": anatomically or 
Rmctionally homologous structures?) It is 
now clear that complex CNS linkages form 
the CPGs for rhythmic jaw movement and 
swallowing in mammals (Dellow and 
Lund, 1971; Jean, 1990), although how 
linked tongue rhythmic behavior fits into 
those identified CPGs is not yet known. 
However, we do know that there is a com- 
plex, and probably itself experientially 
modulated, web of interconnectivity be- 
tween the sensory and motor nuclei in the 
pons and medulla for all the cranial nerves 
(CN V, VII, IX, X, XI, XII) involved. The 
importance of smell and taste in the selec- 
tion of foods for transport and processing 
(Gilbertson, 1998) cannot be ignored: in- 
put from those sensors clearly affects feed- 
ing behavior. At a time when the sources 
of sensory input from the mammalian oro- 
facial complex modulating rhythmic motor 
output are a matter of dispute, broader- 
brush inferences for the evolution of these 

functional behaviors and their control in 
terrestrial tetrapods are intriguing. Nev- 
ertheless, as such, they hopefully can serve 
as a stimulus to further, and difficult, re- 
search focused on the CNS rather than on 
the qualitative analysis of functional be- 
haviors as the basis for modeling CNS con- 
trol mechanisms. 


This review is dedicated to A. W. 
Crompton (A. W. C), without whose sup- 
port and encouragement (especially for K. 
M. H. in the late 1960s through early 
1970s) much of this body of research could 
never have been accomplished. Almost all 
of the authors cited in the Appendix were 
participants or beneficiaries of the Yale 
and then the Museum of Comparative Zo- 
ology research effort that A. W C. over- 
saw. Since 1991, Syracuse University Bio- 
engineering undergraduates electing Se- 
nior Thesis Research in K. M. H.'s labo- 
ratory at the Institute for Sensory 
Research have made much of the analysis 
reported here possible (1992-1999). We 
also acknowledge the superb technical 
support afforded by Xuezhen Wu and 
Chune Yang in Jeffrey Palmer's laboratory 
at Johns Hopkins University. U.S. Public 
Health Ser\dce Awards, first NIH DE 
05738, and later NIH DC 02123, aug- 
mented by institutional resources (Syra- 
cuse University and Johns Hopkins Uni- 
versity), supported this research. 


Anapol, F. 1988. Morphological and videographic 
study of the hyoid apparatus and its function in 
the rabbit (Onjctolagus cuniculus). Journal of 
Morpholog)', 195: 141-157. 

Bramble, D. M., and D. B. Wake. 1985. Feeding 
mechanisms of lower vertebrates, pp. 230-261. 
In M. Hildebrand, D. M. Bramble, K. F. Liem, 
and D. B. Wake (eds.). Functional Vertebrate 
Morphology. Cambridge, Massachusetts: Belk- 
nap Press of Hai"vard University Press. 430 pp. 

Bremer, F 1923. Physiologie nerveuse de la masti- 
cation chez le chat et lapin. Archives Interna- 
tionale Physiologique, 21: 309-352. 

CORTOPASSI, D., AND Z. MuHL. 1990. Videofluoro- 
graphic analysis of tongue movement in the rab- 

216 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

bit {Onjctolagus cuniculus). Journal of Morphol- 
ogy, 204: 139-146. 

Crompton, a. W. 1989. The evolution of mammalian 
mastication. In D. B. Wake and G. Roth (eds.). 
Complex Organismal Function: Integration and 
Evolution in Vertebrates. New York: John Wiley, 
xiii + 451 pp. 

Crompton, A. W, P. Cook, K. M. Hiiemae, and A. 
J. Thexton. 1975. Movement of the hyoid ap- 
paratus during chewing. Nature, 258: 69-70. 

Crompton, A. W, A. J. Thexton, P. Parker, and 
K. M. Hiiemae. 1977. The activity of the hyoid 
and jaw muscles during chewing of soft food in 
the opossum, pp. 287-305. In B. Stonehouse and 
D. Gilmore (eds.). The Biology of Marsupials. 
London: Macmillan. viii + 486 pp. 

Dellow, p., and J. Lund. 1971. Evidence for the 
central timing of rhythmical mastication. Journal 
of Physiology (London), 215: 1-13. 

DE Gueldre, G. D., and F. D. de Vree. 1984. 
Movements of the mandible and tongue during 
mastication and swallowing in Pteropus giganteus 
(Megachiroptera). Journal of Morphology, 179: 

Doran, G. 1975. Review of the evolution and phy- 
logeny of the mammalian tongue. Acta Anatom- 
ica, 91: 118-129. 

Franks, H. A., A. W. Crompton, and R. Z. Ger- 
man. 1984. Mechanism of intraoral food trans- 
port in macaques. American Journal of Physical 
Anthropology, 65: 275-282. 

Franks, H. A., A. W. Crompton, R. Z. German, 
AND K. M. Hiiemae. 1985. Mechanisms of in- 
traoral transport in a herbivore, the hyrax {Pro- 
cavia .syriaciis). Archives of Oral Biology, 30: 

German, R. Z., and H. A. Franks. 1991. Timing in 
the movements of the jaws, tongue and hyoid 
during feeding in the hwax (Procavia sijriacus). 
Journal of Experimental Zoology, 257: 34^2. 

German, R. Z, S. Sake, A. W. Crompton, and K. 
M. Hiiemae. 1989. Mechanism of food move- 
ment through the anterior oral cavity in anthro- 
poid primates. American Journal of Physical An- 
thropology. 80: 765-775. 

GiLBERTSON, T 1998. Peripheral mechanisms of 
taste, pp. 1-28. In R. W A. Linden (ed.). The 
Scientific Basis of Eating. Frontiers of Oral Bi- 
ology. Vol. 9. Basel, Switzerland: Karger. vii -I- 
244 pp. 

Hiiemae, K. M. 2000. Feeding in Mammals, pp. 
399-436. In K. Schwenk (ed.). Feeding: Form, 
Function, Phylogeny in Tetrapod Vertebrates. 
San Diego, Califonia: Academic Press, xv + 537 

Hiiemae, K., and A. Crompton. 1985. Mastication, 
food transport and swallowing, pp. 262-290. In 
M. Hildebrand, D. M. Bramble, K. Liem, and 
D. B. Wake (eds.). Functional Vertebrate Mor- 
phology. Cambridge, Massachusetts: Belknap 
Press of Harvard University Press. 430 pp. 

Hiiemae, K. M., M. R. Heath, G. Heath, E. Ka- 
ZOGLU, J. Murray, D. Sapper, and K. Ham- 
BLETT. 1996. Natural bites, food consistency and 
feeding behaviour in man. Archives of Oral Bi- 
ology, 41: 175-189. 

Hiiemae, K. M., and J. B. Palmer. 1999. Food 
transport and bolus formation during complete 
feeding sequences on foods of different initial 
consistency. Dysphagia, 14: 31-42. 

Hiiemae, K. M., A. Reese, and S. Hayenga. 1995. 
Patterns of tongue and jaw movement: a cineflu- 
orographic study of feeding in the macaque. Ar- 
chives of Oral Biology, 40: 229-246. 

Hiiemae, K. M., A. J. Thexton, and A. W Cromp- 
ton. 1978. Intra-oral food transport: a funda- 
mental mechanism of feeding?, pp. 181-208. In 
D. S. Carlson and J. A. McNamara, Jr (eds.). 
Muscle Adaptation in the Craniofacial Region. 
Ann Arbor, Michigan: Center for Human 
Growth and Development, The University of 
Michigan, x + 252 pp. 

Jean, A. 1990. Brainstem control of swallowing: lo- 
calization and organization of the central pattern 
generator, pp. 294-321. In A. Taylor (ed.). Neu- 
rophysiology of the Jaws and Teeth. London: 
Macmillan. xii + 397 pp. 

Kier, W, and K. K. Smith. 1985. Tongues, tentacles 
and trunks: the biomechanics of movement in 
muscular hydrostats. Zoological Journal of the 
Linnean Society, 83: 307-324. 

Livingstone, R. 1956. Some observations on the 
natural histoiy of the tongue. Annals of the Royal 
College of Surgeons of England, 19: 185-200. 

Oron, U., and a. W Crompton. 1985. A cineradio- 
graphic and electromyographic study of masti- 
cation in Tenrec ecaudatus. Journal of Moqihol- 
ogy, 185: 155-182. 

Palmer, J. B., K. M. Hiiemae, and J. Lui. 1997. 
Tongue— jaw linkages in feeding: a preliminary vi- 
deofluorographic study. Archives of Oral Biology, 
42: 429-441. 

Palmer, J. B., N. Rudin, G. Lara, and A. W. 
Crompton. 1992. Coordination of mastication 
and swallowing. Dysphagia, 7: 187—200. 

Schwenk, K. 2001. Intrinsic versus extrinsic lingual 
muscles: a false dichotomy? Bulletin of the Mu- 
seum of Comparative Zoology, 156: 219—235. 

Smith, K. K. 1994. Are neuromotor systems con- 
served in evolution? Brain, Behavior and Evo- 
lution, 43: 293-305. 

Smith, K. K., and W. Kier. 1989. Tnmks, tongues 
and tentacles: moving with skeletons of muscle. 
American Scientist, 77: 29—35. 

Thexton, A. J., and A. W Crompton. 1989. Effect 
of sensoiy input from the tongue on jaw move- 
ment in normal feeding in the opossum. Journal 
of Experimental Zoology, 250: 233-243. 

. 1998. The control of swallowing, pp. 168- 

222. In R. W A. Linden (ed.). The Scientific Ba- 
sis of Eating. Frontiers of Oral Biology. Vol. 9. 
Basel, Switzerland: Karger vii -I- 244 pp. 

Tongue-Jaw Linkages • Hiiemae and Palmer 217 

Thexton, a. J., AND K. M. Hiiemae. 1997. The ef- 
fect of food consistency on jaw movement in the 
macaque: a cineradiographic study. Journal of 
Dental Research, 76: 552-560. 

Thexton, A. J., K. M. Hiiemae, and A. W. Cromp- 
TON. 1980. Food consistency and particle size as 
regulators of masticatory behavior in the cat. 
Journal of Neurophysiology, 44: 456-474. 

Tongue movement of the cat during lapping. Ar- 
chives of Oral Biology, 33: 331-339. 

. 1989. Tongue movement in the cat during 

the intake of solid food. Archives of Oral Biology, 
34: 239-248. 

AND A. W. Crompton. 1982. Hyomandibular re- 
lationships during feeding in the cat. Archives of 
Oral Biology, 27: 793-801. 

TuRNBULL, W. 1970. Mammalian masticator)' appa- 
ratus. Fieldiana: Geology, 18: 153-356. 


The following lists those mammals for 
which data on tongue and hyoid move- 
ment in feeding are available, with the 
sources used in preparing this review. The 
order represents the approximate chronol- 
ogy of these studies. 

Opossum {Didelphis virginiana) 

CFG (lateral projection), EMG, tongue 
and hyoid luarkers. 

Sources. Crompton, 1989; Crompton et 
al., 1977; Hiiemae and Crompton, 1985; 
Thexton and Crompton, 1989, 1998. 

Cat (Felis domesticus) 

CFG (lateral projection), tongue and 
hvoid markers. 


Sources. Hiiemae et al., 1978; Thexton 

and McGarrick, 1988, 1989; Thexton et al., 
1980, 1982. 

Fruit bat {Pteropus giganteus) 

CFG (lateral and dorsoventral projec- 
tion), tongue (10) and hyoid markers. 
Source, de Gueldre and De Vree, 1984. 

Macaque {Macaca fascicularis) 

CFG (lateral projection), EMG, tongue 
and hyoid ixiarkers. 

Sources. Franks et al., 1984; German et 
al., 1989; Hiiemae and Crompton, 1985; 
Hiiemae et al., 1995. 

Tenrec {Tenrec ecaudatus) 

CFG (lateral and dorsoventral projec- 
tions), EMG, tongue and hyoid markers. 
Source. Oron and Crompton, 1985. 

Hyrax {Procavia syriacus) 

CFG (lateral projection), tongue and 
hyoid markers. 

Sources. Franks et al., 1985; German 
and Franks, 1991. 

Rabbit {Oryctolagus cuniculus) 

VFG (lateral and dorsoventral projec- 
tion), hyoid marker, tongue and hyoid 

Sources. Anapol, 1988; Cortopassi and 
Muhl, 1990. 

IVIan {Homo sapiens sapiens) 

VFG (lateral and posteroanterior pro- 
jection), EMG, tongue and hyoid markers. 

Sources. Hiiemae and Palmer, 1999; 
Palmer et al., 1992, 1997. 



Abstract. The muscular tongue of amniote verte- 
brates is traditionally described as a composite of two 
muscle types: extrinsic muscles originate outside the 
tongue and insert within it; intrinsic muscles arise and 
insert completely within the tongue. Whole-tongue 
movements are attributed to the former, lingual 
shape change to the latter. This dichotomous view of 
tongue structure and function has endured since the 
mid-19th century, despite persistent indications of its 
inadequacy. A histologic analysis of the musculi ge- 
nioglossus and verticalis in mammals and the mus- 
culus (m.) hyoglossus in lepidosaurian reptiles finds 
that the "extrinsic" m. genioglossus contributes exten- 
sively to the "intrinsic" m. verticalis; the verticalis 
"muscle" is composed of fibers from at least three 
nominally separate muscles, both extrinsic and intrin- 
sic (genioglossus, longitudinahs inferior, intrinsic ver- 
ticalis fibers); and the "extrinsic" m. hyoglossus in lep- 
idosaurs comprises both extrinsic and intrinsic parts, 
which may be histochemically differentiated. Current 
models of the tongue as a muscular hydrostat suggest 
tliat it functions as an integrated functional unit and 
that the traditional atomistic, dichotomous view is in- 
accurate and misleading. The notion of individuated 
"muscles" is inapplicable within the tongue and 
should be replaced by reference to "fiber systems." 

Apart from simplifying matters to the student of 
anatomy, the division of the lingual muscles into 
extrinsic and intrinsic groups is of no proper sci- 
entific significance (Abd-El-Malek, 1938: 26) 


The evolution of tetrapod vertebrates 
from piscine ancestors was attended by the 
appearance of a mobile, muscular tongue. 
The tongue, in effect, assumed the ances- 
tral role of water in the dynamics of feed- 
ing and is used by tetrapods today to cap- 
ture, support, manipulate, transport, and 

' Department of Ecology and Evolutionary Biology, 
University of Connecticut, Storrs, Connecticut 

swallow prey in the terrestrial environ- 
ment. These functions depend on the ca- 
pacity of the tongue, and the associated 
hyobranchial apparatus, to generate com- 
plex movements in three dimensions, 
within the mouth and without. The fonn 
and internal anatoiny of the tongue vary 
widely among tetrapods, as does the na- 
ture and extent of its movements (e.g., 
Livingston, 1956; Schwenk, 2000a). Mam- 
mals, lepidosaurian reptiles and terrestrial 
turtles possess the most inuscular and ar- 
chitecturally intricate tongues among tet- 
rapods, and these evince the greatest com- 
plexity of motion. Contradictions about the 
muscular constituents of these amniote 
tongues and their role in generating 
tongue movement is the subject of this pa- 

The tongue of inost nonarchosaurian 
amniotes is a large, inuscular mass, often 
with little or no internal skeletal support. 
The corpus of the tongue comprises or- 
thogonal arrays of interweaving muscle fi- 
bers, the pattern of which is taxonomically 
variable. Despite extensive comingling of 
muscle fibers within the tongue, early 
anatomists treated the tongue like any oth- 
er part of the musculoskeletal system and 
partitioned it into nominally discrete mus- 
cles. Initially, virtually all tongue muscles 
were thought to arise from elements of the 
skeleton outside the tongue (e.g., the man- 
dible, hyobranchium, and styloid process 
of the skull in mammals) and the muscles 
were divided on the basis of their separate 
origins (Barnwell, 1976). However, by the 
early 19th century it was recognized that 
some inuscle fibers arise and insert entire- 

Bull. Mus. Comp. Zool., 156(1): 219-235, October, 2001 219 

220 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

ly within the tongue and several "intrinsic 
muscles" were thus recognized in addition 
to the better known "extrinsic muscles" 
(Barnwell, 1976). Intrinsic muscles were 
identified on the basis of their direction 
(transverse, longitudinal, or circular) and 
position (superior/dorsal or inferior/ven- 
tral) within the tongue. 

The distinction between extrinsic and 
intrinsic muscles was formalized by Salter 
(1852; in Barnwell, 1976) who, in addition, 
attributed different kinds of tongue move- 
ment to the two muscle types: extrinsic 
muscles were said to move the whole 
tongue by virtue of dieir external skeletal 
attachments, whereas intrinsic muscles 
were thought to "move the tongue on it- 
self." Thus, by the mid-19th century, two 
parallel dichotomies were established in 
the literature: an anatomical division of the 
tongue into extrinsic and intrinsic muscles, 
and a functional division relating the for- 
mer to w^hole tongue movements and the 
latter to changes in tongue shape. 

Remarkably, this dichotomous view of 
tongue form and function has endured in 
the literature to the present time (e.g., 
Sonntag, 1925; McGregor, 1938; Bennett 
and Hutchinson, 1946; Cooper, 1953; Liv- 
ingston, 1956; Oelrich, 1956; Sondhi, 
1958; Bowman, 1968; Perkell, 1969; War- 
wick and Williams, 1973; Miyawaki, 1974; 
Barnwell et al., 1978b; Langdon et al., 
1978; Hellstrand, 1980, 1981; Tanner and 
Aveiy, 1982; Schwenk, 1986; Smith, 1988; 
Delheusy et al, 1994; Herrel et al., 1995). 
For example, in describing a lizard tongue, 
Oelrich (1956: 54) stated: "The extrinsic 
muscles, genioglossus and hyoglossus, con- 
trol the motions of the tongue; the intrin- 
sic muscles control its shape." Hellstrand 
(1980: 187) began his paper on the cat 
tongue by pointing out that it "is provided 
with muscles termed extrinsic or intrinsic 
according to whether they run partly or 
totally within the tongue. Functionally, the 
extrinsic muscles are usually classified as 
protruders or retractors and the intrinsic 
as shaping or modeling agents." In Gray's 
Anatomy (Warwick and Williams, 1973: 

1239, 1240) it is noted that within the hu- 
man tongue, "there are two sets of mus- 
cles, extrinsic and intrinsic; the former 
have attachments outside the tongue, the 
latter are contained within it." Each ex- 
trinsic muscle is said to move the tongue 
in some way, that is, retract, depress, or 
elevate it. In contrast, it is held that the 
intrinsic muscles, in toto, are "mainly con- 
cerned in altering the shape of the 

Despite the persistence of the dichoto- 
mous descriptive convention, those who 
have investigated tongue anatomy in some 
detail have often questioned the accuracy 
or appropriateness of the dichotomy — an- 
atomically, functionally, or in both ways 
(e.g., Bennett, 1935; Abd-El-Malek, 1938; 
Bennett and Hutchinson, 1946; Sondhi, 
1958; Barnwell et al., 1978a; Langdon et 
al., 1978; Cave, 1980; Lowe, 1980; Kier 
and Smith, 1985; Schwenk, 1986; Smith, 
1986, 1992; Smith and Kier, 1989; Sokoloff 
and Deacon, 1992; Napadow et al, 1999). 
Some authors have expressed doubt, even 
while beginning with the conventional 
view. Oelrich (1956: 55), for example, not- 
ed that extrinsic and intrinsic fibers inter- 
lace within the tongue and admitted that 
the intrinsic muscles "do not maintain 
their integrity throughout, but at some lev- 
els are intermingled to such an extent that 
their identity is obscured." Schwenk 
(1986: 137) pointed out that in tuatara 
(Sphenodon), the "distinction is not always 
demonstrable in eveiy part of the tongue 
because both intrinsic and extrinsic fibers 
interlace complexly." Barnwell et al. 
(1978a: 8) concluded that the nominally 
intrinsic musculus (m.) longitudinalis su- 
perior of the huinan tongue "is comprised 
of both intrinsic and extrinsic fiber 
groups." It is telling that many writers 
seem compelled to state the conventional 
view, despite their evident dissatisfaction. 
This ambivalence is clear, for example, in 
a textbook account that virtually contra- 
dicts itself within the space of two sen- 
tences: "Generally, 'movements' other 
than those that basically alter the shape of 

Tongue Muscles • Schwenk 221 

the tongue are the result of contractions of 
the extrinsic muscles, though one group 
seldom functions alone. The overlapping, 
intermingling, and decussating nature of 
the intrinsic and extrinsic inuscle groups 
permit the fine coordinated effort so nec- 
essary in speech" (Hiatt and Gartner, 
1982: 239-240). 

Others have rejected the traditional di- 
chotomy altogether. For example, in his 
study of cat tongue anatomy and function, 
Abd-El-Malek (1938) concluded with the 
remark quoted at the outset of this paper, 
suggesting that the dichotomy is no more 
than a convenience, without scientific 
merit. He particularly rejected the func- 
tional dichotomy, suggesting that "most, if 
not all, of the intrinsic muscles are in- 
volved in every movement of the tongue. 
Indeed, in many movements both intrinsic 
and extrinsic muscles so called, are work- 
ing together." Other authors take the rad- 
ical view that all putative intrinsic muscle 
fibers are nothing more than extensions of 
extrinsic muscles (Lesbre [1922] in Cave 
[1980] for the horse, Eqiiiis; Cave [1980] 
for the rhinoceroses. Rhinoceros, Cera- 
totheriiini, and Diceros; and Sondhi [1958] 
for the monitor lizard, Varanus). Accord- 
ing to Cave (1980: 128): "The so-called in- 
trinsic tongue muscles are not therefore, 
morphological entities but merely contin- 
uations of the extrinsic muscles." Sondhi 
(1958: 175) concluded: "While there can 
be no doubt that the 'intrinsic muscles' can 
be distinguished from each other in cer- 
tain regions of the tongue, the fact that 
they arise directly as a result of the change 
in course of certain bundles of the [extrin- 
sic] hyoglossus fibres indicates that they do 
not deserve the status of independent 
muscles." It is worth noting that the stud- 
ies of Lesbre and Cave were based on 
gross dissection without the benefit of his- 
tologic sections, and Sondhi's sections 
were of poor quality. 

The purpose of this paper is to explore 
the anatomical relationship between "ex- 
trinsic" and "intrinsic" lingual muscles in 
representative mammals and lepidosaurian 

reptiles. The mammalian genioglossus and 
verticalis muscles, and lepidosaurian hyo- 
glossus muscle are investigated in detail 
and used as exemplars. Results of the mor- 
phologic investigation are considered in 
light of current models of tongue function 
in order to assess the structural and func- 
tional validity of the dichotomous view. 
Based on this analysis, a synthetic view of 
tongue form and function is proposed. 


Reference material included complete, 
serial paraffin sections of tongues from two 
mammalian species (domestic cat. Fells ca- 
tus, three specimens; crab-eating ma- 
caque, Macaca fascicularis, three speci- 
mens) and more than 100 lepidosaurian 
reptile species (one to three specimens 
each), including tuatara (Sphenodon piinc- 
tatus; Schwenk, 1986) and representatives 
of every major squamate clade (Schwenk, 
1988, 2000b, unpubfished data). The em- 
phasis here is on the cat (Camivora, Feli- 
dae) and generalized lizards (Squamata, 
Iguanidae) that putatively retain the ple- 
siomorphic condition (Schwenk, 1986, 
1988, 2000b). Histologic results were com- 
pared to literature accounts of tongue 

Sections were prepared using standard 
paraffin techniques (Presnell and Schreib- 
man, 1997). Whole-tongue specimens 
were sectioned whenever possible to facil- 
itate fiber tracing. Some previous studies 
have suffered from the myopic view of- 
fered by partial or fragmented specimens. 

Both transverse and sagittal sections (6— 
10 |jLm) were prepared for most species, 
but only transverse sections were available 
for Sphenodon and some squaiuates (note 
that transverse tongue sections correspond 
to 'coronal' sections in the parlance of hu- 
man anatomy). Sections were stained with 
heiTiatoxylin and picro-ponceau or hema- 
toxylin and eosin (Presnell and Schreib- 
man, 1997). 

In addition to paraffin sections, frozen 
tissue sections were available for several 
squamate species. These were stained for 

222 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

myosin adenosine triphosphatase and sue- rupted across its width by invading bands 

cinic dehydrogenase using standard tech- of transversus. However, interpreting sec- 

niques, as part of an ongoing histochemical tions here is difficult because the sheets 

study of tongue muscle fiber types apparently do not run in a plane, becom- 

(Schwenk and Anapol, in preparation), ing curved or cup-shaped, instead. Finally 

These data are preliminary and referred to some vertical fibers cross anteroposteriorly 

here only in passing. between adjacent sheets to form anasto- 
moses (especially evident in the free, an- 

RESULTS tenor part of the tongue; e.g.. Fig. 3). De- 

Genioglossus and Verticalis Muscles in ^P^*^ these complications, throughout 

thp Pflt most or the tongue tlie extreme regularity 

of the alternating vertical and transverse 

Preliminary observations indicate that sheets is its most striking feature. The 
the findings reported here for the cat are sheets of vertical and transverse fibers con- 
equally valid for the macaque. Based on stitute the nominal intrinsic muscles, m. 
consideration of the literature, the findings verticalis and m. transversus, respectively, 
are likely to apply as well to many mam- In the cat they are separated into left and 
mals with generalized tongues, including right moieties by a complete median sep- 
other carnivorans, opossums, and humans, tum. 

However, with the exception of humans. The m. transversus fibers originate from 

detailed information is lacking for these the median septum and run radially across 

and other species, including many with the width of the tongue to insert into the 

highly divergent tongue forms (e.g., mono- lamina propria of the lingual tunic dorsally 

tremes: Doran and Baggett, 1970; Doran, laterally, and ventrolaterally. Some m. ver- 

1973; and nectar-feeding bats: Greenbaum ticalis fibers originate from the lamina pro- 

and Phillips, 1974; T Griffiths, 1978). pria of the tongue's ventral surface and nan 

Therefore, the results of this study cannot dorsomedially to the lamina propria of the 

necessarily be extrapolated to mammals as dorsal surface (but see below). In the lat- 

a whole. eral part of the tongue, dorsolaterally nm- 

When viewed in transverse section, ning transversus fibers cross dorsomedially 
most mammalian tongues are divisible into running verticalis fibers in an X-like pat- 
cortical and medullary regions. The cortex tern (i.e., peipendicular, but approximately 
is distinguished by the presence of longi- 45° to the vertical). Toward the midline 
tudinal fibers, whereas the core of the verticalis fibers become more nearly ver- 
tongue is filled with transverse and vertical tical (see below). 

fibers. "Transverse" and "vertical" are con- The m. genioglossus is one of the major 
venient descriptors for these more-or-less extrinsic muscles of the tongue and is as- 
perpendicular sets of fibers, but fibers of sumed to be its principal protractor and 
both groups are often quite oblique, protruder. In the cat, the m. genioglossus 
Throughout most of the medullary zone, originates medially from two heads on the 
vertical and transverse fibers are organized mandible near the symphysis. A ventral 
into thin sheets of muscle that nm across head gives rise to fibers that run posteri- 
the width of the tongue, alternating one orly, inserting onto the anteroventral sur- 
after the other along the tongue's length face of the basihval. Fibers from a larger 
(Fig. 1). Tracing individual verticalis sheets dorsal head run posterodorsally in a fanlike 
through serial sections confirmed that in array, penetrating the tongue midventrally 
most of the tongue each sheet runs across along the posterior two thirds of its length, 
the full width of the medullary core. How- The most anterior of these fibers curve 
ever, in the anteriormost part of the shaiply dorsad as they enter the tongue's 
tongue, a verticalis sheet may be inter- medullaiy core and nm vertically to the 

Tongue Muscles • Schwenk 223 

Figure 1 . Alternating sheets of vertical (musculus [m.] verticalis) and transverse (m. transversus) muscle fibers in the medullary 
core of the cat midtongue. Note extreme regularity of alternating sheets. (A) Parasagittal section; verticalis sheets are thicker 
than transversus sheets in this region. Anterior is to the left. Scale bar = 0.2 mm. (B) Transverse section near the midline, in 
the ventromedial portion of the tongue's right half; medial septum to the right. Section is slightly oblique relative to plane of 
transversus and verticalis sheets so that it passes through several adjacent layers. As for (A), note extreme regularity of alter- 
nating pattern. Darker-staining tissue at the margins of each sheet is collagenous connective tissue of the thin fascial plane 
separating each sheet. Note that the ventralmost transverse fibers run ventrolaterally and the dorsalmost fibers run laterally. 
Dorsal to these, out of the photographic frame, the transverse fibers run dorsolaterally, that is, the transverse fibers radiate 
laterally from their midline origin on the median septum. Scale bar = 0.2 mm. 

lamina propria of the dorsal surface, but 
posteriorly the fibers become progressively 
more longitudinal, cuiAdng gently dorsad at 
their distal ends to meet the posterior sur- 
face of the tongue as it slopes downward 
to the root of the tongue. In (fetal) hu- 
mans, the anteriormost fibers form a third 
bundle that turns sharply anterior as it en- 
ters the tongue, running to the tip (Lang- 

don et al., 1978), but in the cat (Abd-El- 
Malek, 1938; this study) and several other 
mammals (Doran and Baggett, 1972), no 
comparable bundle is found and the most 
anterior genioglossus fibers run more or 
less vertically. The unattached, anterior 
part of the tongue is therefore devoid of 
genioglossus fibers. 

In reconstructions of the genioglossus. 

224 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

its fibers are often shown in sagittal view 
to end at the base of the tongue before 
penetrating the medullary core, or to run 
obliquely across the alternating sheets of 
vertical and transverse fibers toward the 
dorsal and posterior surfaces of the tongue 
(e.g., Kallius, 1910; Abd-El-Malek, 1938; 
Warwick and Williams, 1973; Crouch, 
1978; Walker and Homberger, 1992). In- 
dependence of the genioglossus from the 
medullary, intrinsic fiber system is implied 
in these and other descriptions. In actu- 
ality, as genioglossus fibers turn dorsally 
into the tongue, they become confluent 
with the serially arranged sheets of verti- 
calis fibers (Fig. 2A). As such, for most of 
the tongue's length, the medial portion of 
the verticalis inuscle comprises genioglos- 
sus fibers. In other words, a large portion 
of the "intrinsic" verticalis muscle is com- 
posed of "extrinsic" fibers. 

Although the previous observations 
seem to support the radical position of 
Cave (1980) and others suggesting that in- 
trinsic fibers are merely extensions of ex- 
trinsic muscles, this view is vitiated by a 
full consideration of m. verticalis anatomy. 
Although the medial portion of each ver- 
ticalis sheet comprises extrinsic genioglos- 
sus fibers, its lateral portion derives from 
purely intrinsic fibers that originate on the 
lamina propria of the ventral surface (Fig. 
2B). Although medial genioglossus fibers 
are relatively vertical and lateral intrinsic 
fibers are oblique (running dorsomedially; 
see above), fibers of both sorts blend in- 
sensibly across the breadth of a single ver- 
ticalis sheet to form a continuous structur- 
al unit. These units are repeated serially 
along the length of the tongue, alternating 
with sheets of m. transversus. Extrinsic 
and intrinsic components of the verticalis 
are also clearly evident in the opossum, 
Monodelphis (Smith, 1994, fig 2a). 

Given the absence of genioglossus fibers 
in the anterior, free part of the tongue, one 
might expect the sheets of verticalis in this 
region to be uniformly intrinsic. Indeed, 
this is true laterally, as elsewhere in the 
tongue (Fig. 3A). However, in place of ge- 


Figure 2. The relationship between musculus (m.) verticalis 
and m. genioglossus fibers in the cat tongue. (A) Parasagittal 
section near the midline, anterior to the left. Ventrally, the ge- 
nioglossus muscle (g) penetrates the tongue's medullary core 
where it curves dorsad and is separated into separate layers 
by intervening sheets of transversus fibers (t). Thus, in this 
medial portion of the tongue the vertical fibers of the medulla, 
nominally m. verticalis (v), are actually contributed by the ex- 
trinsic genioglossus. However, note that some verticalis fibers 
continue a more vertical descent through the genioglossus to 
an intrinsic point of origin (curved arrow). Scale bar = 0.2 mm. 
(B) Transverse section through the ventrolateral part of the 
tongue's right side showing longitudinal fibers of lingual cortex 
and the origin of intrinsic verticalis fibers from the ventrolateral 
lamina propria (arrows). These lateral, purely intrinsic fibers, 
form an uninterrupted continuum within a single verticalis 
sheet with the medial genioglossus fibers. Scale bar = 0.2 

Tongue Muscles • Schwenk 225 

Figure 3. The free, anterior portion of the cat tongue in sag- 
ittal section. The tongue tip is toward the left. Note that longi- 
tudinal fibers run beneath the tongue's dorsal and ventral sur- 
faces forming the cortex, whereas the medulla is filled by the 
alternating sheets of vertical and transverse fibers. These 
sheets are not so regularly disposed as they are posteriorly 
(Fig. 1) and anastomoses between verticalis sheets are fre- 
quent. (A) Section through lateral part of the tongue. Note that 
vertical fibers penetrate the cortex to arise and insert from dor- 
sal and ventral laminae propria, that is, they are intrinsic fibers. 
The dark-staining structure just above the ventral longitudinal 
fibers is a nerve, kinked to permit extension during hydrostatic 
tongue elongation. The white areas are vascular spaces slight- 
ly distended by perfusion of the tongue. Scale bar = 0.2 mm. 
(B) Section more medial to (A), near to midline. In contrast to 

nioglossus fibers, the medial portion of 
each verticalis sheet is here occupied by 
ventral longitudinal fibers that turn dorsad 
into the verticalis (Fig. 3B). The ultimate 
origin of all ventral longitudinal fibers has 
not been traced with certainty, but most 
clearly belong to the intrinsic m. longitu- 
dinalis inferior. Among the ventral longi- 
tudinal fibers of the lingual cortex, longi- 
tudinalis inferior fibers are generally the 
most median (e.g., Barnwell et al., 1978b). 
Nonetheless, longitudinal fibers of the cor- 
tex are notoriously difficult to segregate 
according to source and it remains possi- 
ble that some of the fibers contributing to 
the verticalis in the anterior part of the 
tongue derive from the extrinsic styloglos- 
sus muscle. Sections show that styloglossus 
fibers course anteroventrally along the 
sides of the tongue, joining the ventral lon- 
gitudinal system anteriorly, but it is not 
certain that these fibers extend far enough 
anteriorly and medially to contribute to 
the verticalis in the free part of the tongue. 
Although (extrinsic) m. hyoglossus fibers 
are said to nan \\dthin the ventral longitu- 
dinal system of some inammals (e.g., hu- 
mans, Barnwell et al., 1978b), my sections 
(and those of Abd-El-Malek, 1938) indi- 
cate that in the cat, all hyoglossus fibers 
run anterodorsally into the dorsal longitu- 
dinal system. 

In conclusion, the extrinsic genioglossus 
muscle makes a substantial contribution to 
the putatively intrinsic verticalis muscle. 
Anteriorly, the verticalis also includes fi- 
bers of a second intrinsic muscle, the lon- 
gitudinalis inferior, and possibly extrinsic 
fibers of the st\'loglossus. The verticalis 
"muscle" thus includes fibers from three, 
or possibly, four different sources: intrinsic 
vertical fibers; intrinsic longitudinal fibers; 
extrinsic genioglossus fibers; and possibly. 

(A), vertical fibers are here contributed by the extensively de- 
veloped ventral longitudinal system. These fibers represent 
musculus longitudinalis inferior, although it is possible that ex- 
trinsic styloglossus fibers also contribute. Compare to Figure 
2A. Scale bar = 0.2 mm. 

226 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 4. Transverse section througli the midtongue of a lizard (Holbrookia texana, Iguanidae). The circular structure in the 
center is the lingual (entoglossal) process of the hyobranchium, which is surrounded by fibers of the midline intrinsic muscle, 
musculus (m.) verticalis (V). Dorsal to the verticalis is the m. transversalis (T) and on either side are the paired hyoglossus 
bundles (H); the lateral part of each bundle is cut off in the figure. Note that within each hyoglossus bundle the fibers are 
separated into two parts: a dense, more vertically oriented dorsolateral portion (1) and a more loosely organized, more longi- 
tudinally oriented ventromedial portion (2). Scale bar = 0.2 mm. 

extrinsic styloglossus fibers. An individual 
sheet of verticalis occupies a transverse 
plane across the width of the tongue com- 
prising a continuum of vertically oriented 
muscle fibers, yet within a given sheet, a 
large proportion of the fibers are contrib- 
uted by a nominally separate muscle, usu- 
ally from outside the tongue. Further- 
more, the muscles making this contribu- 
tion vary along the length of the tongue. 
Thus, the nominal m. verticalis satisfies 
neither the definition of "intrinsic," nor 
even the usual notion of a "muscle." None- 
theless, the serial coherence of the verti- 
calis is maintained throughout the med- 
ullaiy core, despite the disparate sources 
of its constituent fibers. 

The Hyoglossus Muscle in Lepidosaurlan 

Lepidosaurs include the tuatara of New 
Zealand (Sphenoclon) and the squamates, 
comprising lizards, snakes, and amphisba- 
enians. The lepidosaurian tongue, with few 
exceptions, is a highly mobile organ that 
rivals that of mammals in its internal com- 
plexity. However, unlike mammals, the 

principal longitudinal muscles of the 
tongue lie within its core and not its pe- 
riphery. These are the hyoglossus muscles, 
evident in transverse section as two large, 
cylindrical or subcylindrical bundles (Fig. 
4). In a few taxa (notably gekkotans) they 
subdivide anteriorly into multiple bundles, 
but in the vast majority of species they re- 
main paired for the length of the tongue 
(Schwenk, 1988, 2000b). 

As in maminals, the hyoglossus is one of 
the major extrinsic muscles of the lepido- 
saurian tongue. It is traditionally described 
as originating on the first ceratobranchial 
of the hyobranchial apparatus and insert- 
ing within the tongue near its tip (e.g., 
Gnanamuthu, 1937^ Oelrich, 1956; Del- 
heusy et al., 1994; Herrel et al, 1997, 
1999) and is regarded as the principal re- 
tractor of the tongue. 

Some studies have indicated that hyo- 
glossus anatomy is more complex than sug- 
gested by conventional descriptions. In his 
figures of transversely sectioned Sphen- 
oclon embryos, Edgeworth (1935) identi- 
fied a separate ventromedial bundle within 
the hyoglossal mass, which he called the 

Tongue Muscles • Schwenk 227 








— <c/^^- ""-« 


^ ^ 

<."^~ '*^. 




— ^-S^ 




:::^'^ -i 





: % 

"- 4j 



* » 

^'^- ^.,. 

. :.'/' A 



K '• 







m . *'-v .t-^ — z* 



^; ii ..-. 

.. .^ 

Figure 5. Transverse section through the midtongue of two 
iguanid lizards. (A) Opiums sebae. showing the lingual pro- 
cess and musculus (m.) verticalis medially, as in Figure 4. On 
either side of the verticalis are the hyoglossus bundles (trun- 
cated laterally by the photograph) showing subdivision of the 
muscle into three portions (1, 2, 3) indicated by differences in 
fiber density and orientation. Dorsally, parts 1 and 2 appear to 
be separated by a thin fascial plane, but ventrally the three 
zones blend insensibly. Scale bar = 0.4 mm. (B) Stenocercus 

longitudinalis linguae. Schwenk (1986) did 
not report such a di\dsion in his study of 
an adult specimen, but the sections reveal 
different fiber orientations in the ventro- 
medial and dorsolateral portions of the 
hyoglossus bundle, supporting the notion 
of a subdivided hyoglossus in Sphenodon. 
Smiths (1988) transverse sections of agam- 
id lizard tongues show distinct partitioning 
of the hyoglossus into two or three divi- 
sions. Smith (1986) suggested that each 
hyoglossus bundle within the highly mod- 
ified tongues of inonitor lizards (Varatms) 
comprises a series of shorter fibers lam- 
ning obliquely within the longitudinal bun- 
dle, possibly spiraling, and inserting into 
the surrounding epimysium along the 
length of the bundle. She disagreed with 
Sondlii's (1958) suggestion that hyoglossus 
fibers in this genus turn within the tongue 
to form the circular fibers surrounding the 
hyoglossus bundles. Indeed, she showed 
that the circular muscles were, themselves, 
composed of shorter, oblique fibers ar- 
rayed helically around each hyoglossus 

A cursory survey of tongue sections 
from a variety of species indicates that in- 
ternal hyoglossus partitioning is common- 
place among lepidosaurs. Transverse sec- 
tions of iguanian lizards, in particular, of- 
ten reveal two (Fig. 4) or even three (Figs. 
5A, B) different moieties, as indicated by 
fiber orientation and, sometimes, fascial 
planes. The number and distinctness of 
the hyoglossus divisions vary among spe- 
cies and, notably, along the length of the 
tongue. In some sections it is evident that 
one division contains vertically oriented fi- 

sp., showing three-part subdivision of the hyoglossus, as in 
(A). The dark-staining band separating transverse fibers from 
the verticalis (v) and hyoglossus (1, 2, 3) is the collagenous 
connective tissue of the dorsal transverse septum. The division 
between verticalis and hyoglossus fibers is indicated on one 
side by three small arrows. Fibers of the middle hyoglossus 
subdivision (2) in this part of the tongue are vertically oriented 
and originate on the dorsal transverse septum (large arrows). 
These fibers represent an intrinsic component of the putatively 
extrinsic m. hyoglossus in lepidosaurs. Scale bar = 0.1 mm. 

228 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

I mi] 

Figure 6. Parasagittal sections of the tongues in two lizards. (A) Sceloporus sp. (Iguanidae). Anterior to the left. The dorsal 
surface of the tongue is covered by long, filamentous papillae. Beneath the papillary surface is a layer of intrinsic longitudinal 
and transverse fibers. Directly beneath the latter is the dark-staining connective tissue band of the dorsal transverse septum 
(arrows). The plane of section passes through the verticalis (v) posteriorly and the hyoglossus (i and e) anteriorly. Note that the 
longitudinal hyoglossus bundle occupies the depth of the tongue from the transverse septum dorsally to the base of the tongue 
ventrally (compare Figs. 4 and 5). The extrinsic hyoglossus is shown here to consist of an intrinsic component (i), originating on 
the dorsal transverse septum and running anteroventrally toward the tongue tip, and an extrinsic component (e), running anter- 
odorsally from the hyobranchium (the posterior extent of this muscle is not evident in this section). Scale bar = 0.2 mm. (B) 
Gonatodes antillensis (Gekkonidae). Anterior to the right. In gekkotans and other scleroglossan squamates, the dorsal transverse 
septum is weakly developed or absent. Nonetheless, the hyoglossus bundle comprises fibers with an intrinsic origin (i) and an 
extrinsic origin (e), as in (A). Scale bar = 0.2 mm. 

bers taking origin from the connective tis- 
sue plane of the dorsal transverse septum 
(Fig. 5B). Sagittal sections reveal that 
these fibers form a substantial bundle run- 
ning anteroventrally froin the dorsal trans- 
verse septum to the tongue tip, joining an- 
terodorsally directed fibers arriving from 

the hyobranchium (Fig. 6). In transverse 
section these separate groups of fibers are 
evident as the aforementioned subdivi- 
sions within the hyoglossus bundle. In oth- 
er words, the hyoglossus inuscle in most 
lepidosaurs minimally includes two dis- 
tinct components: an intrinsic part, origi- 

Tongue Muscles • Schwenk 229 

nating within tlie tongue from the dorsal 
transverse septum, and an extrinsic com- 
ponent, originating from the hyobran- 

This brief characterization does not cap- 
ture the full complexity of the hyoglossus 
muscle. Changes in cross-sectional shape 
of individual fibers along the length of the 
tongue suggest some degree of spiraling 
within the bundle so that extrinsic and in- 
trinsic fibers are interwoven. As such, the 
number, form, and distinctness of the hyo- 
glossus divisions vary anteroposteriorly. 
However, no matter how great its internal 
complexity, conser\'atively two different 
sources of fibers contribute to the hyo- 
glossus muscle, including one that is fully 

Analysis of preliminary data indicates 
that the intrinsic component of the hyo- 
glossus is histochemically distinct in iguan- 
ids (Sceloporus graciosus and Phnjtiosorna 
platijrhinos) and a gecko (unidentified 
species), but not in a teiid {Ameiva un- 
dulatus). In Varanus niloticus, the hyo- 
glossus is neither subdivided nor does it 
show histochemically distinct compart- 
ments (Schwenk and Anapol, in prepara- 
tion). This is consistent with Smiths (1986) 
interpretation of hyoglossus anatoiny in 
Varanus. The tongues of Ameiva and Var- 
anus are both specialized for rapid length- 
ening of the tongue for chemosensory 
tongue-flicking (Smith, 1984, 1986; 
Schwenk, 2000b), whereas iguanids and 
gekkotans exhibit more variable and com- 
plex lingual shape changes associated with 
feeding and other behaviors. It is tempting 
to relate the different histochemical pro- 
files to these functional differences, but 
this is highly speculative at this point. 

In summary, the putatively extrinsic 
hyoglossus muscle includes both an extrin- 
sic and an intrinsic component in many, if 
not most, lepidosaurian reptiles. The in- 
trinsic component originates from the dor- 
sal, transverse septum in the posterior and 
midtongue and runs anteroventrally to the 
tongue tip. The extrinsic component arises 
from the first ceratobranchial of the hyob- 

ranchiuin and runs anterodorsally to the 
tongue tip. It is likely that these fiber 
groups spiral and are complexly interwo- 
ven, changing orientation along the length 
of the tongue. The intrinsic fiber compo- 
nent is histochemically distinct in basal liz- 
ards with functionally generalized tongues. 


The Anatomical Dichotomy 

The results of this study clearly indicate 
the inadequacy of segregating tongue mus- 
cles into "extrinsic" and "intrinsic" types. 
Indeed, the traditional notion of what con- 
stitutes a "muscle," based as it is on more 
typical components of the musculoskeletal 
system, inay be inapplicable within the 
tongue. Only three muscles were consid- 
ered here in any detail, but comparable 
issues arise for nearly every noininal inus- 
cle in the amniote tongue. The genioglos- 
sus is putatively an extrinsic muscle, but it 
contributes substantially to the vertical fi- 
ber system of the tongue's medullary core, 
presumed to consist of intrinsic fibers. 
Conversely, the putatively intrinsic verti- 
calis muscle is largely composed of extrin- 
sic (genioglossus) fibers, as well as fibers 
from one to two other nominally separate 
muscles (longitudinalis inferior and stylo- 
glossus). The lepidosaurian hyoglossus is 
putatively an extrinsic muscle, but a large 
portion of its fibers arises intrinsically. 
Thus, the dichotomous view of tongue 
musculature is falsified. 

The Functional Dichotomy 

Tongue movement derives from one or 
a combination of three different mecha- 
nisms evident among tetrapods: (1) whole- 
tongue movement coupled to hyobranchial 
movement; (2) whole-tongue movement 
independent of hyobranchial movement; 
or (3) length or shape changes intrinsic to 
the tongue and independent of hyobran- 
chial movement (e.g., Livingston, 1956; 
Schwenk, 2000a,b). In the traditional \dew, 
mechanism two is attributed to the action 
of the extrinsic inuscles and inechanism 

230 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

three to the intrinsic muscles. To exempli- 
fy the differences among mechanisms, 
consider tongue protrusion. The tongue 
can be protruded if it is "pushed" beyond 
the ]aw margins by protraction of the 
hyobranchial apparatus. Alternatively, the 
tongue can be "pulled" out of the mouth 
by muscles linking it to the mandible. In 
this case, movement of the tongue relative 
to a fixed hyobranchium must be possible, 
as occurs, for example, in many lizards 
when the tongue slides along the lingual 
process of the basihyal (Schwenk, 2000b). 
Finally, the tongue can protrude beyond 
the jaw margins if it elongates anteriorly 
by changing its shape, despite being fixed 
to a stationary hyobranchium posteriorly. 
Obviously, these mechanisms might be 
combined to effect tongue protrusion. 

The first two types of tongue movement 
conform to traditional models of muscu- 
loskeletal movement, but the third implies 
a far more complex mechanism. The na- 
ture of this mechanism and its relationship 
to the tongues internal architecture were 
appreciated by Owen (1868: 394), who de- 
scribed it in reference to the giraffe and 
its ability to strip an acacia tree of leaves 
with its highly protiTisible tongue: 

The muscular fibres in the free and flexible part of 
the tongue present an arrangement adequate to all 
its movements. The stylo-glossi and inferior lingua- 
les expand into a layer of longitudinal fibres . . . 
these longitudinal muscles inclose a mass of fibres, 
which run in die transverse direction. The action 
of the transverse, combined with that of several 
short vertical, fibres near the margins, and of those 
forming the thin circular stratum smrounding the 
stylo-glossi at the middle part of the tongue, serves 
to attenuate or diminish the transverse diameter of 
the tongue and increase its length; while thus rig- 
idly extended the apex of the tongue can be cui-ved 
upward or downward by tlie superficial longitudinal 
fibres .... 

The intrinsic mechanism of tongue 
elongation outlined by Owen (1868) and 
others (e.g., M. Griffiths, 1968, 1978; Win- 
kelmann, 1971, in T Griffiths, 1978) was 
formalized by Kier and Smith (1985) in 
their "muscular hydrostat" model of 
tongue movement (also Smith and Kier, 

1989). They modeled the tongue (and sim- 
ilar organs) as a constant-volume cylinder 
filled with incompressible fluid (intracel- 
lular water). They recognized that a re- 
duction in one dimension must cause a 
compensatory increase in another and 
showed that muscular hydrostats are char- 
acterized by orthogonal arrays of muscle 
fibers arranged to modulate the tongues 
diameter and length. In addition, superfi- 
cial longitudinal fibers cause bending and 
helical or oblique fibers cause torsion. Dif- 
ferential, localized activity of these fiber 
systems can potentially create a vast range 
of complex shape changes in tongue form. 

In light of the muscular hydrostat model 
of tongue function, it is clear that the se- 
rial, orthogonal arrangement of muscle fi- 
bers in the medullary region of the mam- 
malian tongue serves to lengthen the 
tongue by decreasing its diameter, regard- 
less of whether individual fibers originate 
intrinsically or extrinsically Likewise, lon- 
gitudinal fibers of the cortex participate in 
tongue shortening, retraction, bending, 
and torsion, regardless of nominal origin. 
In other words, within the tongue, muscle 
fibers behave as organized systems that do 
not correspond to the extrinsic— intrinsic 
dichotomy: within the vertical fiber sys- 
tem, both intrinsic and extrinsic fibers 
function together to reduce the vertical di- 
mension of the tongue. Furthermore, the 
serial vertical system must often act in 
concert with the serial transverse system 
to reduce tongue diameter uniformly 
along its length. By implication, a relatively 
"simple" action, such as tongue elongation, 
involves minimally four nominal "mus- 
cles," including both "extrinsic" and "in- 
trinsic" types. 

Similarly, the lepidosaurian tongue is ca- 
pable of hydrostatic shape change. Al- 
though the m. hyoglossus is assumed to be 
a tongue retractor, its internal complexity, 
as described here, suggests that it may be 
active during "intrinsic" shape changes as 
well. Schwenk (2000b) suggested that the 
foretongue in lizards is somewhat inde- 
pendent of the hind tongue, with the for- 

Tongue Muscles • Schwenk 231 

mer more specialized for hydrostatic elon- 
gation independent of the hyobranchium 
and the latter more tightly coupled to the 
hyobranchium. If so, it is possible (but 
speculative) that the traditional notion of 
whole-tongue retraction attributed to the 
hyoglossus resides in its extrinsic compo- 
nent and the anterior, intrinsic component 
functions predominantly to retract the 
foretongue after it is hydrostatically elon- 
gated (as during tongue-flicking behavior 
or lapping). This interpretation implies a 
functional heterogeneity within the hyo- 
glossus muscle, which is consistent with its 
histochemical partitioning in some spe- 
cies — a hypothesis that could be tested 

The Tongue as a Functional Unit 

Based on the traditions of musculoskel- 
etal anatomy, our expectation for muscles 
is that they are discrete entities with an 
origin, an insertion, and an action. Al- 
though we accept that individual fibers 
might not run the full course of a muscle 
(e.g., Loeb et al., 1987), the muscle, as a 
whole, is clearly delimited by its origin and 
insertion, coherence, and separation from 
adjacent muscles. Furthermore, the clear 
relationship between these anatomical at- 
tributes and a muscle's action(s) reinforces 
our sense that muscles are discrete inor- 
phologic units whose homologous counter- 
parts can be identified in other species. As 
such, muscles satisfy the requirements of 
a "character" in the traditional neo- 
Darwinian sense: they are "quasi-indepen- 
dent" (Lewontin, 1984) and "quasi-auton- 
omous" (Wagner, 1999) parts of the phe- 
notype. In other words, they are capable 
of evolutionary change somewhat indepen- 
dent of change in other characters and are 
developmentally autonomous units indi- 
viduated from other such units. 

There is no denying that several lingual 
"muscles" have "typical" origins outside 
the tongue, but once within the tongue 
they deviate from the traditional concep- 
tion of a muscle and seem to be subject to 
a different set of rules. These rules are dic- 

tated by the muscular hydrostat model of 
tongue movement. Fibers are arrayed 
within the tongue to generate global or lo- 
cal reductions in diameter, reductions in 
length, lateral bending, dorsoventral bend- 
ing, torsion, and a nearly infinite variety of 
shape changes — all by virtue of the incom- 
pressibility of intracellular fluid and the 
principle of compensatoiy deformation in 
a constant volume stmcture — regardless of 
the extrinsic or intrinsic origin of the fi- 

In light of muscular hydrostatic func- 
tion, the atomization of the tongue into in- 
dividuated muscles is insupportable. In- 
stead, the tongue is better viewed holisti- 
cally as a single functional unit (Schwenk, 
2001). This does not imply that the tongue 
is a single evolutionary character, but rath- 
er emphasizes that the components of the 
tongue are more tightly integrated func- 
tionally with each other than they are with 
other parts of the oral apparatus. As such, 
proper tongue function depends on the 
closely coordinated action of its many parts 
acting in concert to achieve a given func- 
tional output. An action such as tongue 
elongation emerges as the instantaneous 
manifestation of countless vertical and 
transverse muscle fiber contractions, re- 
gardless of their point of origin or nominal 
affiliation. Indeed, the highly organized in- 
ternal architecture of the tongue is literally 
the incarnation of this extreme functional 
coherence, a coherence that belies its dis- 
parate anatomical components. In a mam- 
mal, for example, it is important that 
sheets of vertical fibers are arranged along 
the length of the tongue; it is not impor- 
tant whether the fibers within each sheet 
derive extrinsically from the genioglossus, 
or intrinsically from the lamina propria or 
the longitudinalis inferior. Similarly, Bam- 
well et al. (1978a) noted that, despite ex- 
trinsic and intrinsic contributions, the dor- 
sal longitudinal muscle layer functions as 
"a whole." Thus, rather than "muscles," 
within the tongue it is more accurate (and 
I believe preferable) to refer to fiber sys- 
tems. Vertical and transverse fiber systems 

232 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

contribute to reduction in tongue diame- 
ter, regardless of which "muscles" provide 
them with fibers. 

A holistic view of tongue form and func- 
tion is furtlier supported by tongue devel- 
opment in mammals. From the earliest 
stages, orientations of all muscle fibers 
within the tongue are evident (Smith, 
1994). In Monodelphis, Smith (1994: 158) 
observed, "adjacent cells simultaneously 
orient into one of the three mutually per- 
pendicular planes, so that the three-di- 
mensional arrangement of muscle fibers in 
the tongue is present at the earliest stages 
observed." As such, sheets of vertical and 
transverse fibers begin to organize within 
the tongue in situ without regard to the 
subsequent connection or insertion points 
of their constituent fibers. For example, a 
given vertical fiber, ultimately connects to 
the lamina propria within the tongue, 
whereas an adjacent cell connects extrin- 
sically within the genioglossus. Another 
nearby cell orients transversely and con- 
tributes to the adjacent sheet of transverse 
fibers (K. K. Smith, personal communica- 

I have here focused on the anatomy of 
the tongue and completely neglected the 
issue of motor control. Such a discussion 
is beyond the scope of this paper. How- 
ever, it is worth noting that, although a 
muscle such as the genioglossus is mor- 
phologically continuous from its origin on 
the mandible to its insertion into the lam- 
ina propria of the tongues dorsal surface, 
individual fibers may not run its full 
length. As such, some fibers might have an 
intrafascicular origin within the muscle 
mass (Loeb et al., 1987). Thus, the fiber 
population actually penetrating the tongue 
possibly has an intrafascicular origin within 
the genioglossus. Such an arrangement 
would allow segregation of the muscle into 
extrinsic and intrinsic motor units, thereby 
simplifying control. Strain mapping data 
from the human tongue in vivo are consis- 
tent with this hypothesis (Napadow et al., 
1999), although simplifying assumptions of 
the analysis somewhat weaken its support. 

The ability of the tongue to form localized 
deformations and to protnide, for exam- 
ple, in cylindrical, spatulate and cupped 
conformations (personal observation) 
strongly implies separate motor control of 
both vertical and transverse fiber systems, 
as well as regional control of these systems 
along the length of the tongue. As such, 
functional integrity of lingual fiber systems 
need not imply coarseness of control. Fi- 
nally, finite element analyses that model 
the tongue as a series of regional elements 
with certain contractile and viscoelastic 
properties, regardless of fiber source, are 
effective in predicting patterns of tongue 
shape change (Wilhelms-Tricarico, 1995; 
Sanguineti et al., 1997), further supporting 
a holistic conception of tongue function- 


The results of this study refute the di- 
chotomous view of the lingual muscula- 
ture, both anatomically and functionally. 
The distinction of "intrinsic" from "extrin- 
sic" muscles may serve as a convenience 
for description or instruction, as suggested 
by Abd-El-Malek (1938), but our current 
understanding of tongue function suggests 
that the convention is no longer tenable. 
It acts only to reinforce an outmoded, at- 
omistic view of the tongue that deflects us 
from a more complete conception of this 
remarkable organ. The muscular hydrostat 
model dictates that the tongue acts as a 
functional unit and as such, internal 
tongue form manifests functional unity 
rather than the separate contributions of 
individuated muscles. Thus, within the 
tongue there is a blurring of distinctions 
among muscles, even those arising from 
outside the tongue, so that muscle individ- 
uality is lost. In this context, the notion of 
a lingual "muscle" is not meaningful and 
reference to "fiber systems" more accu- 
rately, if still inadequately, represents the 
tongue's inner workings. Therefore, a ho- 
listic view of tongue form and function 
highlights coherence over separability and 

Tongue Muscles • Schwenk 233 

the precedence of functional integrity over 
anatomical atomization. 


I am deeply indebted to Fuzz Cromp- 
ton, to whom this contribution is dedicat- 
ed, for his friendship, mentorship, and 
support. My two years as a postdoctoral 
fellow in his lab were among the most en- 
joyable and instructive of my career. I 
thank Parish A. Jenkins, Jr., for organizing 
the symposium in honor of Fuzz and for 
inviting me to participate. I am grateful for 
the support and guidance of Karen Hiie- 
mae, who first introduced me to mammal 
tongues. Kathleen Smith shared her exten- 
sive knowledge of tongue stioicture, func- 
tion, and development. Elizabeth Jock- 
usch, Kathleen Smith, and an anonymous 
reviewer critically read the manuscript. 
Various phases of this work were support- 
ed by Marvalee Wake, University of Cali- 
fornia, Berkeley; the Graduate College 
and Department of Oral Anatomy, Uni- 
versity of Illinois at Chicago; A. W. Cromp- 
ton, the Department of Organismic and 
Evolutionary Biology, and the Milton 
Fund, Harvard University; the University 
of Connecticut Research Foundation; and 
grants NIH F32 DE05467 and NSF IBN- 
9601173 to the author. 


Abd-El-MalEK, S. 1938. A contribution to the study 
of the movements of the tongue in animals, with 
special reference to the cat. Journal of Anatomy, 
73: 1.5-30. 

Barnwell, Y. M. 1976. Human lingual musculature: 
an historical review. International Journal of Oral 
Myology, 2: 31^1. 

Barnvvell, Y. M., K. Klueber, and H. L. Lang- 
don. 1978a. The anatomy of the intrinsic mus- 
culature of the tongue in the early human fetus: 
Part 1, M. longitudinalis superior. International 
Journal of Oral Myology, 4(3): 5-8. 

Barnwell, Y. M., H. L. Langdon, and K. Klue- 
ber. 1978b. The anatomy of the intrinsic mus- 
culature of the tongue in the early human fetus: 
Part II. M. longitudinahs inferior. International 
Journal of Oral Myology, 4(4) .5-8. 

Bennett, G. A. 1935. Die anatomische Grundlage 
der halbseitigen Zungenlahmungen. Sitzungsber- 

icht der Gesellschaft fur Morphologie und Phy- 
siologie (Miinchen), 44: 1-6. 

Bennett, G. A., and R. C. Hutchinson. 1946. Ex- 
perimental studies on the movements of the 
mammalian tongue. II. The protrusion mecha- 
nism of the tongue (dog). Anatomical Record, 
94: 57-83. 

Bowman, J. P. 1968. Muscle spindles in the intrinsic 
and extrinsic muscles of the rhesus monkey's 
(Macaco mulatta) tongue. Anatomical Record, 
161: 483-488. 

Cax'E, a. J. E. 1980. The rhinoceros Ungual intrinsic 
musculature. Mammalia, 44: 12.3-128. 

Cooper, S. 1953. Muscle spindles in the intrinsic 
muscles of the human tongue. Journal of Physi- 
ology, 122: 19.3-202. 

Crouch, J. E. 1978. Functional Human Anatomy, 
third edition. Philadelphia: Lea & Febiger. xxi -I- 
663 pp. 

Delheusy, v., G. Toubeau, and V. Bels. 1994. 
Tongue structure and function in Opiums cuvieri 
(Reptilia: Iguanidae). Anatomical Record, 238: 

Doran, G. a. 1973. The lingual musculature of the 
echidna, Tacfu/glossus aculeatus. Anatomischer 
Anzeiger, 133:' 468-476. 

Doran, G. A., and H. Raggett. 1970. The vascular 
stiffening mechanism in the tongue of die echid- 
na (Tachyglossus aculeatus). Anatomical Record, 
167: 197-204. 

. 1972. The genioglossus muscle: a reassess- 
ment of its anatomy in some mammals, including 
man. Acta Anatom'ica, 83: 403^10. 

Edgeworth, F. H. 1935. The Cranial Muscles of 
Vertebrates. Cambridge, United Kingdom: Cam- 
bridge University' Press, viii + 493 pp. 

GnanAMUTHU, C. p. 1937. Comparative study of the 
hyoid and tongue of some typical genera of rep- 
tiles. Proceedings of the Zoological Society of 
London, Series B, 1937: 1-63 

Greenbaum, I. F, and C. J. Phillips. 1974. Com- 
parative anatomy and general histology of 
tongues of long-nosed bats (Leptonycteris san- 
bomi and L. nivalis) with reference to infestation 
of oral mites. Journal of Mammalogy, 55: 489- 

Griffiths, M. 1968. Echidnas. O.xford: Pergamon. l\ 
+ 282 pp. 

. 1978. The Biology of the Monotremes. New 

York: Academic Press, viii -I- 367 pp. 

Griffiths, T. A. 1978. Muscular and vascular adap- 
tations for nectar-feeding in the glossophagine 
bats Monophijllus and Glossophaga. Journal of 
Mammalogy, 59: 414^18. 
HellsTRAND, E. 1980. Morphological and histo- 
chemical properties of tongue muscles in cat. 
Acta Phvsiologica Scandinavica, 110: 187-198. 

. 1981. Contraction times of the cat's tongue 

muscles measured bv light reflection. Innerva- 
tion of individual tongue muscles. Acta Physiol- 
ogica Scandinavica, 111: 417^23. 

234 BtiUetin Museum of Comparative Zoology, Vol. 156, No. 1 

Herrel, a., p. Aerts, J. Fret, and F. De Vree. 
1999. Moi-phology of the feeding system in 
agamic! lizards: ecological correlates. Anatomical 
Record, 254: 496-507. 

Herrel, A., J. Cleuren, and F. De Vree. 1995. 
Prey capture in the lizard Agama stellio. Journal 
of Morphology, 224: 313-329. 

. 1997. Quantitative analysis of jaw and hy- 

olingual muscle activity during feeding in the liz- 
ard Agama stellio. Journal of Experimental Bi- 
ology, 200: 101-115. 

HiATT, j' L., AND L. P. Gartner. 1982. Textbook of 
Head and Neck Anatomy. New York: Appleton- 
Century-Crofts. xiv -I- 350 pp. 

KalLIUS, E. 1910. Beitrage zur Entwickelung der 
Zunge. in. Teil. Siiugetiere. 1. Sus scrofa dom. 
Anatomische Hefte. Arbeiten aus Anatomischen 
Instituten. 123/124: 176-337. 

Kier, W. M., and K. K. Smith. 1985. Tongues, ten- 
tacles and trunks: the biomechanics of movement 
in muscular- hydrostats. Zoological Journal of the 
Linnean Society, 83: 307-324. 

Langdon, H., K. Klueber, and Y. Barn\\^ll. 
1978. The anatomy of m. genioglossus in the 15- 
week human fetus. Anatomy and Embryology, 
155: 107-113. 

Lewontin, R. 1984. Adaptation, pp. 234-251. In E. 
Sober (ed.). Conceptual Issues in Evolutionary 
Biology. Cambridge, Massachusetts: MIT Press, 
xiv -I- 725 pp. 

Livingston, R. M. 1956. Some observations on the 
natural history of the tongue. Annals of tlie Royal 
College of Surgeons, England, 19: 185-200. 

LOEB, G.E., C. A. Pratt, C. M. Chanaud, and F 
J. R. Richmond. 1987. Distribution and inner- 
vation of short, interdigitated muscle fibers in 
parallel-fibered muscles of the cat hindlimb. 
Journal of Morphology, 191: 1-15. 

Lowe, A. A. 1980. The neural regulation of tongue 
movements. Progress in Neurobiology, 15: 295- 

McGregor, G. 1938. Comparative anatomy of the 
tongue. Annals of Otology, Rhinology and Lar- 
yngology, 47: 196-211. 

MiYAWAKI, K. 1974. A study on the musculature of 
the human tongvie. Annual Bulletin (Research 
Institute of Logopedics and Phoniatrics, Univer- 
sity of Tokyo), 8: 23-50. 

Napadow, V. J^, Q. Chen, V. J. Wedeen, and R. J. 
Gilbert. 1999. Intramural mechanics of die hu- 
man tongue in association with physiological de- 
formations. Journal of Biomechanics, 32: 1-12. 

Oelrich, T M. 1956. The anatomy of the head of 
Ctenosaura pectinata (Iguanidae). Miscellaneous 
Publications of the Museum of Zoology, Univer- 
sity of Michigan, 94: 1-122. 

Owen, R. 1868. On the Anatomy of Vertebrates, vol. 
3. Mammals. London: Longmans, Green, and 
Co. X + 915 pp. 

Perkell, J. S. 1969. Physiology of Speech Produc- 
tion: Results and Implications of a Quantitative 

Cineradiographic Study. Cambridge, Massachu- 
setts: MIT Press. 

mason's Animal Tissue Techniques, fifth editon. 
Baltimore: The Johns Hopkins University Press. 
.\i -I- 572 pp. 

1997. A control model of human tongue move- 
ments in speech. Biological Cybernetics, 77: 11- 

SCHWENK, K. 1986. Moi-phology of the tongue in the 
tuatara, SpJienodon punctatus (Reptilia: Lepido- 
sauria), with comments on function and phylog- 
eny. Journal of Morphology, 188: 129-156. 

. 1988. Comparative morphology of the lepi- 

dosaur tongue and its relevance to squamate 
phylogeny, pp. 569-598. In R. Estes and G. Pre- 
gill (eds.), Phylogenetic Relationships of the Liz- 
ard Families. Stanford, California: Stanford Uni- 
versity Press, xiv + 631 pp. 

. 2000a. An introduction to tetrapod feeding, 

pp. 21-61. 7/1 K. Schwenk (ed.). Feeding in Tet- 
rapods: Form, Function, Phylogeny. San Diego, 
California: Academic Press, x-v -I- 537 pp. 

. 2000b. Feeding in lepidosaurs, pp. 175-291. 

In K. Schwenk (ed.). Feeding in Tetrapods: 
Form, Function, Phylogeny. San Diego, Califor- 
nia: Academic Press, x-v + 537 pp. 

2001. Functional units and their evolution. 

pp. 165-198. In G. P Wagner (ed.). The Char- 
acter Concept in Evolutionaiy Biology. San Di- 
ego, California: Academic Press, xxiii -I- 622 pp. 

Smith, K. K. 1984. The use of the tongue and hyoid 
apparatus during feeding in lizards (Ctenosaura 
similis and Tupinambis nigropunctatus). Journal 
of Zoology, London, 202:1 15-143. 

. 1986. Morphology and function of the 

tongue and hyoid apparatus in Varanus (Varani- 
dae, Lacertilia). Journal of Morphology, 187: 

. 1988. Form and function of the tongue in 

agamid lizards with comments on phylogenetic 
significance. Journal of Morphology, 196: 157- 

. 1992. The evolution of the mammaHan phar- 
ynx. Zoological Journal of the Linnean Society, 
104: 313-349. 

1994. Development of craniofacial nuiscu- 

lature in Monodelphis domestica (Marsupialia, 
Didelphidae). Journal of Morpholog)', 222: 149- 

Smith, K. K., and W. M. Kier. 1989. Trrmks, 
tongues, and tentacles: moving with skeletons of 
muscle. American Scientist, 77: 29-35. 

SOKOLOFF, A. J., AND T W. DEACON. 1992. Muscu- 
lotopic organization of the hypoglossal nucleus in 
the cvnomolgus monkey, Macaca fascicularis. 
Journal of Comparative Neurology, 324: 81-93. 

SONDHI, K. C. 1958. The hvoid and associated stnic- 

Tongue Muscles • Schwenk 235 

tures in some Indian reptiles. Annals of Zoology, 
2: 155-239. 

SONNTAG, C. F. 1925. The comparative anatomy of 
the tongues of Mammalia. XII. Summary, clas- 
sification and physiolog)-. Proceedings of the Zoo- 
logical Society of London, 21: 701-762. 

Tanner, W. W., and D. F. Avery. 1982. Buccal floor 
of reptiles, a summary. Great Basin Naturalist, 
42: 273-349. 

Wagner, G. P. 1999. A research programme for test- 
ing the biological homology concept, pp. 125- 
140. In G. R. Bock and G. Cardew (eds.). Ho- 
mology (Novartis Foundation Symposium No. 

222). Chichester, United Kingdom: John Wiley 
& Sons. \dii + 256 pp. 

Walker, W. F., Jr., and D. G. Homberger. 1992. 
Vertebrate Dissection, eighth edition. Fort 
Wortli, Texas: Saunders College PubHshing. xii + 
459 pp. 

War\\ick, R., and p. L. Williams. 1973. Grays 
Anatomy, 35th edition. Edinburgh: Longman, xvi 
+ 1471 pp. 

WilhelmS-TricaricO, R. 1995. Physiological mod- 
eling of speech production: methods for model- 
ing soft-tissue articulators. Journal of the Acous- 
tical Society of America, 97: 3085-3098. 



ABSTRACT. Gular pumping in monitor lizards is wicz, 1996). In a gular pump cycle, a mon- 

knovvn to play an important role in lung ventilation, ^^^^ i-^^j.^ flj-st draws fresh air into its large 

but its evolutionary origin has not yet been addressed. , i / i \ m, j 4-1 ^ 

To determine whether Ae g^ilar pump derives from phaiyngea (gukr) cavity and then COn- 

die buccal pump of basal tetrapods or is a novel in- tracts its throat muscles to create positive 

vention, we investigated the electromyographic activ- pressure, thus forcing air into the lungS. 

it\- associated with gular pumping in savannah mon- Multiple gular pumps may OCCUr ill SUC- 

itor lizards (Varanus exanthenmticus). Electrodes ^^^^;^^^ between COStal breaths. Gular 

were implanted in hyobranchial muscles, and their .11 1 ^  n l.^ 

activity' patterns were recorded synchronously with pumping has been shown to Slgmhcantly 

hyoid kinematics, respiratory airflow, and gular pres- increase both minute ventilation and aei- 

sure. Movement of the highly mobile hyoid apparatus obic capacity of savannah monitor lizards 

effects large-volume airflows in and out of the gular (Varauiis exanthematicus) during exercise 

cavity. The stemohyoideus and branchiohyoideus de- ,„ 1 . , IQQQ') 

press, retract, and abduct die hyoid, thus expanding ('^werKOWlCZ et ai. i^yy; 

the gular cavit>'. The omohyoideus, constrictor colli. Little IS known about the actual mecli- 

intermandibularis, and mandibulohyoideus elevate, anism of gular pumping and its evolution. 

protract, and adduct the hyoid, thus compressing the \ highly expandable pharynx, supported 

gular cavity. Closure of the choanae by the sublingual , ^^ elaborate hyobranchial basket, is a 

plicae precedes gular compression, allowang positive / . . r r •. 1- j 

pressure to be generated in the gular cavity to force characteristic feature of monitor lizards. 

air into the lungs. The large size and compressibility of the 

The gular pump of monitor lizards is found to ex- pharynx make it ideally suited for puinping 

hibit a neuromotor pattern similar to the buccal j^ volumes of air. Smith (1986) studied 

pump of extant amphibians, and both mechanisms m r .  r . 1 . 1 4- ^^ ; ^ 17 

f ,^ , 1 ^ 1 Ti  „„ f fi. .f ^u^ the function oi throat musculature in V. 

involve homologous muscles. This suggests that the j r j j 

gular pump may have been retained from the ances- exanthenuiticilS during feeding, and re- 

tral buccal pump. This hypothesis remains to be test- corded hyobranchial muscle activity and 

ed by a broad comparative analysis of gular pumping hyoid movements. Bels et al. (1995: 99) 

among the amniotes. provided a kinematic analysis of the throat 

threat display in Varanus griseus, "a ven- 

INTRODUCTION tilatory bucco-pharyngeal breathing pump 

Monitor Hzards (genus Varanus) have (VBPBP)," consisting of alternating expan- 

recently been found to supplement their sions and compressions of the gular cavity, 

lung ventilation with gular pumping dur- These studies have documented the ex- 

ing locomotion (Brainerd and Owerko- treme excursions of the hyoid apparatus 

during a wide repertoire of monitor be- 

,. „ .■ ^ 1 u 1 IT haviors. However, the electromyographic 

'Museum ot Comparative Zoology, Harvard Uni- ,-^^ ,^s  r ^ 1 y o r 

versity, 26 Oxford Street, Cambridge, Massachusetts (EMG) Signature of the gular pump, as a 

02138. ventilatory mechanism, has not been re- 

' Department of Biology and Organismic and Evo- ported, 

lutionaiy Biology Program, University of Massachu- ^j^.^" ^^^ 1 ^-^^ ^^ determine how throat 

setts, Amherst, Massachusetts 01003. , ,.-^... i- . i 1 • 1 

^ Department of Biology, University of Utali, Salt muscle activity IS coordinated during gular 

Lake City, Utali 84112. pumping ill V. exantheniaticus. Which 

Bull. Mus. Comp. Zool., 156(1): 237-248, October, 2001 237 

238 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

muscles are responsible for gular expan- 
sion, and which ones are responsible for 
compression? How is pressure generated 
in the gular cavity? What controls the air- 
flow pattern? 

The evolutionary origin of the gular 
pump also presents an interesting dilem- 
ma. Squamates have been presumed to 
ventilate their lungs solely by means of 
costal aspiration (Gans, 1970; Carrier, 
1987; but see Deban et al., 1994); they rely 
on contraction of intercostal muscles to 
create negative pressure in the pleural cav- 
ity to suck air into the lungs. The use of a 
pressure pump was thought to be reserved 
for air-breathing fish and amphibians 
(Liem, 1985; Brainerd et al, 1993). In this 
mechanism, air is gulped into the buccal 
(mouth) cavity, and subsequently the hyoid 
apparatus generates positive pressure, 
forcing air into the lungs. The gular pump 
of inonitor lizards clearly qualifies as a 
pressure pump and therefore breaks with 
this traditional phylogenetic separation of 
breathing mechanisms. However, the gular 
pump differs from the buccal pump be- 
cause the hyoid apparatus of monitor liz- 
ards is positioned posteriorly in the throat, 
whereas the hyoid apparatus of amphibi- 
ans resides in the mouth cavity between 
mandibular rami. The presence of gular 
pumping behavior has not been rigorously 
investigated in other lizard genera, and it 
is unclear whether gular pumping is a 
uniquely derived trait of monitor lizards or 
whether its ancestry can be traced back to 
buccal pumping of basal tetrapods. 

Therefore, this study attempts to ad- 
dress the origin of gular pumping from a 
functional perspective. How similar is the 
EMG pattern of the gular pump to that of 
the amphibian buccal pump? Is the gular 
pump a case of neuromotor conservatism 
in evolution, or has it evolved de novo in 
monitor lizards? 



Experiments were performed on four 
savannah monitor lizards (230-2,400 g) 

during and immediately after locomotion 
on a motorized treadmill at speeds of 1—5 
km/h. The animals were maintained at 25— 
40°C on a 14:10 hour light: dark photo- 
period and were fed a diet of mice. 


This study follows the terminology of 
Smith (1986) in her description of the os- 
teology and myology of the varanid gular 


Videos of the lizards were taken with a 
Sony DCR VXIOOO digital camcorder (60 
fields/s at 1/250 s shutter speed) and the 
Siemens X-ray fluoroscope at the Museum 
of Comparative Zoology Laboratories at 
Harvard University. Video recordings were 
made separately in lateral and dorsoventral 
projections. Select video fields were im- 
ported into Adobe Photoshop on a Power 
Macintosh computer. 

To better visualize inovements of the 
floor of the mouth relative to the skull and 
hyoid, lead markers (1.6 X 0.5 mm) were 
placed unilaterally in the left sublingual 
plica and in the anterior epithelial border 
of the left choana. Marker implantation 
was performed percutaneously with a 20- 
gauge needle and plunger while the ani- 
mals were under 1—2% halothane anesthe- 


Two animals were used to measure air- 
flow during gular pumping. A lightweight 
mask, fashioned from clear acetate and ep- 
oxy, was custom-fitted and taped over the 
lizard's snout to enclose the mouth and 
nostrils. A bias flow of humidified air 
(1,200 ml/min) was drawn through the 
mask. A pneumotachograph (8421 series 
0-5 LPM, H. Rudolph, Kansas City, Mis- 
souri), connected to a differential pressure 
transducer (MP 45-1-871, Validyne, Noitli- 
ridge, California) downstreain from the 
animal, measured airflow through the 
mask. The system was calibrated against 

EMG Pattern of Gular Pump • Owerkowicz et al. 239 

repeated injections of measured aliquots Pennsylvania). The signals were amplified 

of air into the mask. 5,000-20,000 times (as appropriate for 

each channel) with Grass P511J amplifiers 

Pressure Recordings (Quincy, Massachusetts), with a bandpass 

Two animals (not used for airflow re- of 100-1,000 Hz (with the 60-Hz notch 
cordings) were instrumented to measure filter in). The signals were acquired at 
gular pressure during gular pumping. With 5,000 Hz and analyzed on a Power Mac- 
the animal under 1-2% halothane gas an- intosh computer using AcqKnowledge 
esthesia, a skin incision was made between (BioPac Systems, Santa Barbara, Califor- 
the ceratohyal and ceratobranchial. The nia) and Igor Pro (WaveMetrics, Inc., 
underlying muscles were carefully blunt- Lake Oswego, Oregon) software. Electro- 
dissected and a 13-gauge needle was in- myographic and pressure signals were 
serted through the pharyngeal epithelium temporally synchronized with video re- 
into the gular cavity. A 20- to 30-cm-long cordings by means of a light-emitting di- 
polyethylene cannula (1.14-mm inner di- ode trigger (Thexton Unlimited, London, 
ameter, 1.57-mm outer diameter), previ- United Kingdom), 
ously sterilized and heat-flared at the distal 
end, was threaded through the hole and RESULTS 
securely sutured to the lateral wall of the Kinematics 
gular cavity. Before each recording session, 

the cannula was cleared of mucus to pre- Four stages of hyobranchial movement 

vent capillarity artefacts and connected to characterize each gular pumping cycle 

a differential pressure transducer (PX138- (Fig. 1): resting, active expansion, early 

0.3D5V, Omega Engineering, Inc., Stam- closure, and compression. The cycle usu- 

ford, Connecticut). The pressure transduc- ally begins with the expansion of the hyob- 

er was calibrated against 10 cm H2O after ranchial basket. From its resting position 

each recording session. (Fig. lA), the basihyoid and the lingual 

process are drawn slightly ventrad and 

Electromyography caudad. The anterior processes are ab- 

Electromyographic activity was record- ducted and, pivoting on the basihyoid, 

ed in the following muscles: constrictor swung laterad. The ceratohyals and cera- 

colli, intermandibularis, mandibulohyoid- tobranchials rotate to assume a more ver- 

eus, omohyoideus, sternohyoideus, and tical orientation (Fig. IB), 

branchiohyoideus. Patch and sew-through The compressive phase of the pump cy- 

bipolar electrodes, as detailed by Loeb and cle begins with the elevation of the buccal 

Cans (1986) and Carrier (1996), respec- floor to abut the palate, thereby closing off 

tively, were constructed of 0.28-mm-di- the mouth and nostrils. As highlighted by 

ameter Teflon-coated stainless steel wire radiopaque markers, each choana is oc- 

(Cooner Wire Co., Chatsworth, California) eluded by its ipsflateral sublingual pfica. 

and silastic-reinforced sheeting (Specialty The glottis is located in the gular cavity. 

Manufacturing, Inc., Saginaw, Michigan), while the hyoid remains in a retracted and 

In each electrode, the exposed portions of expanded configuration (Fig. IC). Finally, 

wires were 2 mm long and 1 mm apart, the basihyoid is protracted and elevated. 

With the animal anesthetized, electrodes while the trachea slides rostrad in the gu- 

were surgically implanted and firmly an- lar cavity. The anterior processes are ad- 

chored in the desired muscles. The elec- ducted, and the ceratohyals and cerato- 

trode wires were then passed subcutane- branchials fold horizontally (Fig. ID). The 

ously under the pectoral girdle to a single hyobranchium then relaxes and passively 

exit on the tiTink's dorsum and soldered to returns to its resting state (Fig. lA); an- 

connectors (Microtech, Inc., Boothwyn, other pumping cycle may follow. 

240 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 1. Still frames from an X-ray positive video (shutter speed 1/250 s) showing one gular pump cycle in Varanus exanthe- 
maticus (body mass = 600 g). The left column shows the animal in lateral projection, the right column in dorsoventral projection. 
Although not recorded simultaneously, frames in each row portray corresponding stages of a pump cycle. (A) Gular cavity relaxed. 
(B) Gular cavity expanded. (C) Buccal floor elevated. (D) Gular cavity compressed. Radiopaque markers (arrows) are placed at 
the choana (Ch), the sublingual plica (SP), and the glottis (G). Note the closure of the mouth and nares, which seals the gular 
cavity, as demonstrated by proximity of markers Ch and SP in (C). 

EMG Pattern of Gular Pump • Owerkowicz et al. 241 

IM post 



Time (s) 


Time (s) 

Figure 2. Raw electromyograms of hyobranchial muscles during gular pumping in Varanus exanthematicus. (A) Three gular pump 
cycles recorded simultaneously with airflow at the snout. Gular expansion (negative airflow) is initially passive because of the gravitational 
drop of the hyoid, and then active as the branchiohyoideus and stemohyoideus begin to contract. Gular compression occurs during the 
zero-flow plateau, as air is pressed into the lungs. Minimal air leakage occurs — note the spikes showing early closure of the nares (N) 
and the late closure of the glottis (G). The asterisk indicates unanticipated activity of the omohyoideus due to neck bending. (B) A costal 
breath followed by four successive gular pump cycles, recorded simultaneously with gular pressure. Gular expansion generates little 
negative pressure, but peak compressive pressures up to 15 cm HJD have been recorded. Abbreviations: CC, m. constrictor colli; IM, 
m. intermandibularis; MH, m. mandibulohyoideus; BH, m. branchiohyoideus (ceratohyoid head); BH ap, m. branchiohyoideus (comuhyoid 
head); SH, m. stemohyoideus; OH, m. omohyoideus; ant, anterior; post, posterior. 

Active expansion was not observed in 
every pump cycle. The aniinals often 
chose to proceed to closure and compres- 
sion directly from the resting stage, in 
which case the volume of air pumped was 
not as great as when preceded by active 
expansion. Furthermore, resting gular vol- 
ume was dependent on the positioning of 
the hyobranchial apparatus in the neck of 
the animal, which varied with the animal's 
posture and muscle tone of associated pec- 
toral musculature (stemohyoideus and 
oinohyoideus — see below). 


The airflow pattern at the aniinal's 
mouth and nostrils (Fig. 2A) is consistent 
with a biphasic nature of each gular pump, 
whereby a filling (expansive) phase is im- 

mediately followed by an emptying (com- 
pressive) phase. Average filling volumes 
ranged froin 6 to 15 inl, but individual gu- 
lar inspirations up to 33 ml were also re- 
corded. Early in the pump cycle, animals 
were observed to rely on passive filling of 
the gular cavity by gravitational drop of the 
hyoid (correspondent to the return of the 
hyoid from compressed to resting stage). 
Passive filling was followed by active ex- 
pansion, as indicated by concomitant activ- 
ity in the branchiohyoideus and stemohy- 
oideus, which resulted in faster (more neg- 
ative) airflow and greater filling volumes. 

Early in the filling phase, airflow into 
the gular cavity was rapid, as shown by the 
steep descent of the airflow trace to its 
minimum value. Late in the filling phase, 
inspiratory airflow gradually diininished to 

242 Bulletin Museum of Comparative Zoologij, Vol. 156, No. 1 

nil. A slight overshoot of the zero-flow line milliseconds by almost simultaneous acti- 
(N in Fig. 2A) marked early closure of the vation of the stemohyoideus and mandi- 
mouth and nostrils, and the beginning of bulohyoideus III. The intensity of their fir- 
the emptying phase. A zero-flow plateau ing remained roughly constant while these 
was recorded for the duration of the einp- muscles were active. All three muscles were 
tying phase, as air was forced from the gu- turned off before the end of the filling 
lar cavity into the lungs. The emptying phase, with activity in die branchiohyoideus 
phase terminated in a small expiratory persisting for up to 150 milliseconds after 
spike (0.5—1.0 ml), after the inferred clo- the offset of the stemohyoideus and man- 
sure of the glottis (G in Fig. 2A) and open- dibulohyoideus III. 

ing of the mouth and nares. Unless anoth- Gular compression began with a burst 

er pumping cycle occurred immediately of activity in the intermandibularis anteri- 

afterwards, passive filling followed and a or, which was followed within 100 milli- 

longer apnoeic period ensued. seconds by contraction of the constrictor 

colli, intermandibularis posterior, and all 

rressure three heads of mandibulohyoideus. The 

Each costal breath was followed by as intensity of their firing generally increased 

many as five gular pumps, with successive from onset to offset. Activity ceased in all 

pump cycles generating increasingly posi- muscles together, once gular pressure had 

tive gular pressures (Fig. 2B). Peak pres- peaked. 

sures up to 15 cm H2O were recorded. The activity of the mandibulohyoideus 
With the onset of each compressive phase. III was biphasic; it contracted during gular 
gular pressure cliiTibed steeply and expansion and compression, but not con- 
reached peak pressure within 200 millisec- tinuously (Fig. 2B). It was briefly (50—100 
onds. For most pumping cycles, gular milliseconds) silent between the two phas- 
pressure hovered within 5% of peak value es of the pumping cycle, 
for less than 100 milliseconds, although The omohyoideus was usually quiescent 
gular pressure would soinetimes remain during gular pumping. When participating 
elevated at peak for up to 400 millisec- in gular pumping, the omohyoideus was 
onds. Thereafter, gular pressure dropped predominantly active during gular com- 
rapidly (within 100 milliseconds), often pression. However, it might occasionally 
dipping to subatmospheric levels, and then fire during gular expansion (* in Fig. 2A), 
equilibrated with the atmospheric pres- which tended to decrease the gular filling 
sure. volume. In such instances, activity of the 

Gular pressure varied little from atmo- omohyoideus was correlated with changes 

spheric pressure during the filling phase of in posture and/or neck bending by the an- 

a pump cycle. Active gular expansion gen- imal. 

erated only slightly subatmospheric pres- During locomotion, regular phasic activ- 

sures (—0.2 to —0.4 cm H^O). ity was observed in most hyobranchial 

muscles, but was especially prominent in 

Electromyographic Activity the stemohyoideus and omohyoideus. The 

Recordings from hyobranchial muscles rate of discharge was the same as the foot- 
produced a consistent activity pattern dur- fall frequency, and in the latter two mus- 
ing gular pumping (Fig. 2A, B). Although cles the signal amplitude was greater than 
they remained quiescent during passive fill- in stationary animals, 
ing, branchiohyoideus (both cerato- and niQn iQQinM 
cornuhyoid heads), stemohyoideus, and 

mandibulohyoideus III were tumed on Hyobranchial Muscle Function 

during active gular expansion. Onset of the The role of individual muscles involved 

branchiohyoideus was followed in 50—200 in the gular pump can be inferred by con- 

EMG Pattern of Gular Pump • Owerkowicz et al. 243 



MH III -♦ 

MHII -♦ 

MH I ^ 

OH scap 

OH scap 

■► OH clav 

Figure 3. The hyobranchial apparatus of Varanus exanthematicus, in ventral (top row) and lateral (bottom row) views. Force 
vectors (gray arrows) show the sites of insertion and lines of action of the hyobranchial muscles, as measured in dissected 
specimens. (A) The hyoid in its expanded configuration; contraction of musculi (mm.) mandibulohyoideus and omohyoideuswill 
result in hyoid compression. Constrictor colli and intermandibularis are not shown, because they do not insert directly on the 
hyoid. (B) The hyoid in its compressed configuration; contraction of the branchiohyoideus, sternohyoideus, and mandibulohyoi- 
deus III will result in hyoid expansion. Abbreviations: ap, anterior process; bh, basihyoid; cb, ceratobranchial; ch, ceratohyal; 
clav, clavicular; Ip, lingual process; scap, scapular; others as in Figure 2. 

sidering their insertion sites on the hyoid 
apparatus (Fig. 3), the patterns of their 
EMG activity, and the kinematics of the 
hyoid elements during each gular pump 

Gular expansion results from the coor- 
dinated activity of the sternohyoideus, 
branchiohyoideus, and mandibulohyoideus 
III. The sternohyoideus pulls the cerato- 
branchial in a caudoventrad direction, giv- 
ing it a more vertical orientation. This ac- 
tion tends to retract and depress the ba- 
sihyoid, while pointing the lingual process 
ventrad. With the ceratobranchial stabi- 
lized by the sternohyoideus, the bran- 
chiohyoideus tends to abduct the anterior 
process and retract the ceratohyal. How- 
ever, the proximal end of the anterior pro- 

cess pivots about the basihyoid, and the 
ceratohyal has its distal end anchored to 
the lower jaw by the mandibulohyoideus 
in. The broad, loose nature of articulation 
between the ceratohyal and the anterior 
process allows sliding to occur between 
these two elements. Thus, as the anterior 
process swings laterad, it pushes against 
the ceratohyal, the angle between these 
two elements increases, and the ceratohyal 
assumes a more vertical orientation. Alto- 
gether, these muscles tend to expand the 
gular cavity in the lateral and dorsoventral 

Gular compression begins with the early 
closure of mouth and nostrils. The inter- 
mandibularis anterior, positioned under 
the sublingual plicae, contracts to elevate 

244 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 4. Cross section of the snout of Varanus exanthe- 
maticus at tlie anterior border of the choanae. Each nasal pas- 
sage connects to its ipsilateral choana just posterior to this 
plane. Elevation of the mouth floor allows the sublingual plica 
to plug the choana above. The congruent outlines of the lower 
and upper jaws allow an airtight closure of the mouth, sealed 
with saliva from labial and sublingual glands. Abbreviations: 
Ch, choana; LG, labial gland; LJ, lower jaw; N, nasal passage; 
SG, sublingual gland; SP, sublingual plica; UJ, upper jaw. 

the buccal floor and push the pHcae 
against the choanae (SP -^ Ch in Fig. 4), 
thus blocking airflow through the nasal 
passages. Although no recordings were 
taken from the adductor mandibulae, this 
muscle is most likely to be responsible for 
keeping the mouth closed, while the thick, 
fleshy gums (Auffenberg, 1981) seal the 
oral margins with secretion of salivary 
glands lining both lingual and labial as- 
pects of the lower jaw. The gular cavity is 
effectively made airtight for the duration 
of gular compression. 

Muscles responsible for gular compres- 
sion include the omohyoideus, constrictor 
colli, intermandibularis posterior, and 
mandibulohyoideus. The clavicular and 
scapular heads of the omohyoideus attach 
the ceratobranchials to the respective parts 
of the pectoral girdle, and their contrac- 
tion retracts and elevates the basihyoid. 
The constrictor colli encircles the floor and 
sides of the gular cavity, forming a sleeve 
in which the hyobranchial basket is sus- 
pended. Together with its anterior exten- 
sion, the intermandibularis posterior, this 
superficial transverse muscle elevates the 

hyoid apparatus. On the other hand, the 
mandibulohyoideus connects the hyoid to 
the lower jaw and, along with the sterno- 
hyoideus and omohyoideus, controls its 
anteroposterior position in the neck. As 
the basihyoid is protracted, the ceratohyals 
and ceratobranchials assume a more hori- 
zontal orientation, in effect folding the 
hyobranchial basket. The orthogonal ori- 
entation of muscle fibres in the constrictor 
colli and intermandibularis to the mandi- 
bulohyoideus means that their synchro- 
nous activity will squeeze the air out of the 
gular cavity. 

With the onset of locomotion, most 
hyobranchial muscles show bursts of activ- 
ity in phase with the footfall pattern. This 
is particularly pronounced in the sterno- 
hyoideus and omohyoideus; their firing in- 
tensity during locomotion is several times 
greater than at rest. Yet it is hard to imag- 
ine their acting as locomotor muscles. It is 
more likely that with every retraction of 
the forelimb these muscles are stretched 
and fire reflexively to stabilize the hyoid, 
keeping it in position for pumping. Low- 
intensity phasic activity detected in other 
hyobranchial muscles is probably an arte- 
fact of cross-talk from the underlying axial 
muscles of the neck, participating in the 
lateral bending of the neck during loco- 

Airflow and Pressure Changes 

The pneumotachograph and the pres- 
sure transducer provide complementary 
information about the patterns of airflow 
and pressure generation outside and inside 
the gular cavity. During gular expansion, 
pressure drops only slightly below the at- 
mospheric level. The lizard's mouth and 
nares are wide open (Fig. IB) and provide 
little resistance to airflow into the gular 
cavity. Gular expansion, whether passive or 
active, is too slow for gular pressure to 
turn more negative; instead, it quickly 
equilibrates with atmospheric pressure. 

Inspiratory airflow ceases when the 
mouth and nares are shut (see above). The 
"corking" of the choanae by the sublingual 

EMG Pattern of Gular Pump • Owerkowicz et al. 245 

plicae expels an aliquot of air from the nar- rived varanids, do not necessarily possess 
ial passages (N in Fig. 2A). The gular pres- the same ensemble of hyobranchial mus- 
sure increases from this point on and air cles found in Lissamphibia, and even 
is pressed through the open glottis into the among the latter, the muscular organiza- 
trachea and lungs. The plateau at zero- tion of the throat is highly variable. How- 
flow indicates the efficacy of the oral seal, ever, muscle homologies can be estab- 
Only at the very end of compression lished with a fair degree of certainty by 
does a puff of air leak out of the mouth determining their anatomical relations and 
and nares at a high flow rate (G in Fig. motor nerve supply (Fiirb ringer, 1888, in 
2A). The hyobranchial muscles have al- Cunningham, 1890). The constrictor colli 
ready turned off by this time, yet their of lizards and the interhyoideus of am- 
contraction clearly persists for approxi- phibians seem to be homologous, by virtue 
mately 120 milliseconds (a reasonable time of having a common precursor in the con- 
period for slow-twitch fibers in isometric strictor hyoideus, as found among the Dip- 
contraction) and gular pressure remains noi (Edgeworth, 1935). Except for its la- 
elevated. This expiratory "gular leakage" teralmost third head with a disparate mo- 
possibly represents excess air, which was tor innervation and therefore origin (Riep- 
not pressed into the lungs. The fact that pel, 1978), the mandibulohyoideus is 
this occurs at the end of every pumping clearly a highly differentiated version of 
cycle indicates that the glottis always closes the geniohyoideus, ubiquitous among the 
before the nares open. Such carefully co- vertebrates. The rectus cervicis of caeci- 
ordinated timing suggests that this may be lians is homologous with the stemohyoid- 
a hard-wired mechanism designed to pre- eus in both frogs and monitor fizards. 
vent air escaping from the lungs, which Lacking a pectoral girdle, caecilians lack 
have been pressurized by gular pumping. an omohyoideus. 

The branchiohyoideus is found in nei- 

Origin of Gular Pumping j-j-^^^j. anurans nor caecilians. Its putative 

The gular pump of monitor lizards bears homologue, subarcualis rectus I (not the 

striking resemblance to the buccal pump larval branchiohyoideus extemus; Edge- 

of extant amphibians, in function and in worth, 1935, contra Smith, 1920), is pre- 

mechanism. Both pumps generate positive sent in urodeles, but the EMG activity of 

pressure to force air from the phaiyngeal this muscle during buccal pumping is yet 

cavity into the lungs. Both employ the to be investigated. Nevertheless, EMG ev- 

hyobranchial apparatus to produce volu- idence from feeding studies in Ambystonia 

metric changes of the buccal and gular (Lauder and Shaffer, 1985; Reilly and Lau- 

cavdties. This similarity suggests that buc- der, 1991) shows that the subarcualis rec- 

cal pumping and gular pumping may be tus I is active during buccal expansion in 

homologous behaviors. However, the de- aquatic and terrestrial prey capture, and its 

rived moq^hologv' of the monitor lizards role in generating buccal expansion during 

and their deeply nested position in the suction feeding has been deduced from its 

squamate phylogeny (Estes et al., 1988) anatomical position and fiber orientation 

suggest the possibility that gular pumping, in various salamanders (Erdman and Cun- 

as ventilatory behavior, may have evolved dall, 1984; Lauder and Shaffer, 1988; Lo- 

independently in monitor lizards. This al- renz-Elwood and Cundall, 1994). This 

temative hypothesis would be supported suggests that the subarcuaHs rectus I func- 

by finding that the gular and buccal pumps tions in much the same way in the buccal 

are powered by nonhomologous muscles, pump of salamanders as does the bran- 

or that the patterns of their activation are chiohyoideus in the gular pump of lizards, 

markedly different. Having established the homology of 

Lepldosaurs, especially the highly de- hyobranchial muscles of amphibians and 

246 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 



E C 



(not measured) 



E C 







Figure 5. Comparison of hyobranchial muscle activity patterns during a single pressure pump cycle in (A) a caecilian {Der- 
mophis), (B) a frog (Rana), and (C) a monitor lizard (Varanus). Homologous muscles are shown in the same row. Each pumping 
cycle (E + C) has been scaled to the same duration time (the scale bars are 0.2 seconds). Pressures (bottom trace) have been 
scaled to the same peak value; maximum peak pressures range from 4 cm HjO (frog) to 15 cm Hfi (caecilian and monitor 
lizard). Abbreviations: E, active expansion; C, compression; GH, m. geniohyoideus; IH, m. interhyoideus; RC, m. rectus cervicis; 
others as in Figure 2. (A) is modified from Carrier and Wake (1995); (B) is a composite of de Jongh and Gans (1969) and West 
and Jones (1974); and (C) is from this study. 

lizards, it is possible to directly compare 
their activity patterns in the gular pump of 
V. exanthematicus and in the buccal pump 
of a caecilian (Carrier and Wake, 1995) 
and an anuran (de Jongh and Gans, 1969; 
West and Jones, 1974). Differences in 
pressure profiles aside, all three neuro- 
motor patterns clearly are similar (Fig. 5). 
This suggests that homologous hyobran- 
chial muscles function in much the same 

way in these distantly related clades. 
Therefore, the homology of the gular and 
buccal pumping behaviors cannot be re- 

Nevertheless, analysis of the present 
data does not allow us to conclude with 
confidence that the gular and buccal 
pumps are homologous behaviors. Neuro- 
motor similarity is not sufficient to claim 
that a behavioral mechanism has been 

EMG Pattern of Gular Pump • Otverkowicz et al. 247 

conserved in evolution (Smith, 1994). Un- 
like other situations in which the neuro- 
motor pattern is conser\'ed despite func- 
tional divergence (e.g., in the evolution of 
terrestrial and aerial locomotion; Goslow 
et al., 1989), the gular pump may repre- 
sent functional convergence with the buc- 
cal pump by using homologous structures. 
The question of homology in the case of 
the gular pump is made even more com- 
plex by the fact that cycles of gular expan- 
sion and compression are also used in 
feeding (Smith, 1986), gular flutter (Heat- 
wole et al., 1973), and throat displays (Bels 
et al., 1995). One or more of these behav- 
iors could have retained the ancestral mo- 
tor pattern for hyobranchial movement 
and this pattern could have been co-opted 
for lung ventilation (with appropriate 
modification of narial and glottal vaKdng). 
The next study undertaken to explore the 
homology of buccal and gular pumping 
should be a broad comparative analysis to 
map the character of gular pumping (its 
presence or absence) on the phylogeny of 
Amniota. Preliminaiy investigations within 
Squamata indicate that gular pumping is 
widespread among nonserpentine squa- 
mates (Deban et al., 1994; Al-Ghamdi et 
al., 2001; Brainerd and Owerkowicz, per- 
sonal obsei-vation). This result, combined 
wdth our finding of neuromotor similarity 
in the pumping mechanisms of monitor 
lizards and amphibians, suggests that the 
gular pump of lizards may have been re- 
tained continuously from a buccal pump- 
ing ancestor. 


Fuzz Crompton provided the original 
inspiration for this study by explaining the 
nuance between "gular" and "buccal." Un- 
der Fuzz's guidance, T. O. honed his ex- 
perimental skills in radiographic, electro- 
myographic, and histologic techniques 
used in this study. T. O. wishes to express 
his gratitude to Fuzz for his mentorship 
and support over the last decade. We 
thank C. Farmer and L. Claessens for 
lending a hand wdth the surgeries, K. 

Schwenk for his invaluable critique of the 
early version of the manuscript, and the 
Festschrift editors F. A. Jenkins, Jr., and 
M. D. Shapiro for their infinite patience. 
We appreciate L. Meszoly for his render- 
ing of the hyoid apparatus, and C. Musin- 
sky for applying finishing touches to the 
figures. This work was supported by the 
Chapman Fellowship (Harvard University) 
to T. O., and by National Science Foun- 
dation grants IBN-9875245 to E. L. B. and 
IBN-9807534 to D. R. C. 


Al-Ghamdi, M. S., J. F. X. Jones, and E. W. Tay- 
lor. 2001. Evidence of a functional role in lung 
inflation for the buccal pump in the agamid liz- 
ard, Uromastyx aegijptiiis microJepis. Journal of 
Experimental Biolog>', 204: 521-531. 

AUFFENBERG, VV. 1981. The Behavioral Ecolog)' of 
the Komodo Dragon. Gainesville, Florida: Uni- 
versity Presses of Florida, x + 406 pp. 

Eels, V. L., J. -P. Gasc, V. Goosse, S. Renous, and 
R. VERNET. 1995. Functional analysis of the 
throat display in the sand goanna Varanus griseus 
(Reptilia: Squamata: Varanidae). Journal of Mor- 
pholog)', 235: 95-116. 

Br.\inerd, E. L., J. S. Ditelberg, and D. M. 
Bramble. 1993. Lung ventilation in salamanders 
and the evolution of vertebrate air-breathing 
mechanisms. Biological Journal of the Linnaean 
Society, 49: 163-183. 

Brainerd, E. L., and T. Owerkowicz. 1996. Role 
of the gular pump in lung ventilation during re- 
coverv from exercise in Varanus exanthematicus. 
American Zoologist, 36: 88. 

Carrier, D. R. 1987. The evolution of locomotor 
stamina in tetrapods: circumventing a mechani- 
cal constraint. Paleobiolog), 13: 32.5-.341. 

. 1996. Function of the intercostal muscles in 

trotting dogs: ventilation or locomotion? Journal 
of Experimental Biologv', 199: 145.5-1465. 

Carrier, D. R., and M. H. Wake. 1995. Mechanism 
of lung ventilation in the caecilian Dennophis 
mexicamis. Journal of Morpholog}', 226: 289- 

Cunningham, D. J. 1890. Value of nerve supply in 
the determination of muscular homologies and 
anomahes. Journal of Anatomy, 25: 31-40. 

Deban, S. M., J. C. O'Reilly, and T. Theimer. 
1994. Mechanism of defensive inflation in the 
chuckwalla, Saiiromahts obesits. Journal of Ex- 
perimental Zoology, 270: 451-459. 

DE JONGH, H. J., and G. G.\NS. 1969. On the mech- 
anism of respiration in the bullfrog, Raua catcs- 
beiana: a reassessment. Journal of Morphology, 
127: 2.59-290. 

Edgeworth, F. H. 1935. The Cranial Muscles of 

248 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Vertebrates. Cambridge, United Kingdom: Cam- 
bridge University Press, viii + 493 pp. 

Erdman, S., and D. Cundall. 1984. The feeding 
apparatus of the salamander Amphiurna tridac- 
tijlum: moiphology and behavior. Journal of Mor- 
phology, 181: 175-204. 

Phylogenetic relationships within Squamata, pp. 
119-281. In R. Estes and G. Pregill (eds.), Phy- 
logenetic Relationships of the Lizard Families. 
Stanford, California: Stanford University Press, 
xii -I- 631 pp. 

Cans, C. 1970. Strategy and sequence in the evolu- 
tion of the external gas exchangers of ectother- 
mal vertebrates. Forma et Functio, 3: 61-104. 

GosLow, G. E., Jr., K. P. Dial, and F. A. Jenkins, 
Jr. 1989. The avian shoulder: an experimental 
approach. American Zoologist, 29: 287-301. 

Heatwole, H., B. T. Firth, and G. J. W. Webb. 
1973. Panting tliresholds of lizards. Comparative 
Biochemistry and Physiology, A, 46: 711-826. 

Lauder, G. V, and H. B. Schaffer. 1985. Func- 
tional morphology of the feeding mechanism in 
aquatic ambystomatid salamanders. Journal of 
Morphology, 'l85: 297-326. 

. 1988. Ontogeny of functional design in tiger 

salamanders (Ambystoma tigrinum): are motor 
patterns conserved during major morphological 
transformations? Journal of Morphology, 197: 

Liem, K. F. 1985. Ventilation, pp. 185-209. In M. 
Hildebrand, D. M. Bramble, K. F. Liem, and D. 
B. Wake (eds.). Functional Vertebrate Morphol- 

ogy. Cambridge, Massachusetts: Belknap Press. 
430 pp. 

LOEB, G. E., AND C. Gans. 1986. Electromyography 
for E.xjDcrimentalists. Chicago, Illinois: University 
of Chicago Press, xx -I- 373 pp. 

Morphology and behavior of the feeding appa- 
ratus in Cnjptobranchus alleganiensis (Amphib- 
ia: Caudata). Journal of Morphology, 220: 47-70. 

OwERKOwicz, T, C. G. Farmer, J. W. Hicks, and 
E. L. Brainerd. 1999. Contribution of gular 
pumping to lung ventilation in monitor lizards. 
Science, 284: 1661-1663. 

Reilly, S. M., and G. V. Lauder. 1991. Experimen- 
tal morphology of the feeding mechanism in sal- 
amanders. Journal of Morphology, 210: 33—44. 

RiEPPEL, O. 1978. The throat musculature of Sphen- 
odon, with comments on the primitive character i 
states of the throat muscles in lizards. Anatom- 
ischer Anzeiger, 144: 429-440. 

Smith, K. K. 1986. Morphology and function of the 
tongue and hyoid apparatus in Varanus (Varani- 
dae, Lacertilia). Journal of Morphology, 187: 

. 1994. Are neuromotor systems conserved? 

Brain, Behavior and Evolution, 43: 293-305. 

Smith, L. 1920. The hyobranchial apparatus of Spe- 
lerpes bislineatus. Journal of Morphology, 33: 

West, N. H., and D. R. Jones. 1974. Breathing 
movements in the frog Rana pipiens. I. The me- 
chanical events associated with lung and buccal 
ventilation. Canadian Journal of Zoology, 53: 



Abstract. Suckling and drinking are rh>thmic ac- 
tivities, widi an electromyographic pattern that is 
characterized by a significant coactivation of muldple 
muscles that are not obvious svoiergists. If die rh\i;hm 
and coactivation were due to conventional excitation 
of motor neurons by a single, central source of 
rhvthm generation then, within the periods of coac- 
tiv ation, there should also be a degree of sy:ichroni- 
zation of muscle action potentials, similar to that es- 
tablished for respiration. Both respiration and suck- 
ling are activities that can persist in die absence of 
the cerebral hemispheres. However, in a preliminary 
study of suckling, evidence for synchronization was 
not characteristic of most of the coactive muscles. 
Marked rhythmic coactivation of muscles is also char- 
acteristic of the cerebrally directed, more mature ac- 
tivit)' of drinking. Synchronization of action potentials 
in other conscious movements suggests that cere- 
brally directed drinking might differ from suckling 
with respect to the level of synchronization. We test- 
ed this hvpothesis in miniature pigs by cross-corre- 
lation analysis of the electromyographic activitv' re- 
corded in multiple oral and submandibular muscles 
during suckling and drinking in the same animals. 
Evidence of intraburst synchronization was found in 
only a few muscles (sternohyoid, omohyoid, and ge- 
niohyoid) during suckling, but that evidence was ab- 
sent in drinking in exactly the same muscles in the 
same animals. A tentative ex-planation for the paucit\' 
of evidence for sjiichronization in suckling, despite 
the coactivation of muscles, is that the rhythmic mus- 
cle activitv is generated by plateau potentials in in- 
dividual motor neurons; this is a phylogenetically 
primitive mechanism with respect to mammals. The 
absence of evidence for svnchronization in drinking 
may reflect the fact that the neural mechanisms for 
drinking precede those for suckling phylogenetically, 
although not ontogenetically. 


Rhythmic oral movements can be gen- 
erated by brainstem mechanisms alone. 

' Physiology Department, Kings College London 
(School of Biomedical Sciences of Guy's, King's and 

Suckling occurs in the anencephalic hu- 
man infant (Hall, 1833; Gamper, 1926; 
Monnier and Willi, 1953) and rhythmic 
chewing or lapping movements can be 
elicited in decerebrate animals (Bremer, 
1923). These reports are consistent with 
the subsequent localization of a central 
pattern generator (CPG) for rhythmic oral 
movements within the brainstem (Dellow 
and Lund, 1971). Nevertheless, cerebro- 
cortical centers have been described for 
both mastication and for suckling (Iriki et 
al., 1988). When these higher centers are 
functional, they may simply supply drive to 
the CPG (Bremer,' 1923; Lund and Del- 
low, 1971) but their activity may also in- 
teract with the brainstem mechanisms in 
other ways, including direct monosynaptic 
connections to trigeminal motor neurons 
(Mishima et al, 1982; Moriyama, 1987; 
Ohta et al, 1989; Ohta and Saeki, 1989). 
The pattern of activity of cortical motor 
neurons can, in the case of the limbs, be 
directly related to the amplitude and to 
the direction of a resultant limb movement 
(Georgopoulos, 1995) so that a similar ac- 
tion in directing oral activity towards ex- 
traorally located food is to be expected in, 
for example, drinking. 

When the infant pig locates on an arti- 
ficial teat, rhythmic suckling is only trig- 
gered when milk is present (German et al., 
1997) and suckling then continues as a re- 

st. Thomas's Hospitals), St. Thomas's Campus, Lam- 
beth Palace Road, London SEl 9RT. 

- Department of Biological Sciences, Universitv' of 
Cincinnati, Cincinnati, Ohio 45221-0006. 

^ To whom correspondence should be addressed. 

Bull. Mus. Comp. ZooL, 156(1): 249-256, October, 2001 249 

250 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

sponse to the intraoral delivery of the liq- the drive to the inotor neurons during 

uid. Conversely, drinking is directed to the suckling does not produce simultaneous 

acquisition of an extraorally located liquid intraburst spikes during EMG activity in 

and the maintained direction of this activ- most of the coactive muscles (but does in 

ity is likely to involve cortical activity. The just a few hyoid muscles). This contrasts 

postnatal maturation of cerebral functions with studies of rhythmic somatic activity 

might then relate to the ability to change (Sears and Stagg, 1976) in adult animals 

from suckling to drinking. and with studies of cortically directed ac- 

Consequently, the aim of this study was tivity (Datta and Stephens, 1990) where 
to test the proposition that some aspect of synchronization can be regularly demon- 
the generation or control of the pattern of strated. Consequently, the initial hypoth- 
electromyographic (EMG) activity in esis was that, in drinking, action potentials 
drinking differed from that in suckling. As within bursts of EMG activity of some 
a first stage in this investigation, a form of pairs of muscles might show greater de- 
analysis was adopted that had originally grees of synchronization, compared to the 
been developed in relation to the rhythmic more immature acti\dty of suckling, 
activity of respiration (Sears and Stagg, /^ o 
1976). Respiration is a similar activity to MATERIALS AND METHODS 
suckling and drinking in the sense that res- Three miniature pigs (Sii.s scrofa), each 
piration is generated by a CPG and as such 3 weeks old, were trained to feed in a box 
involves a number of different muscles from an artificial teat. At this age pigs are 
that exliibit simultaneous bursts of EMG at the start of weaning; some individuals 
activity. Each of these bursts of EMG ac- will repeatedly change their mode of in- 
tivity consists of a number of individual gestion from suckling to drinking and back 
spikes or muscle action potentials, hence- again within a few minutes (depending 
forth referred to as intraburst spikes. Some upon how the food is made available). 
of these intraburst spikes occur at the Consequently, data comparisons can be 
same time in the muscles of different in- made in which there is no significant time 
tercostal spaces, that is, the different mus- delay during which electrodes can move or 
cles are not only coactivated but show syn- deteriorate or during which other longer- 
chronous firing within the periodic bursts, term maturational changes can confuse the 
indicating that some of the motor neu- picture. 

rones receive identical synchronous acti- The methods used in these experiments 

vation. have been described in detail elsewhere 

The demonstration of synchronization (Thexton et al., 1998) so that only a brief 
of spikes within bursts of EMG activity oc- description is given here. Under general 
curring at the same time (in two different anesthesia (halothane/oxygen), the sub- 
but coactive or synergistic muscles) only mandibular musculature was exposed and 
indicates that a proportion of the action individual muscles identified. Bipolar fine 
potentials arriving presynaptically at the wire electrodes were then inserted into the 
motor neurons are also synchronous. Cor- muscles and the leads passed via a sub- 
tically originating activity acting on two cutaneous tunnel to a multipole connector 
synergistic muscles is often associated with on the back. After recovery, the animals 
a degree of such synchronization in the were again fed in the box, allowing the 
EMGs. Similarly, a CPG supplying two multipole connector to be connected to 
synergistic muscles will also produce syn- amplifiers. The amplified signals were then 
chronization of a proportion of the spikes recorded on tape for off-line digitization, 
in the two EMGs. All analysis was carried out on the digi- 

In contrast, pilot studies on infant pigs tized data. 
(Banks and Thexton, 1999) indicated that The EMG activity in the relevant mus- 

Synchronization in Muscles During Suckling • Thexton and German 251 

cles was first half-wave rectified and the 
spike peaks identified. The spike peaks 
were then converted to point events and 
all events, wdth an amplitude less than 10% 
of the maximum, were eliminated as being 
due largely to noise, that is, they were set 
to zero; all the other events were given a 
\alue of one. A pairwise correlation be- 
tween the two series of point events was 
then calculated. This was then repeated 
for different time shifts of one data set rel- 
ative to the other, for example, from f,, — 
40 milliseconds to ^o + 40 milliseconds, 
where t^ is zero time shift between the two 
data sets (Fig. 1). If any tendency exists 
for intraburst spikes to occur either syn- 
chronously in the two data streams or with 
a regular time lag in one data stream with 
respect to the other, they come to corre- 
spond with each other at one of the time 
shifts and so produce a larger correlation 
at that time shift. 

The EMG activities were analyzed to 
determine if any difference existed in the 
level of synchronization occurring in suck- 
ling and drinking. The hypothesis was that 
if the two activities arose or were centrally 
influenced in different ways, as indicated 
in the Introduction, there might be differ- 
ent levels of synchronization. 


The data analyzed in this study were the 
multichannel EMG activities recorded in 
a variety of muscles during suckling and 
during drinking. The EMG activity record- 
ed during a short period of drinking is 
shown in Figure 2. The period contained 
just over six cycles of movement including 
three cycles in which swallowing was pre- 
sent; the three swallowing cycles were as- 
sociated with the three higher-amplitude 
bursts of activity in the hyoglossus. Simul- 
taneous bursts of EMG activity occurred 
in a number of muscles that are innervated 
by different motor nuclei, for example, di- 
gastric (V), stylohyoid (VII), and hyoglos- 
sus (XII). This coactivation of sets of mus- 
cles is characteristic of the activity in feed- 

ing in the weaning pig, whether suckling 
or drinking (Thexton et al., 1998). 

When the EMG signals obtained from 
the submandibular muscles during suck- 
ling were processed, they produced a se- 
ries of correlograms that, for the most 
part, were flat (Figs. 3A-C). Across all 
three animals the expected central peaks, 
indicative of short- or medium-term syn- 
chronization, were generally absent, as 
previously reported (Banks and Thexton, 
1999). The only cross-correlograms with 
visible central peaks, which were consis- 
tent with synchronization, were those that 
were obtained by the crosscorrelation of 
geniohyoid, sternohyoid, and omohyoid 
activity (Figs. 3D, E); of these, the cross- 
correlation between sternohyoid and omo- 
hyoid activity provided the clearest evi- 
dence (in two out of three animals). 

When the same analyses were applied 
to drinking data, the results were similar 
in so far as the cross-correlograms were 
generally flat. However, the results dif- 
fered in that no obvious central peaks oc- 
curred in the cross-correlations derived 
from the geniohyoid, sternohyoid, or omo- 
hyoid signals. Direct intra- animal compar- 
isons were also made between the cross- 
correlations of omohyoid, sternohyoid, and 
geniohyoid activity in suckling and the 
same cross-correlations in drinking. In two 
animals (that within minutes would change 
from suckling to drinking and back again) 
the evidence for intraburst synchroniza- 
tion, which was clearly evident in muscles 
such as sternohyoid and omohyoid during 
suckling (Fig. 3E), was always absent in 
drinking (Fig. 3F). 


One of the earliest studies that provided 
evidence of intraburst synchrony among 
the motor units of cocontracting muscles 
(Sears and Stagg, 1976) was carried out on 
the respiratory system, that is, on a rhyth- 
mic activity produced by a brainstem pat- 
tern generator. Suckling and drinking sim- 
ilarly involve cocontracting muscles and, 
on current evidence, are also generated by 

252 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

A Data 2 shifted time units with respect to Data 1 

1 T 

1 -^ 





Products: 00001 000000000 


E Data 2 shifted +2 time units with respect to Data 1 

I Mill 

1 -L 




B Data 2 shifted -1 time units with respect to Data 1 

1 T 

1 -L 



Products: 01000010101000 


C Data 2 shifted -2 time units with respect to Data 1 

1 - 



1 - 






D Data 2 shifted +1 time units with respect to Data 1 

I Mill 

1 -L 


Products: 000000001 01 000 


Figure 1. Cross-correlation using two ultrashort sections of synthetic data in binary form. Such data sets are produced If the 
peaks of half-wave rectified electromyographic (EMG) activity are reduced to point processes indicating the presence or the 
absence of a spike peak; the example shown could represent the spikes in a single burst of EMG activity. In (A), the two data 
series (shown both as binary and as graphical data) are initially correctly aligned in time (zero time shift or Q. At each point in 
time, the product of the values in the two series is obtained and this is then summed. In (B), data set 2 is moved one unit back 

Synchronization in Muscles During Suckling • Thexton and German 253 












Time - sec 

Figure 2. Raw electromyogram (EMG) signals recorded from representative muscles during drinking. In several muscles two 
sets of electrodes were inserted to establish that the signals recorded were not obviously site dependent within the muscle. 
Abbreviations used: GH, geniohyoid; GG, genioglossus; OMO, omohyoid; STH, sternohyoid; HYOG, hyoglossus; DIG, digastric; 
subscript numbers identify two electrodes in the same muscle. All muscles are unilateral. 

a brainstein pattern generator. Conse- 
quently, the assumption was that synchro- 
ny would also be present between coacti- 
vated muscles during suckling and drink- 
ing. However, most cross-correlograms be- 
tween inuscle activities showed no 
evidence for intraburst synchronization 

during either suckling or drinking (Fig. 3; 
Banks and Thexton, 1999). Nevertheless, 
peaks that were consistent with synchro- 
nization were evident in the cross-corre- 
lograins of sternohyoid, geniohyoid, and 
omohyoid activities, with the cross-corre- 
lation between stemohvoid and omohyoid 

in time (t ,) relative to data set 1; again the products of the values in the two series are obtained and summed. In (C), the 
process is repeated with data set 2 moved back two units back in time (f^s) relative to data set 1 . In (D), the process is repeated 
with a time shift of +1 time units (f,) and in (E) with a time shift of +2 time units (y. In (F), the sum of cross products (a count 
of those spikes that coincided at a particular time shift) is plotted against the time shift. In these synthetic data, the maximum 
correspondence between spikes in the two data sets occurred with a shift of f ,. In practice two data sets each at least 15,000 
time units long, containing many hundreds of spike locations, are cross-correlated. 

254 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

DIG-GH (suck) 

GH-OMO (suck) 


200 T 


I I I I I I I I I I I 



I I I I I I I I I I I I I I I I 

-40 +40 


HYOG-STYLO (suck) 

STH-OMO (suck) 



80 T 


GG-GH (suck) 

300 T 


I I I I I I I I I I I M I I I 

40 +40 


STH-OMO (drink) 


1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

40 +40 



Figure 3. Cross-correlations between signals detected by different pairs of electrodes in one animal. Abbreviations used: DIG, 
digastric; GH, geniohyoid; HYOG, hyoglossus; STYLO, styloglossus; GG, genioglossus; OMO, omohyoid; STH, sternohyoid. 
The horizontal axis indicates time shift relative to zero time shift (fg). Examples of cross-correlations that were commonly found 
in suckling are shown in (A), (B), and (C). Some evidence of short-term synchronization (elevation of central region ± 2 milli- 
seconds) in suckling is present in (D) and stronger evidence is present in (E). In (F), the activities in the same muscles as in 
(E) were recorded during drinking and were cross-correlated but with no sign of a central peak. 

providing the clearest evidence of the 
presence of synchronization. 

Before deahng with the significance of 
this finding, it is necessary to comment 
briefly on its vahdity. First, the results 
shown in Figure 3E indicate that the 
cross-correlation method used in this pa- 
per was intrinsically capable of detecting 
synchronization within the recorded EMG 

signals. However, motor neurons are acti- 
vated froin multiple sources within the 
central and peripheral nei'vous systems 
and not all the motor neurons supplying 
units in a given muscle are subject to the 
same neural influences. Consistent with 
this, some motor units have different tasks 
from others within the same muscle. If 
EMG activity is recorded with an intra- 

Synchronization in Muscles During Suckling • T/j^xfon and German 255 

muscular wire electrode (which only re- 
cords from a limited number of motor 
units) the record will represent a restricted 
sample of the different motor units within 
that muscle. If, by chance, the sample 
from one of the two muscles being cross- 
correlated does not include motor units 
that receive drive synchronous with the 
sample in the other muscle, then the 
cross-correlograms will be flat. This sug- 
gests that those motor units, the activity of 
which generated flat correlograms, were 
essentially firing randomly in time. How- 
ever, such correlograms do not prove that 
other motor units within the muscles 
would not exliibit synchrony. Therefore, 
the general failure to detect signs of syn- 
chrony in this study is best interpreted as 
indicating a scarcity of synchronized units 
in most coactive muscles, with synchro- 
nized units simply being more frequently 
found in the coactive geniohyoid, sterno- 
hyoid, and omohyoid muscles during suck- 

One possible scenario is that a large 
number of neurons and synapses intervene 
between the CPG neurons generating the 
rhythmic activity and most of the motor 
neurons receiving that rhythinic drive. Be- 
cause of the nonlinear input-output rela- 
tions at each synapse, considerable tem- 
poral dispersion of the signals would then 
occur and the original synchronous drive 
would become diluted by other influences 
also acting at those synaptic relays. Thus, 
although the rhythmic drives to the motor 
neurons of the different coactive muscles 
would still be expected to wax and wane 
together, little of the intraburst spike gen- 
eration would still be synchronous in the 
coactive muscles. However, this mecha- 
nism would not be consistent with the 
abrupt "on" and "off" of the bursts seen 
in the recorded EMG activity (Fig. 2). 

Thus, a problem may exist with the con- 
cept of a CPG producing the activities of 
suckling or drinking purely by supplying a 
rhythmically fluctuating drive to the dif- 
ferent motor neuron groups producing the 
EMG pattern seen in Figure 2. In fact, the 

absence of consistent evidence for syn- 
chrony (Fig. 3) suggests that the drive 
from the CPG may not directly supply se- 
ries of action potentials that produce a cu- 
mulative depolarization of most of the mo- 
tor neurons and, therefore, most of the 
spike production in the coactive muscles. 
However, if the central pattern generator 
functioned as an on-off switch, simply 
triggering the motor neurons into becom- 
ing their own independent burst genera- 
tors, no synchronization would occur be- 
tween intraburst spikes in coactive mus- 
cles. That situation can arise when motor 
neurons are simply triggered to produce 
long-lasting depolarizations known as pla- 
teau potentials. Although such potentials 
are well known in invertebrates (Croll et 
al., 1985), they have only relatively recent- 
ly been described as occurring in mam- 
mals. It is of particular interest that these 
potentials have been found in mammalian 
motor nuclei involved in generating the 
movements of feeding (Mosfeldt Laursen 
and Rekling, 1989; Rekling and Feldman, 

Whatever the mechanisms involved in 
generating the rhythmic movements, the 
results clearly do not support the initial 
ontogenetic hypothesis, that the EMG sig- 
nals in the more mature and probably cor- 
tically directed activity of drinking would 
show more synchronization than in suck- 
ling. However, suckling is an evolutionary 
novelty (Clark and Smith, 1993; German 
and Crompton, 2000), whereas mammali- 
an drinking is probably a substantially old- 
er evolutionary trait, functionally resem- 
bling drinking in other nonmammalian tet- 
rapods. Consequently, the neural basis of 
drinking could reflect the evolutionary 
heritage of mammals and, in that sense, be 
less derived than suckling. The limited ev- 
idence for synchrony during suckling, 
which is absent in drinking, might reflect 
the contemporaneous evolution of suck- 
ling and the more developed central ner- 
vous system that characterizes maminals. 

256 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


Banks, D., and a. J. Thexton 1999. Evidence for 
pre-synaptic synchronization of electromyo- 
graphic activity in oral musculature during suck- 
ling in mini pigs. British Neuroscience Associa- 
tion Abstracts, 15: 44. 

Bremer, F. 1923. Physiologie nerveuse de la masti- 
cation chez le chat et le lapin. Archive intema- 
tionale de Physiologie, 21: 309-352. 

Clark, C. T., and K. K. Smith. 1993. Cranial oste- 
ogenesis in Monodelphis dornestica (Didelphi- 
dae) and Macropus eugenii (Macropodidae). 
Journal of Morphology, 215: 103-114. 

Croll, R. p., M. R Kovac, and W. J. Davis. 1985. 
Neural mechanisms of motor program switching 
in the mollusc Pleurobranchaea. II. Role of tlie 
ventral white cell, anterior ventral, and B3 buccal 
neurons. Journal of Neuroscience, 5: 56—63. 

Datta, a. K., and J. A. Stephens. 1990. Synchro- 
nization of motor unit activity during voluntary 
contraction in man. Journal of Physiology, 422: 

DelloW, p. C, and J. P. Lund. 1971. Evidence for 
central timing of rhythmical mastication. Journal 
of Physiology" 215: 1-13. 

Camper, E. 1926. [In the film of:] The Human Mid- 
Brain. British Film Institute. London. 

Ceorgopoulos, a. p. 1995. Current issues in direc- 
tional motor control. Trends in Neurosciences, 
18: 506-510. 

Cerman, R. Z., and a. W. Crompton. 2000. Ontog- 
eny of feeding in mammals, pp. 449^57. In K. 
Schwenk (ed.). Feeding: Form, Function and 
Evolution in Tetrapod Vertebrates. San Diego, 
CA: Academic Press, xv + 537 pp. 

Cerman, R. Z., A. W. Crompton, D. W. Hertweck, 
AND A. J. Thexton. 1997. Determinants of 
rhythm and rate in suckling. Journal of Experi- 
mental Zoology, 278: 1-8. 

Hall, M. 1833. On die reflex function of the medulla 
oblongata and medulla spinalis. Philosophical 
Transactions of the Royal Society, London, 132: 

IRIKI, A., S. NOZAKI, AND Y. Nakamura. 1988. Feed- 
ing behavior in mammals: corticobulbar projec- 

tion is reorganised during conversion from suck- 
ing to chewing. Developmental Brain Research, 
44: 189-196. 

Lund, J. P., and P. G. Dellow. 1971. The influence 
of interactive stimuli on rhythmical masticatory 
movements in rabbits. Archives of Oral Biology, 
16: 215-223. 

Mishima, K., K. Sasamoto, and M. Ohta. 1982. 
Amygdaloid or cortical facilitation of antidromic 
activity of trigeminal motoneurons in the rat. 
Comparative Biochemistry and Physiology. A, 
Comparative Physiology, 73: 355—359. 

Monnier, M., and H. Willi. 1953. Die integrative 
Tatigkeit des Nervensystems beim mesorhom- 
benspinalen Anencephalus (Mittelhernwesen). 
Monatsschrift der Psychiatric und Neurologie, 
126: 239-273. 

MORIYAMA, Y. 1987. Rhythmical jaw movements and 
lateral ponto-medullary reticular neurons in rats. 
Comparative Biochemistry and Physiology. A, 
Comparative Physiology, 86: 7-14. 

Electrophysiological properties of hypoglossal 
motoneurons of guinea-pigs studied in vitro. 
Neuroscience, 30: 619—637. 

Ohta, M., S. Ishizuka, and K. Saeki. 1989. Corti- 
cotrigeminal motor pathway in the rat — II. An- 
terio- and retrograde HRP labeling. Comparative 
Biochemistry and Physiology. A, Comparative 
Physiology, 94: 405-414. 

Ohta, M., and K. Saeki. 1989. Corticotrigeminal 
motor pathway in the rat — I. Antidromic activa- 
tion. Comparative Biochemistry and Physiology. 
A, Comparative Physiology, 94: 99-104. 

Rekling, J. C, AND J. L. FELDMAN. 1997. Calcium- 
dependent plateau potentials in rostral ambiguus 
neurons in the newborn mouse brain stem in vi- 
tro. Journal of Neurophysiology, 78: 2483-2492. 

Sears, T.A., and D. Stagg. 1976. Short term syn- 
chronization of intercostal motoneurone activity. 
Journal of Physiology, 263: 357-381. 

Thexton, A. J., A. W. Crompton, and R. Z. Ger- 
man. 1998. Transition from suckling to drinking 
at weaning: a kinematic and electromyographic 
study in miniature pigs. Journal of Experimental 
Zoology, 280: 327-343. 



Abstract. To improve our measurements of me- 
chanical power output of fl>'ing birds, we examined 
the congruency between two recording techniques of 
pectoralis muscle-length change (sonomicrometric 
and kinematic) used in empirical measures of the me- 
chanical power output of black-billed magpies flying 
over their fufl range of flight speeds (0-14 m/s) in a 
variable-speed wdnd tunnel. Simultaneous recordings 
of pectoralis muscle force (obtained from strain gaug- 
es attached to the muscle's humeral insertion) were 
integrated with the bA'O recording technicjues to gen- 
erate a work-loop for each wingbeat. Although the 
overall shapes of the power curves obtained by the 
h\'o techniques were similar, estimates of muscle 
lengtli change using sonomicrometric data were, on 
average, very similar to those obtained by one-axis 
kinematics and slightly lower than those obtained by 
two-axis kinematics. Given these small differences, 
our sonomicrometry measurements indicate that ki- 
nematic estimates of muscle length change, when 
combined with humeral strain measurements, can 
provide an accurate estimate of pectoralis work and 
power output during bird flight. However, a key ad- 
vantage of sonomicrometry is that it provides a direct 
measure of the lengthening and shortening of the 
muscle's fascicles relative to their rest length, which 
is not possible to derive from wing motion alone. 


The cost of flight is central to under- 
standing the biology of any fl>Tng animal. 
In particular, without knowing the meta- 
bolic costs of flight, conclusions obtained 
from time and energy budget studies of 
birds regai-ding selection for migration 
habits, optimal foraging behaviors, and re- 
productive strategies will remain unclear 

' Division of Biological Sciences, University of 
Montana, Missoula, Montana 59812. 

- Concord Field Station, Harvard University, Bed- 
ford, Massachusetts 01730. 

^ To whom reprint requests should be addressed. 

(Welham and Ydenberg, 1993; Heden- 
strom and Alerstam, 1995). 

Various approaches have been taken to 
address the metabolic power requirements 
of flight. The mechanical cost of flight may 
be modeled using aerodynamic theory 
(Pennycuick, 1975; Rayner, 1979), and the 
metabolic cost of flight can then be in- 
ferred using estimates of the muscular ef- 
ficiency of converting metabolic energy to 
mechanical work (Pennycuick, 1989; 
Thomas and Hedenstrom, 1998). Meta- 
bolic power has been measured directly 
over a limited midrange of flight speeds by 
means of oxygen consumption studies of 
birds flying in a wind tunnel (Tucker, 1968, 
1972; Berger et al, 1970; Torre-Bueno 
and LaRochelle, 1978; Rothe et al, 1987). 
Other methods have included measure- 
ments of doubly-labeled water (Hails, 
1979; Flint and Nagy, 1984) and heart rate 
(Berger et al, 1970; Butler et al, 1977) of 
birds in flight. More recently, infrared im- 
aging has also been used to assess meta- 
bolic power (Speakman et al., 1997; Ward 
et al., 1997). However, many of these stud- 
ies have been restricted to a narrow range 
of flight speeds and flight behavior, pre- 
venting a complete picture of metabolic 
cost versus flight speed. Thus, models of 
mechanical power are the most useful 
means currently available for predicting 
flight costs over a wide range of speeds. 
The wide acceptance and use of such me- 
chanical power models (Welham and 
Ydenberg, 1993; Pennycuick, 1997) sug- 
gests that it may prudent to provide an in- 
dependent test of their accuracy. 

In a previous study (Dial et al., 1997), 

Bull. Mus. Comp. ZooL, 156(1): 257-268, October, 2001 257 

258 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

the mechanical power output of black- 
billed magpies (Pica pica) was measured 
using force-calibrated strain gauges at- 
tached to the deltopectoral crest (DPC), 
which serves as the insertion site of the 
primary downstroke flight muscle (mus- 
culus pectoralis) on the humerus. These 
force measurements were integrated with 
kinematics of the wing to estimate the me- 
chanical power output of the pectoralis 
and, hence, the whole bird. When appro- 
priate kinematic parameters (Tobalske and 
Dial, 1996; Tobalske et al, 1997) are in- 
corporated in models based on aerody- 
namic theory, predicted power output 
agrees reasonably well with our earlier 
measureinents of mechanical power out- 
put of magpies (Rayner, 1999). Our pre- 
vious measurements used kinematic esti- 
mates of pectoralis length change. How- 
ever, some questions remain regarding es- 
timates of the magnitude and timing of 
pectoralis length change inferred from dis- 
tal wing movement in the work of Dial et 
al. (1997). In the present study, we reas- 
sess the mechanical power output of mag- 
pies based on direct sonomicrometric 
measurements of pectoralis fascicle length 
changes and compare these data directly 
with corresponding kinematic estimates 
based on wing excursion. 


Bird Training and Wind Tunnel Trials 

Three black-billed magpies were trained 
to fly in a wind tunnel (Tobalske and Dial, 
1996) over a range of flight speeds (0—14 
m/s) that represented the full extent that 
the animals were willing to fly. Birds were 
housed in the University of Montana's an- 
imal facility and given food and water ad 
libitum. The protocol for all facilities, care, 
and surgical procedures was approved by 
the Institutional Animal Care and Use 
Committee established at the University of 

Force-Strain Measurements 

Birds were anesthetized (25 mg/kg ke- 
tamine and 2 mg/kg xylazine, supplement- 

ed as needed) and the feathers removed 
over the left shoulder and the middle of 
the back between the scapulae. A 15-mm 
incision was made in the skin overlying the 
DPC, which was then exposed by gently 
parting the fascicles of the deltoid muscle. 
The dorsal surface of the DPC was pre- 
pared by lightly scraping away the perios- 
teum and then swabbing the underlying 
bone surface with methyl-ethyl ketone to 
remove any residual tissue and to dry the 
site. A strain gauge (single-element metal 
foil type FLE-05-11, Tokyo Sokki Kenk- 
yujo, Ltd., Tokyo, Japan) was then at- 
tached to the dorsal surface of the DPC 
using self-catalyzing cyanoaciylate adlie- 
sive, with its principal axis aligned approx- 
imately 15° proximal to the perpendicular 
axis of the humeral shaft (Dial and Biew- 
ener, 1993). Strain gauge lead wires (36 
gauge. Teflon insulated; Micromeasure- 
ments Inc.) ran beneath the deltoid and 
subcutaneously to a miniature connector 
plug (Microtech FG-6 [X2]) that was 
mounted on the back of the bird by su- 
turing the plug's epoxy base securely to the 
intervertebral ligaments using silk. The 
sldn was drawn snugly around the protrud- 
ing connector plug, and the surrounding 
skin was covered with elastic surgical tape. 

The DPC strain signals were transmit- 
ted to bridge amplifiers (Vishay model 
2120A, Micromeasurements Inc.) via two 
light-weight shielded cables running 
through a small hole in the top of the wind 
tunnel test section. Raw in vivo DPC 
strains, sonomicrometry signals, and elec- 
tromyograms (see below) were sampled at 
5,000 Hz by a Keithley Instruments A/D 
converter and stored in a computer. To 
monitor the quality of the recordings dur- 
ing the experiment, live data from each tri- 
al were printed on a Gould 2400 chart re- 

In past studies, the tensile strains ex- 
perienced by the DPC during flight were 
calibrated to pectoralis force in situ after 
flight trials. However, in all three birds, af- 
ter flights at all eight speeds had been re- 
corded, either the bonding of the strain 

Mechanical Power Output of Magpie Flight • Warrick et al. 259 

gauge to die bone or the strain gauge itself 0.74-mm offset was added to all length 

failed. Rather than attempting to reattach measurements. Total fascicle length 

the strain gauge to the DPC in the same change (L) for the pectoralis was calculat- 

position or use another strain gauge for ed as: 

calibration, we chose to leave the DPC j _ j tr j /i\ 

strain data uncalibrated as raw voltages. i ir r. 

Although this precluded a quantification of where L, is the change in distance between 

pectoralis force and calculation of muscle the sonomicrometry crystals, L\^ is the local 

work, relative work-loops derived from resting length between the crystals, and L^ 

raw DPC strain voltages coupled with fas- is the total (average) resting length of the 

cicle length changes (from sonomicrome- pectoralis muscle fascicles measured to 

try, next section) provide valuable infor- with 0.5 mm using digital calipers, 
mation concerning the relative change in 

mechanical power as a function of flight vVOrK-lOOps 

speed (henceforth "uncalibrated power"). The work done throughout one wing- 
To average the data for all three birds, the beat cycle (W^^.^) 'was calculated by inte- 
uncalibrated power required for flight at grating the change in DPC strain with the 
all speeds was normalized relative to the change in fascicle length: 
maximum ("relative") power recorded for ^^^ _ v-p /j _ j \i (n\ 
each individual, which for all three birds "^^ ~ ^ "^ " "^''^ ^ ' 
was observed during hovering flight. where F is the DPC strain voltage and L 

is the pectoralis fascicle length at the same 

Sonomicrometry point in the time series n. Only that part 

Sonomicrometiy crystals (2-mm SL-2, of the wingbeat cycle during which posi- 

Triton Inc.) were implanted in the anterior tive strain voltages were recorded (i.e., 

region of the left pectoralis muscle of each when the gauge measured tensile strain, 

bird in a position thought to represent best rather than the compressive strain experi- 

the contraction of the entire muscle (as enced by the DPC during upstroke) was 

suggested in Shigeoka [1999]). Each ciys- used to calculate relative pectoralis work, 

tal was mounted on a stainless steel wire Relative work (volt X mm) was then divid- 

featuring two anchoring points ("loops"), ed by the cycle time {t) for each individual 

After a skin incision was made to expose work-loop to obtain an estimate of relative 

the pectoralis muscle, the muscles fasci- muscle power (Fig. 1). Measurements of 

cles were parted using surgical scissors, relative power were obtained over a range 

creating two openings into which the crys- of speeds, based on averages at each speed 

tals were inserted. Once aligned, the ciys- obtained from a minimum of two wing- 

tals were held in place by suturing the two beats (for magpie 3 during hovering flight) 

anchoring loops to the surrounding muscle and a maximum of 23 wingbeats (magpie 

tissue and fascia. The wires for the crystals 3 at 8 m/s). 
were passed subcutaneously to the dorsal 

side of the bird and connected to the back Kinematics 

plug. After closing the incision, the birds Each trial was videotaped from lateral 

were allowed to recover from the anesthe- (Panasonic S-VHS) and caudal (Sony Hi-8 

sia for approximately 24 hours. The resting Handicam) views. Video from each trial 

length between the crystals (resting fasci- was captured on computer (Iomega Buz 

cle length, LJ was taken while the bird Video Capture), and bitmap still images 

was standing at rest, with its wings held (60 fields per second [fps]) were digitized 

against its body. To compensate for the using NIH Image. The x and ?/ coordinates 

faster transmission of the ultrasound pulse of the bird's eye, tail base, and wing tip 

through the epoxy lens of each ciystal, a were scaled using a 2-cm grid background 

260 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Magpie 3 

14 ms" 




Flight speed (m s' ) 

Figure 1 . Work-loops (insets) created by plotting the change in tensile strain (y axis, in volts not calibrated to force) of the 
deltopectoral crest by the fiber length (x axis, mm). The area of the work-loop above zero volts was integrated to calculate work 
done by the pectoralis. Dividing this area by the wingbeat cycle time yields a measure of relative power. The mean relative 
power output (± SDs) of magpie 3 is shown over the range of flight speeds. 

located behind the bird, adjusted for cam- 
era parallax, and imported into Microsoft 
Excel. The eye and base of the tail were 
used to establish the orientation of the 
body from which the excursion of the wing 
tip was measured (Fig. 2). From lateral 
views, the excursion of the wing tip was 
calculated both one dimensionally (in the 
dorsoventral plane, following Dial et al. 
[1997]; Fig. 2A) and two dimensionally 
(dorsoventral and anteroposterior planes; 
Fig. 2B). Wing-tip excursion was used to 
estimate DPC excursion and, hence, 
change in pectoralis fiber length by the 
formula L = sin~^{h/b)r, where h is die 
wing-tip excursion (either one or two di- 
mensionally), b is wing length ineasured 
from the wing tip to the glenoid, and r is 

the distance from the glenoid to the DPC. 
For those wingbeats in which simulta- 
neous sonomicrometric and kinematic data 
were available (n = 20; magpie 1, n = 9; 
magpie 2, n = 7; magpie 3, n = 5), fascicle 
length changes obtained using the two 
methods were compared using reduced 
major axis regression. 


Sonomicrometry and strain-gauge data 
were obtained from magpies 2 and 3 over 
the entire speed range, whereas, because 
of equipment failure, ineasurements of 
magpie 1 were only obtained over speeds 
from to 8 m/s. 

As was the case in our earlier study 
(Dial et al, 1997), all three birds exiiibited 

Mechanical Power Output of Magpie Flight • Warrick et al. 


10 12 

Sonomicrometry (mm) 




16 T 

10 12 

Sonomicrometry (mm) 



Figure 2. Wing kinematics (insets) as measured from lateral video images (60 fields per second). The one-axis (dorsoventral 
excursion only) wingbeat amplitude (2A inset) was measured as in Dial et al. (1997), whiereas thie two-axis excursion (2B inset) 
also incorporated thie anteroposterior movement of the wing. (A) Correlation of one-axis kinematic estimate of fiber length change 
(i.e., deltopectoral crest excursion) with fiber length change as measured by sonomicrometry. (B) Correlation of two-axis kine- 
matic estimate of fiber length change with fiber length change as measured by sonomicrometry. 

the greatest muscle power output during 
hovering flight (Fig. 1). Minimum power 
speeds differed between the three birds, 
being 6 m/s, 4 in/s, and 10 m/s, respec- 
tively, for magpies 1, 2, and 3. However, 
the low power output of magpie 3 at 10 
m/s was probably due to a ceiling effect, 
because this bird insisted on flying close 
(<30 cm) to the ceiling of the test cham- 
ber. This likely resulted in the disruption 
of tip vortices and consequent reduction 

in drag, causing the low observed power 

Strain estimated from single-axis kine- 
inatics was positively correlated with strain 
ineasured for the same muscle contraction 
using sonomicrometry. Although signifi- 
cant, the correlation was not strong (r- = 
0.36; ij = 0.98x- + 0.78; F = 0.003, df = 
18; Fig. 2A). Fiber length changes esti- 
inated from two-axis kinematics and son- 
omicrometry exliibited a stronger correla- 

262 Btilletin Museum of Comparative Zoology, Vol. 156, No. 1 

tion (r2 = 0.52, ij = 1.02.T + 2.4; P < 
0.0001, df = 18; Fig. 2B). 

Our sonomicrometric measurements of 
fascicle strain reported here were 10% 
greater than those estimated in our pre- 
vious study of magpie flight based on 
wingbeat kinematics (Dial et al., 1997). 
Across all speeds, mean fascicle strain (L/ 
L/lOO) obtained from sonomicrometry av- 
eraged 27.6% for magpie 1, 36.2% for 
magpie 2, and 35.0% for magpie 3. The 
greatest fascicle strain was recorded dur- 
ing hovering flight for magpies 1 and 3 
(32.2 ± 1.1% and 46.0 ± 0.9%, respec- 
tively), whereas magpie 2 exliibited the 
greatest strain at 14 m/s (43.5 ± 0.3% 

Averaging the relative muscle power 
among birds (by first normalizing each in- 
dividual's uncalibrated relative power out- 
put at each speed to the maximum for that 
individual) yielded relative power curves 
similar to those that we previously report- 
ed (Figs. 4A, B). When the maximum rel- 
ative power obtained for magpies in this 
study is set to the maximum power (W) 
reported in Dial et al. (1997), Figure 4B 
reveals a similar shape for the power 
curves derived from the two approaches 
(sonometric versus kinematic). This is par- 
ticularly the case over the intermediate 
speed range (8-12 m/s), where a marked 
increase in mean power output at 8 m/s is 
observed in both sets of data. In the pre- 
sent study two of the three birds exliibited 
this increase, whereas power increased at 
8 m/s for all three individuals in the pre- 
vious study. 


Our new results for magpies, obtained 
from direct measurements of muscle 
length change using sonomicrometry, but 
relying on a relative estimate of pectoralis 
force and, thus, power output, generally 
agree well with the measurements of me- 
chanical power output that we obtained in 
our previous study using kinematic esti- 
mates of muscle length change (Dial et al., 
1997). For both data sets, power output 

was highest during hovering, dropped rap- 
idly (—40%) at 2 m/s, reached minimum 
values at intermediate speeds, and then in- 
creased again modestly (—20%) at the fast- 
est speeds (12 and 14 m/s). As was the case 
in our earlier study (Dial et al., 1997), min- 
imum power speed varied among birds (6, 
4, and 10 m/s for magpies 1, 2, and 3, re- 
spectively), and no bird exliibited mini- 
mum power at 8 m/s (Fig. 3A). Conse- 
quently, our sonomicrometry measure- 
ments provide support for the use of ki- 
nematic data (accounting for both humeral 
elevation— depression and protraction— re- 
traction) to estimate muscle length change 
and, when combined with DPC strain 
measurements, to calculate muscle work 
and power output. 

Given that all three birds possessed fair- 
ly uniform wing and tail dimensions, the 
variance in power curves among birds was 
most likely due to differences in flight be- 
havior. As noted above, differences in the 
birds' flight positions in the test section of 
the wind tunnel also likely affected our es- 
timated power output. For example, the 
low power obtained at 10 m/s for magpie 
1 may well have resulted from its flying in 
close proximity to the ceiling of the tunnel. 
Individual differences may also reflect dif- 
fering behavioral reactions that the birds 
exliibited having to fly at a fairly uniform 
speed because of the space restrictions of 
the wind tunnel. Magpies (or any bird, for 
that matter) rarely fly at steady speeds in 
the wild, instead exliibiting gait modula- 
tion and intermittent flight; when flying in 
a wind tunnel, magpies adopt a more reg- 
ular pattern of gait modulation than they 
exliibit outdoors (Tobalske et al., 1997). It 
is likely that some birds are more adept 
(i.e., capable of assuming body postures 
and wing presentations that reduce their 
flight power requirement) than others at 
flying with greater modulation of their 
normal flight pattern. Nevertheless, our 
findings of a fairly flat power curve over 
much of this species' speed range are gen- 
erally consistent with aerodynamic theoiy 
(Rayner, 1999; Fig. 4B; Appendix), existing 

Mechanical Power Output of Magpie Flight • Warrick et al. 




10 12 


Flight speed (m s' ) 
I Sonomicrometry D Kinematic (Dial et al. '97) 

Magpie 3 sonomicrometry vs 
Dial et al. '97 Kinematic 

8 10 12 


Flight speed (m s' ) 

 Sonomicrometry D Kinematic (Dial et al '97) 

Figure 3. (A) Mean percent (standard deviations) pectoralis muscle strain (change in fiber length/resting fiber length x 100) of 
the three magpies over the range of flight speeds. (B) Mean pectoralis muscle strain for magpie 3 compared with pectoralis 
muscle strain estimated from distal wing kinematics from our previous study. 

metabolic data (Ellington, 1991), and our 
earlier findings based on force measure- 
ments and wing kinematics. 

Despite the general consistency of our 
measurements of pectoralis length change 
with those based previously on wing ki- 
nematics, sonomicrometry has a clear ad- 

vantage by providing a direct measure of 
the lengthening and shortening of the 
muscle's fascicles relative to their resting 
length. This enables an assessment of the 
fractional length change of the muscle as- 
sociated with active lengthening versus ac- 
tive shortening. Consistent with earlier ki- 

264 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Relative "Power" 

0.5 - 





Flight speed (m s' ) 

Magpie 1 
"Magpie 3 

- Magpie 2 



10 12 14 16 18 20 22 24 26 28 

Flight speed (m s" ) 
•Sonomicrometry "=-^"Dial et al. '97 

Theoretical, C= 0.05 "" ~ Theoretical, C= 0.4 

Figure 4. (A) Mean relative power (mean uncalibrated power for a bird at a flight speed normalized to the bird's maximum 
power output speed, which was m/s In all cases) of all three magpies over the range of flight speeds. (B) Mean power output 
of magpies flying over the same range of speeds from Dial et al. (1997) and the present study. For purposes of comparing the 
shapes of the power curves, the uncalibrated relative power at hovering was set to equal the mean power at hovering from Dial 
et al. (1997). The theoretical power curve was derived from modified momentum jet theory, using a Q coefficient of either 0.4 
(dashed line upper; Pennycuick, 1989) or 0.05 (dashed line lower; Pennycuick et al., 1996) and an Induced power factor of 1.2 
(see Appendix). Note that although magpies seem to possess sufficient power to fly at speeds well above 20 m/s, their top 
sustainable speed In the wind tunnel was 14 m/s. 

nematic estimates (Dial and Biewener, 
1993) and recent sonomicrometiy mea- 
surements of length changes in pigeons 
(Biewener et al., 1998), we find that pec- 
toralis force development during the 
downstroke of magpies is achieved over a 

surprisingly large range of muscle length 
change (35—50% of resting length, over the 
range of recorded speeds). Moreover, this 
is achieved primarily by the muscle being 
lengthened (mean: 28%) versus being 
shortened (inean: 10%) relative to its rest- 

Mechanical Power Output of Magpie Flight • Warrick et al. 265 

ing length. This large length change is component). For lift to be produced, a 

clearly linked to the pectoralis's function in portion of the wing inust have a positive 

generating considerable mechanical power angle of attack relative to incident air. If, 

by moving the wing through a broad range at faster speeds, the angular velocity of the 

(approximately 80—110°) during the down- wing during downstroke is not sufficient to 

stroke. maintain a positive angle of attack while 

Classical aerodynamic theory predicts a directing the net aerodynamic force for- 
U-shaped power curve for any flying ob- ward (i.e., providing thrust), the animal 
ject. Under steady-state conditions, the will be unable to sustain flight, 
underlying physics of this relationship is It is noteworthy that the empirical curve 
inevitable. However, one conclusion sug- of inechanical power was generally similar 
gested by analysis of our empirical data to predicted curves at slow and interme- 
relative to theoretical power curves (Fig. diate speeds (Fig. 3B), particularly given 
4B) is that magpies may not be able to fly that we used published values for induced 
fast enough to achieve the speeds neces- power factor and drag coefficients in pre- 
sary to incur a significant rise in power re- dieting the mechanical power required for 
quirement, such as a U-shaped curve flight in die magpie (Appendix). We as- 
would predict. Our ineasurements of me- sumed an induced power factor of 1.2, 
chanical power output at and 2 m/s which gives an estimate of induced power 
might seem to suggest that magpies should during hovering that is approximately 30% 
be able to attain speeds of up to 18-28 m/ lower than that that would result from vor- 
s (Fig. 3B). However, the greater power tex theory (Rayner, 1979; Ellington, 1984). 
output achieved at these very low speeds However, a magpie flying in the closed- 
likely reflects nonsustainable flight perfor- section flight chamber could have enjoyed 
mance and a significant anaerobic energy a 27—31% reduction in power costs be- 
supply. Consistent with this interpretation, cause of lift recirculation (Rayner, 1994; 
we found that well-trained birds (before Tobalske and Dial, 1996). Thus, it would 
surgery) rarely achieved, and could never be unwise to use our data to address the 
sustain, 16 m/s in the tunnel. This is also validity of inomentum jet theory versus 
consistent with observations of the maxi- vortex theory as applied to slow flight in 
mum flight speeds recorded for magpies birds. Further research into inechanical 
in the field being only 11 m/s (Tobalske et power output during free flight, in the ab- 
al., 1997), as well as our observation of a sence of ground effects, therefore, would 
low variance in power output at higher be worthwhile. 

speeds (Dial et al., 1997). We have argued At fast speeds, our empirical data are 

tliat decreased variation in wingbeat kine- less than predicted values obtained from a 

matics and power output suggests a more model using a coefficient of body drag = 

constrained flight style as an animal ap- 0.4 (Pennycuick, 1989) and greater than 

proaches its limit of sustainable perfor- those from a more recent model using a 

mance. Consequently, we interpret 14 m/ coefficient of body drag = 0.05 (Penny- 

s as this species' maximum sustained flight cuick et al., 1996). Assuming our estimates 

speed. of induced and profile power are correct. 

An alternative explanation for the ob- this comparison suggests that the coeffi- 

served maximum flight speed may be that cient of body drag for the magpie should 

the inagpies were unable to generate suf- be somewhere between 0.4 and 0.05, and 

ficient thrust at speeds greater than 14 m/ that the magpie has a relatively "dirty" or 

s. A bird engaged in flapping flight gen- nonstreamlined shape compared to soine 

crates lift with its wings that provides both bird species (Pennycuick et al., 1996). 

weight support (vertical component of With regard to predictions of minimum 

aerodynamic force) and thrust (horizontal cost of transport (work per unit distance. 

266 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

or power divided by flight speed), which is 
generally expected by ecologists and oth- 
ers to define maximum range speed, or i;,,,^, 
it can be inferred from Figure 4B that pre- 
dicted u,„r would differ greatly depending 
upon parasite drag. Using our empirical 
data, we calculate u^ir to be 12 m/s. In con- 
trast, v„^^ was 10 m/s using a coefficient of 
body drag of 0.4 and 18 m/s using a co- 
efficient of body drag of 0.05. This brief 
analysis shows that a specific prediction of 
t;,^r for a given bird species should be re- 
garded with caution, if not distrust, unless 
direct measures of mechanical or meta- 
bolic power are used to support the pre- 


K. P. D. and A. A. B. would like to thank 
A. W. "Fuzz" Crompton for his input and 
extraordinarily positive attitude toward our 
research efforts and life in general. We 
thank F. A. Jenkins, Jr., for the invitation 
to the Crompton Symposium and for the 
review by his editorial staff of this manu- 
script for publication. This project was 
funded by NSF-IBN-9507503 to K. P. D. 


Berger, a. J., J. S. Hart, and O. Z. Roy. 1970. 
Respiration, oxygen consumption and heart rate 
in some birds during rest and flight. Zeitschrift 
fiir Vergleichende Physiologie, 66: 201-214. 

BiEWENER, A. A., W. R. Corning, and B. W. To- 
BALSKE. 1998. In vivo pectoraHs muscle force — 
length behavior during level flight in pigeons 
(Colli inba livia). Journal of Experimental Biolo- 
gy, 201: 3293-3307. 

Butler, P J., N. H. West, and D. R. Jones. 1977. 
Respiratoiy and cardiovascular responses of the 
pigeon to sustained level flight in a wind tunnel. 
Journal of Experimental Biolog)', 71:7-26. 

Dial, K. R, and A. A. Biewener. 1993. Pectoralis 
muscle force and power output during different 
modes of flight in pigeons (Cohimba livia). Jour- 
nal of E.xjDerimental Biology, 176: 31-54. 

Dial, K. P., A. A. Biewener, B. W. Tobalske, and 
D. R. Warrick. 1997. Mechanical power output 
of bird flight. Nature, 390: 67-70. 

Ellington, C. P. 1984. The aerodynamics of hov- 
ering insect flight. V. A vortex theory. Philosoph- 
ical Transactions of the Royal Society of London, 
Series B, Biological Sciences, 305: 115-144. 

. 1991. Limitations on animal flight perfor- 

mance. Journal of Experimental Biology, 160: 

Flint, E. N., and K. P. Nagy. 1984. Fhght energet- 
ics of free-living sooty terns. Auk, 101: 288-294. 

Hails, C. J. 1979. A comparison of flight energetics 
in hirundines and other birds. Comparative Bio- 
chemistry and Physiology A, 6: 581-585. 

Hedenstrom, a., and T. Alerstam. 1995. Optimal 
flight speed of birds. Philosophical Transactions 
of the Royal Society of London, Series B, Bio- 
logical Sciences, 48: 471-487. 

NORBERG, U. M. 1990. Vertebrate Flight. Berlin: 
Springer-Verlag. 291 pp. 

PennycuicK, C. J. 1975. Mechanics of flight, pp-l- 
75. In D. S. Famer and J. R. King (eds.). Avian 
Biology 5. London: Academic Press, xxii + 523 

. 1989. Bird Flight Performance: A Practical 

Calculation Manual. Oxford, United Kingdom: 
Oxford University Press. 153 pp. 

1997. Actual and "optimum" flight speeds: 

field data reassessed. Journal of Experimental Bi- 
ology, 200: 2355-2361. 

Pennycuick, C. J., M. Klaassen, a. Kvist, and a. 
LiNDSTROM. 1996. Wingbeat frequency and the 
body drag anomaly: wind tunnel observations on 
a thrush nightingale (Luscinia luscinia) and a teal 
{Anas crecca). Journal of Experimental Biology, 
199: 2757-2765. 

Rayner, J. M. V. 1979. A new approach to animal 
flight mechanics. Journal of Experimental Biol- 
ogy, 80: 17-54. 

. 1994. Aerodynamic corrections for the flight 

of birds and bats in wind tunnels. Journal of Zo- 
ology, London, 234: 537-563. 

. 1999. Estimating power curves for flying ver- 
tebrates. Journal of Experimental Biology, 202: 

Rothe, H.-J., W. Biesel, and W Nachtigall. 
1987. Pigeon flight in a wind tunnel. II. Gas ex- 
change and power requirements. Journal of 
Comparative Physiology B, 157: 99-109. 

Shigeoka, C. 1999. Regional muscle activity and 
contraction dynamics of the a\dan pectoralis dur- 
ing flight. M.S. thesis. University of Montana, 
Missoula. 83 pp. 

Speakman, J. R., S. Ward, U. Moller, D. M. Jack- 
son, J. M. V. Rayner, and W Nachtigall. 
1997. Thermography: a novel method for mea- 
suring the energy cost of flight? Journal of Mor- 
phology, 232: 326. 

Thomas, A. L. R., and A. Hedenstrom. 1998. The 
optimum flight speeds of flying animals. Journal 
of Avian Biology, 29: 469^77. 

Tobalske, B. W, and K. P Dial. 1996. Flight ki- 
nematics of black-billed magpies and pigeons 
over a wide range of speeds. Journal of Experi- 
mental Biology, 199: 263-280. 

Tobalske, B. W, N. E. Olson, and K. P Dial. 
1997. Flight style of the black-billed magpie: var- 
iation in wing kinematics, neuromuscular con- 

Mechanical Power Output of Magpie Flight • Warrick et al. 267 

trol, and muscle composition. Journal of Exper- 
imental Zoolog>', 279: 313-329. 

The metabolic cost of flight in unrestrained 
birds. Journal of Experimental Biology, 75: 223- 

Tucker, V. A. 1968. Respiratory exchange and evap- 
orative water loss in the flying budgerigar. Jour- 
nal of Experimental Biology, 48: 67-87. 

. 1972. Metabolism during flight in the laugh- 
ing gull. Lams atriciUa. American Journal of 
Physiology, 222: 237-245. 

Van den Berg, C, and J. M. V. Ravner. 1995. The 
moment of inertia of bird wings and the inertial 
power requirement for flapping flight. Journal of 
E.xperimental Biolog)', 198: 1655-1664. 

Ward, W., U. Moeller, J. M. V. Rayner, D. M. 
Jackson, W. Nachtigall, and J. R. Speakman, 
1997. Metabolic power requirement for starling 
Stunius vulgaris flight. Journal of Morphology, 
232: 338. 

Welham, C. V. J., AND R. C. Ydenberg. 1993. Ef- 
ficiency-maximizing flight speeds in parent black 
terns. Ecology, 74: 1893-1901. 



Numerous assumptions are intrinsic to 
mathematical models of mechanical power 
in animal flight, and it is beyond the scope 
of this paper to compare the family of 
curves that could be generated using dif- 
ferent aerodynamic theories (Pennycuick, 
1975; Rayner, 1979; Ellington, 1984) and 
their associated estiinates of lift and drag 
on a body and flapping wings. As an alter- 
native, we modeled aerodynamic power 
using equations and simplifying assump- 
tions in Norberg (1990), which were mod- 
ified froin the work of Pennycuick (1975) 
and Rayner (1979). Total power was cal- 
culated as the sum of induced, parasite, 
and profile power; thus, we neglected in- 
ertial power (Van den Berg and Rayner, 
1995) and ventilation and circulation fac- 
tors (Pennycuick, 1975). We selected an 
induced power factor of 1.2 (Pennycuick, 
1975, 1989) even though induced power 
factors of 1.5 or 1.7 are predicted from 
vortex theory (Rayner, 1979; Ellington, 
1984). For drag coefficients on the body 
and wings, we employed values from pub- 
lished sources (Pennycuick, 1975, 1989; 

Pennycuick et al, 1996; Rayner, 1979), 
and we treated the body as a flat plate 
(Pennycuick, 1975) rather than a tilted-cyl- 
inder (Rayner, 1979). Our predicted me- 
chanical power curves should not be in- 
terpreted to represent the only, or best, 
curves that can be synthesized from exist- 
ing theory. 

Symbols and assumed values: 

b = wing span = 0.573 m 

Cb = drag coefficient for body, 0.4 (Pen- 
nycuick, 1989) or 0.05 (Pennycuick 
et al, 1996) 

C,^ = drag coefficent for wdngs, 0.02 (Ray- 
ner, 1979) 

g = gravitational acceleration, 9.805 m 

k = induced power factor, 1.2 (Penny- 
cuick, 1975, 1989) 

m = body inass = 0.174 kg 

Pi = induced power during forward 

Fji, = induced power during hovering 

Fp^ = parasite power 

^pro ~ profile power, assuming constant 
wing area and mean resultant veloc- 
ity (Norberg, 1990). 

P{ = total mechanical power during for- 
ward flight 

Fi, = total mechanical power during hov- 

S = coinbined surface area of both 
wings = 0.064 m"- 

Sb = frontal area of body 0.00238 m'^ 
(Pennycuick, 1989) 

Sj = disk area, 0.2578 m^ 

S, = strip area at distance r from wing 

T = wingbeat duration, s, assumed 
wingbeat frequency of 7.5 Hz 

t = proportion of wingbeat in down- 
stroke (assuined 0.55) 

V = flight velocity 

V, = induced velocity 

V^ = resultant velocity on wing at strip 
distance r from wing root 

W = body weight, 1.706 N 

(f) = wing amplitude in radians, assumed 
to be 1.57 ( = 90°) 

268 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

p = air density, 1.115 kg m ^. To calculate induced, parasite, and profile 


To calculate total mechanical power dur- p _ uti/2/o .c /cr\ 

ing hovering flight: ' 

PiH = \^wy{2pv,s,r' (6) 

Fk = P.H + Pp. + Pp. (3) ^^^ ^ 0.5pt;3S,C, (7) 

And during forward flight: ^J^ 

P,ro = 2 [(1-77 - 10-3) pb'^sp.yr'T^] 

r = 0.01 

P,= P, + F,,„ + P,„ (4) (8) 



Abstract. This paper tests the hypothesis that cor- 
tical bone growth (modeUng) and repair (Haversian 
remodeling) responses to exercise-induced mechani- 
cal loading vary according to loading and position 
within the skeleton. Higher rates of modeling and 
Haversian remodeling are predicted to occur in re- 
sponse to loading, but a trade-off is predicted be- 
tween modeling and Haversian remodeling, with pro- 
portionately higher Haversian remodeling rates at 
distal than proximal midshafts, and proportionately 
higher modeling rates at proximal than distal mid- 
shafts. The hypothesis is tested with cross-sectional 
and histologic data from juvenile sheep (Ovis aries) 
who trotted at low speed (4 km/h) for 60 min/d on a 
treadmill for 90 days, compared with sedentary con- 
trols. Exercised sheep had higher periosteal modeling 
and Haversian remodeling rates than controls. In 
both groups, midshaft periosteal growth rates were 
higher in proximal than distal elements in inverse 
proportion to the area-normalized inertial cost of ac- 
celerating mass; midshaft Haversian remodeling rates 
were higher in distal than proximal elements in pro- 
portion to the same energetic cost. The results sug- 
gest that growing animals modulate modeling versus 
remodeling responses to loading at different skeletal 
locations in order to optimize cross-sectional strength 
relative to the kinetic energy cost of accelerating add- 
ed mass. 


This paper tests a hypothesis initially 
proposed by Lieberman and Crompton 
(1998) about the processes by which bones 
optimize in vivo responses to mechanical 
loading. The observation that bones adjust 
dynamically to their functional environ- 
ment (Wolffs law) has been well substan- 
tiated over the last 100 years despite a 
poor understanding of the high degree of 

' Department of Anthropology, Harvard University, 
Cambridge, Massachusetts 02138. 

- Department of Anthropology, University of New 
Mexico, Albuquerque, New Mexico 87131. 

variability that characterizes osteogenic re- 
sponses to mechanical loading (see Lanyon 
and Rubin, 1985; Bertram and Swartz, 
1991). Such variations are of special inter- 
est for understanding the general trend 
among most cursorially adapted mammals 
for distal limb elements to have smaller di- 
ameters than proximal elements, which 
gives the limb a tapered shape (Smith and 
Savage, 1956; Alexander, 1980, 1996; Cur- 
rey, 1984). Limb tapering is almost cer- 
tainly an adaptation to minimize the ki- 
netic energetic cost of limb acceleration. 
At a given angular velocity, this cost for a 
limb segment is proportional to the prod- 
uct of its mass and the square of its mo- 
ment arm (Hildebrand, 1985). However, 
limb tapering has a structural cost. By re- 
ducing bone mass, limb tapering decreases 
the second moment area (/) available to 
counteract bending forces that may ac- 
count for approximately 75-95% of mid- 
shaft strains (Bertram and Biewener, 
1988). If applied forces are similar in prox- 
imal and distal elements, then tapered dis- 
tal elements will experience higher strains 
than proximal elements, potentially lead- 
ing to the generation and propagation of 
microcracks and other structural damage 
that ultimately contribute to mechanical 
failure (Currey, 1970, 1984; Carter et al., 
1981a,b; Martin and Burr, 1982, 1989; 
Burr et al, 1985; Burr, 1993; Mori and 
Burr, 1993). As a result, distal elements 
may have lower safety factors than proxi- 
mal elements (Currey, 1984; Alexander, 

Some mammals, most notably equids, 
have evolved structural adaptations to 

Bull. Mus. Comp. Zool., 156(1): 269-282, October, 2001 269 

270 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

avoid the potentially higher stresses that 
may result from limb tapering, including 
orientating distal elements in line with 
ground reaction forces to decrease bend- 
ing (Gambaiyan, 1974; McMahon, 1975; 
Biewener, 1983a, 1989, 1990; Biewener et 
al., 1988), shortening of distal elements to 
reduce their bending moments (Gambar- 
yan, 1974; Alexander, 1977), and de- 
creased curvature of distal elements to 
minimize compressive bending and buck- 
ling (Biewener, 1983a,b; Pauwels, 1980; 
Bertram and Biewener, 1992). However, 
many cursorially adapted mammals, in- 
cluding most ungulates, are characterized 
by tapered limbs with long distal elements 
that have high excursion angles (Gambar- 
yan, 1974). These species are predicted to 
experience high functional strains in their 
distal elements and must compensate 
through other means. 

The two most important in vivo pro- 
cesses by which distal and proximal ele- 
ments can adapt differentially to functional 
loading are growth (referred to as model- 
ing) and repair (Haversian remodeling). 
Modeling responses are the best studied. 
Modeling increases resistance to bending 
stresses by augmenting I so that a given 
force generates less strain (Wainwright et 
al., 1976). Bones model in response to me- 
chanical loads through increases in peri- 
osteal apposition (Chamay and Tchantz, 
1972; Goodship et al., 1979; Lanyon et al., 
1982; Lanyon and Rubin, 1984; Rubin and 
Lanyon, 1984a,b, 1985; Biewener et al., 
1986; Raab et al, 1991), and through in- 
hibition of endosteal resorption (Woo et 
al., 1981; Ruff et al, 1994). The optimal 
way for bones to minimize mass and max- 
imize / (which is a fourth-power function 
of bone radius), is to add mass periosteally 
and remove mass endosteally. Because 
marrow is 50% as dense as bone, the op- 
timum diameter to thickness (D/t) ratio in 
mammals is predicted to be 4.6 (Alexan- 
der, 1981; Currey, 1984). One study of ter- 
restrial mammals (Currey, 1984: 109-111) 
found the median D/t ratio to be approx- 
imately 4.4 (albeit with considerable vari- 

ation), with a higher median value for the 
femur (5.6) and correspondingly lower val- 
ues for the humerus and other more distal 
limb elements. 

The other mechanism by which bones 
adapt to high functional loads is to in- 
crease the frequency of bone repair re- 
sponses through Haversian remodeling. 
Haversian remodeling is a sequential ac- 
tivation of vascularborne osteoclasts on a 
resorption surface that cuts a channel 
through old bone, followed by circumla- 
mellar deposition of new bone by osteo- 
blasts around a central neurovascular 
channel (Frost, 1963; Martin and Burr, 
1989). Haversian remodeling was once 
thought to be a mechanism for maintain- 
ing calcium homeostasis (de Ricqles et al., 
1991), but most calcium exchange occurs 
in osteocyte canaliculi and through remod- 
eling of trabecular bone (Parfitt, 1988b). 
Instead, Haversian remodeling in cortical 
bone is probably an adaptation to strength- 
en bone by removing weakened tissue or 
reorienting its structure (Currey, 1970, 
1984; Carter and Hayes 1976a,b, 1977a,b; 
Carter et al., 1981a,b; Martin and Burr, 
1982; Schaffler et al, 1989, 1990). Al- 
though secondary osteonal (Haversian) 
bone is weaker in vitro than young primary 
osteonal bone (Currey, 1959; Carter et al., 
1976; Carter and Hayes 1977a,b; Vincen- 
telli and Grigorov, 1985; Schaffler and 
Burr, 1988), secondary osteonal bone is 
stronger than microcrack-damaged prima- 
ry bone (Schaffler et al, 1989, 1990). Hav- 
ersian remodeling may also increase elas- 
ticity and halt microfracture propagation 
(Currey, 1984). Several studies demon- 
strated that remodeling preferentially oc- 
curs in older regions of bones that have 
presumably accumulated more damage 
than younger bone (Frost, 1973; Bouvier 
and Hylander, 1981; Currey, 1984), and 
that loading significantly increases remod- 
eling rates (Hert et al., 1972; Bouvier and 
Hylander, 1981, 1996; Burr et al, 1985; 
Schaffler and Burr, 1988; Mori and Burr, 
1993; Lieberman and Crompton, 1998). 
However, the effectiveness of Haversian 

Trade-Off Responses to Loading in the Mammalian Limb • Lieberman and Pearson 271 

Kinetic energy 

Metabolic energy 

Radius of Gyration 
Proximal ► Distal 

Figure 1 . Predicted variation in kinetic versus metabolic en- 
ergy costs of adding mass as a function of moment arm length 
(radius of gyration). Kinetic energy costs increase exponen- 
tially as a function of moment arm, but ttie metabolic energy 
cost of growing or repairing bone is predicted to be constant. 

remodeling for counteracting functional 
strains has three major constraints. The 
basic multicellular unit through which 
Haversian remodeling occurs requires 
blood supply and cannot take place in 
avascular bone (de Ricqles et al, 1991). 
Haversian remodeling also weakens bone 
by increasing porosity, although presum- 
ably less so than the effects of microfrac- 
ture accumulation (Parfitt, 1988a; Schaf- 
fler and Burr, 1988; Martin, 1995). Finally, 
Haversian remodeling is an ineffective 
short-term solution to high in vivo strains 
because the remodeling sequence time of 
a basic multicellular unit is approximately 
30-40 days from activation to termination 
(Martin and Burr, 1989). 

Trade-Off Model 

Modeling and remodeling are the most 
labile osteogenic responses to mechanical 
loading that generate phenotypically plas- 
tic variations in bone shape and strength. 
However, the effects of modeling and re- 
modeling on limb bones are predicted to 
vary because of different costs and bene- 
fits, the most important of which are the 
energetic costs of moving bone tissue and 
the metabolic costs of growing new bone 
tissue (illustrated in Fig. 1). The dominant 
long-term cost of adding bone mass 
through modeling is most likely the kinetic 

energy E^,) of accelerating and decelerating 
the added mass. During the swing phase, 
limbs function like pendulums in which E^^ 
increases exponentially in more distal 
bones as a function of m^Rf, where m^ is 
the mass of a given bone, i, and R^ is its 
radius of gyration (Hildebrand, 1985). In 
particular, limb acceleration becomes in- 
creasingly costly as the frequency of loco- 
motion diverges inore from the natural 
frequency of oscillation of the limb in 
which gravitational potential energy at the 
beginning of swing phase is converted to 
kinetic energy at midswing and then back 
to gravitational potential energy at the end 
of the swing phase. Therefore, the ener- 
getic costs of weight increments are influ- 
enced both by the location of added mass 
and by the divergence of stride frequency 
from the natural oscillation frequency of 
the limb. Thus, at a given speed, modeling 
responses to strain are more expensive (to 
the second power) in terms of kinetic en- 
ergy in distal than in proximal elements 
(Myers and Steudel, 1985; see also Taylor 
et al, 1974). Such costs may be consider- 
able in some species. Taylor and col- 
leagues estimate that approximately 66% 
of E^. required for a horse to nm at 15 m/ 
s is spent to accelerate and decelerate the 
three distal limb segments, whose mass is 
approximately 80% bone tissue (Fedak et 
al., 1982, Heglund et al, 1982a,b). 

An additional cost of modeling and re- 
modeling is the metabolic energy cost (£,„) 
of growing and repairing bone tissue. Al- 
though cortical bone is probably expensive 
to synthesize, £,„ for both primary and 
Haversian bone is likely to be constant re- 
gardless of skeletal location (see Fig. 1). 
However, Haversian remodeling is pre- 
dicted to incur long-term higher metabolic 
costs than modeling because it leaves a 
bone insufficiently strong to resist further 
strain damage; in such cases. Haversian re- 
modeling must reoccur regularly. There- 
fore, where mass and size do not impose 
high kinetic energetic costs, changes in 
cross-sectional geometry generated by 
modeling are probably more effective re- 

272 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

sponses to loading than Haversian remod- 
eling because they leave the bone perma- 
nently stronger and avoid any potential in- 
creases in porosity that Haversian remod- 
eling causes. Further research is necessary 
to establish the inetabolic costs of cortical 
bone growth and turnover. 

On the basis of the above-defined costs 
and benefits, Lieberman and Crompton 
(1998) proposed a trade-off model to pre- 
dict variations in modeling and Haversian 
remodeling rates in response to mechani- 
cal loading with respect to the nature of 
the applied forces (exercise) and skeletal 
location. In order to compare functionally 
comparable (hoinotypic) sites, the model 
focuses solely on bone midshafts, which 
are usually the location of highest bending 
forces (Biewener et al., 1986; Biewener 
and Taylor, 1986). Although both modeling 
and Haversian remodeling rates at bone 
midshafts are predicted to be higher in an- 
imals whose limbs experience more load- 
ing, the relative frequency of modeling 
and Haversian remodeling rates are ex- 
pected to vary as a function of their radius 
of gyration (R; see Fig. 2). Because the 
energetic cost of modeling increases ex- 
ponentially with increasing R, but the en- 
ergetic cost of remodeling is hypothesized 
to remain constant, midshafts of more 
proximal elements are predicted to re- 
spond to applied forces with proportion- 
ately higher modeling rates than midshafts 
of more distal elements. In contrast, mid- 
shafts of more distal elements are predict- 
ed to respond to applied forces with pro- 
portionately higher Haversian remodeling 
rates than midshafts of more proximal el- 

Although Lieberman and Croinpton 
(1998) provided preliminary data in sup- 
port of the trade-off model, their sample 
size was small and their analysis did not 
include kinematic or kinetic data. This 
study tests the model more completely by 
comparing the cross-sectional geometry 
and histology of the midshaft femur, tibia, 
and metatarsal in a larger sample of exer- 
cised and sedentary control subjects. Two 

specific hypotheses are tested with kine- 
matic, kinetic, and morphometric data. 
First, osteogenic responses to strains are 
predicted to vary according to loading and 
to position within the skeleton. In partic- 
ular, higher rates of both modeling and 
Haversian remodeling are predicted to oc- 
cur at homotypic sites in exercised animals 
than in matched controls. Second, a trade- 
off is predicted between modeling and 
Haversian remodeling responses, with pro- 
portionately higher Haversian remodeling 
rates at distal than at proximal midshafts, 
and proportionately higher modeling rates 
at proximal than at distal midshafts (stan- 
dardized by body mass and element 
length). Forthcoming publications will 
present data on variations in peak strains 
at distal and proximal midshafts, and dif- 
ferences in trade-off responses between 
juvenile, subadult, and adult subjects. 


Subjects and Exercise Training 

Eleven juvenile rams (Ovis aries; Dor- 
set) were divided into control (n = 5) and 
exercise (n = 6) groups. One animal in the 
exercise group (lamb 10) had a respiratory 
infection for 4 weeks and was excluded 
froin all calculations because it gained al- 
most no weight during that period. Sub- 
jects were 40 days old at the start of the 
experiment, which lasted 89 days. For 1 
week before the start of the experiment, 
the exercise group animals were habituat- 
ed to run in an enclosed box on a Mar- 
quette 1800 treadmill (GE Medical Sys- 
tems, Milwaukee, WI). Exercise group an- 
imals ran every day at a horizontal incli- 
nation for 60 minutes at a Froude speed 
of 0.5 (approximately 4 km/h), generating 
approximately 6,000 loading cycles per da) 
per limb. At this speed, which is well be- 
low maximum running speed, the sheep's 
gait is a slow trot. Subjects were housed in 
raised 1.0-m- cages, limiting additional 
loading to minor locoinotor activity and 
sedentary weight support. All subjects 
were fed the saine quantity of hay per day 

Trade-Off Responses to Loading in the Mammalian Limb • Liebennan and Pearson 273 



Radius of Gyration 

Proximal ' Distal 

Radius of Gyration 

Proximal ' Distal 

Figure 2. Predicted trade-off between periosteal growth 
(modeling) and Haversian remodeling (repair) rates in cortical 
bone as a function of the radius of gyration. 

and water ad libitum. Fluorescent dyes 
that incorporate into bone mineral were 
administered by intraperitoneal injection 
at the start of the experiment (calcein, 20 
mg/kg), on day 30 (oxytetracyline, 50 mg/ 
kg), and on day 63 (xylenol orange, 25 mg/ 
kg). Body mass was measured weekly. At 
the end of the experiment, total limb 
length from the greater trochanter of the 
femur to the ground was measured for 
each subject. The subjects were then eu- 
thanized, and their limbs bones were re- 
moved and defleshed. The radii of gyration 
from the femoro-acetabular joint to the 
midshaft of the femur, tibia, and metatar- 
sal were measured at simulated midstance 
using a plastic tape measure (accurate to 1 
mm) on each animal after death. Articular 
lengths of the femur, tibia, and metatarsal 
were measured using digital calipers (ac- 
curate to 0.01 mm). Femoral length was 
measured from the most proximal point on 
the femoral head to the line connecting 
the two distal condyles. Tibial length was 
measured from the center of the lateral 
condylar surface to the center of the distal 
articular surface. Metatarsal length was 
measured from the center of the proximal 
articular surface to the most distal point of 
the distal articular surface. 

Cross-Sectional Geometry and Histologic 

A 2-cm section was cut from the limb 
midshaft of each left femur, tibia, and 
metatarsal, and the inarrow was removed. 

After measuring their densities, sections 
were fixed in 100% ethanol, and then si- 
multaneously stained and dehydrated in a 
solution of 1% basic fuchsin in denatured 
ethanol under 20 mm Hg vacuum. Sample 
solution was changed every 24 hours for 1 
week. Samples were then embedded in 
poly-methyl methacrylate polymer (Osteo- 
bed®, Polysciences Inc, Warrington, 
Pennsylvania). Two sections were cut from 
each embedded midshaft using an Iso- 
met® 1000 low-speed saw (Buehler Ltd., 
Lake Bluff, Illinois), affixed to glass shdes 
with Epotek® 301 epoxy (Epoxy Technol- 
ogy Inc., Billerica, Massachusetts), ground 
to approximately 100 ixm thick using a 
Hillquist® 1005 thin-section machine 
(Hillquist Inc, Fall City, Washington), pol- 
ished using a Hillquist® 900 grinder, and 

Cross-sections were analyzed using an 
Olympus® SZH 10 stereozoom micro- 
scope with cross-polarized light and a 
Highlight® 3000 Fluorescence (Olympus 
America, Mellville, New York). Digitized 
images were captured using a Pro- Series® 
CCD technical video camera (Media Cy- 
bernetics, Silver Spring, Maryland) con- 
nected to a Macintosh® 4440 computer 
wdth a Scion® LG-3 capture board (Scion 
Corp., Frederick, Maryland). A version of 
NIH Image, v 1.61 (W. Rasband, National 
Institutes of Health) with a macro written 
by M. Warfel (Cornell University), was 
used to calculate cortical area, medullary 
area, the maximum and minimum second 
moment areas (Z^^„ I„,J, and the polar mo- 
ment of inertia (/). These measurements 
were averaged for the two sections from 
each inidshaft. 

Haversian systems were counted and 
measured using NIH Image. Only com- 
plete secondary osteons with reversal lines 
were counted. Haversian density was com- 
puted as the total number of complete sec- 
ondary osteons divided by cortical bone 
area. Linear periosteal growth was mea- 
sured along the anterior, posterior, medial, 
and lateral axes of each section as the dis- 
tance from the line of the calcein dye (ad- 


274 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

ministered on day 1 of the experiment, see 
above) to the periosteal surface. Periosteal 
modeling rate (PMR) was calculated as the 
average of periosteal growth from the four 
axes divided by the number of days in the 
experiment. Haversian density and PMR 
were averaged for the two sections from 
each midshaft. 

Kinematic Analysis 

To acquire data on limb element ori- 
entation and angular velocity, three sub- 
jects were recorded at the end of the ex- 
periment (after surgery in which strain 
gauges were attached to the limbs; strain 
data are reported in a forthcoming paper). 
The subjects were videotaped in lateral 
view using a SONY® DCR VX-1000 Han- 
dycam (Sony Corp., Tokyo, Japan) at 60 
fields/s. Subjects were recorded approxi- 
mately 4 and 24 hours after surgery, and 
ran with a normal gait (e.g., with full 
weight-bearing on all limbs and no signs 
of favoring one limb over another, showing 
no signs of lameness, distress, or discom- 
fort). At least five complete gait cycles 
were digitized for each animal at 4.0 km/ 
h. NIH Image was used to measure the 
orientation of the longitudinal axes of the 
femur, tibia, and metatarsal relative to the 
horizontal plane of the treadmill in each 
frame. These data were used to calculate 
angular velocity (w) following the method 
of Winter (1990). 


All cross-sectional dimensions were 
standardized by body mass (calculated as 
mean mass during the final 3 weeks), and 
second moment areas were standardized 
by both body mass and element length. In 
order to evaluate osteogenic responses to 
mechanical loading in terms of kinetic en- 
ergy costs, the area-normalized inertial 
cost (ANIC) of each midshaft cross-section 
(0 was estimated (following Winter, 1990) 
in joules as 

ANIC = 0.57,0)2 


7, = 7?,2.CAD 

in which co is the angular velocity in radi- 
ans, K, is the radius of gyration in meters, 
CA is cortical area expressed in m-, and D 
is bone density in kg/m^. Note that ANIC 
is not an estimate of the kinetic energy 
cost of accelerating the limb segment as a 
whole, but instead estimates the kinetic 
energy cost of accelerating the midshaft of 
each limb segment. 

Because of the small sample sizes, and 
to avoid assuming normal distribution of 
the data, all tests of significance between 
elements and between groups were cal- 
culated using nonparametric methods (in 
most cases, Wilcoxon two-sample test). 


Cross-Sectional Geometry 

Cross-sectional properties of the femur, 
tibia, and metatarsal in the exercised and 
control sheep are summarized in Table 1, 
along with data on body mass and element 
length at the end of the experiment used 
to standardize comparisons. Little differ- 
ence in element length and body mass 
(less than 1% in all cases) was found be- 
tween exercise and control groups. No sig- 
nificant effects were found of exercise on 
overall cross-sectional area and shape: 
mass-adjusted medullary area (MA), CA, 
and midshaft shape as measured by 7^^ 
7„,in did not differ significantly between 
groups; CA/kg in the tibia was nearly sig- 
nificant at the a = 0.05 level (F = 0.08). 
The value of 7 differed significantly be- 
tween exercise and control groups in cer- 
tain elements: distal elements had abso- 
lutely smaller and weaker midshafts than 
the femur, and the difference in 7 between 
exercise and control groups was more pro- 
nounced in the tibia and metatarsal than 
it was in the femur. When standardized for 
element length and body mass, 7,,,.^^, 7„,i,„ 
and/ were 25—29% higher in the exercised 
than in control groups in the tibia (P < 
0.05); 20% higher in the metatarsal (P = 
0.05-0.08); and 13-16% higher in the fe- 
mur (P = 0.17-0.25). 


Trade-Off Responses to Loading in the Mammalian Limb • Lieberman and Pearson 275 

Table 1. Standardized cross-sectional properties. 





% Difference t 




1 SD 


Mean ± 

1 so 

P (Wilcoxon) 

Body mass 







38.82 ± 






length (mm) 






174.8 ± 










2.73 ± 










3.57 ± 










0.66 ± 










0.49 ± 




*- max * min 






1.35 ± 










1.15 ± 





Articular length (mm) 






193.7 ± 










1.05 ± 










3.55 ± 










0.38 ± 










0.27 ± 




■* ina.v ■* min 






1.41 ± 










0.65 ± 






length (mm) 






135.8 ± 










1.01 ± 










2.66 ± 










0.30 ± 










0.27 ± 




■* ma.v ■* min 






1.14 ± 










0.57 ± 




° MA, mass-adjusted medullary area; CA, cortical area; /,„„, maximum second moment area; 
ength; 7^,„„ minimum second moment area; J, polar moment of inertia. 
t Calculated as [(exercised group mean — control mean)/control mean] X 100. 
\ Average of the last 3 weeks of body mass. 

L, articular 

[nergetic Costs 

The ANIC of each inidshaft cross-sec- 
don is useful for evaluating osteogenic re- 
sponses to mechanical loading in teriTis of 

nergetic costs (see above). This estiinate 
requires data on cortical area (see Table 1) 
as well as D, R,, and co for each eleinent 
^suminarized in Table 2). As Table 2 in- 
dicates, no significant differences existed 
in D between runners and controls for any 
slements. Peak co in the femur was consid- 

rably lower than that of the tibia and 
metatarsal, which were roughly siinilar. 
Maxiinum angular excursion was approxi- 
mately 47° for the femur, 68° for the tibia, 
and 62° for the metatarsal; these ranges 
are similar to those of other midsized un- 
gulates (e.g., cervids) but are considerably 

greater than those reported for larger cur- 
sorially specialized mammals such as hors- 
es (Gambaryan, 1974; Biewener et al., 
1988). The above-described differences in 
CO, combined with the substantial differ- 
ences in CA (see Table 1) and R, (Table 2) 
resulted in estimates of area normalized 
inertial cost that were roughly 10 times 
higher for the tibia than for the femur, and 
roughly twice as high in the metatarsal as 
in the tibia. No statistically significant dif- 
ferences in ANIC were detected between 
the exercise and control groups. 

Modeling and Haversian Remodeling 

Table 3 summarizes histologic data on 
average periosteal growth rates (PGR) in 
|xm/d for the femur, tibia, and inetatarsal. 

276 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Table 2. Kinematic and density data used to calculate inertial cost of midshaft cross sections 

Density {kg/m') 

Angular \'elocity (w, radiaus/s) 






Tibia Metatarsal 

Controls (n = 5) 
Exercised (n = 5) 

1,976 ± 66 
1,974 ± 104 

1,995 ± 83 
1,902 ± 119 

1,897 ± 152 
1,846 ± 186 

6.89 ± 1.3 

12.6 ± 1.8 11.2 ± 0.7 

" ANIC, area-normalized inertial cost, calculated as 0.05CAdensity-R,^-a)^ using average value of w for the 
three exercised animals. CA, cortical area; K,, radius of gyration of a given bone, i. 
t NA, not available. 

Analysis of these data indicated that PGR trols at the midshaft of the femur, tibia, 
is significantly higher in the exercised than and inetatarsal; these animals also had sig- 
the in control group for all three elements, nificantly higher secondaiy osteonal den- 
In addition, although PGR for the femur sities than controls at the midshaft of the 
is significantly higher (F < 0.05) than for tibia and metatarsal, but not the femur, 
the tibia in both groups, PGR is signifi- Therefore, this study supports the findings  
cantly higher in the tibia than in the meta- of previous studies that indicated that 
tarsal in the exercised (P = 0.02) but not growth (periosteal modeling) and repair 
in the controls {P = 0.31), largely because (Haversian remodeling) in cortical bone 
of the markedly higher tibial PGR in the occur in response to the functional strains 
exercised versus control groups. There- generated by luechanical loading (e.g., 
fore, modeling in response to exercise oc- Lanyon et al., 1982; Lanyon and Rubin, 
curs more rapidly at the midshaft in prox- 1984; Rubin and Lanyon, 1984a,b, 1985; 
imal versus distal elements of the limb in Burr et al., 1985; Biewener et al, 1986; 
proportion to the kinetic energy cost of ac- Raab et al, 1991; Bouvier and Hylander, 
celerating added mass, with an especially 1996). However, further research is nec- 
pronounced effect on the tibial midshaft, essary to establish the nature of the strain 
This relationship is illustrated in Figure signal that induces these responses (e.g., 
3 A, which shows that midshaft modeling magnitude, frequency), and the extent to 
rates in the hind limb are inversely pro- which osteogenic responses to mechanical 
portional to estimates of midshaft ANIC. loading change with age. These results ad- 
Analysis of data on Haversian remodeling ditionally support the trade-off model for 
rates in Table 3 also indicated that second- growing animals of Lieberman and 
ary osteon density is higher in distal than Crompton (1998). Cortical bone growth 
in proximal midshafts, with more pro- and repair mechanisms in juvenile sheep 
nounced differences between exercised vary inversely at homotypic sites as a func- 
and control groups in distal than in prox- tion of the kinetic energy cost of acceler- 
imal midshafts. Analysis of these results in- ating additional mass. In particular, mid- 
dicated that Haversian remodeling in re- shaft periosteal growth rates are signifi- 
sponse to exercise occurs at higher levels cantly higher in proximal than in distal el- 
at the midshaft in distal versus proximal ements in inverse proportion to the 
elements of the limb in proportion to es- estimated ANIC of accelerating mass, and 
timates of midshaft ANIC, as shown in midshaft Haversian remodefing rates are 
l^igure 3B. significantly higher in distal than in proxi- 
nicjpi iQQinM '^^^ elements in proportion to the same 

energetic cost. 
Growing sheep that exercised at mod- Optimization of modeling and remod- 
erate levels for 90 days had significantly eling responses in distal and proximal el- 
higher periosteal modeling rates than con- ements in growing animals is a probable 

Trade-Off Responses to Loading in the Mammalian Limb • Lieberman and Pearson 277 

Table 2. Extended. 

Radius of g)'ration (m) 

ANIC (joules)" 







0.09 ± 0.01 
0.09 ± 0.00 

0.20 ± 0.01 
0.20 ± 0.01 

0.33 ± 0.03 
0.33 ± 0.02 

0.05 ± 0.01 
0.05 ± 0.01 

0.20 ± 0.01 
0.19 ± 0.01 

0.33 ± 0.03 
0.32 ± 0.02 

adaptation for limb-bone tapering to en- 
able distal elements to remain lighter than 
proximal elements, thereby reducing the 
kinetic energy costs of locomotion (Hil- 
debrand, 1985). By modulating growth 
and repair responses to loading between 
proximal and distal elements, aniixials save 
kinetic energy, albeit at the expense of 
higher metabolic energy costs over the 
long term. Currently, no data exist on the 
metabolic cost of Haversian remodeling, 
but this cost is predicted to be less in the 
long term than the kinetic energy cost of 
accelerating additional distally located 
mass that might otherwise be necessary. 
One important limitation of the trade-off 
model tested here is that differential mod- 
eling versus Haversian reinodeling re- 
sponses to inechanical loading may not oc- 
cur in all regions of the skeleton or in tra- 
becular bone, and are likely to change with 
age. As aniinals inature, periosteal growth 
rates in response to a given strain stiinulus 
decline, although endosteal gro\vth inay 
increase, leading to stenosis (Woo et al., 
1981; Ruff et al, 1994), and Haversian re- 

modeling rates probably increase regard- 
less of skeletal location. These hypothe- 
sized effects can be tested by comparing 
the above results with modeling and re- 
modeling responses in adult sheep sub- 
jected to the same mechanical loads. 

The variable responses to mechanical 
loading documented here suggest that any 
relationships between loading and bone 
cross-sectional dimensions are more com- 
plex than is sometimes assumed. In partic- 
ular, the higher periosteal growth rates of 
the exercised juvenile sheep influenced 
overall midshaft I more than cortical or 
medullary areas (see Table 1), providing 
experimental support for coinparative 
studies that indicate tliat / and not cortical 
areas should be used to estimate biome- 
chanical adaptations from bone cross sec- 
tions (Jungers and Minns, 1979; Ruff and 
Hayes, 1983; Ruff, 1989; Ruff and Runes- 
tad, 1992; Lieberman, 1997). In addition, 
proximodistal differences in modeling ver- 
sus Haversian remodeling rates suggest 
tliat inferences about biomechanical ad- 
aptations to loading from cross-sectional 

Table 3. Histologic data. 

Element or variable 






Mean ± 1 SD 


Mean ± 1 SD 



11.03 ± 0.71 


13.92 ± 1.59 




6.99 ± 0.72 


11.04 ± 1.07 




6.43 ± 0.96 


8.28 ± 1.14 




0.42 ± 0.45 


0.51 ± 0.51 




2.34 ± 1.13 


4.67 ± 2.79 




7.89 ± 2.26 


16.31 ± 4.71 



Femur: average growth ratet (jjuin/d) 
Tibia: average growth ratet (|xm/d) 
Metatarsal: average growth ratef (jjuni/d) 
Femur osteon densityl 
Tibia osteon density* 
Metatarsal osteon density! 

" Calculated as [(exercised group mean — control mean)/control mean] X 100. 

t Average growtli rate (periosteal modeling rate) calculated as mean growth rate (|jLm/d) of anterior, pos- 
terior, medial, and lateral cortices. 

I Osteon density was calculated as the total number of secondary osteons in the entire section divided by 
the cortical area (mm^). 

278 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

geometry of the midshaft may be less re- 
hable in distal elements, especially the 
metatarsal, than in more proximal ele- 
ments. One unexpected result from this 
study is that / discriminates better be- 
tween exercise and control groups at the 
tibial midshaft than in the femur or meta- 
tarsal. The greater percent difference be- 
tween treatment groups in the tibia than 
in the femur or metatarsal may be a func- 
tion of different local muscle forces, dif- 
ferences in element orientation, or be- 
cause the femur has a higher safety factor 
than more distal elements. Bertram and 
Biewener (1992) also found more variation 
in limb bone curvature relative to body 
mass in the femur than in more distal el- 
ements among a wide range of mammals, 
possibly suggesting that the femur may be 
relatively stronger and therefore experi- 
ences lower strains than more distal ele- 
ments. This hypothesis needs to be tested 
with in vivo strain data. 

Perhaps the most important problem 
raised by the trade-off between modeling 
and Haversian remodeling documented 
here is the mechanism by which bones 
modulate proximodistal differences in os- 
teogenic responses to mechanical loading. 
Several hypotheses merit further study. 
First, different modeling and Haversian 
remodeling responses to mechanical load- 
ing may be a function of differences in 
strain magnitude, strain polarity, or other 
aspects of strain energy histoiy. Frost 
(1990) and Martin and Burr (1989) pro- 
posed that remodeling and modeling are 
mutually exclusive responses below or 
above specific strain thresholds (mininnum 
effective strains). However, preliminary 
data (to be presented in a subsequent pa- 
per) on cross-sectional strains normal to 
the tibia and metatarsal midshaft in the 
sheep indicate that the tibia and luetatarsal 
experience either similar bending strains, 
or that the metatarsal experiences slightly 
more bending in the sagittal plane than 
the tibia. Another hypothesis is that Hav- 
ersian remodeling rates may be positively 
correlated with strain magnitudes or strain 

energy history. Secondary osteonal bone is 
not as strong as primary osteonal and cir- 
cumlamellar bone (Currey, 1959; Carter 
and Hayes, 1977a; Vincentelli and Grigo- 
rov, 1985), so any differential response to 
elevated strain levels might actually exag- 
gerate the apparent trade-off between 
modeling and remodeling in proximal ver- 
sus distal bones. However, these and other 
hypotheses can only be evaluated with in 
vivo strain data from the midshaft of the 
femur, tibia, and metatarsal in conjunction 
with histologic data on growth and remod- 
eling rates. 

Another possibility is that variations in 
vascular or cellular density may modulate 
growth versus repair responses to loading 
in different limb elements. Bone vascular 
density must constrain remodeling re- 
sponses to some extent because bone mor- 
phogenetic units require arteries to supply 
nutrients and cells (Winet et al., 1990; 
Singh et al., 1991). Avascular cortical bone 
tissue is rarely if ever reconstructed by os- 
teoclasts (de Ricqles et al., 1991), and thus 
has little potential to respond to loading 
through Haversian remodeling. Moreover, 
the vascular density of bone may be partly 
rate-dependent (de Ricqles, 1975), possi- 
bly explaining the correlation between re- 
modeling rates and vascular density 
(Green et al., 1987). Further research is 
needed to test if a correlation exists be- 
tween the density of vascular channels in 
the primary cortex of the femur, tibia, and 
metatarsal and Haversian remodeling 
rates. An additional possibility is that os- 
teocytes act as strain transducers (Lanyon, 
1993; Turner et al., 1994, 1995), poten- 
tially limiting Haversian remodeling re- 
sponses in primary compact bone in pro- 
portion to their density. Osteocytes possi- 
bly play a role in sensing and transducing 
information on strains and/or microcracks 
to precursor stem and mesenchymal cells 
(Lanyon, 1993), which may limit the ability 
of acellular cortical bone to undergo Hav- 
ersian remodeling. Differences in osteo- 
cyte density have been implicated as a pos- 
sible contributing factor to osteoporosis 

Trade-Off Responses to Loading in the Mammalian I^imb * Liebemmn and Pearson 279 






Midshaft Area Normalized Inertial Cost (joules *10 ) 



t;5 15- 


i 10 










'■ I 


; ( 





' 1 ' 

' 1 





Midshaft Area Normalized Inertial Cost (joules *10 ) 

Figure 3. Observed trade-off between modeling (A) and Haversian remodeling (B) in the hindlimb as a function of area-nor- 
malized inertial cost of accelerating mass at the midshaft of the femur, tibia, and metatarsal. Density ellipses are plotted for 10% 
(shaded) and 65% (unshaded) within-group variation for exercised (solid line) and control (dashed line) groups. 

(Mullender et al., 1996), but these remain 
to be tested in the sample here. 

The observed trade-off between mod- 
ehng and remodehng in cortical bone sug- 
gests that a more complete understanding 
of WolfPs law will require integration of 
both intraindividual as well as interindivid- 

ual variations in the mechanisins by which 
bones respond dvnamically to their func- 
tional environment. In particular, further 
research needs to address the effects of 
age, variations in strain levels, and the in- 
terinediary factors that modulate the ap- 
parent trade-off between growth and re- 

280 Bulletin Museum of Comparative Zoologij, Vol. 156, No. 1 

pair processes that occurs between proxi- 
mal and distal elements in the limb and 
presumably elsewhere in the skeleton. 


Thank you. Fuzz, for inspiring the se- 
nior author's interest in experimental skel- 
etal biology and functional morphology, 
and for many years of training, support, 
and humorous encouragement. We also 
thank Andy Biewener, Brigitte Demes, 
Brian Richmond, and Bernard Wood for 
comments on the manuscript; Robin Bern- 
stein for helping to make the thin sections; 
and Mike Toscano for running the sheep. 
This research was supported by National 
Science Foundation IBN 96-03833 and by 
funding from the American Federation for 
Aging Research. 


Alexander, R. McN. 1977. Terrestrial locomotion, 

pp. 168-203. In R. McN. Alexander and G. 

Goldspink (eds.). Mechanics and Energetics of 

Animal Locomotion. London: Chapman and 

Hall, xii + 346 pp. 
. 1980. Optimum walking techniques for 

quadnipeds and bipeds. Journal of the Zoological 

Society (London), 173: .549-573. 
. 1981. Factors of safety in the structure of 

animals. Science Progress, Oxford, 67: 109-130. 
. 1996. Optima for Animals. Princeton, New 

Jersey: Princeton University Press, viii + 169 pp. 
1998. Symmoi"phosis and safety factors, pp. 

28-35. In E. Weibel, C.R. Taylor 'and L. Bolis 
(eds.). Principles of Biological Design: The Op- 
timization and Symmorphosis Debate. Cam- 
bridge: Cambridge University Press, xx + 314 

Bertram, J. E. A., and A. A. Biewener. 1988. Bone 
curvature: sacrificing strength for load predict- 
ability? Journal of Theoretical Biology, 131: 75- 

. 1992. Allometry and cuivature in the long 

bones of quadrupedal mammals. Journal of the 
Zoological Societ)' (London), 226: 455-467. 

Bertram, J. E. A., and S. M. Swartz. 1991. The 
'Law of Bone Transformation': a case of crying 
Wolff? Biological Review, 66: 245-273. 

Biewener, A. A. 1983a. Allometry of quadrupedal 
locomotion: the scaling of duty factor, bone cur- 
vature and limb orientation to body size. Journal 
of Experimental Biology, 105: 147-171. 

. 1983b. Locomotory stresses in the limb 

bones of two small mammals: the ground squirrel 

and the chipmunk. Journal of Experimental Bi- 
ology, 103: 131-154. 

. 1989. Scaling and body support in mammals: 

limb posture and muscle mechanics. Science 
245: 45-48. 

. 1990. Biomechanics of mammalian terrestri- 

al locomotion. Science, 250: 1097-1103. 

Biewener, A. A., S. M. Swartz, and J. E. A. Ber- 
tram. 1986. Bone modeling during growth: Dy- 
namic strain equilibrium in the chick tibiotarsus. 
Calcified Tissue International, 39: 390—395. 

Biewener, A. A., and C. R. Taylor. 1986. Bone 
strain: a determinant of speed and gait? Journal 
of Experimental Biology, 123: 383^00. 

Biewener, A. A., J. J. Thomason, and L. E. Lan- 
YON. 1988. Mechanics of locomotion and jump- 
ing in horse Eqiius; in vivo stress in the tibia and 
metatarsus. Journal of Zoology (London), 214: 

Bouvier, M., and W. L. Hylander. 1981. Effect of 
bone strain on cortical bone structure in ma- 
caques Macaca mulatto. Journal of Morphology, 
167: 1-12. 

. 1996. The function of secondary osteonal 

bone: mechanical or metabolic? Archives of Oral 
Biology, 41: 941-950. 

Burr, D. B. 1993. Fatigue and bone remodeling. 
Calcified Tissue International, 53: S75— S81. 

Burr, D. B., R. B. Martin, M. B. Schaffler, and 
E. L. Radin. 1985. Bone remodefing in response 
to in vivo fatigue microdamage. Journal of Bio- 
mechanics, 18: 189-200. 

Carter, D. R., W. E. Caler, D. M. Spengler, and 
V. H. FRANKEL. 1981a. Fatigue behavior of adult 
cortical bone — the influence of mean strain and 
strain range. Acta Orthopaedica Scandinavica, 
55: 481-490. 

. 1981b. Uniaxial fatigue of human cortical 

bone. The influence of tissue physical character- 
istics. Journal of Biomechanics, 14: 460—470. 

Carter, D. R., and W.C. Hayes. 1976a. Fatigue fife 
of compact bone. I. Effects of stress amplitude, 
temperature and density. Journal of Biomechan- 
ics, 9: 27-34. 

. 1976b. Bone compressive strength: the influ- 
ence of density and strain rate. Science 194: 

. 1977a. Compact bone fatigue damage — a mi- 
croscopic examination. Clinical Orthopedics and 
Related Research, 127: 265-274. 

1977b. Compact bone fatigue damage I. Re- 

sidual strength and stiffness. Journal of Biome- 
chanics, 10: 323-337. 

Carter, D. R., W. C. Hayes, and D. J. Schurman. 
1976. Fatigue hfe of compact bone — II. Effect 
of microstructure and density. Journal of Bio- 
mechanics, 9: 211-218. 

Chamay, a., a.nd p. Tchantz. 1972. Mechanical in- 
fluences in bone remodeling. Experimental re- 
search on Wolffs law. Journal of Biomechanics, 
5: 173-180. 

Trade-Off Responses to Loading in the Mammalian Limb • Liebennan and Pearson 281 

CURREY, J. D. 1959. Differences in the tensile 
strengths of bone of different histological types. 
Journal of Anatomy, 93: 87-95. 

. 1970. The mechanical properties of bone. 

Clinical Orthopedics and Related Research, 73: 

. 19S4. The Mechanical Adaptations of Bones. 

Princeton, New Jersey: Princeton University 
Press, viii + 294 pp. 

Fedak, M. a., N. C. Heglund, and C. R. Taylor. 
1982. Energetics and mechanics of terrestrial lo- 
comotion 2. Kinetic energy changes of the limbs 
and body as a function of speed and body size in 
birds and mammals. Journal of Experimental Bi- 
ology, 79: 23-40. 

Frost, H. M. 1963. Bone Remodehng Dynamics. 
Springfield, Illinois: Charles C. Thomas. 175 pp. 

. 1973. Bone Remodeling and Its Relationship 

to Metabolic Bone Diseases. Springfield, Ilhnois: 
Charles C. Thomas, ix -I- 210 pp. 

. 1990. Skeletal structural adaptations to me- 
chanical usage SATMU: 2. Redefining WolfPs 
law: the remodeling problem. The Anatomical 
Record, 226: 414-422. 

Gambaryan, P. P. 1974. How Mammals Run: Ana- 
tomical Adaptations. New York: Wiley, xiv + 367 

GOODSHIP, A. E., L. E. Lanyon, and J. H. MacFie. 
1979. Functional adaptation of bone to increased 
stress. Journal of Bone and Joint Surgery, 61A: 

Green, J. R., J. Reeve, M. Tellez, N. Veall, and 
R. Wootton. 1987. Skeletal blood flow in met- 
abolic disorders of the skeleton. Bone, 8: 293- 

Heglund, N. C, G. a. Cavagna, and C. R. Tay- 
lor. 1982a. Energetics and mechanics of teiTCS- 
trial locomotion. III. Energy changes of the cen- 
ter of mass as a function of speed and body size 
in birds and mammals. Journal of Experimental 
Biology, 79: 41-56. 

. 1982b. Energetics and mechanics of terres- 
trial locomotion. IV. Total mechanical energy 
changes as a function of speed and body size in 
birds and mammals. Journal of Experimental Bi- 
ology, 79: 57-66. 

Hert, J., E. Prybylova, and M. Liskova. 1972. Re- 
action of bone to mechanical stimuli, part 3. Mi- 
crostructure of compact bone of rabbit tibia after 
intermittent loading. Acta Anatomica, 82: 218- 

HildebranD, M. 1985. Walking and running, pp. 
38-57. In M. Hildebrand, D. M. Bramble, K. F. 
Liem, and D. B. Wake (eds.). Functional Verte- 
brate Morphology. Cambridge, Massachusetts: 
Harvard University Press. 430 pp. 

Jungers, W. J., AND R. J. Minns. 1979. Computed 
tomography and biomechanical analysis of fossil 
long bones. American Journal of Physical An- 
thropology, 50: 285-290. 

Lanyon, L. E. 1993. Osteocyte, strain detection. 

bone modeling and remodeling. Calcified Tissue 
International, 53: S102-S107. 

Lanyon, L. E., A. E. Goodship, C. J. Pye, and H. 
MacFie. 1982. Mechanically adaptive bone re- 
modehng. Journal of Biomechanics, 15: 141- 

Lanyon, L. E., and C. T. Rubin. 1984. Static versus 
dynamic loading as an influence on bone remod- 
ehng. Journal of Biomechanics, 17: 897-906. 

. 1985. Functional adaptation in skeletal struc- 
tures, pp. 1-25. In M. Hildebrand, D. M. Bram- 
ble, K. F. Liem, and D. B. Wake (eds.). Func- 
tional Vertebrate Morphology. Cambridge, Mas- 
sachusetts: Harvard University Press. 430 pp. 

Lieberman, D. E. 1997. Making behavioral and phy- 
logenetic inferences from fossils: considering the 
developmental influence of mechanical forces. 
Annual Review of Anthropology, 26: 185-210. 

Lieberman, D. E., and A. W Crompton. 1998. Re- 
sponses of bone to stress, pp. 78-86. 7)7 E. Wei- 
bel, C. R. Taylor, and L. Bolis (eds.). Principles 
of Biological Design: The Optimization and Sym- 
morphosis Debate. Cambridge: Cambridge Uni- 
versity Press. XX -I- 314 pp. 

Martin, R. B. 1995. Mathematical model for repair 
of fatigue damage and stress fracture in osteonal 
bone. Journal of Orthopedic Research, 13: 309- 

Martin, R. B., and D. B. Burr. 1982. A hypothet- 
ical mechanism for the stimulation of osteonal 
remodeling by fatigue damage. Journal of Bio- 
mechanics, 5: 137-139. 

. 1989. Structure, Function, and Adaptation of 

Compact Bone. New York: Raven Press, xii + 
275 pp. 

McMahON, T. a. 1975. Using body size to under- 
stand the structural design of animals: quadru- 
pedal locomotion. Journal of Applied Physiology, 
39: 619-627. 

Mori, S., and D. B. Burr. 1993. Increased intra- 
cortical remodeling following fatigue damage. 
Bone, 16: 103-109. 

Mullender, M. G., D. D. van der Meer, R. Huis- 
KES, and p. Lips. 1996. Osteocyte density chang- 
es in aging and osteoporosis. Bone, 18: 109—113. 

Myers, M. J., and K. Steudel. 1985. Effect of limb 
mass and its distribution on the energetic cost of 
running. Journal of Experimental Biology, 116: 

ParFITT, a. M. 1988a. Bone remodeling: relationship 
to the amount and structure of bone, and the 
pathogenesis and prevention of fractures, pp. 45— 
93. In B. L. Riggs and L. J. Melton (eds.). Os- 
teoporosis: Etiology, Diagnosis and Manage- 
ment. New York: Raven Press, xii -I- 501 pp. 

. 1988b. Osteonal and semi-osteonal remod- 
eling: the spatial and temporal framework for 
traffic signal in adult human bone. Journal of 
Cell Biomechanics, 55: 273-286. 

Pauwels, F. 1980. Biomechanics of the Locomotor 
Apparatus: Contributions on the Functional 

282 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Anatomy of the Locomotor Apparatus. Berlin: 
Springe r-Verlag. vdii + 518 pp. 

Raab, D. M., T. D. Crenshaw, D. B. Kimmel, and 
E. L. Smith. 1991. A histomorphometric study 
of cortical bone activity during increased weight- 
bearing exercise. Journal of Bone and Mineral 
Research, 6: 741-749. 

DE RiCQLES, A. 1975. Evolution of endothermy: his- 
tological evidence. Evolutionary Theory, 1: 51- 

H. Francillon-Vieillot. 1991. Comparative 
microstructure of bone, pp. 1-78. In B. K. Hall 
(ed.). Bone. Vol. 3: Bone Matrix and Bone Spe- 
cific Products. Boca Raton, Florida: CRC Press. 
X + 333 pp. 

Rubin, C. T, and L.E. Lanyon. 1984a. Dynamic 
strain similarity in vertebrates: an alternative to 
allometric limb bone scaling. Journal of Theo- 
retical Biology, 107: 321-327. 

. 1984b. Regulation of bone formation by ap- 
plied dynamic loads. Journal of Bone and Joint 
Surgery, 66: 397-102. 

1985. Regulation of bone mass by mechani- 

cal strain magnitude. Calcified Tissue Interna- 
tional, 37: 411-^17. 

RUFF, C. B. 1989. New approaches to structural evo- 
lution of limb bones in primates. Folia Primato- 
logica, 53: 142-159. 

Ruff, C. B., and W. C. Hayes. 1983. Cross-sectional 
geometry of Pecos Pueblo femora and tibiae — a 
biomechanical investigation: I. Method and gen- 
eral patterns of variation. American Journal of 
Physical Anthropology, 60: 359-381. 

Ruff, C. B., and J. A. Runestad. 1992. Primate 
limb bone structural adaptations. Annual Review 
of Anthropology, 21: 407^33. 

Ruff, C. B., A. Walker, and E. Trinkaus. 1994. 
Postcranial robusticity in Homo III: ontogeny. 
American Journal of Physical Anthropology, 93: 

SCHAFFLER, M. B., AND D. B. BURR. 1988. Stiffness 
of compact bone: effects of porosity and density. 
Journal of Biomechanics, 21: 13-16. 

1989. Mechanical and morphological effects of 

strain rate on fatigue of compact bone. Bone, 10: 

1990. Long-term fatigue behavior of com- 

pact bone at low strain magnitude and rate 
Bone, 11: 321-326. 

Singh, I. J., H. S. Sandhu, and M. S. Herskovits. 
1991. Bone vascularity, pp. 141-164. In B. K. 
Hall (ed.). Bone. Vol. 3: Bone Matrix and Bone 
Specific Products. Boca Raton, Florida: CRC 
Press, xii -I- 333 pp. 

Smith, J. M., and R. J. G. Savage. 1956. Some lo- 
comotory adaptations in mammals. Zoological 
Journal of the Linnaean Society, 42: 603-622. 

Taylor, C. R., A. Shkolnik, R. Dmi'el, D. Bahar- 
AV, and a. Borut. 1974. Running in cheetahs, 
gazelle and goats: energy costs and limb config- 
uration. American Journal of Physiology, 227: 

Turner, C. H., M. R. Forwood, and M. W. Otter. 
1994. Mechanotransduction in bone: do bone 
cells act as sensors of fluid flow? FASEB Journal, 
8: 875-878. 

Turner, C. H., I. Owan, and Y. Takano. 1995. Me- 
chanotransduction in bone: role of strain rate. 
American Journal of Physiology, 269: E438- 

Vincentelli, R., and M. GrigoROV. 1985. The ef- 
fect of Haversian remodeling on the tensile prop- 
erties of human cortical bone. Journal of Bio- 
mechanics, 18: 201-207. 

Wainwright, S. a., W. D. Biggs, J. D. Currey, and 
J. M. GOSLINE. 1976. Mechanical Design in Or- 
ganisms. Princeton, New Jersey: Princeton Uni- 
versity Press, xii + 423 pp. 

WiNET, H., J. Y. Bag, and R. Moffat. 1990. A con- 
trol model for tibial cortex neovascularization in 
the bone chamber. Journal of Bone and Mineral 
Research, 5: 19-30. 

Winter, D. A. 1990. Biomechanics and Motor Con- 
trol of Human Movement, second edition. New 
York: Wiley, xiii + 227 pp. 

Woo, S. L. Y, S. C. KuEi, D. AMiEL, M. A. Gomez, 
W C. Hayes, F C. White, and W. H. Akeson. 
1981. The effect of prolonged physical training 
on the properties of long bone: a study of Wolffs 
law. Journal of Bone and Joint Surgery, 63: 780- 



Abstract. The function of the avian hind limb dur- 
ing running has received considerable attention, par- 
ticularly as a potential analog for locomotor function 
in extinct bipeds. Comparisons of limb kinematics in 
avian rrmners and mammalian quadrupeds have re- 
vealed consistent differences in the pattern of joint 
excursions, presumably related to the constraints of 
bipedal support in birds. The present study asks 
whether these kinematic differences are paralleled by 
differences in muscle forces and stresses developed 
in hind limb locomotor muscles in birds and quadru- 
pedal mammals. High-speed video and force-plate 
analyses along with anatomical measurements were 
used to estimate muscle forces and stresses in the 
locomotor muscles of small dogs and wild turkeys 
during running. Turkeys and dogs developed remark- 
ably similar patterns of force in hind limb muscles, 
despite large differences in the magnitude of ground 
reaction force moments. It was expected that differ- 
ences in absolute muscle force in hind limb muscles 
would be matched by differences in cross-sectional 
area of muscle available to produce force, to maintain 
similar muscle stress. Instead, muscle stresses varied 
widely between homologous joints in dogs and tur- 
keys, and between joints within species. The distri- 
bution of muscle stress between joints may reflect 
differences in the design of the avian and mammalian 
limb for high -power locomotor activities. 


Avian and inainmalian runners move 
their limbs in very different ways. Mam- 
malian runners and most quadrLipeds re- 
tract their hind limbs during stance phase 
primarily by a large extension of the femur 
at the hip; the knee flexes and extends over 
only a small angle (Goslow et al., 1981). 
Hip extension in birds is negligible at low 
speeds and increases with running speed 
(Gatesy, 1999a). At all speeds, the knee 
undergoes substantial flexion in birds and 

^ Oregon State University, Department of Zoology, 
3029 Cordley Hall, Corvallis, Oregon 97331-2914. 

accounts for a greater angle change than 
the hip, providing inost of the inovement 
for retraction of the limb (Storer, 1971; Ja- 
cobson and Hollyday, 1982; Gatesy, 
1999a). The suggestion has been inade 
that these differences in limb moveiTient 
patterns are necessitated by the constraints 
of balance in bipedal birds (Storer, 1971; 
Gatesy, 1990). The horizontally oriented 
femur of birds positions the foot beneath 
the center of mass for standing and slow 
movement. The result is that the restricted 
inovement of the avian femur leaves the 
knee to act, at least kineinatically, as the 
functional equivalent of the mammalian 
hip (Gatesy and Biewener, 1991). 

It is unclear how these kinematic differ- 
ences, or the requirements of balance in a 
biped, are reflected in the pattern of force 
development in locomotor inuscles. Be- 
cause muscle forces are critical to the en- 
ergetics of running and the mechanical 
stresses on muscles, tendons, and bones, 
the most important consequence of varia- 
tion in limb morphology may be variation 
in timing and magnitude of muscle forces 
(Biewener, 1989; Kram and Taylor, 1990; 
Roberts et al., 1998). It has recently been 
shown that dogs and wild turkeys produce 
similar total muscle forces per unit ground 
reaction force (GRF) during running 
(Roberts et al., 1998). The present study 
addresses two questions about the design 
of the avian and mainmalian hind limb for 
locomotor force production: is the timing 
of flexor and extensor forces the same at 
homologous joints in dogs and turkeys, 
and are the stresses developed in the lo- 

Bull. Mus. Comp. Zool., 156(1): 283-295, October, 2001 283 

284 Btilletin Museum of Comparative Zoology, Vol. 156, No. 1 

comotor muscles equivalent at equivalent tional area of muscle to maintain similar 

speeds? stress. The similar stress hypothesis is sup- 

The muscle forces required to balance ported by the observation that kangaroo 
gravitational and inertial forces during rats and white rats maintain the same level 
running are determined primarily by the of peak stress in their ankle extensors, de- 
mechanical advantage for force production spite a fourfold difference in peak GRF 
against the ground. The mechanical advan- per limb (Perry et al., 1988). 
tage at any given joint can change signifi- In the present study, GRF-based joint 
cantly during the stance phase, as the po- moments are measured in dogs and tur- 
sition of the GRF, and its leverage, chang- keys running over a force plate. These data 
es with respect to the joint (Biewener, are combined with anatomical data to cal- 
1989; Carrier et al., 1998). These changes culate the required muscle forces and the 
in muscle mechanical advantage mean that total stress in major locomotor muscles at 
the timing of the maximum muscle force homologous joints in dogs and turkeys, 
may be independent of the timing of the Higher hind limb muscle forces are to be 
GRF. Forces at all joints must be carefully expected in turkeys compared with dogs 
coordinated during running to maintain because they support their weight on two 
balance. Presumably, the significant differ- limbs rather than four. To maintain similar 
ences in limb posture and running kine- stress, turkeys should have a greater cross- 
matics in bipedal birds and quadrupeds sectional area of hind limb muscle in pro- 
might be associated with distinct differ- portion to higher absolute muscle forces, 
ences in the timing and coordination of 
joint moments and muscle forces. MATERIALS AND METHODS 

One possible morphologic indicator of Animals 
the forces that are developed during lo- 
comotion is the cross-sectional area of ex- Three small terriers (Canis famiUaris) 
tensor muscles at a joint. The range of pos- and three wild turkeys (Meleagris gallo- 
sible joint forces is limited by the total V^^o) were used in these studies. These 
cross-sectional area of muscle available species were chosen because they were 
and the maximum stress that can be de- approximately the same body mass, 4.5 ± 
veloped by verirebrate skeletal muscle. It l^ kg and 5.3 ± 2.3 kg for the dogs and 
has been hypothesized that locomotor turkeys, respectively. The methods for an- 
muscles should undergo the same peak mial training and the data collection meth- 
stresses during equivalent movements in ods for dogs and turkeys have been de- 
terrestrial runners (Peny et al., 1988). This scribed in detail previously (Roberts et al., 
hypothesis is based upon the idea that the 1998), and will be presented only briefly 
muscular system, like the skeletal system, here. 

should be neither over- nor underbuilt, ,, , r- >< 
u ^. 4.U 4_u 4- j-u ■4_ r r „ Muscle ForcG Measurements 
but rattier tliat the capacity tor torce gen- 
eration should be matched to the demand High-speed video and force-plate mea 

(Biewener, 1990). Because the maximum surements were made as animals ran freely 
stress that can be developed is a nearly in- over a 15-m track. Measurements were 
variant property of skeletal muscle, the made at the speeds that were initially free- 
equivalent stress hypothesis states that a ly chosen by the animals, approximately 
similar fraction of the capacity for force 3.5 m/s for the turkeys and 2.0 m/s for the 
production in the extensors of a joint dogs. Joint moments of force were calcu- 
should be used at equivalent speeds. Thus, lated from the magnitude and position of 
any differences in force requirements at a the GRF vector relative to the joint cen- 
joint in different species will be matched ters of rotation as determined by force- 
by a proportional difference in cross-sec- plate and high-speed video analysis (Rob- 

Muscle Stress During Running * Roberts 285 

erts et al., 1998). Moments were deter- 
mined for the hip, knee, and ankle (dog 
and turkey), as well as the shoulder, elbow, 
and wrist (dogs only). Measurements in- 
cluded only the GRF-based moments. 
Moments due to accelerations of limb seg- 
ments relative to the center of mass also 
contribute to the required muscle force. 
Limb inertia— based moments are usually 
small in birds and quadrupedal mammals 
(Clark and Alexander, 1975; Pandy et al., 
1988), but may be significant at more prox- 
imal joints. Muscle forces (F,„) were cal- 
culated from the measured joint moments 
and the average muscle moment arm mea- 
sured at each joint: 



where the joint moment is the product of 
GRF magnitude F, (N) and moment arm 
R (m), and f (m) is the average muscle 
moment arm at that joint. 

Calculation of muscle force and stress 
requires anatomical measurements of 
muscle moment arms and cross-sectional 
areas for muscles active during support. 
The average moment arm (f) was calcu- 
lated for the extensor muscles at a joint 
from an average weighted by the cross- 
sectional area of each individual muscle 
(Roberts et al., 1998). This measure as- 
sumed that the relative contribution of a 
given muscle to the total force produced 
at a joint was proportional to the muscle s 
force-generating capacity. Muscle physio- 
logic cross-sectional area (A„) was calcu- 
lated as: 

A„ — 

m-cos 6 


where m is the mass of the muscle (g), 6 
is the fiber pinnation angle, p is the density 
of muscle (g/cm^) , and € is muscle fascicle 
length (cm). 

Measurements were included for mus- 
cles that are primarily extensors of the 
joint (or flexors in the case of the wrist), 
and are active during stance phase based 

on electromyographic activity (Tokurild, 
1973; Goslow et al., 1981; Gatesy, 1999b). 
Muscles included for turkeys were ilioti- 
bialis lateralis pars postacetabularis, iliofi- 
bularis, flexor cruris lateralis, flexor cruris 
medialis, puboischiofemoralis, and ischio- 
femoralis (hip extensors); femorotibiales 
(knee extensors); and gastrocnemius, fi- 
bularis longus, flexor hallucis longus, flexor 
digitorum longus, flexor perforatus digiti II 
and III, and flexor perforans et perforatus 
digiti II, III, and IV (ankle extensors). 
Muscles included for dogs were gluteus 
medius, gluteus superficialis, biceps fe- 
moris, and semimembranosus (hip exten- 
sors); vastus lateralis, medialis, and inter- 
medins (knee extensors); gastrocnemius, 
plantaris, and deep digital flexors (ankle 
extensors); supraspinatus (shoulder exten- 
sor); triceps brachii (elbow extensor); and 
flexor digitorum profundus, flexor digito- 
rum superficialis, flexor carpi ulnaris, and 
flexor carpi radialis (wrist flexors). Some of 
the anatomical measurements were taken 
from specimens available from unrelated 
experiments. These animals were of simi- 
lar weight and condition. Small corrections 
to the muscle dimension measurements 
were made for differences in the body 
mass of the experimental subjects by as- 
suming geometric scaling (Roberts et al., 

Contribution of Biarticular Muscles to 
Extensor Muscle Forces 

In addition to the muscle forces re- 
quired to overcome GRF based moments, 
extensor muscles must produce force to 
balance antagonist flexor muscle forces. 
Cocontraction of muscles that are primar- 
ily joint flexors was assumed to be negli- 
gible during stance. However, many biar- 
ticular (or multiarticular) extensor muscles 
also act as flexors. The flexor forces pro- 
duced by biarticular extensor muscles ac- 
tive during stance were included as part of 
the calculation of total muscle force re- 
quired at a joint. Thus, part of the calcu- 
lated extensor moment at several joints in- 
cludes a component of force necessary to 

286 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

counteract cocontraction of multiarticular 
muscles. The contribution of a biarticular 
muscle to a joint flexor moment was cal- 
culated from the extensor force it devel- 
oped at its extensor articulation, assuming 
that force was distributed among extensor 
muscles in proportion to their cross-sec- 
tional area. The flexor moment was then 
calculated from the force produced in the 
biarticular muscle and its flexor moment 

Flexor forces in dogs included gastroc- 
nemius flexion of the knee and flexion of 
the shoulder by the long head of triceps 
brachii. In turkeys, extensor muscle forces 
were required to balance flexor moments 
produced at the knee by the hip extensors 
flexor cruris lateralis and medialis, and il- 
iofibularis. The following ankle extensors 
also produced a flexor moment at the knee 
in turkeys: gastrocnemius lateralis and me- 
dialis, and the digital flexors flexor hallucis 
longus, flexor perforatus digiti II, III, and 
IV, and flexor perforans et perforatus digiti 
II and III. Mulitiarticular muscles also act- 
ed to reduce extensor forces in two cases. 
Some of the digital flexors that contribute 
to flexion of the wrist also act to extend 
the elbow in dogs (flexor digitorum pro- 
fundus, flexor digitorum superficialis, flex- 
or carpi radialis, and flexor carpi ulnaris). 
In turkeys, iliotibialis lateralis pars posta- 
cetabularis was included primarily as a hip 
extensor but it also contributes an extensor 
moment at the knee. 

The presence of inultiple biarticular 
muscles that both flex and extend the knee 
and the hip makes calculation of muscle 
forces an indeterminate problem, because 
it is impossible to know the degree of co- 
contraction of antagonist muscle groups 
(Winter, 1990). This problem was avoided 
in the present study by omitting the con- 
tribution of the rectus feinoris to knee ex- 
tension and hip flexion in dogs. Electro- 
myographic studies suggest that this mus- 
cle is active during the second two thirds 
of stance phase in trotting dogs (Tokuriki, 
1973). Rectus femoris is approximately 
35% of the total cross-sectional area of the 

extensor muscles at the knee; omission of 
this muscle may lead to an underestimate 
of hip extensor forces late in stance and a 
small overestimate of knee inuscle stress 
in dogs. 

Muscle Stress Measurements 

Muscle stress was determined by divid- 
ing the peak measured muscle force by the 
total physiologic cross-sectional area of ex- 
tensor muscles at a joint. This measure is 
equivalent to the muscle stress in each in- 
dividual muscle at a joint only if the force 
is evenly distributed among muscles in 
proportion to each muscle s cross-sectional 
area. Although this assumption has proven 
accurate for hopping kangaroo rats (Biew- 
ener et al., 1988), force buckle measure- 
ments in cats suggest that the relative con- 
tribution of individual ankle extensors to 
total extensor force can change with speed 
of movement and locomotor activity 
(Walmsley et al., 1978). Although variation 
in distribution of muscle force between 
muscles may make it difficult to calculate 
stresses in individual muscles, the measure 
of muscle stress of a whole muscle group 
will still reflect the fi'action of the capacity 
for force production that is being used in 
the muscle group. 

Statistical Analyses 

To calculate an average force versus 
time plot for all of the trials it was neces- 
sary to normalize for variation in both 
stride time and total force between indi- 
vidual trials. Moment and force measure- 
ments in an individual trial were divided 
by the peak GRF for that trial. To control 
for variation in ground contact time, each 
original stance period was intei"polated to 
a wave of 30 points using the cubic spline 
inteipolation function in the coinputer 
software application Igor (Wavemetrics, 
Lake Oswego, OR). The interpolation also 
applied a smoothing function to the data. 
Average muscle moment and force curves 
are represented as the means and standard 
deviations for all of the running trials. 

Analysis of variance was used to test for 

Muscle Stress During Running • Roberts 287 

significant differences at P < 0.05. Values 
presented are means ± one standard de- 
viation unless indicated othei-wise. All in- 
dicated differences are significant at F < 

Muscle Forces 

Dogs and turkeys produce remarkably 
similar patterns of force at homologous 
joints in the hind limb. Figure 1 presents 
joint moments (FyR) during stance phase 
normalized to the peak GRF produced 
during the step. Hip moments were great- 
est early in stance and relatively small dur- 
ing the second half of stance in both dogs 
and turkeys. Ankle moments were also at 
their maximum early in the stance phase. 
In both the dog and the turkey, the knee 
musculature produced a net flexor mo- 
ment early in stance, followed by an ex- 
tensor moment. This cyclic flexor— extensor 
moment pattern also occurred at the 
shoulder of the dog. The highest knee mo- 
ments occurred during the second half of 
stance phase in both dogs and turkeys. The 
peak GRF, indicated by an arrow, occurs 
earlier in stance in the turkey hind limb 
than in the dog hind limb or forelimb. 

The hind limb muscles of turkeys pro- 
duce greater joint moments per unit GRF 
compared with dogs (Fig. 1). Hip mo- 
ments per unit GRF were 3.1-fold greater 
in the turkey (turkey, 0.085 ± 0.013; dog, 
0.027 ± 0.013), knee moments were great- 
er by a factor of 2.4 (turkey, 0.034 ± 0.012; 
dog, 0.014 ± 0.004), and ankle moments 
were 1.8-fold greater in the turkey (turkey, 
0.082 ± 0.012; dog, 0.046 ± 0.006). These 
differences suggest that the turkey's rela- 
tively horizontally oriented femur and long 
tibiotarsus and tarsometatarsus result in 
greater GRF moment arms (R) at the hip, 
knee, and ankle. 

Despite the large differences observed 
in muscle moments (Fig. 1), muscle forces 
produced per unit GRF were similar in 
the knee and the hip in dogs and turkeys 
(Fig. 2). Muscle forces presented in Fig- 

ure 2 were calculated from joint moments 
by dividing by the average muscle moment 
arm (P; Equation 1), and including mo- 
ments produced by biarticular agonists or 
antagonists. This similarity in muscle forc- 
es at the knee and hip indicates that mus- 
cle moment arms (r) at these joints are 
proportionately larger in the turkeys to 
compensate for the higher moments re- 
sulting from differences in limb posture. 
The average muscle moment arm at the 
ankle was similar in dogs and turkeys; 
muscle forces were larger in turkeys, as 
were the muscle moments. 

The contribution of biarticular muscle 
agonists and antagonists to the extensor 
muscle forces calculated at the knee, el- 
bow, and shoulder is represented in Figure 
2 by the difference between the total cal- 
culated extensor force (bold line) and the 
extensor muscle force calculated neglect- 
ing the contribution of biarticular muscles 
(dotted line). Interestingly, flexor forces 
produced by biarticular muscles seem to 
provide the flexor moment necessary to 
balance the GRF early in stance at the 
knee and shoulder. Later in stance the ex- 
tensor force necessary is increased by the 
action of two-joint muscles, assuming that 
these muscles are fully active throughout 

Muscle Stresses 

Muscle stresses were quite variable be- 
tween joints in dogs and turkeys (Fig. 3) 
rrmning at the speeds measured in this 
study. The values presented in Figure 3 
were obtained by dividing the peak force 
produced by the total physiologic cross- 
sectional area of the extensor muscles at a 
joint. Stress varied by more than fivefold 
between joints in dogs, from 59 ± 28 kN/ 
m- in the hip extensors to 309 ± 59 kN/ 
m- in the shoulder extensors. Generally, 
hind limb stresses were lower than fore- 
limb stresses, and stresses were higher 
moving distally in the dog hind limb. Tur- 
keys showed less variation in muscle stress 
between joints and generally operated 

288 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 








hind limb 



1 • • • - 



Dog forelimb 



Stance time (%) 

Figure 1. Muscle moments during the stance phase normalized to the peak ground reaction force (GRF). Mean values and 
standard deviations are presented for 12 strides of three animals. Positive moments represent extensor muscle moments, with 
the exception of the wrist; wrist flexor muscle moments are represented as positive. Arrows indicate the time of the peak GRF. 

with higher hind limb muscle stresses than 
did dogs. 

Higher muscle stresses at a joint can be 
due to either higher muscle forces or a 
smaller cross-sectional area of muscle 
available. Table 1 allows a comparison of 
the contribution to differences in muscle 
stress of muscle mechanical advantage, the 
total GRF produced, and muscle cross- 
sectional area. The muscle force produced 
per peak GRF is determined by the mus- 

cle mechanical advantage, and is signifi- 
cantly different in the dog and turkey only 
at the ankle. The peak GRF on a single 
hind limb in turkeys was equal to 2.4 X 
body weight. Dogs developed a total peak 
GRF of 2 X body weight, but only 41% of 
this was developed by the hind limbs; 
therefore, the dog hind limb GRF was 0.8 
X body weight. When this difference in 
hind limb GRF is included, it is apparent 
that, at equivalent speeds, turkeys produce 






Muscle Stress During Running • Roberts 289 

Dog forelimb 

Dog hind limb 






• ■i'i..i..l..'rn 




Stance time (%) 

Figure 2. Muscle forces during tine stance phase in dogs and turkeys. Muscle force values are normalized to the peak ground 
reaction force (GRF) developed during the stance phase to give a dimensionless ratio. Solid lines represent the muscle forces 
necessary to balance the GRF moment and to balance flexor moments produced by two-joint muscles. Dotted lines indicate the 
muscle forces required to balance GRF moments only. Arrows indicate the time of peak GRF. Means and standard deviations 
are presented for 12 runs of three animals. 

much higher absolute muscle forces at all 
joints than do dogs (Table 1). At the ankle, 
for example, turkeys produced muscle 
forces equivalent to 16.9 X body weight, 
whereas dogs produce only 3 X body 
weight with ankle extensor muscles. At the 
knee and the hip, the differences in mus- 
cle cross-sectional area are too small to 
compensate for the large differences in 
force, and therefore stresses in turkey 

muscles are much higher. However, at the 
ankle, the large difference in muscle cross- 
sectional area compensates for both the 
turkeys higher hind limb GRFs and its 
poorer mechanical advantage; despite a 
5.6-fold difference in muscle force, stress- 
es are not significantly different at the an- 
kle in dogs and turkeys. 

The speeds measured in this study were 
approximately 2.0 m/s in the dogs and 3.5 

290 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 


S 200 

Hind limb 

Fore limb 


Shoulder Elbow Wrist 

Figure 3. Stress (force/cross-sectional area [/4J) developed in limb muscles of dogs and turkeys. Stresses are presented for 
the extensors at the joint indicated, with the exception of the wrist values, which represent flexor stresses. Values are means 
and standard deviations. 

m/s in the turkeys. Despite the large dif- 
ference in absolute speed, the average 
GRF developed during stance was similar 
in the two species, 1.3 ± 0.1 X body 
weight in the bird and 1.1 ± 0.1 X body 
weight in the dog. This suggests that the 
animals were operating at similar duty fac- 
tors, one of the criteria that has been used 
to determine equivalent speeds in running 
mammals (Biewener, 1983). The turkeys 
were able to run at faster speeds with duty 
factors similar to those of dogs because 

they have longer legs. The small difference 
in average GRF at the speeds used in this 
study indicates that some of the difference 
in estimated muscle stress may be due to 
a 15—20% difference in the magnitude of 
GRF between the dogs and turkeys. 


Patterns of Force Development and Limb 

The results of the present study dein- 
onstrate that the timing of muscle force 

Table 1. Muscle force per unit peak ground reaction force (GRF), total GRF and muscle 



Muscle force per GRF 

Peak GRF per BW 

Muscle force per BW 

A,., (CU1-) 

Muscle stress (kN/m-) 



2.3 ± 0.3 

2.4 ± 0.2 

5.8 ± 0.8 

15.9 ± 1.0 

180 ± 28 


1.8 ± 1.0 

0.8 ± 0.1 

1.6 ± 0.8 

12.8 ± 2.0 

59 ± 28 









2.3 ± 0.8 

2.4 ± 0.2 

5.7 ± 2.0 

11.0 ± 0.6 

248 ± 65 


1.4 ± 0.4 

0.8 ± 0.1 

1.2 ± 0.3 

7.8 ± 0.7 

71 ± 20 









6.9 ± 0.7 

2.4 ± 0.2 

16.9 ± 2.3 

50.6 ± 8.4 

163 ± 21 


3.6 ± 0.5 

0.8 ± 0.1 

3.0 ± 0.5 

11.4 + 1.5 

125 ± 26 







* Denotes significant difference between dogs and turkeys, p < 0.05. 

Muscle Stress During Running • Roberts 291 

development is remarkably similar at ho- produced per unit GRF were similar at the 
mologous joints in the hind limbs of dogs knee and the hip in dogs and turkeys. Tur- 
and turkeys. This siinilarity exists despite keys produced higher joint moments at the 
differences in the timing and magnitude of knee and hip because they have larger 
joint angular excursions in bipedal birds muscle moment arms, rather than higher 
and quadrupedal mammals (Gatesy and muscle forces. Thus, differences in mus- 
Biewener, 1991), as well as differences in culoskeletal anatomy can compensate for 
limb posture. Both dogs and turkeys pro- differences in limb posture to achieve sim- 
duce peak net extensor muscle moments ilar mechanical advantage for force pro- 
early in the stance phase at the ankle and duction. It has recently been demonstrat- 
the hip, and late in the stance phase at the ed that the pattern of joint excursions 
knee. In both species, the knee muscula- changes markedly with increases in walk- 
ture produces a net flexor moment early ing or running speed in guinea fowl (Ga- 
in stance, followed by a net extensor mo- tesy, 1999a). There appears to be little 
ment for approximately the last two thirds change with speed in the average inechan- 
of stance phase. The similarity in joint mo- ical advantage with which muscles gener- 
ment timing between animals that have ate force in running quadrupedal mam- 
significantly different limb morphology mals (Biewener, 1989), but it remains to 
and have independently evolved cursori- be seen whether significant changes in the 
ality tempts the speculation that this pat- pattern of force development occur across 
tern of joint moments is a common feature speed in running bipeds, 
of hind limb dynamics in runners. How- 
ever, human runners exliibit patterns of Biarticular Muscles and Joint Moment 
joint moments different from those ob- rattems 

served in the present study. In particular. Several functional advantages have been 

the knee musculature produces an exten- proposed for muscles that articulate across 

sor moment diroughout stance and a net two or more joints. Two- joint muscles can 

flexor muscle moment is produced at the operate to maintain a uniform length due 

hip during the second half of stance phase to compensating displacements at two 

(Winter, 1983). Also, studies of larger dogs joints (Goslow et al., 1973; McCleam, 

have found that, at the hip, the net muscle 1985), and can function to transfer power 

extensor moment during early stance is from distal to proximal limb segments 

followed by a significant net muscle flexor (Gregoire et al., 1984). It has also been 

moment during the second half of stance shown that two-joint muscles play an im- 

(Carrier et al., 1998). portant role in controlling the orientation 

The suggestion has been made that the of the GRF during various movements in 

differences in limb posture and joint ex- humans (Jacobs and van Ingen Schenau, 

cursion patterns during locomotion be- 1992; van Ingen Schenau et al., 1992). At 

tween bipedal birds and quadrupedal the same time, it can be argued that most 

mammals are associated with the problem biarticular muscles are energetically dis- 

of bipedal support in birds ( Store r, 1971; advantageous because they produce flexor 

Gatesy, 1990). The horizontally oriented forces that must be balanced by cocon- 

femur in birds allows them to position the traction of antagonist extensor muscles, 

point of support under the center of mass The present results suggest that two-joint 

during standing or slow movement (Storer, muscles, in both birds and dogs, provide 

1971). The present study demonstrates an important mechanism for balancing al- 

that in turkeys, this difference in limb ori- temating flexion and extension moments at 

entation results in greater muscle mo- a joint. The flexor moments produced at 

ments per unit GRF compared with qua- the knee in dogs and turkeys appear to be 

drupedal dogs. However, muscles forces quantitatively matched to the knee flexion 

292 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

forces contributed by biarticular hip and 
ankle extensors (Fig. 2). Thus, early in 
stance biarticular muscles at the knee pro- 
duce force that supports body weight at 
both the joint they act to extend (hip or 
ankle) and the joint they act to flex. The 
long head of triceps brachii also appears to 
function early in stance to provide support 
through both its flexor function at the 
shoulder and its extensor function at the 
elbow. Later in the stride biarticular mus- 
cles tend to increase the extensor force re- 
quired at the knee and shoulder. However, 
electromyographic studies indicate that 
most of the biarticular inuscles considered 
in the present study are most active early 
in stance in dogs (Tokuriki, 1973; Goslow 
et al., 1981), allowing for the possibility 
that energetically disadvantageous cocon- 
traction may be reduced during the sec- 
ond half of the stance phase. It may be 
that the important role that biarticular 
muscles appear to play in maintaining the 
proper orientation of the GRF (Jacobs and 
van Ingen Schenau, 1992; van Ingen 
Schenau et al., 1992) explains in part the 
similarity in joint moment patterns at the 
knee in dogs and turkeys. 

Muscle Stress and Strategies for Force 
and Power Development 

Many of the muscle stresses estimated 
in the present study are near the upper 
end of the range of muscle stresses typi- 
cally measured in vitro for vertebrate skel- 
etal muscle. Peak isometric stress mea- 
surements typically range from 100 to 300 
kN/m^ (Josephson, 1993). Muscles can de- 
velop stresses above peak isometric when 
actively lengthening (Katz, 1939), and 
stress is reduced as relative shortening ve- 
locity increases (Hill, 1938). Thus, differ- 
ences in muscle stress between joints are 
proportional to differences in fraction of 
recruited cross-sectional area only if the 
muscles are operating at similar relative 
shortening velocities. The high stresses es- 
timated in the present study at the knee 
of the turkey and the wrist and shoulder 
of the dog suggest that these muscles may 

be operating near their limit of force pro- 
duction, or undergoing sufficient length- 
ening to operate at high active muscle 
stresses. Stresses measured at the wrist 
likely overestimate actual muscle stresses, 
because some of the wrist moment is bal- 
anced by force produced in stretched lig- 
aments (Alexander, 1974). Shoulder stress 
measurements may also overestimate ac- 
tual muscle stress if moments due to limb 
inertia are significant. Previous measure- 
ments of stresses in running quail reported 
lower values than those reported here for 
turkeys (80-100 kN/m^; Clark and Alex- 
ander, 1975). However, the low speed and 
GRF (approximately 1.3 X body weight 
peak GRF) measured in quail make direct 
comparison of stress values with the pre- 
sent study difficult. It is important to note 
that the present measurements of muscle 
cross-sectional areas do not account for 
the volume of noncontractile elements 
within inuscle (mitochondria, capillaries, 
and so on), which can be as much as 40% 
of the inuscle volume in bird pectoralis 
muscle (James and Meek, 1979; Conley et 
al., 1987). Differences in noncontractile 
element components between muscles or 
between species represent a potential 
source of error in comparisons of muscle 
stress. However, the variation in voluine of 
nonmyofibrillar components between 
these largely oxidative muscles is unlikely 
to be very large, and could not account for 
the several-fold differences in muscle 
stress found between inuscle groups with- 
in dogs and between dogs and turkeys. 

The similar stress hypothesis for inuscle 
was originally formulated upon the idea 
that animals should operate their limb 
muscles with the same reserve capacity for 
force development at equivalent speeds 
(Perry et al., 1988). The hypothesis was 
supported by observations of similar stress 
in the ankle extensors in a bipedal hopper, 
the kangaroo rat, and a quadrupedal run- 
ner, the white rat (Perry et al., 1988). The 
similar scaling of locomotor forces and 
muscle cross-sectional area across body 
size also supports the similar stress hy- 

Muscle Stress During Running * Roberts 293 

pothesis (Biewener, 1990). The results versus force development. The large mass 
from the present study for the ankle joint of largely parallel-fibered muscles located 
provide another example of how large dif- proximally in the limb is well suited to pro- 
ferences in muscle morphology can com- viding high power outputs, whereas the 
pensate for differences in muscle forces to short pinnate fibers of the ankle in con- 
produce similar stress in different animals, junction with their long tendons function 
However, a comparison of stresses be- well as springs (Alexander, 1974). 
tween joints in the dog reveals that the If differences in muscle stress between 
similar stress hypothesis is not supported joints reflect specializations of different 
for comparisons between different muscle muscle groups for power versus force de- 
groups operating at different joints. velopment, then a Comparison of muscle 
Why do some muscle groups operate stresses in dogs and turkeys may provide 
with a greater reserve capacity for force insight into how avian and luammalian 
development (i.e., low stress) during run- limbs are designed to provide work for 
ning? This variation in reserve capacity high-power activities. Turkeys operate 
may reveal a specialization of some mus- with the same reserve capacity for force at 
cles for locomotor activities that involve the hip and the ankle. Furthennore, the 
high mechanical power outputs. The cross- muscle mass distribution is strikingly dif- 
sectional area of active muscle required to ferent in birds compared with mammals, 
produce force during high-power— output The muscle mass available to extend the 
activities such as acceleration, jumping, or ankle in birds is greater than that available 
incline running can be greater than that to extend the hip (compared with a hip 
required during level running because of muscle mass greater than three times the 
an increase in GRF or a change in muscle ankle muscle mass in dogs; Roberts et al., 
mechanical advantage. Perhaps more im- 1998). The mechanical power that can be 
portantly, the cross-sectional area of active developed at a joint is proportional to the 
muscle must also increase if the relative muscle mass (assuming equivalent muscle 
shortening velocity of the muscle increas- properties); thus the large muscle mass at 
es, as has been demonstrated for the lat- the ankle in birds represents a significant 
eral gastrocnemius muscle in turkeys run- potential source of power for movement, 
ning on an incline (Roberts et al., 1997). The present ineasureinents of muscle 
It might be expected that muscle groups stress suggest that turkeys may power ac- 
that operate with a high reserve capacity celerations and jumps as inuch with the 
during level running are the most impor- ankle musculature as with the hip. In fact, 
tant power producers during acceleration measurements of turkeys running uphill 
or jumping. It has been generally assumed suggest that the gastrocnemius increases 
that it is the muscles that articulate be- muscle shortening, recruitinent, and pow- 
tween the limb and the body (extrinsic er output to provide the power to run up- 
muscles), and particularly those in the hill (Roberts et al., 1997). Thus, differenc- 
hind limb, that are the primary sources of es in muscle mass distribution and limb 
power for acceleration (Gray, 1968). Like- design between running birds and mam- 
wise, it has been shown that jumping dogs mals may reflect differences in strategies 
produce the greatest muscle work and for powering accelerations, rather than dif- 
power at the hip with a smaller contribu- ferences in muscle function during steady- 
tion at the knee and almost no change in speed running, 
ankle muscle function froin running to ^....^..,. ^^^. .^».-^^ 
jumping (Alexander, 1974). Alexander ACKNOWLEDGMENTS 
(1974) suggested that differences in mus- I thank M. S. Chen for help with ex- 
cle architecture in proximal versus distal periments and data analysis, and the late 
muscles reflects a specialization for power C. R. Taylor for his help and support with 

294 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

the project. The project was supported in 
part by National Institutes of Health grant 
ROl AR18140. 


Alexander, R. McN. 1974. The mechanics of jump- 
ing by a dog {Canis familiaris). Journal of Zool- 
ogy, London, 173: 549-573. 

BlEWENER, A. A. 1983. Allometry of quadrupedal lo- 
comotion: the scaling of duty factor, bone cur- 
vature and limb orientation to body size. Journal 
of Experimental Biology, 105: 147-171. 

. 1989. Scahng body support in mammals: 

limb posture and muscle mechanics. Science, 
245: 45-48. 

. 1990. Biomechanics of mammalian terrestri- 

al locomotion. Science, 250: 1097-1103. 

BlEWENER, A. A., R. Blickhan, a. K. Perry, N. C. 
Heglund, and C. R. Taylor. 1988. Muscle 
forces during locomotion in kangaroo rats: force 
platform and tendon buckle measurements com- 
pared. Journal of Experimental Biology, 137: 

Carrier, D. R., C. S. Gregersen, and N. Silver- 
ton. 1998. Dynamic gearing in running dogs. 
Journal of Experimental Biology, 201: 3185- 

Clark, J. and R. McN. Alexander. 1975. Mechan- 
ics of running quail (Cotumix). Journal of Zool- 
ogy, London, 176: 87-113. 

CONLEY, K. E., S. R. Kayar, K. Rosler, H. Hop- 
peler, E. R. Weibel, and C. R. Taylor. 1987. 
Adaptive variation in the mammalian respiratory 
system in relation to energetic demand. IV. Cap- 
illaries and their relationship to oxidative capac- 
ity. Respiration Physiolog\; 69: 47-64. 

Gatesy, S. M. 1990. Caudofemoral musculature and 
the evolution of theropod locomotion. Paleobi- 
ology, 16: 170-186. 

. 1999a. Guineafowl hind limb function. I: 

cineradiographic analysis and speed effects. Jour- 
nal of Moi-phology, 240: 115-125. 

. 1999b. Guineafowl hind limb function. II: 

electromyographic analysis and motor pattern 
evolution. Journal of Morphology, 240: 127-142. 

Gatesy, S. M., and A. A. Biewener. 1991. Bipedal 
locomotion: effects of speed, size and limb pos- 
ture in birds and humans. Journal of Zoology, 
London, 224: 127-147. 

GosLow, G. E. J., R. E. Reinking, and D. G. Stu- 
art. 1973. The cat step cycle: hind limb joint 
angles and muscle lengths during unrestrained 
locomotion. Journal of Morphology, 141: 1^2. 

Goslow, G. E., H. J. Seeherman, C. R. Taylor, M. 

N. MCCUTCHIN, AND N. C. Heglund. 1981. 
Electrical activity and relative length changes of 
dog limb muscles as a function of speed and gait. 
Journal of Experimental Biology, 94: 15^2. 

Gray, J. 1968. Animal Locomotion. New York: W. W. 
Norton and Company, vii -I- 479 pp. 

Gregoire, L., H. E. Beeger, p. a. Huijing, and G. 
J. van Ingen Schenau. 1984. Role of mono- 
and bi-articular muscles in explosive movements. 
International Journal of Sports Medicine, 5: 

Hill, A. V. 1938. The heat of shortening and the 
dynamic constants of muscle. Proceedings of the 
Royal Society of London, B, 126: 136-195. 

Jacobs, R., and G. J. van Ingen Schenau. 1992. 
Control of external force in leg extensions in hu- 
mans. Journal of Physiology, 457: 611—626. 

Jacobson, R. D., and M. HollyDAY. 1982. A be- 
havioral and electromyographic study of walking 
in the chick. Journal of Neurophysiology, 48: 

James, N. T, and G. A. Meek. 1979. Stereological 
analyses of the structure of mitochondria in pi- 
geon skeletal muscle. Cell Tissue Research, 202: 

JOSEPHSON, R. K. 1993. Contraction dynamics and 
power output of skeletal muscle. Annual Review 
of Physiology, 55: 527-546. 

Katz, B. 1939. The relation between force and speed 
in muscular contraction. Journal of Physiology, 
96: 45-64. 

Kram, R., and C. R. Taylor. 1990. Energetics of 
running: a new perspective. Nature, 346: 265— 

McClearn, D. R. 1985. Anatomy of raccoon (Pro- 
cyon lotor) and coati {Nasiia narica and N. na- 
sua) forearm and leg muscles: relations between 
fiber length, moment-arm length, and joint-angle 
excursion. Journal of Morphology, 83: 87-115. 

Pandy, M. G., V. Kumar, N. Berme, and K. J. 
WaldrON. 1988. The dynamics of quadrupedal 
locomotion. Journal of Biomechanical Engineer- 
ing, 110: 230-237. 

Perry, A. K., R. Blickhan, A. A. Biewener, N. C. 
Heglund, and C. R. Taylor. 1988. Preferred 
speeds in terrestrial vertebrates: are they equiv- 
alent? Journal of Experimental Biology, 137: 

Roberts, T. J., M. S. Chen, and C. R. Taylor. 
1998. Energetics of bipedal running II: hmb de- 
sign and running mechanics. Journal of Experi- 
mental Biology, 201: 2753-2762. 

Roberts, T. J., R. L. Marsh, P. G. Weyand, and C. 
R. Taylor. 1997. Muscular force in running tur- 
kevs: the economy of minimizing work. Science, 
275: 1113-1115. 

Storer, R. W. 1971. Adaptive radiation in birds, 
149-188. In D. S. Earner and J. R. King (eds.). 
Avian Biology. New York: Academic Press, xiii -I- 
586 pp. 

TOKURIKI, N. 1973. Electromyographic and joint-me- 
chanical studies in quadrupedal locomotion II: 
trot. Japanese Journal of Veterinary Science, 35: 

Muscle Stress During Running • Roberts 295 


Groot, R. J. Snackers, and W. W. L. M. van 
WOENSEL. 1992. The constrained control of 
force and position in multi-joint movements. 
Neuroscience, 46: 197—207. 
Walmsley, B., J. A. Hodgson, and R. E. Burke. 
1978. Forces produced by medial gastrocnemius 
and soleus muscles during locomotion in freely 

moving cats. Journal of Neurophysiology, 41: 

Winter, D. a. 1983. Moments of force and mechan- 
ical power in jogging. Journal of Biomechanics, 
16: 91-97. 

. 1990. Biomechanics and Motor Control of 

Human Movement, 2nd Edition. New York: 
John Wiley & Sons. 296 pp. 



Abstract. The aim of the present study was to in- long bones in rats (Markel et al., 1991). In 

vestigate the effect of low-energy laser irradiation ^^.j^g^. models, bone repair was investigated 

(LELI) on the process of skeletal muscle regenera- . .1 j 11 -1, r 4-1 ^ 4--U- . ,fi-^^ 

L J u  ,- A A ^ T->^ft„^^ ;.. in the medullary cavity ot the tibia atter 

tion and bone repair m toads and rats. Denned m- J J r 1 

juries (partial excision or cold injury) to the gastroc- ablation or removal ot marrow trom long 

nemius muscle and hole injury to the tibia in rats bones. Ill this model, regeneration was 

were performed. The rate of regeneration in skeletal found tO be preceded by a local phase of 

muscles was analyzed by quantitative^ histomoipho- ^^^05^^^! bo^e formation (Liang et al., 

metric methods, and bone repair was determmed us- ,„„^. „ . p ,1 c^ 

inghistomorphometric methods. Low-energy laser ir- 1992). Promotion of the procesS of boiie 

radiation (He-Ne laser) was applied directly to the repair has been achieved in the past by 

injured sites for 2 minutes at different time intervals autOgeilOUS bone grafting, application of 

postinjuiy. The rate of skeletal muscle regeneration ^^rious growth factors (mainly bone mor- 

was enhanced Uvo- and eightfold in the rat and toad, , ^ . re 4- „1 

respectively, whereas bone^epair was enhanced two- phogenetlC proteins [Spencer et al., 

fold by the LELI. Although mechanism of the laser 1991 J), USe ot low-mtensity pulsed ultra- 

irradiation is not yet clearly understood, it is associ- souild and electromagnetic fields (Cane et 

ated with gene activation in the ceU and the trigger- .^\ 1993)^ and by low-energy laser irradi- 

ing of a cascade of intracellular events initiating ^^.^^^ (LELI) in vivo (Kusakari et al., 

changes in physiologic processes in tlie celi. OQO^ 

,- ,T._,_r->i 1,^-ri^M Low-energy laser irradiation has recent- 

INTRODUCTION ,1 r 1 . j 1 . lv.i^^ 

ly been lound to modulate various biolog- 

The process of skeletal muscle regen- ical processes in tissue cultures and animal 

eration after injuiy has been well docu- models (Belkin et al., 1988; Karu, 1989, 

mented and reviewed in mammals (All- 1998; Galletti et al., 1992; Conlan et al., 

brook, 1981; Carlson and Faulkner, 1983), 1996). For example, laser irradiation was 

and to a much lesser extent in amphibians found to increase mitochondrial respira- 

(Carlson, 1970). The possible regulatory tion and adenosine triphosphate (ATP) 

mechanisms associated with the process of synthesis (Morimoto et al., 1994; Zhu et 

muscle regeneration, and the stimulation al., 1997) and to modulate oxidative me- 

of the dormant satellite cells after injury tabolism in the mitochondria (Yu et al., 

are not yet fully understood. 1997). Low-energy laser irradiation also 

Osteogenesis and bone repair after trau- was demonstrated to elevate activity of an- 

ma are regulated by such mechanisms as tioxidant enzymes such as superoxide dis- 

growth factors and hormones, among oth- mutase and catalase (Zhu et al., 1997) and 

ers (Marks and Popoff, 1988). Several ex- to enhance nitric oxide (NO) production 

perimental models have followed the heal- in rat lymphocytes (Chi et al., 1995). The 

ing of osseous defects (usually fractures) in biostimulatoiy effect of LELI was also in- 
vestigated in several experimental animal 

r^ , ^ o A, 7- T- 1^ models after injury. Helium-neon laser ir- 

' Department of Zoologv, George S. Wise Faculty -,. . -^i ..1l i 

of Life Sciences, Tel A^v University, Ramat Aviv radiation was demonstrated to elevate 

69978, Israel. coinpouiid action potential after crush in- 
Bull. Mus. Comp. Zool., 156(1): 297-303, October, 2001 297 

298 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

jury to the ischiadic nerve (Rochkind et al., ture (3-4 months old, 300-350 g body 

1987), as well as slow Wallerian degener- weight) Sprague-Dawley male rats (Lev- 

ation in the injured optic nerve (Schwartz instein Inc., Yoknean, Israel). Rats were 

et al., 1987; Assia et al., 1989). The inflam- anesthetized with Avertin (1 ml/100 g body 

matory response after cold injury to toad weight intraperitoneally). A longitudinal 

muscle also was demonstrated to be mark- incision was made through the skin and 

edly decreased by laser irradiation (Bibi- muscles to expose the tibia on its proximal 

kova and Oron, 1993), and neoformation and medial surfaces about 10—15 mm dis- 

of blood vessels in the injured zone was tal to the knee joint. A hole was drilled in 

demonstrated to be elevated by laser ir- the cortical bone of the tibia in the center 

radiation (Bibikova et al., 1994). Recently, gf its medial aspect about 9 inm distal to 

we have shown that LELI causes indue- the knee joint (diaphyseal region) using 

tion of cell cycle regulatory proteins in sat- hand-driven dental drills of increasing di- 

ellite cells from skeletal muscles due to ac- ameter in order to obtain a final hole di- 

tivation of early cell cycle regulatory genes ameter of 1.6 mm. The hole was created 

(Ben-Dov et al., 1999). in such a manner as to penetrate the cor- 

The aim of the present study was to in- tical bone and damage the trabeculae in 

vestigate the effect of LELI on skeletal the medullary canal, but not to damage the 

muscle regeneration in rats and toads and contralateral cortical bone. Care was taken 

bone repair in rats. not to injure the periosteum in the vicinity 

of the hole. After the inuscles and sldn 

MATERIALS AND METHODS were sutured, the rats were injected intra- 

Surgical Procedures muscularly with Penicillin G Sodium 

(leva, retach- 1 ikva, Israel) at a dose ot 

Foity-nine male toads {Bufo viridis) and 100 U/g body weight, followed by 3 days 

30 mature male laboratory rats (Charles of bioxin (Solmycin 500, Teva, Petach-Tik- 

River) were used for the experiments. Par- va, Israel) in the drinking water (1 g/L). In 

tial excision to the gastrocneinius inuscle the above model a spontaneous partial 

of the rats was performed as previously de- healing of the gap injury (probably due to 

scribed (Roth and Oron, 1985). Cold in- its low diameter relative to the tibia width) 

jury to the toads' muscles was performed occurs at about 3 weeks after injury, and a 

by placing the end of a copper rod (1.8- complete healing occurs at longer time in- 

mm diameter), prechilled in liquid nitro- teivals. 
gen, against the muscle for 10 seconds. 

This created an injured zone of about 4 Laser Irradiation 

mm in diameter. Four to six injured ex- An He-Ne laser (Ealing, Electro-Optics, 

perimental (laser-irradiated) and control Holliston, Massachusetts, USA) was em- 

(injured and irradiated by red light) toads ployed at 632 nm, 5.3-mW power output, 

or rats were used for each time interval (9, and 1.9-miTi beam diameter. Laser irradi- 

14, and 30 days for toads; 3, 8, and 11 days ation was always applied directly on the 

for rats) postinjury. In brief, after removal injured zones of the skeletal muscles of the 

of the skin and biceps femoris muscle the toads and rats after removal of the skin 

gastrocnemius was exposed. A special de- and muscle. In order to cover the total 

vice with two fixed scalpel blades was ap- area of the injured zone (which was larger 

plied to create a fixed excision (5X2 mm) than the laser beam diameter) in the mus- 

in the middle part of the lateral belly of cles, the laser was applied several times to 

the gastrocnemius muscle. The biceps fe- each of the visibly injured zones. Laser ir- 

inoris and the skin were then closed. radiation to toad injured zones in the mus- 

Surgery to create a fixed injuiy to the cles was applied every alternate day, from 

tibia was performed on a total of 52 ma- 4 days until 14 days post— cold injury, for 

Regeneration of Bone and Skeletal Muscle • Oron 299 

2.3 minutes each time (31 J/cm^). Laser ume fraction of typical stnictures such as 
irradiation to the rats' skeletal muscles was inononucleated cells, inyotubes, and 
applied as for the toad, but only on the young myofibers in the injured area. The 
second and third day postinjury. In the laser irradiation during the process of 
case of the hole injury in the tibia, laser inuscle regeneration caused a significant 
irradiation was applied on days 5 and 6 change in the voluine fraction of these 
postinjury once a day for 2.3 minutes (31 structures in the trauinatized area at all 
J/cm-) directly on the hole injury after time intervals after injury. Young myofi- 
careful removal of the sutures in the skin bers populated 15.5 ± 7.9% and 65.0 ± 
and muscles above it. In all control (sham- 9.5% of the muscle regenerates in the in- 
operated) experimental animals, the laser jured zone in the laser-irradiated toad 
was applied but was not connected to a muscles at 9 and 14 days, respectively; 
power source. whereas in control nonirradiated inuscle 

regenerates young myofibers were not ev- 

Histology and HiStomorphometry i^ent at 9 days postinjury, and comprised 

At various time intervals postinjury the only 5.3 ± 2.9% of the area at 14 days 
rats were anesthetized with chloroform, (Figs. 1, 2a). The process of muscle regen- 
and the gastrocnemius muscle exposed, re- eration in toads was almost completed af- 
moved, fixed in Bouin's fixative, and em- ter 30 days (90% of the regenerated area 
bedded in paraffin. Serial sections were was occupied by mature muscle fibers) in 
prepared from each muscle and stained the laser-treated muscles, whereas in con- 
with hematoxylin and eosin and Masson's trol muscles only young myofibers still 
trichrome stain. Morphometric measure- populated a large part (75.7 ± 13.2%) of 
ments were per^formed on the entire in- the injured zone (Fig. 2b). In the case of 
jured zones of four to six randomly chosen the rat gastrocnemius muscle, young myo- 
sections per each muscle, using the point fibers were not evident in the injured zone 
mounting method. The volume fraction of either experimental or control muscles 
(% of the total volume of injured zone) at 3 days postinjury. Their volume fraction 
was calculated for each of the structures was twofold significantly higher in LELI- 
analyzed. The results were finally statisti- treated rat muscles as compared to control 
cally analyzed using the three-level nested muscles at 8 and 11 days postsurgeiy, re- 
analysis of variance. spectively (data not shown). 

Injured tibial bone from six rats was tak- The morphometric analysis of the tissue 

en at each time interval (10, 13, and 15 components of the hole in the rat tibia in- 

days) postinjury. This bone was fixed, de- dicated that at all time intenals (10, 13, 

calcified, and processed for histology as and 15 days) the relative areas occupied 

described above. Histomorphometry was by compact bone in the hole injury were 

performed to determine the area fraction higher in the LELI-treated rats than in the 

of various structures (woven bone, com- control (Figs. 3, 4). At 15 days postinjury 

pact bone, and so on) in the gap created to the tibia, this value comprised 92 ± 9% 

in the tibia out of the total area of the gap in the LELI-treated rats, which was sig- 

(injured) zone. This process was per- nificantly (P < 0.01) higher than the area 

formed with a microscope and the aid of that the compact bone occupied (58 ± 

a video camera and screen using Sigma 8%) in the control nonirradiated rats. 

Scan software. __ ^ 


ntbULIb 'pj^g results of the present study clearly 

The process of muscle regeneration in indicate that the process of muscle regen- 

control muscles after cold injury was char- eration after partial excision injury or cold 

acterized by sequential changes in the vol- injury is markedly promoted by direct ex- 

300 Bulletin Museum of Comparative Zoology, Vol. 156, No. 1 

Figure 1 . Light micrographs of regenerated area in control nonirradiated (a) and laser-irradiated (b) toad gastrocnemius muscle 
14 days postinjury. Note mainly myotubes (MT) and mononucleated (MN) cells in (a) as compared to mainly young myofibers 
(YU) with large diameter in (b). x40. 












Histological structures 


Days after injury 


Figure 2. Volume fraction of young myofibers at various time intervals post-cold injury to the toad gastrocnemius muscle injury 
(a) and of mononucleated cells (MN), young myofibers (YM), and mature myofibers (MP) at 30 days postinjury (b). Values are 
mean ± SEM of control nonirradiated (dashed columns) and laser-irradiated (solid columns) muscles. 

Regeneration of Bone and Skeletal Muscle • Oron 







10 13 15 

Days After Injury 

Figure 3. Volume fraction of compact bone at the site of in- 
jury to the rat tibia at different time intervals postinjury. Note 
less compact bone in the control nonirradiated rats (dashed 
columns) as compared to low-energy laser-irradiated rats (sol- 
id columns). *P < 0.05; **P < 0.01. 

posure to He-Ne laser irradiation (during 
the regeneration process) in both toads 
and rats, although in the toads the pro- 
motion effect was higher. We have recent- 
ly shown that LELI promotes proliferation 
of satellite cells from skeletal muscle origin 
grown in culture (Ben-Dov et al., 1999). 
These cells are considered to be the stem 
cells for muscle regeneration and, there- 

fore, their enhanced proliferation by LELI 
in vitro may explain the enhanced kinetics 
of regeneration in vivo, as demonstrated in 
the present study. Because we have pre- 
viously found that LELI also proinotes 
new blood vessel formation (angiogenesis) 
at the site of skeletal inuscle regeneration 
(Bibikova et al., 1994), it might be postu- 
lated that a better oxygen and nutrient 
supply will also contribute to enhancement 
of the regeneration. Assia et al. (1989) sug- 
gested that in the case of optic nerve re- 
generation after crush injury, LELI acts to 
slow degenerative processes rather than 
promoting regeneration, and that the ef- 
fect of laser irradiation is transient and 
subsides markedly when irradiation ceases. 
In the present work, the effect of laser ir- 
radiation was not transient because the 
ainount of newly formed young myofibers 
was twofold higher in the injured zone of 
experimental rats after two subsequent la- 
ser irradiations. Tissue reaction to injury in 
cranial nerves and skeletal muscles is most 
probably entirely different and they may 
react differently to laser irradiation. The 
results of the present study also indicate 

Figure 4. Light micrographs of longitudinal section of the injured zone in the tibia of control (a) and laser-irradiated (b) rat at 
13 days postinjury. Note the gap in the tibia (marked by arrows) that is filled only with woven bone (WB) in the control nonir- 
radiated rat and partially filled with compact bone (CB) that bridges the gap in the laser-irradiated rat. x120. 

302 Bulletin Museum of Comparative Zoologtj, Vol. 156, No. 1 

that LELI may promote, to a much higher hances synthesis of ATP. Changes in plas- 

extent, biological processes in cells with ma membrane conduction and transient 

lower metabolic rates such as in the toads increase of calcium flux into the cells have 

(as compared to the rats). These results also been observed in cells irradiated in 

corroborate the notion that during a high vitro by low energy lasers (Lubart et al., 

state of metabolism and proliferation, cells 1997). These changes are photochemical 

are affected to a lesser extent by laser ir- in nature; the energy is probably absorbed 

radiation (Karu, 1989, 1998). in intracellular chromophores and con- 

This communication also indicates by verted into metabolic energy involving the 

direct measurements that the rate of bone respiratory (cytochrome) chain (Karu, 

repair in the cortical part of the tibia in an 1989, 1998; Galletti et al, 1992). 

experimental model in the rat is enhanced We have recently shown (Ben-Dov et 

by LELI. Kusakari et al. (1992) previously al., 1999) that LELI causes induction of 

reported stimulation of DNA and protein cell cycle regulatory proteins in satellite 

synthesis, and alkaline phosphatase activity cells from skeletal muscles, due to activa- 

in osteoblastlike tissue culture by LELI. tion of early cell cycle regulatory proteins. 

These results corroborate the results of the Furthermore, it was recently found (un- 

present study indicating that the enhance- published data, this laboratoiy) in the 

ment of bone repair may be caused by en- same system that LELI induces activation 

hanced proliferation of osteoblasts at the of receptors on the cell plasma membrane 

injured site. and other components associated with cell 

The present study demonstrates that proliferation (kinase enzymatic activity) in 
LELI can promote skeletal muscle regen- a certain signal transduction pathway in 
oration as well as bone repair, indicating the cell. Thus, it may be postulated that 
that the biostimulation may have a com- the LELI biostimulates the cell via certain 
mon mechanism that triggers certain cell mechanisms at the molecular level. How- 
processes that, in turn, will enhance cell ever, the precise interaction of the laser 
proliferation and differentiation in the irradiation with cellular components or 
skeletal tissues. Thus, it may be hypothe- molecules that trigger a cascade of intra- 
sized that the mechanism of biostimulation cellular processes that eventually lead to 
by LELI may be general for various tissues changes in physiologic processes in the 
and in different groups of vertebrates, cells will have to be elucidated by further 
However, parameters such as energy, studies, 
wavelength, and timing of laser may differ 

for different tissues and according to the LITERATURE CITED 

physiologic state of the cells to be biosti- Allbrook, D. 1981. Skeletal muscle regeneration, 

mulated. Muscle and Nerve, 4: 234-245. 

The precise regulatory mechanisms as- ^^su, E. M., M. Rosner, M. Belkin, A. Solomon, 

. . J -.1 1 • X.- 1 1. rr i. r AND M. Schwartz. 1989. Temporal parameters 

sociated with biostimulatory eirects oi r, i  ^- .■ r ; i j i 

-^ or low energy laser irradiation tor optimal delay 

LELI are not yet fully understood. Laser ofpost traumatic degeneration of rat optic nerve. 

irradiation probably induces changes in Brain Research, 476: 205-212. 

cellular homeostasis, initiating a whole cas- Belkin, M., B. Zaturunsky, and M. Schwartz. 

cade of reactions. A number of respiratoiy J^^^- ^ ^"^^^^ f ^f ^ f !°^ f "^""^ '^f 1^°^^" 

, . ,. in tects. Laser Light Ophthalmology, 2: 63-/1. 
Cham components (i.e., cytochromes, fla- ben-Dov, N., G. Shefer, A. Irintchev, a. Wer- 
vine dehydrogenases, and so on) may be nig, U. Oron, and I. O. Hale\t. 1999. Low 
primaiy photoacceptors or chromophores energy laser irradiation affects cell proliferation 
of laser photons and thus able to absorb ^"^ differentiation in vitro. Biochimica et Bio- 
1 1- lA. J. 0.1 1 J.1 T-i • physica Acta, 1448: 372-380. 
laser light at the proper wavele^igths. This bibikova. A., A. Belkin, and U. Oron. 1994. En- 
causes short-term activation of the respi- hancement of angiogenesis in regenerating gas- 
ratory (cytochroine) chain that, in turn, en- trocnemius muscle of the toad (Bttfo viridis) by 

Regeneration of Bone and Skeletal Muscle • Oron 


low energy laser irradiation. Anatomy and Em- 
bryology, i90: 597-602. 

BiBIKOVA, A., AND U. OrON. 1993. Promotion of 
muscle regeneration in the toad {Bitfo viridis) 
gastrocnemius muscle by low energy laser irra- 
diation. Anatomical Record, 235: 374-380. 

Cane, V, P. Bom, and S. Soana. 1993. Pulsed mag- 
netic fields improve osteoblast activity during the 
repair of an experimental osseous defect. Journal 
of Orthopaedic Research, 11: 664-670. 

Carlson, B. M. 1970. Histological observations in 
the regeneration of mammalian and amphibian 
muscle, and myogenesis, pp. 38-74. In A. Mauro, 
S. A. Shafiq, and A. T. Milhorat (eds.). Regen- 
eration of Striated Muscle, and Myogenesis; Pro- 
ceedings of the International Conference con- 
vened by the Muscular Dystrophy Associations 
of America at the Institute of Muscle Disease, 
New York, March 28-29, 1969. Amsterdam: Ex- 
cerpta Medica. x -I- 299 pp. 

Carlson, B. M., and J. A. Faulkner. 1983. The 
regeneration of skeletal muscle fiber following 
injury. A review. Medical Sciences in Sports and 
Exercise, 15: 107-198. 

Chl L. H., W. Yu, J. O. Naim, and R. J. Lanzafame. 
1995. Increased synthesis of nitric oxide by laser 
irradiation in sepsis. Lasers in Surgery and Med- 
icine, 7(Suppl.): 19. 

Conlan, M. J., J. W. Rapley, and C. M. Cobb. 1996. 
Biostimulation of wound healing by low energy 
laser irradiation. A review. Journal of Clinical 
Periodontics, 23: 492-496. 

Gallettl G., L. Bolognani, and G. Ussia. 1992. 
Laser Applications in Medicine and Surgery. Bo- 
logna: Monduzzi Editore. 552 pp. 

Karu, T. 1989. Photobiology of low power laser ef- 
fects. Health Physics, 56: 691-704. 

Karu, T. 1998. Photobiology of Low Power Laser 
Therapy. Amsterdam: Gordon and Breach Pub- 
lications. 282 pp. 

Kusakarl H., N. Orisaka, and H. Tanl 1992. Ef- 
fects of low lasers on wound healing of gingiva 
and bone, pp. 49-56. In G. Galletti, L. Bolog- 
nani, and G. Ussia (eds.). Laser Applications in 
Medicine and Surgery. Bologna: Monduzzi Edi- 
tore. 552 pp. 

Liang, C. T., J. Barnes, J. G. Seeder, H. A. Quar- 


RODAN. 1992. Impaired bone activity in aged 
rats: alternation and the cellular and molecular 
levels. Bone, 13: 435^41. 

LuBART, R., H. Friedman, N. Grossman, N. Co- 
hen, AND H. Breitbart. 1997. Reactive oxygen 
species and photobiostimulation. Trends in Pho- 
tobiology, 4: 277-283. 

Markel, M. D., M. a. Wikenheiser, and E. Y. S. 
Chao. 1991. Formation of bone in tibial defects 
in a canine model. Journal of Bone and Joint Sur- 
gery, 73A: 914-923. 

Marks, S. C, and S. N. Popoff. 1988. Bone cell 
biology: the regulation of development structure 
and function in the skeleton. American Journal 
of Anatomy, 183: 1-44. 

MORIMOTO, Y, T. Arai, M. Kikuchi, S. Nakajima, 
and H. Nakamura. 1994. Effect of low intensity 
argon laser irradiation on mitochondria respira- 
tion. Lasers in Surgery and Medicine, 15: 191- 

Rochkind, S., L. Barr Nea, a. Bartal, M. Nissan, 
R. LUBART, and N. RazON. 1987. New methods 
of treatment of severely injured sciatic nerve and 
spinal cord. Acta Neurochirurgia, 43: 91-93. 

ROTH, D., and U. ORON. 1985. Repair mechanisms 
involved in muscle regeneration following partial 
excision of the rat gastrocnemius muscle. Exper- 
imental Cell Biolog)', 53: 107-114. 

Schwartz, M., A. Doron, M. Ehrlich, V. Lovic, 
S. Benbasat, M. Belkin, and S. Rochkind. 
1987. Effect of low energy He-Ne laser irradia- 
tion on post traumatic degeneration of adult rab- 
bit optic nerve. Lasers in Surgery and Medicine, 
7: 51-55. 

Spencer, F. M., C. C. Liu, R. C. C. Si, and G. A. 
Howard. 1991. In vivo action of insulin-like 
growth factor I (IGF-I) on bone formation and 
resorption in rats. Bone, 12: 21-26. 

Yu, W., M. McGowan, K. Iffolito, and R. J. Lan- 
zafame. 1997. Photomodulation of oxidative me- 
tabolism and electron chain enzymes in rat liver 
mitochondria. Photochemistry Photobiology, 66: 

Zhu, Q., W. Yu, X. Yang, G. L. Hicks, R. J. Lan- 
zafame, and T. Wang. 1997. Photo-irradiation 
improved functional preservation of the isolated 
rat heart. Lasers in Surgery and Medicine, 20: