o /\(/xl u ur APR 2 11977 
B R E V I O R A^ 
Miuseiim of Comparative Zoology 
us ISSN 0006-9698 
CambridCxE, Mass. 8 April 1976 Number 436 
STUPENDEMYS GEOGRAPHICUS, 
THE WORLD'S LARGEST TURTLE 
Roger Conant Wood^ 
Abstract: Stupendemys geographicus, a gigantic fossil pelomedusid turtle 
from the late Tertiary (Huayquerian) Urumaco Formation of northern 
Venezuela is described. Stupendemys was evidently a highly aquatic form. 
Whether it was a fresh water or marine turtle, however, cannot be determined 
with certainty on the present evidence. One or perhaps even both pairs of 
limbs may have been modified into flippers, and the head may not have 
been fully retractable in the usual pleurodiran manner. Comparisons with 
records of other enormous chelonians reveal that the carapace of Stupen- 
demys is larger than that of any other turtle, fossil or recent. 
INTRODUCTION 
Paleontologists are occasionally fortunate enough to make 
totally unexpected discoveries. Such was the case during the 
summer of 1972, when a Harvard paleontological expedition 
working in late Tertiary deposits of northern Venezuela un- 
earthed the remains of several huge fossil turtles. These cer- 
tainly attained greater size than any other extinct chelonians yet 
known; they also appear to be larger than any living ones and 
hence the largest turtles that ever existed. The purpose of this 
paper is to describe these gargantuan creatures. 
The following abbreviations are used : 
AMNH: American Museum of Natural History, herpetological 
collections 
MCNC: Museo de Ciencias Naturales, Caracas 
MCZ: Museum of Comparative Zoology: (H), herpetological 
collections; (P), paleontological collections 
PU: Geology Museum, Princeton University 
iFaculty of Science and Mathematics, Stockton State College, Pomona, N. J. 
08240. 
2 BREVIORA No. 436 
SYSTEMATICS 
Order Testudines 
Suborder Pleurodira 
Family Pelomedusidae 
Sttipendemys^ gen. nov. 
Plate 1 and Figures 1-3, 5, 6, and 9 
Type species. S. geographicus^ sp. nov. 
Distribution. Huayquerian, Venezuela 
Diagnosis. Shell gigantic; carapace depressed, with irregular 
nodular contours on external surface and deep median notch at 
front; anterior border of nuchal bone thickened and moderately 
to strongly upturned; posterior peripheral bones moderately 
scalloped along margins; neurals arranged in uninterrupted 
sequence, numbers two through six hexagonal, the seventh 
pentagonal. Mesoplastra hexagonal to subcircular, largely con- 
fined to bridge; lateral ends of pectoral-abdominal scute sulci 
terminating just anterior to axial notches of shell. 
Cervical vertebrae (probably seventh and eighth) with saddle- 
shaped articulations; neural arches relatively high in relation 
to anteroposterior lengths of centra; angle of neural arch of 
presumed eighth cervical with horizontal plane greater than in 
any other pelomedusid; articular facets of postzygapophyses of 
both cer\dcals forming acute angle of less than ninety degrees 
with respect to each other; prezygapophyses of presumed eighth 
cervical directed more perpendicularly than in other pelomedu- 
sids; thin, bladelike spine on anterior face of eighth neural arch; 
no ventral keel on eighth centrum. 
Angle of divergence between two ventral processes of scapu- 
locoracoid roughly ninety degrees; ventromedial process of scap- 
ula dorsoventrally flattened; coracoid greatly thickened along 
medial edge; glenoid socket facing forward rather than laterally. 
Humerus squat, massive, lacking ectepicondylar groove of 
foramen; deep bicipital fossa between radial and ulnar articular 
facets on ventral surface; prominent ridge traversing ventral 
surface of shaft from ulnar process to distal end, terminating 
iThe generic name alludes to the astonishing size of this turtle, and the 
species is named in honor of the National Geographic Society in recognition 
of its generous support of my research on turtles. 
1976 world's largest turtle 3 
just above radial condyle; ulnar condyle broadest at anterior 
end; ulnar and radial condyles facing somewhat more ventrally 
than in other pelomedusids; entepicondyle and supinator process 
strongly developed, resulting in distal expansion of humerus 
almost as great as that of proximal end; shaft triangular in 
cross-section rather than circular. 
Femur squat, massive, greatly flattened dorsoventrally ; 
breadth of tibial condyle approximately one-third total length 
of bone. 
Stupendemys geographictis sp. nov. 
Type. MCNC 244, medial portion of the carapace with asso- 
ciated left femur, frag'ments of a scapulocoracoid and a cervical 
vertebra, probably the eighth. 
Hypodigm. The type, and MCZ(P) 4376, much of the cara- 
pace, fragments of the plastron, a cervical vertebra (probably 
the seventh), both scapulocoracoids and a caudal vertebra; 
MCNC 245, a plastron lacking the epiplastra and entoplastron, 
two nearly complete pleurals, several peripherals, and one neu- 
ral, all from the same individual; MCZ(P) 4377, a cervical, 
probably the eighth; and MCZ(P) 4378, a left humerus. 
Horizon and localities. "Capa de huesos" (also known as 
"Capa de tortugas"), upper member of the Urumaco Forma- 
tion, Huayquerian (which is probably of Pliocene age; see, for 
example, Simpson, 1974: 5). 
Outcrops of the Urumaco Formation are restricted to a rela- 
tively small area in the northwestern part of the state of Falcon, 
centering around the now-abandoned El Mamon oil field (lat. 
11°13'N, long. 70°16'W), just north of the town of Urumaco. 
The type was found immediately west of Ouebrado Tio Gre- 
gorio, near its mouth. Other specimens were found as follows: 
MCZ(P) 4376 — one-half km north of Ouebrado Picacho and 
50' m east of the Chiguaje fault; MCZ(P) 4377 — three and 
one-half km north 30° west of El Picacho on the up side of the 
Chiguaje fault; MCZ(P) 4378 — as for the previous specimen, 
but "about 15 m higher in the section; MCNC 245 — three- 
quarters km north of Kilometer 153 on the oil pipeline running 
from Punta Gorda to the Paraguana Peninsula (same locality 
as MCNC 238, a trionychid; Wood and Patterson, 1973). 
Diagnosis. As for the genus. 
4 BREVIORA No. 436 
DESCRIPTION 
Shell. The most complete carapace is that of MCZ(P) 4376 
( Plate 1 ) , which lacks some of the anterior peripherals on the 
right side, as well as peripherals from the bridge region on both 
sides. Scute sulci are deeply impressed onto the external surface 
but, as in many giant chelonians, most of the bone sutures have 
become largely fused and the pattern of these cannot be traced 
with any degree of certainty. The carapace is low-arched in 
the manner typical of aquatic turtles, and its dorsal surface, 
rather than being smooth, is somewhat nodose. There is a 
strong median indentation at the anterior margin of the cara- 
pace that is unique among pelomedusids (and perhaps even 
among turtles in general) in having the bone of this region 
curled up into a thickened, collarlike structure. Posterior to the 
bridges, the peripheral bones have mildly scalloped margins. 
The sacral region of this specimen is fairly well preser\'ed. There 
are four sacral ribs abutting against the attachments of the ilia 
onto the visceral surfaces of the eighth pair of peripherals; the 
distal ends of the last two of these are fused together. This is 
essentially the same pattern as reported by Zangerl ( 1 948 : 
30-31 and pi. 4, fig. 3) for the largest living South American 
pelomedusid, Podocnemis expansa. There is a slight postero- 
medial overlap of the iliac scars onto the suprapygal. Whether 
these also extended forward onto the under surface of the 
seventh pair of pleurals (and if so, to what extent) is uncertain 
because the course of the suture between the seventh and eighth 
pairs of pleurals cannot be determined. Measurements of this 
carapace are given in Table 1. 
The carapace of the type specimen ( Fig. 1 ) differs in several 
respects from that of the one just described and moreover pro- 
vides information about the shape and arrangement of the neu- 
ral bones not revealed by the more complete specimen. Meas- 
urements of the vertebral scutes of the two carapaces (Table 1) 
indicate that the type was somewhat larger, roughly by five 
per cent. Its midline length, therefore, would have been in the 
neighborhood of ten to twelve centimeters longer, giving an 
estimated midline length of as much as 230 centimeters. The 
curling and thickening of bone at the anteromedian indentation 
is less pronounced in the type than in MCZ(P) 4376. The out- 
lines of six neural bones can be traced on this specimen. The 
pattern revealed is typical for South American pelomedusids; 
the last neural, which I believe to be the seventh, is pentagonal 
1976 world's largest turtle 5 
while those anterior to it (presumably the second through sixth) 
are hexagonal. The neurals, again typically, tend to become 
progressively broader in relation to their anteroposterior length 
toward the rear of the series (Table 2). As far as can be de- 
termined, the neurals were arranged in an uninterrupted se- 
quence. Behind the last neural, part of the seventh and all of 
the eighth pair of pleurals meet in the midline. 
An isolated neural bone from another specimen (MCNC 
245 ) adds further information about the structure of the median 
part of the carapace. The bone is hexagonal and somewhat 
longer than broad (Table 2), indicating that it comes from the 
anterior part of the series. Because the first neural of pelomedu- 
sids is usually elongate and rectangular or oval, it seems reason- 
able to assume that the specimen in question is either the second 
or third. The bone was obviously in direct contact with neurals 
both to the front and rear. This reinforces the impression al- 
ready given by the type carapace that the neural series was 
continuous, and, in fact, if the neural is actually the second 
rather than the third, proves the point. A notable feature of 
this neural is its exceptional thickness in proportion to its length 
and width; at various places around the periphery the bone 
measures 2.8, 2.6^ and 2.4 centimeters dorsoventrally. In gen- 
eral, pelomedusid neurals tend to be proportionately much 
thinner. Although it is not feasible to measure the thickness of 
the individual neurals of the type carapace, it is possible to state 
that the carapacial bone does appear to be disproportionately 
thick, even for a turtle of such exceptional size. Perhaps the 
unusual thickness of the shell should be considered a diagnostic 
character of the taxon. 
There is nothing remarkable about the carapace scute pattern 
of S. geographicus. It is virtually indistinguishable from that of 
any of the living South American pelomedusids which, except 
for minor variations, are all very similar. 
No identifiable plastral remains are associated with the type 
specimen. However, the mesoplastra, hyoplastra, and right 
hypoplastron of MCZ(P) 4376 were recovered; these had been 
crushed down into and molded against the shallow bowl-shaped 
depression formed by the visceral surface of the carapace (the 
shell was found lying upside down) and unfortunately preserve 
little in the way of detail. Nevertheless, the presence of meso- 
plastra in conjunction with pelves that were clearly fused to the 
shell leaves no doubt that these gigantic turtles are pelomedusids. 
6 BREvioRA No. 436 
The mesoplastra are relatively small, hexagonal to subcircular 
elements, laterally positioned and confined largely to the bridge. 
This is the standard configuration for all known living and 
fossil South American pelomedusids. On the basis of size and 
thickness, I have referred a fairly complete plastron and some 
miscellaneous carapacial fragments ( MCNC 245 ; Fig. 2 ) to 
Stupendeynys. Although very large by ordinarv' pelomedusid 
standards (Table 3), this plastron is relatively small in com- 
parison to the carapaces described above. Presumably it repre- 
sents a young adult. The forward portion of the anterior lobe is 
missing. This is regrettable because it is this part of the pelo- 
medusid shell that is generalh' the most useful for taxonomic 
purposes. Nevertheless, some interesting characteristics are evi- 
dent. The bridge is considerably longer at its base than the 
posterior plastral lobe (Table 3). The bone is exceptionally 
thick in proportion to its length and breadth. And, most notably, 
the lateral ends of the pectoral-abdominal scute sulci terminate 
just in front of the bases of the shell's axial notches, on the 
edges of the anterior plastral lobe. This position is in contrast 
to other South American fossil and recent pelomedusids in which 
these sulci typically meet marginal scute sulci on the forward 
third of the bridge, usually just in front of the anterior meso- 
plastral bone sutures. The plastral formula, insofar as it can 
be determined, is: femoral > abdominal > anal. 
Axial skeleton. The three cervical vertebrae that have been 
recovered (MCZ[P] 4376, MCZ[P] 4377, and MCNC 244) 
belong to three different individuals and represent only two of 
the eight bones in the series. Measurements of these are given 
in Table 4. Because of the unique morphology of these verte- 
brae, it is difficult to be certain as to their positions in the series. 
In the cervicals of living pelomedusids, the neural arches be- 
come increasingly prominent from front to rear, that of the 
eighth always having the greatest height in relation to the length 
of the centrum (Table 4). The two morphologicallv identical 
fossil cervicals (MCZ[P1 4377 and MCNC 244) have neural 
spines that are, relatively, even more prominent than that of 
the eighth cervical in living pelomedusids, while the third 
(MCZ[P] 4376) has an arch only slightly less prominent 
(Table 4). On this basis it would seem likely that we are 
dealing with cervicals at the posterior end of the series, pre- 
sumably the seventh (MCZ[P] 4376) and eighth (MCZ[P] 
4377 and MCNC 244). 
1976 world's largest turtle 7 
However, examination of the central articulations furnishes 
contradictory evidence. Cervicals four, five, and six of all living 
South American pelomedusids have saddle-shaped articulations, 
the seventh is similarly shaped anteriorly but convex posteriorly, 
and the eighth is concave in front and convex behind (Williams, 
1950: 528, 532, 552, and fig. 11). The three known cervicals 
of Stupendemys have saddle-shaped articulations, and hence 
compare in this feature to the fourth through sixth cervicals of 
the extant South American pelomedusids, rather than to the 
seventh or eighth. (Undescribed fossil pelomedusid cervicals 
from the late Cretaceous of Brazil, which I have been able to 
examine through the courtesy of Dr. L. I. Price, are indistin- 
guishable from those of living South American representatives 
of the family.) In living African pelomedusids, the centra of 
cervicals three through eight are uniformly procoelous (Williams, 
ibid. ) . Cervicals are known for only one African fossil pelo- 
medusid (Wood, 1971), and these differ from both living Afri- 
can and South American forms in having articular surfaces 
intermediate in shape between the saddle joints of the latter and 
the procoelous condition of the former. No cervicals have been 
reported for fossil pelomedusids from continents other than 
Africa and South America, the only regions, together with 
Madagascar, where the family still survives. The cervical artic- 
ulations of Stupendemys are therefore most closely comparable 
to those of its South American relatives. 
Because the trend of anteroposteriorly increasing neural spine 
height seems to be consistent in all pelomedusids, whereas the 
pattern of cervical articulation varies somewhat, I am inclined 
to place more reliance in the former feature as a means for 
determining the relative position of the Stupendemys neck 
vertebrae in the cervical series. As Table 4 shows, the height/ 
length ratio of the eighth cervical is always the greatest for any 
individual. Moreover, as shell size increases, the height/length 
ratio also increases, so that it is greater for the eighth cervical 
of Podocnemis expansa than for that of the much smaller Pelo- 
rhedusa subrufa. Given these observations, and in view of the 
fact that the height/length ratios of MCZ(P) 4376 and MCNC 
244 are considerablv s^reater than those recorded for anv of the 
Recent species, while that of MCZ(P) 4376 is about the same 
as the greatest ratio for the largest Recent specimen measured, 
it seems that the cervicals of Stupendemys are from the pos- 
terior part of the series, probably representing the seventh and 
eighth. 
8 BREvioRA No. 436 
If the cervicals of Stupendemys are, in fact, the seventh and 
eighth, then they are unique among known pelomedusids by 
virtue of their saddle-shaped articulations. There are, in addi- 
tion, several other features of these vertebrae that reinforce this 
impression. One of the most obvious is that the neural arch of 
the eighth cervical of Stupendemys makes a much less acute 
angle with the anteroposterior axis of the centrum than do those 
of the comparable cervical in other pelomedusids. (In the 
cervical series of Recent pelomedusids that I have examined, the 
neural arch of the eighth cervical always makes the greatest 
angle to the horizontal plane.) In posterior view, the articular 
facets of the postzygapophyses form an acute angle of less than 
ninety degrees with each other. Those of other pelomedusids 
are nearly horizontal to the dorsoventral axis of the vertebrae 
(fig. 4; see also WilHams, 1950, fig. 11). Viewed laterally, 
the shafts of the prezygapophyses of the presumed eighth cervi- 
cals of Stupendemys are directed much more perpendicularly 
than those of other pelomedusids. Although impossible to meas- 
ure precisely, the angle made with the horizontal plane in the 
specimens of Stupendemys seems to be roughly sixty to seventy 
degrees, whereas in others it is closer to thirty degrees (cf. 
figs. 3 and 4). The thin, median, bladelike spine on the anterior 
face of the neural arch of the presumed eighth cervical of 
Stupendemys is also unlike anything seen on comparable parts 
of other pelomedusid cervicals. In most pelomedusids, the 
ventral surfaces of the cervical centra are typically bowed up- 
wards, sometimes quite strongly, along the anteroposterior axis. 
The one exception known to me is the eighth cervicals of 
South American representatives of Podocnemis. In these, a flat 
blade of bone projects downward from the ventral surface 
( Fig. 4 ) . But in both examples of the presumed eighth cervical 
of Stupendemys, the ventral surface is neither bowed upwards 
nor downwards; it is, instead, flat. Unfortunately, the bottom of 
the presumed seventh cervical vertebra (MCZ[P] 4376) is too 
badly damaged to determine its original shape. 
A single, small caudal vertebra was found in association with 
one of the shells (MCZ[P] 4376). It is poorly preserved and 
reveals no features of special interest. 
Appendicular skeleton. Much of both scapulocoracoids have 
been preserved for MCZ(P) 4376, as well as fragments of one 
belonging to the type. It is not possible to determine with cer- 
tainty the relative lengths of the three prongs making up the 
1976 world's largest turtle 9 
shoulder girdle. The medial tips of the ventromedial portions 
of the scapulae are broken ofT. The dorsal processes of this 
same bone ha\'e been broken at their bases and flattened into 
the same plane as the other two elements. Since their basal 
contacts have been obliterated, it is impossible to determine how 
much (if any) of these processes is lacking. The coracoids, 
however, appear to be complete. Both the left and right ones 
are of essentially the same lengths in MCZ(P) 4376 and are 
considerably lons^er than what remains of the ventromedial 
processes of the scapula, but slightly shorter than the more 
complete of the two dorsal scapular processes that have been 
preserved ( Table 5 ) . These proportions are in accord with 
those of Recent pelomedusids, in which the ventromedial process 
of the scapula is much shorter than the dorsal one, while the 
coracoid is intermediate in length, generally somewhat flattened 
dorsoventrally, and moderately to greatly expanded towards its 
extremity. Despite this incompleteness a number of distinctive 
features are evident. The glenoid socket faces almost directly 
forward in Stupendemys, whereas in typical pelomedusids it 
tends to face in a lateral direction ( Fig. 5 ) . The angle at which 
the two ventral prongs of the scapulocoracoid diverge is con- 
siderably less acute in Stupendemys than in any other known 
pelomedusid (Fig. 5). The shoulder girdle of Stupendemys 
further differs from those of typical Recent South American 
pelomedusids in that the ventromedial process of the scapula is 
dorsoventrally flattened. In specimens of Podocnemis dumerili- 
ana, P. expansa, and P. unifilis that I have examined, this bone 
is anteroposteriorly flattened. The medial side of the coracoid 
of Stupendemys is greatly thickened. This is not true of the 
coracoids in living xA.frican representatives of the family, which 
are uniformly thin, flat, and greatly expanded. In typical South 
American pelomedusids as well as in Podocnemis madagascarien- 
sis, the coracoid is not so expanded but is transversely arched, 
with the apex of the arch on the dorsal side. (The one excep- 
tion of which I am aware is Podocnemis erythrocephala [Mit- 
t'ermeier and Wilson, 1974]; the coracoid of this species does 
not expand at all towards its tip but remains uniformly oval 
along its entire length [e.g., MCZ(H) 10096].) The coracoid 
of Stupendemys may have been similarly arched, if the dorso- 
ventral crushing of this element is taken into account. The 
thickness of bone along its medial edge, however, still seems to 
set it apart from the other South American forms. The dorsal 
scapular process in Stupendemys appears somewhat flattened, 
10 BREvioRA No. 436 
whereas in Recent pelomedusids it is more oval in cross-section. 
This flatness, however, may result from crushing in the hori- 
zontal plane; because of my uncertainty about this feature I 
have refrained from listing it as a diagnostic character. 
A nearly complete left humerus (MCZ[P] 4378) is all that 
is known of the forelimb. This specimen is of great interest in 
that it is totally unlike the humerus of any other known chelon- 
ian — let alone pelomedusid — living or fossil. The head as 
well as the terminal portions of the radial and ulnar processes 
are missing, but otherwise the bone is complete ( Fig. 6 ) . This 
humerus is extraordinarily massive, with distal and proximal 
ends both markedly expanded, the latter slightly more so than 
the former (see Table 5 for measurements). The curvature of 
the shaft does not appear to differ appreciably from that of 
living pelomedusids. There is no trace of an ectepicondylar 
groove or foramen on the dorsal surface, a feature present in 
all other pelomedusids (and, indeed, chelonians in general). 
Between the radial and ulnar processes, on the ventral side, is a 
very deep, semicircular depression, the bicipital fossa. This is 
more prominent than in the fossil pelomedusid Bothremys 
barberi ( Zangerl, 1 948 : 34 and fig. 1 3 ; Gaffney and Zangerl, 
1968) or Podocnemis but is developed to about the same 
extent as in Pelomedusa or Pelusios. Immediately above the 
articular facets on the ventral surface at the distal end of the 
shaft is a very deep, triangular fossa. This seems to be a natural 
depression rather than the result of poor preservation of the 
bone and has no equivalent, so far as I have been able to de- 
termine, elsewhere within the order. A thick, prominent ridge 
extends transversely across the ventral surface from the base 
of the ulnar process to a point adjacent to the radial condyle. 
Such ridges are absent in living pelomedusids, although less 
pronounced ones have been reported in fossil pelomedusids, 
Bothremys (Zangerl, 1948) and Taphrosphys (Gaffney, 1975; 
Fig. 8, this paper). Typically, the ulnar condyle in pelomedu- 
sids has a spool-shaped outline, equally expanded at both ends. 
The ulnar condyle of Stupendemys, however, is markedly 
broader at its anterior end than at its posterior limit. A further 
distinctive feature of Stupendemys is that the trochlea extends 
farther onto the ventral surface than in other pelomedusids. 
To either side of the trochlea, the supinator process and 
entepicondyle bulge outwards, the latter especially. Only in 
Taphrosphys is the distal end of the humerus expanded to 
such an extent (distal width over total length equals 0.47 in 
1976 world's largest turtle 11 
Taphrosphys [Gaffney, 1975, p. 16], 0.44 in Stupendemys). 
In cross-section, midway between the ends, the shaft is triangular 
rather than circular or oval, as is typically the case for pelo- 
medusids. 
A left femur (Fig. 9) was found associated with the type 
shell. The head and terminal portions of both trochanters are 
missing, as well as some bone from an area at the distal end 
of the dorsal surface. The distal articular surfaces, however, 
have been largely preserved. If complete, the femur would 
have been of essentially the same length as the only known 
humerus (Table 5). Like the humerus, the femur of Stu- 
pendemys is massive. Its shaft is oval in cross-section and 
greatly flattened dorsoventrally. The shaft of Podocnemis ex- 
pansa is also oval in coss-section but is instead flattened antero- 
posteriorly. As for the humerus of Stupendemys, the curvature 
of its femur does not seem to differ significantly from that of 
living pelomedusids. The distal end of the shaft is markedly 
expanded, much more so than in Podocnemis expansa (distal 
width over total length equals 0.47 in Stupendemys, 0.29 in 
P. expansa [MCZ(H) 4469]). 
DISCUSSION 
Stupendemys has many very unusual anatomical features. 
No modern chelonian is at all comparable to it, nor does it 
closely resemble any of the better known fossil turtles. 
Its systematic position, at least, is clear: it is an aberrant 
member of the Pelomedusidae. This is conclusively demonstrated 
by several characters : 1 ) the presence of mesoplastra ; 2 ) fusion 
of the pelvis to carapace and plastron; and 3) shape of the 
cervical articulations. 
It is when one strives to understand Stupendemys as a living 
animal that difficulties arise. In the following pages I attempt 
a functional analysis of the known parts of the skeleton, search- 
ing for clues to behavior and habitat. 
The relatively low-arched carapace of Stupendemys indicates 
that it was almost certainly a highly aquatic form, as are all 
living pelomedusids and most fossil ones. Pelomedusids (not 
yet formally described) from two different African fossil local- 
ities, one of Oligocene and the other of Miocene age, are the 
only terrestrial members of the family yet known (Wood, 1971). 
These forms had extremely high-domed shells, superficially very 
tortoiselike in appearance. Conversely, the only strictly terres- 
12 BREVIORA No. 436 
trial, flat-shelled turtle is the exotic pancake tortoise of East 
Africa, Malacochersus, and its shell structure represents an 
adaptation to most unusual habits. Shell shape thus seems to 
be a nearly infallible indicator as to whether a chelonian was 
aquatic or terrestrial, and Stupendemys clearly falls into the 
former catesrorv. 
The strong median indentation at the front end of the cara- 
pace is not characteristic of pelomedusids in general, but is 
reminiscent of the condition seen in the unrelated, big-headed 
turtle, Platysternon, of southeast Asia. Platysternon has a xtry 
large head in proportion to the size of its shell; consequently, 
indi\iduals of this genus are not able to withdraw their heads 
into the shell in the typical cryptodiran manner. But the an- 
terior embrasure of the carapace provides a notch into which 
the back of the head fits when retracted to the maximum extent 
possible. The heavily boned dorsal roof of the skull then acts, 
in effect, as an anterior continuation of the carapace and evi- 
dently ser^^es as a reasonably efTective deterrent to predators. 
Stupendemys, too, may have had a proportionately large, heavily 
armored skull which did not have to be swung under the cara- 
pace for protection in the usual pleurodiran fashion, but instead 
was simply lodged against its anterior border when danger was 
imminent. 
I cannot readily account for the significance of the thickened, 
curled-up bone at the anterior margin of the carapace. It might 
represent a variably-expressed secondary sexual character if the 
two carapaces in the available sample represent opposite genders. 
It has, so far as I am aware, no structural equivalent elsewhere 
within the order. 
South American pelomedusids are the only chelonians having 
saddle joints on the articular surfaces of their cervical centra 
(Williams, 1950, appendix 1). But, as pointed out (p. 8), 
the cervical vertebrae of Stupendemys, although possessing the 
characteristic saddle joints, are in detail very different from 
those of any pelomedusid known from that continent or else- 
where. This fact supports the supposition that neck retraction 
in the genus was fundamentally different from that of other 
pleurodires. But if, as suggested above, Stupendemys was com- 
parable to Platysternon in its ability to retract its skull only 
partially, then the similarities in behavior were not paralleled 
by structural resemblances of even the most superficial kind. 
The articular surfaces of the fifth throus^h eis^hth cervical centra 
in Platysternon are generally doubled, the centra themselves are 
1976 world's largest turtle 13 
very broad and flat, the neural arches lack spines, and so on. 
In sum, while it is clear that the cervicals of Slupendemys are 
markedly different from those of any other known turtle, the 
significance of these differences is not readily apparent. 
Re2:rettablv, the relative sizes of the humerus and femur in 
Stupendernys cannot be determined with any degree of cer- 
tainty. This is unfortunate because, for turdes in general, the 
proportions of the fore and hind limbs are good indicators of 
the customary mode of progression. Pelomedusids and most 
aquatic cryptodires rely primarily on their hand limbs for pro- 
pulsion while swimming, hence their femora are larger than 
their humeri. But in tortoises and marine turtles, the opposite 
is true. Thus, for example, if it were possible to establish that 
the humerus of Stupendemys was larger than its femur, this 
might be taken as reasonably good presumptive evidence that 
this peculiar pelomedusid swam in a different way from all 
other pelomedusids ^ perhaps even with flipperlike appendages, 
as in the modern marine turtles. But direct comparisons between 
the humerus and femur of a single specimen of Stupendemys are 
impossible. Moreover, the only known humerus of Stupendemys 
was an isolated find, which therefore cannot be tied to sheU size, 
so that even indirect comparisons (in which limb size is related 
to shell length) cannot readily be made. 
Normally, limb structure is also a good index to the loco- 
motory capabilities of turtles. The highly modified, flippered 
forelimbs of marine cryptodires have a humerus that tends to 
be broad, flat, and relatively straight-shafted. In aquatic (or 
largely aquatic) forms, such as the pleurodires and emydines, 
it is much more gracile, ordinarily more or less circular in cross- 
section, and with a moderate curvature of the shaft. Tortoise 
humeri are stout and often have a strongly bowed shaft. The 
humerus of Stupendemys does not fall satisfactorily into any of 
these broad categories. It is considerably more massive even 
than that of a tortoise, fairly straight in the shaft, but more 
circular than flat in cross-section. The heavy ridge across the 
ventral surface of the shaft almost surely provided an increased 
area for the attachment of hypertrophied antebrachial muscula- 
ture. Such muscles would only be required if the distal ex- 
tremity of the forelimb were for some reason disproportionately 
large, as in marine turtles. While admittedly tenuous, this line 
of reasoning leads me to suspect that the foreHmb of Stupen- 
demys was modified into a paddle, a structure highly efficient 
for swimming but ill adapted to a terrestrial existence of any 
14 BREvioRA No. 436 
sort. Gh'en the absence of direct fossil evidence, however, this 
can only be a very tentati\'e suggestion. 
The humerus of the fossil pelomedusid Taphrosphys (Fig. 8; 
Gaffney, 1975, fig. 12) appears to be intermediate in structure 
between that of Stupendemys and those of typical representa- 
tives of the family. Unfortunately, the humerus is the only part 
of the forelimb of Taphrosphys so far known, so that this taxon 
provides no further insight into the structure and function of the 
Stupendemys forelimb. 
Forms intermediate in femoral structure between Stupen- 
demys and the typical pelomedusids (or turtles in general, 
for that matter) do not exist. Had the femur not been found in 
association with pelomedusid shell remains, its familial allocation 
would have been impossible. Differences between the femur of 
Stupendemys and that of a representative pelomedusid [Podoc- 
nemis expansa) hax^e already been enumerated (p. 11). The 
strongly projecting trochanters, broad intertrochanteric fossa and 
flattened shaft of Stupendemys distinguish it readily from both 
marine cryptodires and tortoises, while the massiveness of the 
bone and the broad, flat shaft together differentiate it from that 
of the other aquatic forms. In these characters, in fact, together 
with the relative straightness of the shaft, the femur of Stupen- 
demys is more like the forelimb of marine turtles than anything 
else. For this reason it is tempting to speculate that the hind 
limbs of Stupendemys may have been modified into paddling 
flippers as large as those possibly present on its forelimb. 
In sum, the available anatomical evidence demonstrates that 
Stupendemys was an aquatic form. In all likelihood, one or 
perhaps even both pairs of limbs were modified as flippers. The 
very size of its shell suggests that Stupendemys must have in- 
habited large, permanent bodies of water which it probably left 
only to lay eggs. Among living aquatic turtles in general, the 
larger the species, the less likely it is to come out of the water 
except for nesting. Size alone probably prevented Stupendemys 
from basking along shores. Flippers, if it had them, would have 
made such an undertaking even more awkward. I suspect that 
Stupendemys was largely if not entirely herbivorous, again 
simply because of its size ; all of the largest living turtles - — land 
tortoises as well as the marine forms — are totally ( or nearly 
totally ) herbivorous. 
Geological evidence, although often helpful in attempting to 
determine the habitat of a fossil, is, in the present case, equivo- 
cal. A variety of different facies are represented in the upper 
1976 world's largest turtle 15 
member of the Urumaco Formation, including near-shore ma- 
rine, brackish, and fresh water deposits. Some of these fresh 
water facies consist largely of platy concretion zones, which are 
probably best interpreted as representing small ephemeral ponds. 
Root casts and locally abundant leaf impressions are also char- 
acteristic of these deposits. Mammalian remains (especially very 
large rodents) tend to be more abundant here, as are certain 
of the reptiles (e.g., Chelus, nettosuchids ) . Other fresh water 
deposits probably represent stream channels and, in some cases, 
swampy areas (as evidenced by localized accumulations of veg- 
etable debris). In general, the vertebrate-bearing sediments 
were evidently laid down in a coastal area over which the posi- 
tion of the shoreline fluctuated back and forth repeatedly. 
Stupendemys could thus have been a marine form that washed 
up on a barrier beach or was stranded in the lagoonal waters 
behind one. Or it may have been a fresh water form carried to 
the delta of a large river system and buried there. Since the 
associated fossil fauna has strong Amazonian affinities and is 
deficient in typical marine components, the latter possibility 
seems strong. But all of the largest known aquatic turtles, both 
living and fossil, are marine forms. This fact, coupled with the 
fairly convincing presumptive evidence that a number of other 
fossil pelomedusids were marine forms,^ prevents categorical 
rejection of the idea that Stupendemys may have been a marine 
turtle. 
The largest of the living pelomedusids (all of which are fresh 
water forms) is Podocnemis expansa, which has a wide distribu- 
tion throughout much of the Amazon and Orinoco River basins 
of South America. This species is sexually dimorphic, the fe- 
males growing to much larger adult size than males (Ojasti, 
1971 ). In a large sample taken from the Orinoco River over a 
period of several years, the maximum carapace length for a male 
was 51 centimeters whereas that for a female was 81 centimeters 
(J. Ojasti, personal communication). The largest shell of this 
species yet reported is 82 centimeters long (Williams, 1954: 
293). Presumably this record is of a female, although the sex 
of this particular specimen was not indicated. With the excep- 
Jlncluded among these are several species of Taphrosphys (Schmidt, 1931; 
Gaffney, 1975; Wood, 1975) , Bothremys (Zangerl, 1948; Gaffney and Zangerl, 
1968) , and a generically indeterminate form from Puerto Rico (Wood, 
1972) . All of these were found in near-shore marine sediments, generally 
under circumstances such that tPicy cannot reasonably be regarded as exotic 
elements washed in from a nonmarine environment. 
16 BREVIORA No. 436 
tion of Stupendemys, no known fossil pelomedusids exceed 
Podocnemis expansa in size, nor do representatives of the only 
other known family of side-necked (pleurodiran) turtles, the 
CheHdae, ever approach P. expansa in size. Thus Stupendemys 
is by far the largest pleurodire, living or fossil, yet known. 
A few species of living fresh water cryptodiran turtles attain 
greater carapace lengths than P. expansa, but none are reliably 
known to approach the size of Stupendemys. A length of nearly 
130 centimeters has been recorded for the carapace of the 
Asiatic trionychid Pelochelys bibroni (Pope, 1935). Another 
Asiatic soft-shelled turtle, Chitra indica, is generally believed to 
have a maximum carapace length of approximately 90 centi- 
meters. One unsubstantiated report indicates that Chitra may 
occasionally reach a carapace length of roughly 180 centimeters 
(Pritchard, 1967:211). No other living or ifossil fresh water 
cryptodires as large as either of these recent trionychids are 
known. 
Some other fossil cryptodiran turtles of enormous size have 
been described, but none of these had shells as large as those of 
Stupendemys. Archelon ischyros, from the Cretaceous of North 
America, is the largest of the fossil marine turtles; its straight- 
line carapace length is 193 centimeters (Wieland, 1909), When 
first described, Geochelone atlas (originally and rather appro- 
priately named Colossochelys) was believed to reach twelve feet 
in carapace length (Falconer and Cautley, 1844). This estimate 
was based on composite reconstructions of fragmentary material 
and has subsequently been modified to a maximum of six feet 
(roughly 180 cm; see Lydekker, 1889, and Auffenberg, 1974: 
173). None of the specimens that have since been referred to 
G. atlas, which is now known from the Pleistocene of India, 
Burma, Java, Celebes, and Timor (Hooijer, 1971; Auffenberg, 
ibid.), appears to have reached or exceeded this length. One 
or more species of Geochelone from the Pleistocene of Florida 
and Texas may also have attained similarly gigantic dimensions 
( W. Auffenberg, personal communication ) . However, no tor- 
toises — li\'ing or fossil — ever seem to have grown any larger. 
In fact, of all known turtles, only the anatomically peculiar 
marine turtle Dermochelys coriacea may rival Stupendemys in 
size. Dermochelys, commonly referred to as the leatherback, is 
reputedly the largest of all turtles, living or fossil. x\dults con- 
sistently attain carapace lengths of over 150 centimeters (Pritch- 
ard, 1971 ). In the only large series of measurements ever made, 
involving 1500 mature female specimens encountered laying eggs 
1976 
WORLD S LARGEST TURTLE 
17 
on the beaches of French Guiana over several field seasons, the 
maximum length recorded was 180 centimeters (three individu- 
als; P. C. H. Pritchard, personal communication). Larger speci- 
mens have occasionally been reported, up to a supposed length of 
3.35 meters, but these are unusual and suspect because they are 
probablv based on estimates rather than actual measurements 
(Carr, 1952:446), and, as Brongersma (1968:38-39) has 
noted, estimates of the sizes of free-swinging marine creatures 
generally tend to be greatly exaggerated. Thus, there do not 
seem to be any reliable records of leatherbacks that equal or 
exceed Stupendemys in carapace length. On the average, cer- 
tainly, carapace lengths of Dermochelys are significantly shorter 
than those of Stupendemys. Moreover, if the known specimens 
are typical representatives of Stupendemys, then adult popula- 
tions evidently tended to be significantly larger than those of 
Dermochelys are today. In sum, it is clear that Stupendemys 
is unquestionably larger than any other previously described 
fossil turtle and it also appears to be larger than any living spe- 
cies. Stupendemys, therefore, is the largest turtle that ever lived. 
TABLE 1 
Measurements (in cm) for carapaces of Stupendemys geographiciis. Di- 
mensions are given as straight-line distances rather than over the curvatures 
of the shells, 
MCNC 244 MCZ(P) 4376 
midline length (as preserved) 
total midline length 
maximum width (estimated) 
maximum parasagittal length 
first vertebral 
second vertebral 
third vertebral 
fourth vertebral 
fifth vertebral 
18 
BREVIORA 
No. 436 
TABLE 2 
Neural bone measurements (in cm) for specimens of Stupendemys geogra- 
phicus. 
Midline Maximum Width/ 
Specimen No. Neural No. Length Width Length 
TABLE 3 
Measurements (in cm) of the plastron (MCNC 245) referred to Stupendemys 
geographicus. 
midline length (as preserved) 
total midline length (estimated) 
width at axial notch 
width at inguinal notch 
anteroposterior length of bridge 
midline length of posterior lobe 
parasagittal length of posterior lobe 
(to tips of xiphiplastra) 
Cleft side 
j right side 
fleft side 
) right side 
57.2 
76 
34.0 
35.3 
35.2 
36.2 
21.0 
25.2 
25.5 
1976 
WORLD S LARGEST TURTLE 
19 
TABLE 4 
Measurements (in cm) of the cervical vertebrae of Stupendemys compared 
with those of adult representatives of each of the three living pelomedusid 
genera, (MCZ [H]44G9, Podocnemis expansa; AMNH 10065, Pelusios sub- 
niger; MCZ[H]1 46146, Pelomedusa subrufn) . 
Height of Neural 
Midline No. in Midline Arch Spine above 
Carapace Cervical Length of Base of Posterior Height/ 
Length Series Centrum End of Centrum Length 
Specimen No. 
iThe bottom of the posterior end of this centrum is somewhat damaged so 
that a precise measurement is impossible; the figure recorded here is an 
estimate. 
20 BREVIORA No. 436 
TABLE 5 
Measurements (in era) of the known appendicular skeletal elements of 
Sttipendemys geographicus. 
SCAPULOCORACOID (MCZ[P]4376) 
Cleft: 36.2 
lengths (as preserved) of dorsal processes of scapulae -: . , oq ^ 
lengths (as preserved, along anterior edge, start- 
ing from lateral side of glenoid fossa) of ventro- ^eft: 25.3 
medial prongs of scapulae bright: 26.9 
Cleft: 37.0 
lengths of covacoids j right: 36.9 
HUMERUS (MCZ[PJ4378) 
length (as preserved) 31.0 
estimated total length 34 
maximum width of proximal expansion (as preserved) 18.0 
maximum width of distal expansion 15.0 
dorsoventral width at middle of shaft 8.3 
anteroposterior width at middle of shaft 6,4 
combined widths of ulnar and radial condyles on ventral surface 10.1 
FEMUR (MCNC 244) 
length (as preserved) 29.5 
estimated total length 33-34 
maximum width of distal expansion 15.7 
dorsoventral width at middle of shaft 6.5 
anteroposterior width at middle of shaft 8.0 
1976 
WORLD S LARGEST TURTLE 
21 
Plate 1. The carapace of Stupendemys geographinis (MCZ[P]4376) , in 
dorsal view. Note especially the strongly curled bone at the base of the 
antero-median indentation. Midline length of this specimen is 218 cm. 
Peripheral bones in the region of the bridge on both sides, some of the more 
anterior peripherals on the right, and the lateral ends of some of the 
pleurals have been restored. 
22 
BREVIORA 
No. 436 
l_ 
cm 
50 
1 
Figure 1. Carapace of the type of Stupendemys geographicus (MCNC 
244) showing the shapes and positions of the second through seventh neural 
bones. 
1976 
WORLD S LARGEST TURTLE 
23 
cm 
50 
Figure 2. Sketch of a plastron (MCNC 245) referred to Stupendemys 
geographicus, showing the unusual position of the pectoral-abdominal scute 
sulcus. The full extent of the abdominal -femoral scute sulci cannot be 
traced. 
24 
BREVIORA 
No. 43& 
Figure 3. The seventh (bottom; MCZ[P]4376) and eighth (top; a com- 
posite based on MCNC 244 and MCZ[P]4377) cervical vertebrae of Stu- 
pendemys geographicus in left lateral (left) , anterior (center) , and posterior 
(right) views. 
Figure 4. The fifth through eighth cervical vertebrae of Podocnemis ex- 
pansa (MCZ[H]4469) in left lateral view. The arrow points toward the- 
anterior end of the neck. Compare with the lateral views of Figure 3. 
1976 
WORLD S LARGEST TURTLE 
25 
I L_ 
cm 
5 
for AMNH 13582 & MCZ(H) 4467 
scapula 
cm 10 
for MCZ(P) 4376 
Figure 5. The ventral elements of the left scapulocoracoid of Stupendemys 
geographicus (MCZ[P]4376; bottom) juxtaposed with comparable bones of 
the Recent pelomedusids Podocnemis unifilis (MCZ[H]4467; middle) and 
Pelusios castaneus (AMNH 13582; top) . The midline axis of the specimens 
to which they belong would be toward the left margin of the page. The 
arrow points anteriorly. The glenoid socket of the fossil faces forward while 
those of the Recent specimens are directed laterally. For clarity, the dorsal 
prong of the scapula and the suture between the scapula and coracoid 
have been omitted. 
26 
BREVIORA 
No. 436 
3 civT^iSiSs^P 
M i l 
Figure 6. The left humerus of Stupendemys geographicus (MCZ[P]4378) 
m dorsal (left) and ventral (right) views. 
1976 
WORLD S LARGEST TURTLE 
27 
3 CM 
Figure 7. The left humerus (top) and left femur (bottom) of Podocnemis 
expansa (MCZ[H]4469) in dorsal (left) and ventral (right) views. Compare 
with Figures 6 and 9. 
28 
BREVIORA 
No. 436 
Figure 8. The right humerus of Taphrosphys sulcatus (PU 18707) in 
ventral view, showing the prominent ridge extending from the base of the 
ulnar process to just above the radial condyle. Compare with Figure 6. 
1976 
WORLD S LARGEST TURTLE 
29 
3 CM 
I I I I 
Figure 9. The left femur of Stupendemys geographicus (MCNC 244) in 
dorsal (left) and ventral (right) views. 
30 BREvioRA No. 436 
ACKNOWLEDGMENTS 
My thanks go first to my colleagues in the field during the 
summer of 1972, Messrs. Bryan Patterson, Arnold Lewis, Daniel 
Fisher, Robert Repenning, and Michael Stanford, all of whom 
helped to collect the various specimens of Stupendemys found 
by the expedition. Without the splendid cooperation of our 
Venezuelan colleagues from the Escuela de Geologia, Universi- 
dad Central de Venezuela (especially Profesora Lourdes de 
Gamero) and the Ministerio de Minas e Hidrocarburos, our 
fossil collecting in Venezuela would have been impossible. 
Funds for our field work were provided by National Science 
Foundation grant no. GB-32489X to Professor Patterson. I 
am, in addition, grateful to the National Geographic Society 
for its support of my research on South American turtles. 
Mr. Arnold Lewis supervised the monumental task of preparing 
the specimens of Stupendemys for study and exhibition in his 
usual capable manner, and I am particularly indebted to him. 
The considerable talents of Messrs. Al Coleman and Laszlo 
Meszoly are responsible, respectively, for plate 1 and figures 
3, 4, and 6-9. For information, access to or loan of specimens 
in their care, I am grateful to: W. AufTenberg; D. Baird; 
D. Fisher; J. Ojasti; L. Price; P. Pritchard; E. Williams; and 
R. Zweifel. Finally, my special thanks go to Professor Patterson 
for critically reading several manuscript versions of this paper. 
REFERENCES CITED 
AuFFENBERG, W. 1974. Checklist of fossil land tortoises (Testudinidae) . 
Bull. Fla. State Mus., Biol. Sci., 18: 121-251. 
Brongersma, L. D. 1968. The Soay Beast. Beaufortia, 15: 33-46. 
Carr, a. 1952. Handbook of Turtles: the Turtles of the United States, 
Canada, and Baja California, Ithaca, Comstock Publishing Associates. 
542 pp. 
Falconer, H. and P. T. Cautley. 1844. On the osteological characters 
and paleontological history of the Colossochelys atlas, a fossil tortoise 
of enormous size from the Tertiary strata of the Siwalik Hills in the 
north of India. Proc. Zool. Soc. London, 12: 54-55. 
Gaffney, E. S. 1975. A revision of the side-necked turtle Taphrosphys 
sulcatus (Leidy) from the Cretaceous of New Jersey. Am. Mus. Novit., 
no. 2571: 1-24. 
Gaffney, E. S. and R. Zangerl. 1968. A revision of the chelonian genus 
Boihremys (PIcurodira: Pelomedusidae) . Fieldiana: Geol., 16: 193-239. 
1976 world's largest turtle 31 
HooijER, D. A. 197L A giant land tortoise, Geochelone atlas (Falconer 
and Caullcy) , from the Pleistocene of Timor. Proc. Koninkl. Nederl. 
Akad. Wet., (B) , 74: 504-525. 
Lydekkfr. R. 1889. On the land tortoises of the Siwaliks. Rec. Geol. Surv. 
India, 22: 209-212. 
MiTTERMEiER. R. A. AND R. A. WiLsoN. 1974. Rcdescriplion of Podocnemis 
ery throe ephala (Spix. 1824) , an Amazonian pelomedusid turtle. Papeis 
Avulsos Zool., 28: 147-162. 
OjASTi, J. 1971. La tortuga arrau del Orinoco. Defensa de la Naturaleza, 
no. 2: 3-9. 
Pope, C. H. 1935. The Reptiles of China: Turtles, Crocodilians, Snakes, 
Lizards. Natural History of Central Asia, 10 (Amer. Mus. Nat. Hist.) : 
604 pp. 
PRITCHARD, p. C. H. 1967. Living Turtles of the World. Neptune City, 
N. J.. T. F. H. Publications, Inc. 288 pp. 
1971. The leatherback or leathery turtle, Dermochelys coria- 
cea. International Union for the Conservation of Nature Monograph, 
1: 1-39. 
Schmidt, K. P. 1931. A fossil turtle from Peru. Field Mus. Nat. Hist., 
Geol. Ser., 4: 251-254. 
Simpson, G. G. 1974. Notes on Didelphidae (Mammalia, Marsupalia) 
from the Huayquerian (Pliocene) of Argentina. Amer. Mus. Novit., 
no. 2559: 1-15. 
WiELAND, G. R. 1909. Revision of the Protostegidae. Amer. Jour. Sci., 27: 
101-130. 
Williams, E. E. 1950. Variation and selection in the cervical central articu- 
lations of living turtles. Bull. Amer. Mus. Nat. Hist., 94: 505-562. 
. 1954. A key and description of the living species of the 
genus Podocnemis {sensu Boulenger) (Testudines, Pelomedusidae) . 
Bull. Mus. Comp. Zool., Ill: 279-295. 
Wood, R. C. 1971. The fossil Pelomedusidae (Testudines, Pleurodira) of 
Africa. Ph.D. thesis. Harvard University. 
. 1972. A fossil pelomedusid turtle from Puerto Rico. Breviora, 
no. 392: 1-13. 
1975. Redescription of "Bantuehelys" eongolensis, a fossil 
pelomedusid turtle from the Paleocene of Africa. Rev. Zool. Africaine, 
' 89: 127-144. 
Wood, R. C. and B. Patterson. 1973. A fossil trionychid turtle from 
South America. Breviora, no. 405: 1-10. 
Zangerl, R. 1948. The vertebrate fauna of the Selma Formation of Ala- 
bama. Part 2. The pleurodiran turtles. Fieldiana: Geol. Mem., 3: 
17-56.