The famed cannibal-Coelophysis hypothesis, although tenable, has never been rigorously examined. Two criteria must be met to unequivocally ascribe cannibalistic behaviour to dinosaurs from ‘stomach’ contents. The remains must be shown to reside in the abdominal cavity and come from the same taxon. This has not occurred. The stomach contents were never prepared to a degree allowing their location in the body to be definitively ascertained. Furthermore, it is well established that most basal archosaurian reptiles, including early dinosaurian taxa, have similar hindlimb morphology, thus leaving open the possibility that the purported stomach contents come from other animals besides C. bauri.

Recent access to the Colbert's Coelophysis specimens allowed an unprecedented opportunity to reanalyse the cannibal-Coelophysis hypothesis. We further prepared the specimens (AMNH FR 7223 and AMNH FR 7224) and used detailed morphological and bone histological analyses to critically examine whether the aforementioned criteria supporting cannibalism in Coelophysis are met.

AMNH FR 7223 (see figure 1c and electronic supplementary material) is a nearly complete adult skeleton (femoral length=209mm) lying on its right side, with purported gut contents consisting of articulated vertebrae, an articulated leg (femoral length=130mm) and various small bone fragments. Other than the articulated leg, none of the purported gut material possesses diagnostic characters, allowing it to be referred to a taxon below the sauropsid level; let alone to a dinosaur such as Coelophysis. Furthermore, it seems unlikely that such a large item, an entire hindlimb (62% adult size hindlimb), could have been ingested in its entirety (Gay 2002; see electronic supplementary material). Whether these materials represent gut contents is also questionable. Although the left dorsal ribs cover all the possible stomach remains laterally, the right dorsal ribs are deflected posteriorly and do not underlie the purported gut contents. This indicates that the abdominal cavity may have been ruptured prior to burial and/or the ribs were disarticulated by fluvial processes (Schwartz & Gillette 1994). In any event, all that can be concluded is that the remains of AMNH FR 7223 came to lie upon the supposed stomach contents. Collectively, no evidence exists to unambiguously conclude that ‘stomach contents’ were preserved in this specimen's abdominal cavity.

Figure 1 Abdominal region of Coelophysis bauri neotype (AMNH FR 7224). (a) Detail of abdominal cavity of AMNH FR 7224 showing posterodorsally intact stomach (dotted line) with preserved contents (AMNH FR 30616) highlighted in yellow. Right femur and left ilium, (more ...)

The neotype of C. bauri (AMNH FR 7224) consists of a nearly complete mediolaterally crushed skeleton (femoral length=203mm) lying on its left side (figure 1). The abdominal cavity contains disarticulated skeletal remains concentrated in the posterodorsal region and articulated remains in the anteroventral region (figure 1). The left and right dorsal ribs surround the posterodorsal portion of the abdominal remains; however, both the right and the left ribs of the anterior dorsal vertebrae lie atop of the articulated series of caudal vertebrae in the anteroventral region. Articulated gastralia just anterior to the pubis indicate that the abdominal cavity remained intact prior to and after burial. Thus, we conclude that the posterodorsal remains, but not the anteroventral ones, are unambiguously contained in the abdominal cavity. The first criterion for these being cannibalized gut contents is met.

The posterodorsal concentration of bones preserves a three-dimensional structure that appears to delimit the stomach cavity (figure 1). Small bone fragments are concentrated near, and the smaller flat bones are oriented parallel to, the surface of the inferred stomach wall. Previous studies (Colbert 1989, 1995; Gay 2002) specifically highlighted the bones of this region as juvenile Coelophysis. We have identified well-preserved remains consisting of left and right proximal ends of femora, a left ilium and a sacral vertebra within the stomach cavity (figure 1). Comparative morphological analysis of this material reveals that these bones lack any synapomorphies of Coelophysis, Theropoda or even Dinosauria (figure 2). However, they are consistent with Crocodylomorpha. For instance, the left ilium bears a closed acetabulum; nearly all dinosaurs (including Coelophysis) possess an open acetabulum. The sacral vertebra shows no indications of fusion with surrounding vertebrae, has an anteroposteriorly short, dorsally expanded, neural spine, and a sacral rib articulating to the centre of the centrum—all characters consistent with crocodylomorphs, but not with Coelophysis. Additionally, the femora lack the typical dinosaurian offset femoral head, anterior trochanter and well-developed articular facet for the antitrochanter (Langer 2004); instead they exhibit proximal condylar folds (C. A. Brochu 1992, Unpublished Master's Thesis; figure 2). A proximal condylar fold is found only in crocodylomorph archosaurs, including Hesperosuchus agilis (figure 2), whose presence is well known from the Chinle Formation and the Ghost Ranch Coelophysis Quarry. Our results show that although stomach contents were remarkably preserved in situ in Ghost Ranch Coelophysis, no evidence for cannibalism exists.

Figure 2 Comparative morphological and histological evidence illustrating Crocodylomorpha affinities of the Coelophysis bauri stomach contents. (a–d) Hesperosuchus agilis (AMNH FR 6758), right proximal femur in (a) lateral, (b) medial and (c) dorsal views, (more ...)

As an independent test of our hypothesis that AMNH FR 7224 consumed a crocodylomorph rather than a juvenile Coelophysis, we conducted comparative histological analyses between the purported cannibalized stomach contents, Coelophysis, and the most common Chinle Formation crocodylomorph—Hesperosuchus (see electronic supplementary material for taxonomic assignment). The metaphyses of the femur from the stomach content specimen and cf. Hesperosuchus show nearly identical histological patterning and structure and hence, attest to comparable developmental histories (figure 2). In the stomach content specimen and cf. Hesperosuchus, the transition zone to the primary bone is not distinct; instead it blends with the compacted cancellous bone. Furthermore, the initial primary bone that was deposited across the zone is avascular. Notably, the transition zones in both also show a pronounced concavity. The rest of the primary bone shows longitudinally or locally semi-radiating vascularization. In stark contrast, the transition zone of the dinosaur Coelophysis is very thin, and distinctive—like the structuring seen when lines of arrested growth form. This suggests that bone deposition completely stopped before the capping of primary bone began. The thick avascular zone seen in the stomach content and cf. Hesperosuchus is not present in Coelophysis. The transition zone is straight, unlike the concave pattern seen in the stomach content and cf. Hesperosuchus specimens. Finally, the vascular canals in Coelophysis are relatively long and show pronounced, inclined, radiating patterns—those stemming more medially incline laterally and those originating laterally incline medially.

Cannibalistic behaviour is very common among carnivorous animals (Polis 1981), and may be expected in non-avian dinosaurs. Then again, among living dinosaurs (birds) this behaviour is not prevalent, being most common in colonial nesting seabirds (Parsons 1971) and birds of prey (Ingram 1959). This brings into question just how prevalent this behaviour was among non-avian dinosaurs. The phylogenetic distribution of cases supposedly attesting to cannibalism suggests that it was prevalent throughout the Theropoda. This study shows that there is no compelling evidence for this behaviour in the famed Coelophysis AMNH ‘cannibals’. What about the other cases? Recent reports of coprolites and cololites (internal intestinal casts), said to be preserved below the base of a Coelophysis tail (Rinehart et al. 2005) and to contain cannibalized manual elements, also cannot be considered as evidence for cannibalism in Coelophysis as the purported digested skeletal material is taxonomically uninformative.

Occurrences of cannibalism in other theropods are also problematic. Jacobsen (1998) reported tyrannosaur bite marks on the remains of tyrannosaurs from the Dinosaur Park Formation of Alberta and used these as evidence for cannibalism. However, given that there are at least two species of tyrannosaurs in the formation (Daspletosaurus and Gorgosaurus; Jacobsen 1998), cannibalism cannot be demonstrated sufficiently. This leaves just bite mark evidence and associated tooth crowns in the large theropod Majungatholus atopus (Rogers et al. 2003) as the only compelling evidence for cannibalism out of the hundreds of known Mesozoic theropod dinosaurs. On the basis of this evidence, cannibalism is not as prevalent as was once supposed in non-avian dinosaurs.

We thank Jack Conrad for his insightful comments while preparing the manuscript. Alex Downs supplied the crocodylomorph femur for histology sectioning. Mick Ellison composed the figures. The photograph appearing in figure 1 is courtesy of the American Museum of Natural History. This work was supported by a National Science Foundation Graduate Research Fellowship and the American Museum of Natural History Division of Paleontology. The histological analyses were supported by National Science Foundation, Division of Earth Sciences grant EAR 0207744 presented to G.M.E. and M.A.N.