The Skull

The skull of IGM 100/1195 preserves most of its elements, except for the anteriormost region of the rostrum and the anterior region of the secondary palate (figs. 4, 9). The snout is moderately low and broad (i.e., platyrostral sensu Busbey, 1994). The rostrum of Shamosuchus djadochtaensis is short, approximately 43% of the anteroposterior skull length (as preserved). Because the anterior tip of the rostrum is missing, this proportion would have been slightly longer. All other Shamosuchus species have a noticeably longer rostrum, varying between 63% and 70% of the total skull length (Efimov, 1983, 1988). This is also the case in other advanced neosuchians, such as Bernissartia fagesii (Buffetaut, 1975), Rugosuchus nonganensis (Wu et al., 2001a), and Isisfordia duncani (Salisbury et al., 2006). However, a relatively short rostrum has also been described for the basal eusuchian Iharkutosuchus makadii (Ösi et al., 2007); a condition probably shared with Hylaeochampsa vectiana. The external nares are not preserved in the holotype of Shamosuchus djadochtaensis and only the posterolateral margins of this opening are present in IGM 100/1195. These margins are vertically oriented suggesting that the external nares faced somewhat anteriorly (or anterodorsally), in contrast to the dorsally facing nares of most neosuchian crocodyliforms. The antorbital fenestra is completely obliterated and there is no sign of an antorbital fossa (fig. 10). The orbits of Shamosuchus djadochtaensis are moderately large, approximately 27% of the preserved skull length (measured from rostral end to the caudal end of the parietal dorsal surface). The supratemporal fenestrae are relatively reduced, being smaller than the orbits and subcircular in shape (fig. 8). The infratemporal fenestrae are subtriangular and longer than high. The suborbital openings are not completely preserved, although they seem to be well developed and anteroposteriorly elongate (fig. 9). The external surface of the skull is deeply ornamented with the pitted pattern characteristic of neosuchian crocodyliforms. The sculpture present in IGM 100/1195 is slightly more developed and extended than in the type specimen of Shamosuchus djadochtaensis (AMNH FR 6412).

The premaxillae are incompletely preserved in IGM 100/1195, lacking the region anterior to the external nares and most of their palatal branches. The lateral margins of the narial opening are vertically oriented and smooth, denoting the presence of a perinarial fossa extending along the ventrolateral region of this opening. The premaxilla also forms part of the dorsal margin of the external nares, deflecting dorsally and contacting the anterior tip of the nasals.

Posterior to the perinarial region, the external surface of the premaxilla is ornamented with subcircular pits, as in most neosuchian crocodyliforms. The premaxilla extends posteriorly forming the anterior fourth of the rostrum. Its external surface has two exposure planes, a ventral region oriented vertically and a dorsal region facing dorsolaterally to dorsally. As in most platyrostral forms, the ventral region is dorsoventrally low while the dorsal region is much more extensive (figs. 8, 10). The premaxilla-maxilla suture is slightly interdigitated, extends vertically from the alveolar margin, and then deflects posteriorly, forming the ventral margin of a well-developed posterodorsal process of the premaxilla. The dorsomedial margin of the premaxilla contacts the lateral margin of the nasal on the dorsal surface of the rostrum. The anterior half of this contact forms a linear and parasagitally oriented suture (fig. 8). The posterior half of this suture is slightly deflected posterolaterally.

The alveolar margin of the premaxilla-maxilla contact is continuous and lacks a notch for the reception of the opposing dentary tooth, unlike most platyrostral neosuchians (except for alligatorids). The palatal branches of the premaxilla are only briefly preserved medial to the posteriormost alveoli. Unfortunately, the palatal sutures between the premaxillae and with the maxillae are not preserved in the type specimen or in IGM 100/1195. Only the two posteriormost alveoli of each premaxilla are preserved, bearing fragments of the subcircular tooth crowns. These alveoli are discrete and subcircular, with the posterior one significantly smaller than the anterior premaxillary alveolus (and the anterior maxillary alveoli).

The maxillae are well preserved in IGM 100/1195, except for their palatal branches (figs. 9, 10). The external surface of the maxilla forms most of the rostrum and is densely ornamented with subcircular pits. As in the premaxilla, the external surface has two planes of exposure, a dorsoventrally low ventral region oriented vertically and an extensive dorsal region facing dorsolaterally. The dorsomedial margin of the maxilla contacts the nasal along a straight and posterolaterally oriented suture. The alveolar margin of the maxilla bears a single outgrowth, projecting ventrally at the level of the third and fourth maxillary enlarged alveoli (fig. 10). At this point, the external surface of the maxilla is slightly bulged, forming a noticeable convexity seen in dorsal and ventral views (figs. 8, 9). Posterior to this region the buccal margin of the maxilla is straight. Variation in alveolar size closely follows the shape of the ventral margin of the maxilla, increasing in size toward the third and fourth teeth and decreasing posterior to these elements. In contrast, most advanced neosuchians (e.g., Rugosuchus nonganensis (Wu et al., 2001a), Goniopholis simus BMNH 41098, Eutretauranosuchus delfsi (Mook, 1967), Bernissartia fagesii (Norell and Clark, 1990), and brevirostran crocodylians) have two “waves” of maxillary tooth size variation (or festooning pattern). There are 12 preserved maxillary teeth that resemble those of most advanced neosuchians in that they are subconical and only slightly compressed lateromedially toward the tip of the crowns, bear a slightly marked constriction between root and crown, and lack serrated margins. The third maxillary alveolus is approximately 35% larger than its adjacent alveoli, which bear the second largest teeth of the maxilla. In contrast, other species referred to Shamosuchus (e.g., S. major, S. gradilifrons, S. ulgicus) have the cheek maxillary alveolus significantly larger than the others (Efimov, 1983, 1988; Storrs and Efimov, 2000). Most advanced neosuchians also have enlarged maxillary teeth located in the fourth and/or fifth position (e.g., Rugosuchus nonganensis [Wu et al., 2001a], Bernissartia fagesii, most crocodylians [Norell, 1988]). The Glen Rose form (USNM 22039), however, shares with Shamosuchus djadochtaensis the presence of a distinctly enlarged third alveolus with respect to the other maxillary alveoli (Brochu, 1999). The posteriormost teeth are not well preserved, although they seem to be more robust and with shorter crowns. The maxillary alveoli are closely packed to each other, as in other neosuchians (e.g., the Glen Rose form, Bernissartia fagesii, Theriosuchus pusillus, Gonipholis simus), but differ from the well-separated alveoli of Rugosuchus nonganensis, as noted by Wu et al. (2001a).

Several small neurovascular foramina are scattered on the lateral surface of the maxilla, close to its alveolar margin. They are evenly spaced and most of them are positioned between adjacent alveoli. The posterior region of the maxilla is laterally overlapped by the anterior process of the jugal on its ventral region and sutured to the lacrimal dorsally. The latter suture is difficult to delimit precisely in IGM 100/1195 due to the heavy ornamentation pattern of this region, but is clearly preserved in the type specimen (see below). Here the maxilla is completely sutured to the lacrimal, lacking an antorbital fossa or opening, as in derived neosuchians. An acute branch projects from the posteromedial region of the maxilla, reaching the anterior tip of the prefrontal, excluding the lacrimal from contacting the nasal, as in Hylaeochampsa vectiana (Clark and Norell, 1992; BMNH R177), Susisuchus krebsi (Salisbury et al., 2003), and some extant eusuchians (e.g., Alligator mississippiensis FMNH 8201).

The nasals form the dorsal surface of the rostrum and bear a similar (although less developed) ornamentation pattern as present on the rest of the skull in IGM 100/1195. In the type specimen (AMNH FR 6412) most of the dorsal surface of the nasals is not ornamented, a difference considered here as likely due to preservational or individual variation. These elements extend from the anterior end of the preserved rostrum to the level of the anterior margin of the orbits. Their lateral margins diverge slightly posteriorly along their contact with the premaxillae and maxillae (fig. 8). The nasals are sutured to each other along their medial margins forming a slightly developed medial crest (fig. 12).

Posterior to their contact with the maxillae, the nasals are sutured to the prefrontals along a laterally concave suture directed posteromedially. Thus, the posterior region of the nasals narrows gradually, being at their posterior end approximately half the width of their breadth at the midpoint of the rostrum (fig. 8). The posterior margins of the nasals are sutured to the frontals along a transversally oriented suture, although details on this suture were not preserved in IGM 100/1195 or in the type specimen AMNH FR 6412.

The lacrimal forms most of the anterior margin of the orbit and is sutured to the prefrontal, maxilla, and jugal. As clearly seen in the type specimen, the lacrimal is rather low dorsoventrally and slightly elongated anteroposteriorly, extending on the lateral surface of the snout (where it faces laterodorsally; figs. 3, 8). The lacrimal extends further anteriorly than the prefrontal as in Rugosuchus nonganensis (Wu et al., 2001a) and Isisfordia duncani (Salisbury et al., 2006), in contrast to the condition of some basal eusuchians such as Hylaeochampsa vectiana (Clark and Norell, 1992) and Iharkutosuchus makadii (Ösi et al., 2007). The anterior margin of this region of the lacrimal is anteriorly convex and sutured to the maxilla. As noted above, in the two specimens of Shamosuchus djadochtaensis studied here, there are no signs of an antorbital opening or fenestra at the maxilla-lacrimal contact. Posteroventrally, the lacrimal narrows markedly and briefly contacts the jugal at the anteroventral corner of the orbit.

The anterior half of the dorsomedial margin of the lacrimal is medially convex and bounded by the maxilla; as a result, it is excluded from contact with the nasal. This morphology contrasts with that of Rugosuchus nonganensis (Wu et al., 2001a) and most other neosuchians (see above), in which the elongated lacrimal contacts the nasal. Posteriorly, this margin of the lacrimal is bordered by the prefrontal, along a similarly curved suture. The posterodorsal surface of the lacrimal is horizontally exposed (lateral to its suture with the prefrontal; figs. 3, 8). In contrast to the rest of the external surface of the lacrimal, this dorsally exposed surface bears a smooth depression bounded by elevated ridges (fig. 13). This morphology is similarly present in AMNH FR 6412 for which Mook (1924) described the lacrimal ridges as diagnostic of Shamosuchus djadochtaensis. Other Shamosuchus species, however, also show prominent lacrimal (and prefrontal) ridges (Efimov, 1983, 1988; Storrs and Efimov, 2000). The depressed surface bounded by these ridges continues medially along the prefrontal's dorsal surface. The posterior margin of this depression is elevated, forming a slightly developed ridge at the anterodorsal margin of the orbit. This depression is located close to the position in which the anterior palpebral articulates in most crocodyliforms. However, the posterior margin of this depression in IGM 100/1195 is separated from the orbital margin by an elevated ridge of the lacrimal, rather than being confluent with the orbit as the palpebral articular facet of other crocodyliforms. This suggests that either this depression is not the palpebral articular facet, or that the anterior palpebral of Shamosuchus djadochtaensis is markedly distinct from that of other Crocodyliformes. Although further material and study is needed to test these options this depression seems to be diagnostic of Shamosuchus djadochtaensis or, alternatively, a putative synapomorphy for the higher taxon Shamosuchus more broadly, given the presence of a transverse prefrontal ridge in other species of Shamosuchus (Efimov, 1983; Wu et al., 2001a).

Figure 13

Orbit region of referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in anterolateral view showing shallow hemispherical depression on the prefrontal and lacrimal (character 277.1).

Inside the orbit, the lacrimal-prefrontal suture is interdigitated and directed ventrally (fig. 14). The posterior opening of the lacrimal duct opens into the orbital cavity, just lateral to this suture and ventromedial to the crest formed by the posterior margin of the lacrimal depression described above. Additionally, there is a much smaller foramen located ventrolaterally to the main lacrimal opening. The ducts that extend anteriorly from these openings merge within the lacrimal bone.

Figure 14

Detail of Shamosuchus djadochtaensis autapomorphic characteristic of a prominent depressed area anterior to the posterior lobe of the squamosal. A, Referred specimen IGM 100/1195, and B, holotype specimen AMNH FR 6412.

The dorsal surface of the prefrontals of Shamosuchus djadochtaensis is anteroposteriorly elongate, extending from the anteroposterior midpoint of the orbit to the level of the posteromedial branch of the maxilla. In IGM 100/1195 most of the dorsal surface of the prefrontals is ornamented, except for the depressed facet described above (fig. 13). The anterior end of the prefrontal is acute, wedging between the nasal and the posteromedial branch of the maxilla. Its anterior tip does not exceed the anterior extension of the lacrimal (fig. 3), but surpasses the anterior end of the frontal, in contrast to the condition seen in longirostrine crocodyliforms (e.g., Terminonaris robusta [Wu et al., 2001b], Sarcosuchus hartii MNN 604, and Rhabdognathus aslerensis CNRST-SUNY 190).

The posterior end of the prefrontal tapers gradually, wedging between the lateral margin of the frontal and the orbital margin. The entire posterolateral region of the dorsal surface of the prefrontals is occupied by the smooth depressed surface described above. This surface extends along the anterior half of the dorsal orbital margin (figs. 3, 8), forming a posteriorly elongate, wing-shaped facet.

Within the orbital cavity, the prefrontal has a well-developed descending process that forms the prefrontal pillars, reaching the dorsal surface of the palate as in other mesoeucrocodylians (Clark, 1994). The dorsal half of the descending process is a flat surface oriented obliquely to the longitudinal axis of the skull, in contrast to the transversally oriented condition seen in basal mesoeucrocodylians (e.g., Araripesuchus gomesii AMNH FR 24450, Notosuchus terrestris MACN-RN 1041; Sphagesaurus huenei RCL-100). The condition of Shamosuchus djadochtaensis actually resembles that of Allodaposuchus precedens (Buscalioni et al., 2001) and extant crocodylians, which also have this surface exposed posterolaterally. The lateral region of this surface contacts the lacrimal along a ventrally directed suture.

Ventromedially, the descending process of the prefrontal forms the prefrontal pillar proper and is sutured to the palate (fig. 14). The prefrontal pillar is laminar, being anteroposteriorly thin and lateromedially expanded, except for its most distal region where it expands anteroposteriorly at the suture with the dorsal surface of the palate. IGM 100/1195 preserves the incomplete base of the prefrontal, which displays a medial process dorsoventrally centered on its medial edge.

As mentioned previously, none of the studied specimens preserve palpebrals. However, the presence of an anterior palpebral in Shamosuchus djadochtaensis cannot be rejected at the moment and further material is needed to assess this issue. The presence of a posterior palpebral that articulates with the postorbital is unlikely, due to the absence of a corresponding facet on the postorbital, which occurs in most advanced neosuchians.

The frontals are completely fused to each other on the midline as in all mesoeucrocodylians (Clark, 1994). The dorsal surface of this element is densely ornamented and bears a well-developed ridge along the sagittal plane (figs. 8, 13). The frontal ridge is clearly present in the type specimen (as described by Mook, 1924) and in IGM 100/1195. This resembles the condition of Rugosuchus nonganensis (Wu et al., 2001a), Isisfordia duncani (Salisbury et al., 2006), and some non-neosuchian crocodyliforms (e.g., Sichuanosuchus shuhanensis IVPP V 10594, Simosuchus clarki UA 8679, Notosuchus terrestris MACN-RN 1041). Lateral to this ridge, the dorsal surface of the frontal is flat, differing from the condition in Rugosuchus nonganensis (as noted by Wu et al., 2001a). The anterior end of the frontal tapers between the prefrontals and is markedly constricted above the orbits (being narrower than the width of the nasals). The lateral margins of the frontal form the posterior half of the dorsal margin of the orbits. Along this region, the lateral edges of the frontal are slightly elevated with respect to the rest of the dorsal surface of the frontal. Thus, the frontal forms supraorbital ridges similar to those of several neosuchian crocodyliforms (e.g., Trematochampsa taqueti MNHN-IBC 231, Theriosuchus pusillus BMNH R48330, Bernissartia fagesii [Buffetaut, 1975], Rugosuchus nonganensis [Wu et al., 2001a], Hylaeochampsa vectiana BMNH R177, and some crocodylians).

The frontal expands at the posterior end of the orbits where it contacts the postorbitals laterally and the parietal posteriorly. The suture with the postorbital is anteroposteriorly directed on the skull roof from the posteromedial margin of the orbit to the anterior edge of the antorbital fossa (fig. 8). This suture extends within the supratemporal fossa, where the frontal meets the anterolateral tip of the parietal and the postorbital is excluded form contacting the parietal by a portion of the laterosphenoid. Thus, although the frontal extends into the supratemporal fossa, it does not form part of the margin of the supratemporal (internal) fenestra, due to the existence of a laterosphenoid-parietal contact. Within the supratemporal fossa, the frontal-parietal suture is interdigitated and oriented transversely, extending onto the skull roof with the same orientation and located just posterior to the anterior margin of the antorbital fossae. Inside the orbital cavity, the frontal is sutured to the postorbital laterally and ventrally to the anterodorsal margin of the laterosphenoid (fig. 15).

Figure 15

Supratemporal fenestra of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195 with accompanying interpretation of sutures. See appendix 5 for abbreviations.

As in all crocodyliforms, the parietals are fused to each other forming an unpaired element. Its dorsal surface is ornamented and rather wide between the supratemporal fossae, forming part of the characteristic skull roof of most Crocodyliformes. The anterior half of this surface is flat, whereas the posterior region bears a prominent sagittal ridge (figs. 8, 15).

The lateral region of the parietal enters into the supratemporal openings, where it forms the medial surface of the supratemporal fossae. This region is smooth and almost vertically oriented (fig. 8). The ventrolateral margins of the parietals are bounded by the laterosphenoid and the dorsal edge of the quadrate. The anterior region of the parietal fails to contact the postorbital, as in most noncrocodylian neosuchians. Posteriorly, the parietal expands slightly laterally, contacting the squamosals. The parietal-squamosal suture interdigitates and extends posteriorly from the posterior region of the supratemporal fossa onto the skull roof, reaching the occipital margin of the skull table. Within the supratemporal fossa, the parietal and squamosal enclose the anterior opening of the orbitotemporal passage. This opening is covered dorsally by a well-developed rim of the external margin of the supratemporal fossa (fig. 16). Therefore, the orbitotemporal opening is not visible in dorsal view, a character noted to be present only in advanced neosuchians (e.g., species of Goniopholis, Bernissartia fagesii, and adult eusuchians; Norell and Clark, 1990; Ortega et al., 2000). This character, however, is subject to marked ontogenetic variation in extant crocodylians and in the goniopholid Sunosuchus junggarensis (Wu et al., 1996a).

Figure 16

Close-up of supratemporal fossa in the referred specimen IGM 100/1195 showing anterior opening of temporo-orbital opening hidden in dorsal view and overlapped by squamosal rim of fossa (character 173.1).

The dorsal surface of the squamosals forms the posterolateral region of the flat skull table characteristic of Crocodyliformes. This dorsal surface of the squamosal surface is triradiate, composed by a medial branch contacting the parietal, an anterior process contacting the postorbital, and a posterolateral process. The medial branch is rather short but anteroposteriorly extensive, well separating the supratemporal fossa from the occipital margin of the skull roof (figs. 3, 8, 15). The anterior branch of the squamosal is slightly curved and meets the postorbital in an interdigitated suture oriented transversely that is located at the anteroposterior midpoint of the supratemporal opening.

The posterolateral process of the squamosal is well developed and slightly curved medially toward its distal end, having a concave occipital margin and a sigmoid lateral margin (fig. 3). The dorsal surface of this process is ornamented with a pitted pattern, as is the rest of the squamosal, and is located at the same level as the skull table (figs. 8, 10). This pattern is in contrast to the ventrally deflected posterolateral process of most basal mesoeucrocodylians (Araripesuchus gomesii AMNH FR 24450, Notosuchus terrestris MACN-RN 1037, Sebecus icaeorhinus AMNH FR 3160). The type specimen of Shamosuchus djadochtaensis (AMNH FR 6412) bears two posteromedially directed grooves on this process. One of them is posteriorly concave, bordered by a weakly developed ridge, and located at the base of the posterolateral process. The other groove is narrow, anteriorly concave, and located at the anteroposterior midpoint of this process (fig. 3). The first of these is present in IGM 100/1195, although the presence of the posterior groove cannot be corroborated in this specimen due to the poor preservation of this region. The presence of a distinct, depressed, and smooth posterolateral “lobe” has been noted as a characteristic of some neosuchians (e.g., atoposaurids) by Clark (1994) (fig. 14). Wu et al. (1996a, 2001a), however, noted that a smooth depressed lobe of the squamosal may have a wider distribution and that it is ontogenetically variable in Sunosuchus junggarensis. Interestingly, Rugosuchus nonganensis has a morphology similar to Shamosuchus djadochtaensis, having the posterolateral region slightly sculpted, depressed, and separated by a weak ridge (Wu et al., 2001a). The presence of a sculpted and slightly depressed squamosal “lobe,” however, may be also present in other species of Shamosuchus (Efimov, 1983, 1988) but not in Shamosuchus major (Wu et al., 2001a).

Medial to the skull roof surface, the squamosal forms the posterolateral region of the supratemporal fossa (figs. 15, 16). This surface is smooth and exposed almost vertically, forming a reduced lateral floor of the supratemporal fossa. The ventral margin of this region is bounded by the anterodorsal process of the quadrate.

The lateral margin of the squamosal overhangs the quadrate and quadratojugal forming a deep otic recess. The anterior half of the squamosal's lateral margin is laterally convex, whereas this edge is concave along the squamosal posterolateral process (fig. 3). The squamosal bears a rather broad discontinuous groove on the external surface of its lateral edge, resembling the attachment point of the movable dorsal earflap of extant crocodylians located above the otic aperture (Shute and Bellairs, 1955). The morphology of this groove in Shamosuchus djadochtaensis is, however, autapomorphic and differs slightly between the type specimen and IGM 100/1195. Both specimens share the presence of a shallow, but broad, groove extending from the anterior region of the lateral edge of the postorbital onto the lateral edge of the squamosal, up to the level of the otic aperture (fig. 14). At this point, the groove tapers posteriorly, disappearing and forming a sharp lateral edge of the squamosal. A similar groove reappears at the posterior edge of the otic aperture, along the lateral edge of the posterolateral process of the squamosal. This posterior region of the squamosal bulges dorsally and overhangs the posteriormost region of the otic recess (fig. 14). This combination of characters seems to be unique among Crocodyliformes. The differences between the type specimen and IGM 100/1195 consist of the ornamentation of the anterior region of the groove (lacking in the holotype) and the depth of the posterior region of the groove, which is notably shallower in IGM 100/1195 (see figs. 3, 14). Within the otic recess, the squamosal bears a descending process that contacts the anterodorsal process of the quadrate and the postorbital. The otic aperture is located between the squamosal and quadrate and these two elements meet posteriorly to close the otic aperture as in other mesoeucrocodylians, except for Allodaposuchus precedens (Buscalioni et al., 2001) and Gonipholis simus (Salisbury et al., 1999), in which the squamosal and quadrate are not sutured posteriorly to the otic aperture and therefore the cranioquadrate passage is laterally open. The squamosal-quadrate contact is located at the dorsoventral midpoint of the posterior edge of the otic aperture, in contrast to the derived condition of alligatorids in which it is located at the posteroventral corner of the otic aperture (Brochu, 1999).

The squamosal is briefly exposed on the occipital surface of the skull. This smooth surface is vertically oriented and is dorsoventrally low. Its dorsoventral depth increases slightly toward its lateromedial midpoint and then tapers laterally along the occipital margin of the posterolateral process of the squamosal (fig. 11). This surface is bounded dorsally by a broad ridge of the skull table and ventrally by the paroccipital process.

The postorbital has a dorsal surface that forms the anterolateral region of the skull roof and a descending process that forms the dorsal half of the postorbital bar as in most mesoeucrocodylians (Clark, 1994). The anteromedial end of the postorbital contacts the frontal on the skull roof and both the frontal and laterosphenoid on the supratemporal fossa. The anterolateral corner of the postorbital bears a slightly developed pointed process, which extends ventrally as a thin, anteriorly convex, low lamina that merges with the anterior margin of the postorbital bar (figs. 13, 14). This process resembles the condition present in some longirostrine taxa (e.g., dyrosaurs, pholidosaurs; Buffetaut, 1979), although it is remarkably less developed than in those forms. The anterior surface of this lamina bears at least one small foramen located inside the orbital cavity. Similar structures are present in some mesoeucrocodylians (e.g., Araripesuchus gomesii AMNH FR 24450, Trematochampsa taqueti MNHN-IBC 231, Rhabdognathus aslerensis CNRST-SUNY 190, Allodaposuchus precedens [Buscalioni et al., 2001], and crocodylians), although the homology of these foramina is poorly understood due to differences in number and precise location.

The descending process of the postorbital is a flat lamina facing posterolaterally (figs. 10, 14). This process is subdivided into a short posterior branch and an elongated anteroventral process that forms the postorbital bar. The short posterior branch extends to the anterior border of the otic recess. The dorsal region of this posterior branch contacts the ascending process of the quadrate and the squamosal. The ventral end of the short posterior branch projects as an acute tip that fits into the dorsal end of the ascending process of the quadratojugal, forming a posteroventrally directed V-shaped suture (fig. 14). The ventral margin of this acute ventral tip forms the dorsal apex of the infratemporal fenestra.

The elongate anteroventral process of the descending process of the postorbital becomes subcylindrical toward its contact with the ascending process of the jugal that delimits the posterior margin of the orbit. The postorbital-jugal suture is interdigitated on its lateral side and is located approximately at the dorsoventral midpoint of the postorbital bar.

The jugal of Shamosuchus djadochtaensis is markedly elongate, exceeding posteriorly the infratemporal fenestra and anteriorly the orbital opening. The external surface of the jugal is heavily ornamented, except for its ascending process that forms the ventral half of the postorbital bar. This element is dorsoventrally low posteriorly and increases gradually in depth toward the orbital region, being approximately twice as deep in the suborbital region as it is in the infratemporal region (as in most mesoeucrocodylians; Clark, 1994). The anterior end of the jugal is acute and overlaps laterally the posterior end of the maxilla (figs. 5, 10). In the orbital region, it dorsally contacts the ventral extension of the lacrimal. The external surface of the jugal bears a moderately developed ridge oriented on the longitudinal axis along the suborbital region.

The base of the ascending postorbital process is located approximately at the anteroposterior midpoint of the jugal (figs. 5, 10). This process is narrow, cylindrical, posterodorsally directed, and inset from the lateral surface of the jugal. The postorbital process of the jugal forms the ventral half of the postorbital bar, separating the orbit from the infratemporal fenestra. At this region the jugal laterally overlaps the descending process of the postorbital.

The jugal bar below the infratemporal fenestra is dorsoventrally narrow and forms the ventral margin of the infratemporal fenestra (except for its posteroventral corner). Along this region, the lateral surface of the jugal is convex and bears a faint longitudinal ridge. Toward the posterior end of the infratemporal opening, the jugal becomes lateromedially flattened. Posterior to this point, the jugal narrows progressively and laterally overlaps the quadratojugal (figs. 5, 10). The posterior end of the jugal is a pointed process that reaches the level of the posterior edge of the otic aperture.

The quadratojugal is a triradiate element, with a notably short and broad anterior branch, an elongate and thin anterodorsal process, and a robust posterior process. The external surface of the quadratojugal is ornamented except for the thin anterodorsal process and the caudal region of the posterior process. The anterior branch forms the anteroventral corner of the infratemporal fenestra and contacts ventrally the posterior end of the jugal, through a posteroventrally directed suture. The quadratojugal ascending process extends dorsally forming the posterior edge of the infratemporal fenestra. This process is moderately thin along its ventral half and bears a slightly developed ridge located close to its anterior margin. The anterior margin of this process has a well-developed knob at the dorsoventral midpoint of the ascending process, resembling the condition of some neosuchians (e.g., Stolokrosuchus lapparenti [Larsson and Gado, 2000], Sarcosuchus imperator [Sereno et al., 2001]). This knob is interpreted as putatively homologous to the quadratojugal spine present in several neosuchians (e.g., Terminonaris robusta [Wu et al., 2001b], Bernissartia fagesii [Norell and Clark, 1990], nonalligatorid crocodylians). Dorsal to the quadratojugal knob, the ascending process of the quadratojugal is markedly narrower and contacts the postorbital (figs. 10, 14).

The posterior margin of the quadratojugal contacts the anterodorsal process of the quadrate along a straight suture. The posteroventral region of the quadratojugal is unsculpted and laterally bulged. It seems to overlap laterally the quadrate, reaching the level of the articular condyles although it apparently does not form part of the craniomandibular articulation (fig. 10), similar to Rugosuchus nonganensis (Wu et al., 2001a) and other advanced neosuchians (e.g., Goniopholis simus BMNH 41098, crocodylians). Although this region is not perfectly preserved in the two specimens studied here, the distal end of the quadratojugal does not seem to form part of the craniomandibular joint.

The quadrates are well preserved in IGM 100/1195 although their articular surfaces are partially covered by the articular. The anterodorsal region of the quadrate is smooth and forms most of the ventral surface of the otic recess (figs. 10, 14). The anterior edge of this process contacts the quadratojugal and its dorsal edge is sutured to the squamosal and parietal. The posterodorsal region of the anterodorsal process of the quadrate forms the anterior and ventral margins of the otic aperture. This opening is anteroposteriorly elongate with a markedly concave anterior margin. As indicated above, this notch is posteriorly closed by the squamosal-quadrate suture (fig. 10). The posteroventral corner of the otic aperture is strongly curved and the quadrate extends as a short dorsal process that bears a dorsoventrally centered protrusion delimiting the dorsal border of the anterior end of the cranioquadrate passage. A small siphoneal foramen is located anteroventrally to the otic aperture. The anterodorsal process of the quadrate bears a slightly depressed area that surrounds the ventral margins of the otic notch and siphoneal foramen (figs. 10, 14).

The distal body of the quadrate is short, robust, and directed posteroventrally. The extension of this region of the quadrate is more developed than in basal crocodyliforms (e.g., Protosuchus richardsoni AMNH FR 3024; Gobiosuchus kielanae ZPAL MgR-II/67), but less than in derived neosuchians (Rhabdognathus sp. CNRST-SUNY 190; Alligator mississippiensis FMNH 8201; Crocodylus niloticus FMNH 1757; Goniopholis stovalli AMNH FR 5782). The distal body of the quadrate is slightly wider than anteroposteriorly long, resembling the neosuchian condition. The posterior surface of the quadrate body is slightly convex and seems to lack a well-developed ridge, a condition noted by Wu et al. (2001a) to be shared with Shamosuchus tersus, Rugosuchus nonganensis, and Bernissartia fagesii. A similar ridge, however, is present in several basal mesoeucrocodylians (e.g., Libycosuchus brevirostris BSP 1912.VIII.574, Sebecus icaeorhinus AMNH FR 3160, Hsisosuchus chungkingensis CNM V 1090). The medial surface of the distal quadrate body is pierced by a small foramen aërum. The quadrate condyles are in their natural position on the articular facets of the craniomandibular joint. The dorsal region of the quadrate body is overhung laterally by the paroccipital process up to the posterior opening of the cranioquadrate passage (fig. 11). The dorsomedial region of the quadrate body contacts the otoccipital through an interdigitated suture. This contact originates near the cranioquadrate passage and runs ventromedially toward the ventral margin of the occiput. Medial to this point, the suture continues along the ventral edge of the occipital surface of the skull, forming a sharp ridge (fig. 11). The ventral surface of the quadrate extends anteriorly to this ridge and is strongly sutured to the posterolateral margins of the basisphenoid and the quadrate processes of the pterygoids. The ventral surface of the quadrates is not fully exposed although the visible region bears a well-developed crest B (sensu Iordansky, 1973) (fig. 18). Details on the quadrate contact with the rest of the braincase are not exposed, although this branch certainly extends dorsally reaching the posterior region of the supratemporal fossa, where it is sutured to the squamosal.

The ectopterygoids are preserved in articulation with the pterygoids in IGM 100/1195 and AMNH FR 6412, being partly occluded by the lower jaw. Their anterolateral end articulates with the maxilla and jugal through well-developed anterior and posterior processes. This element seems to lack the ascending process that extends along the medial surface of the postorbital bar in neosuchians including derived forms such as Hylaeochampsa vectiana (Clark and Norell, 1992), Allodaposuchus precedens (Buscalioni et al., 2001), and nonalligatorid crocodylians (Brochu, 1999). The posteromedial region of the ectopterygoid is sutured to the lateral edge of the pterygoid flange (figs. 4, 9, 18). This suture is sigmoidal and extends from the posterior edge of the suborbital fenestra toward the posterolateral end of the pterygoid flanges. Along this region, the ectopterygoids partially overlap the ventral surface of the pterygoids and fail to reach the posterior end of the pterygoid flanges (as in most mesoeucrocodylians). The ventral surface of these elements is flat, smooth, and broadly exposed on the palate, occupying approximately the lateral third of the lateromedial extension of the pterygoid flanges.

The palatines are partly preserved in the two specimens studied here. The type specimen has preserved the midsection of the palatine secondary palate (fig. 4) and the specimen IGM 100/1195 has preserved its posterior region (fig. 9). These elements are medially sutured to each other constituting a posteriorly extended secondary palate that forms the floor of the nasopharyngeal passage, characteristic of mesoeucrocodylians. The anterior ends of the palatines seem to have extended anteriorly to the suborbital fenestra, between the palatal branches of the maxillae (figs. 4, 9). However, this region is poorly preserved and details on the maxilla-palatine suture or on the degree of extension of this anterior process cannot be determined at the moment. The palatines extend between the suborbital fenestra, forming a relatively narrow palatine bar. Along this region, their lateral margins are concave, producing an hourglass-shaped ventral surface of the palatines (figs. 4, 9).

The posterolateral region of the right palatine of IGM 100/1195 projects a process that contacts the anterolateral region of the pterygoid at the anterolateral corner of the choanal opening. Thus, the choana of Shamosuchus djadochtaensis is enclosed between the palatine and pterygoids (i.e., a “mesosuchian” palate) in contrast to the pterygoid bounded choana of Isisfordia duncani (Salisbury et al., 2006), Hylaeochampsa vectiana (Clark and Norell, 1992), and more derived eusuchians. Mook (1924) has originally interpreted the type specimen as having a eusuchian type choanal opening, but the holotype has this region severely damaged and restored (fig. 4) and the new specimen clearly rejects his original interpretation. The palatine forms the entire anterior margin of the choanal opening and fails to project extensively along the lateral margins of the choana (fig. 9), differing from the condition of all other noneusuchian mesoeucrocodylians (except for Rugosuchus nonganensis and the Glen Rose form). As in Rugosuchus nonganensis (Wu et al., 2001a), the posterolateral end of the palatines does not reach the posterior margin of the suborbital fenestra, resulting in a rather anterior location of the anterior margin of the choanal opening (i.e., rostral to the posterior edge of the suborbital opening). Some of the described species of Shamosuchus have a similar position of the choanal anterior margin (e.g., Shamosuchus ulgicus; Storrs and Efimov, 2000: fig. 20.8). Wu et al. (2001a), however, noted that other species (e.g., Shamosuchus ulanicus) have a more posteriorly located choanal opening, which is the condition found in some advanced neosuchians (e.g., Bernissartia fagesii [Buscalioni and Sanz, 1990], the Glen Rose form [Langston, 1973], Isisfordia duncani [Salisbury et al., 2006]) and Crocodylia. The medial region of the posterior end of the palatines has not been preserved in any of the studied specimens.

The pterygoids are completely fused to each other as in all mesoeucrocodylians. IGM 100/1195 preserves most of the ventral surface of these elements, in contrast to the fragmentary pterygoids of the holotype of Shamosuchus djadochtaensis. The ventral surface of the pterygoids is smooth, laterally flat along the pterygoid flanges and depressed at their medial region (figs. 9, 18). At the anterior end, the medial region has a broad depression that extends laterally up to the caudal end of the suborbital openings (i.e., exceeding in width the lateromedial extension of the palatine bar between the suborbital fenestrae). The lateral walls of the broad depression are subvertically oriented laminae that separate the choana from the suborbital fenestra and their anterior ends contact the posterolateral projections of the palatines (fig. 9). This broad and shallow depression bears a markedly concave choanal groove, which is deep, narrow, and elongate (fig. 9). The narrow choanal groove is more clearly differentiated from the broad medial depression of the pterygoids toward the anterior end of the choanal opening (approximately at the level of the anteroposterior midpoint of the suborbital opening). Posteriorly, the narrow and deep choanal groove gradually opens into the broader medial depression of the pterygoids, at the level of the pterygoid flanges. This region is closed posteriorly by a well-developed buttress located close to the posterior edge of the pterygoid. The presence of a relatively broad choanal opening is a plesiomorphic condition for neosuchians, and clearly differs from the reduced opening present in Hylaeochampsa vectiana and more derived eusuchians (Salisbury et al., 2006). In IGM 100/1195 there are no signs of a pterygoidal choanal septum.

The pterygoids expand laterally forming well-developed pterygoid flanges. These flanges are large, laminar, and directed laterally at their base. Along their lateromedial extension, the pterygoid wings gradually deflect ventrally (figs. 11). The pterygoid flanges seem to lack the air cavities present in some basal crocodyliforms (Edentosuchus tienshanensis GMPKU-P 200101, Araripesuchus buitreraensis MPCA-PV 235, Notosuchus terrestris MACN-RN 1037). The anterior edges of the pterygoid flanges are L-shaped and form the posterolateral and posterior margin of the suborbital opening. As mentioned above, their lateral edges are overlapped by the ectopterygoids. The posterior edge of the pterygoid flanges is directed posteromedially. The medial region of their posterior margin bears a deep notch bounded by the base of the quadrate processes of the pterygoids (fig. 9). The base of these processes is markedly narrow, as in most advanced neosuchians, and extends dorsally contacting the basisphenoid and the quadrates (fig. 11). The posterior surface of this dorsal extension of the pterygoids is strongly concave and contacts the anteromedial region of the basisphenoid. The quadrate processes of the pterygoids extend dorsolaterally, bordering the lateral margins of the basisphenoid through an interdigitated suture. These processes are rather extensive and are dorsally sutured to the pterygoid processes of the quadrates.

The basisphenoid is preserved only in IGM 100/1195. The ventral surface of this element is crescent shaped and has a reduced exposure on the ventral surface of the braincase (figs. 9, 11). The anterior edge of the basisphenoid is convex and contacts the pterygoids through an interdigitated suture. The posterior edge of the basisphenoid is deeply concave and is similarly sutured to the basioccipital (figs. 4, 9, 11). The basisphenoid exhibits elongate dorsolateral processes that are bounded posterodorsally by the basioccipital and the ventromedial extension of the otoccipital and posteroventrally by the pterygoid processes of the quadrate and the pterygoids. The distal end of these processes is slightly expanded with respect to their base, extending for a short distance onto the lateral surface of the braincase (rather than onto the occipital surface of the skull). This expansion could be homologous with the well-developed basisphenoid exposure on the lateral surface of the braincase present in Isisfordia duncani (Salisbury et al., 2006) and crocodylians. The ventral surface of the basisphenoid is flat, except for its posterior edge, which bears a well-developed ridge along its contact with the basioccipital (figs. 4, 9, 11). The foramen intertympanicum is rather large and located medially on the basisphenoid-basioccipital suture. The lateral eustachian foramina are also enclosed between these two bones, but are notably smaller than the foramen intertympanicum and located at the midpoint of the posterolateral processes of the basisphenoid.

The degree of exposure of the basisphenoid is intermediate between the plesiomorphic condition of basal crocodyliforms and the derived condition of crocodylians (fig. 9). The anteromedial region of the basisphenoid is reduced, contrasting with the broad basisphenoid of basal crocodyliforms (e.g., Gobiosuchus kielanae ZPAL MgR-II/67, Sichuanosuchus shuhanensis IVPP V10594; Zosuchus davidsoni IGM 100/1305). The basisphenoid of IGM 100/1195 (in particular its posterolateral process), however, is more exposed in ventral view than in crocodylians (e.g., Alligator mississippiensis FMNH 8021, Crocodylus niloticus FMNH 17157) and most neosuchians (e.g., Rhabdognathus CNRST-SUNY 190, Sarcosuchus hartii MNN 604, Isisfordia duncani [Salisbury et al., 2006], Hylaeochampsa vectiana BMNH R177). The condition of Shamosuchus djadochtaensis, with a moderately exposed anteromedial region and well-exposed posterolateral processes, resembles the morphology of the basal neosuchian Theriosuchus pusillus (Clark, 1986) and several basal mesoeucrocodylians (e.g., Notosuchus terrestris MACN-RN 1037, Simosuchus clarki UA 8679, Baurusuchus pachecoi DGM 299-R, Lomasuchus palpebrosus MOZ-P 4084). The condition of IGM 100/1195, however, seems to be more derived than in these forms in two features. First, the exposure of the basisphenoid of Shamosuchus djadochtaensis is slightly less developed in ventral view. Second, the anteromedial region of the basisphenoid projects further ventrally in Shamosuchus djadochtaensis, so that it is visible in posterior view (fig. 11), as in extant forms. The latter feature can be interpreted as an incipient stage in the development of the strong verticalization of the basicranium that characterizes crocodylians (Tarsitano, 1985).

The basioccipital of Shamosuchus djadochtaensis is subrhomboidal and is laterodorsally bounded by the otoccipitals and ventrally by the basisphenoid. At its dorsal end, the basioccipital forms most of the occipital condyle. The occipital condyle lacks a well-developed neck and is slightly deflected posteroventrally, as is the rest of the basioccipital surface extending ventrally from the condyle. This orientation of the basioccipital surface is also found in Isisfordia duncani (Salisbury et al., 2006), although it is also present (and more developed) in non-neosuchian crocodyliforms (e.g., Gobiosuchus kielanae ZPAL MgR-II/67, Zosuchus davidsoni IGM 100/1305, Notosuchus terrestris MACN-RN 1037, Araripesuchus gomesii AMNH FR 24450, Simosuchus clarki UA 8679). Other neosuchians, however, have a more vertically oriented basioccipital that faces posteriorly (Lomasuchus palpebrosus MOZ-P 4084, Sarcosuchus hartii MNN 604, Rhabdognathus CNRST-SUNY 190, Goniopholis simus BMNH 41098). This condition is also present in eusuchians (including Hylaeochampsa vectiana BMNH R177), due to the strong verticalization of the braincase (Tarsitano, 1985).

Ventral to the occipital condyle, the basioccipital bears a slightly developed medial depression and a moderately developed sagittal ridge extending along the ventral half of the basioccipital. This ridge reaches the posterior margin of the foramen intertympanicum, where it bifurcates surrounding the posterolateral edges of this opening (fig. 9). The lateral region of the basioccipital expands along their contact with the otoccipital, giving the subrhomboid shape to this bone (figs. 9, 11). The basioccipital-otoccipital suture is slightly interdigitated and reaches the ventral margin of the occipital surface of the skull, at the triple contact between the basioccipital, otoccipital, and basisphenoid (located just ventral to the foramen for the internal carotid artery). Ventromedially to this point, the basioccipital is sutured to the basisphenoid as described above. The basioccipital lacks well-developed tubera, a character usually present in long snouted crocodyliforms and consequently it is not expected here.

The otoccipitals are partially preserved in both specimens studied here, although those of IGM 100/1195 are more complete. The otoccipitals extend along most of the occipital surface of the skull (fig. 11). Their medial margins form the lateral and dorsal margins of the foramen magnum and contact each other medially above this opening. Dorsally to this point, the otoccipitals contact the supraoccipital, although details on this contact cannot be determined due to the poor preservation of this region. Most of the lateral extension of the otoccipitals is exposed on two different planes. The ventral half of the otoccipital is exposed posteroventrally, although this region is slightly more vertical than the basioccipital. This morphology resembles the condition of Rugosuchus nonganensis (Wu et al., 2001a) and several other noneusuchian crocodyliforms. The ventral region of the otoccipital narrows ventrally, lacking the enlarged ventrolateral process that characterizes most basal crocodyliforms and thalattosuchians (Clark, 1986). The lateral margin of this region of the otoccipital contacts the quadrate, forming the ventral ridge of the occipital surface of the skull (as described above). The ventral end of the otoccipital is acute and wedges between the basioccipital and the basisphenoid (figs. 9, 11). The ventral end of this surface is pierced by the foramen for the internal carotid artery. Dorsolaterally to this foramen, the otoccipital bears a slightly larger opening for the cranial nerves IX–XI (fig. 11). A single foramen for the cranial nerve XII is located just lateral to the foramen magnum.

The dorsal half of the otoccipital forms a relatively short paroccipital process, which briefly extends laterally to the posterior opening of the cranioquadrate passage. A similarly reduced lateral extension of the paroccipital process is also present in other advanced neosuchians (e.g., Gonipholis stovalli AMNH FR 5782, Isisfordia duncani [Salisbury et al., 2006], Allodaposuchus precedens [Buscalioni et al., 2001], Hylaeochampsa vectiana BMNH R177), but differs from the relatively long paroccipital process of crocodylians. This process is subtabular in shape and vertically oriented. The dorsal margin of this region of the otoccipital is sutured to the occipital flange of the squamosal (as described above). The lateral end of the paraoccipital process is slightly narrower and curved posteriorly, but forms a relatively blunt lateral end as in most mesoeucrocodylians. The ventral margin of the paroccipital process forms a transverse ridge that overhangs the cranioquadrate passage (fig. 11). Above this margin, the paroccipital process displays slightly marked striations.

Only the dorsalmost region of the supraoccipital is preserved in IGM 100/1195 (fig. 11). This region bears a poorly developed sagittal ridge and does not seem to extend onto the skull table, as in other noneusuchian neosuchians (and Alligator; Brochu, 1999). The laterodorsal region of the supraoccipital bulges slightly close to the location of the posttemporal fenestra.

The laterosphenoid forms the anterolateral wall of the braincase. The left laterosphenoid is exposed and well preserved in IGM 100/1195. It is divided into distinct anterior and posterior surfaces by a prominent cotylar crest, with the posterior surface forming the anteromedial wall of the supratemporal fenestra. The laterosphenoid contacts the quadrate posteriorly in a robust suture enclosing the trigeminal foramen. The ventroposterior margin of the trigeminal foramen is obscured, so it is unclear whether, or to what extent the prootic participated in the nerve foramen. Anterior to the trigeminal foramen, the laterosphenoid has a moderate suture with the basisphenoid medially and the pterygoid ventrally. This area corresponds to the location of the laterosphenoid bridge in Crocodylia. In IGM 100/1195 this area is broad and clearly lacks a distinct passage for the ophthalmic branch of cranial nerve V. This is also the case in the basal eusuchian Hylaeochamsa vectiana (Clark and Norell, 1992), which indeed lacks a laterosphenoid bridge. On the posterior side of the cotylar crest, just anterior to the trigeminal foramen there is a shallow groove that may correspond to the path of the ophthalmic branch of cranial nerve V as it passes over the laterosphenoid body. Anteriorly the laterosphenoids converge toward the midline; however, IGM 100/1195 does not preserve details of the notches and/or foramina for the passage of cranial nerves II, III, and IV. Anterodorsally, the laterosphenoid sutures broadly with the frontal. The capitate process is well preserved and contacts the postorbital near the midpoint of the anterior body of the bone.

Posterior to the basisphenoid rostrum, CT imagery indicates small sellae turcicae on the basisphenoid (fig. 19). Although artifacts obscure the exact location of the anterior border of the dorsum sellae, the foramina passing through the dorsum sellae are well preserved and clear in CT images. Posterior to the sellae two pairs of foramina pierce the basisphenoid. The medial pair mark the passage of the anterior carotid artery into the tympanic cavity (fig. 19; see also Colbert, 1946). Lateral to the anterior carotid foramina smaller opening are present corresponding to the exit of cranial nerve VI from the braincase. Ventral to these passages, a large cavity corresponding to the anterior branch for the median Eustachian tube is preserved. Moving posterolaterally, the anterior and posterior branches of the median Eustachian tube open into a large rhomboidal sinus. Dorsal to the posterior border of the rhomboid sinus the metotic fissure is seen dividing the opisthotic from the basioccipital. The opisthotic is generally well preserved, but its exact suture with the exoccipital (forming the otoccipital) is not clearly defined. The opisthotic forms the posteroventral border of the tympanic bulla, delineates the dorsal border of the metotic fissure, and forms the dorsal wall of the perilymphantic foramen. The anterior portion of the lateral semicircular canal passes within the opisthotic. Also, the canal for the foramen vagi is well defined. The canal is undivided, unlike that in living crocodylians (Iordansky, 1973) and passes posterolaterally from the metotic fissure toward its exit from the skull near the posterior carotid foramen.

The area surrounding the prootic is not exposed; thus, it is uncertain whether the bone is visible on the lateral side of the braincase. Internally, it forms the anterior portion of the bulla tympani (cochlear prominence) and part of the lateral wall of the braincase (fig. 19). The prootic contains the common crus of the anterior and posterior semicircular canals, the ventral portion of the anterior semicircular canal, and the anterior half of the lateral semicircular canal, although the lateral semicircular canal is very poorly preserved.

The Mandible

The mandible is dorsoventrally shallow anteriorly and a solid and robust element posteriorly. The external mandibular fenestra is completely obliterated, as in other species described for Shamosuchus and several noncrocodylian neosuchians such as Theriosuchus pusillus (BMNH R48328), Gonipholis simus (Salisbury et al., 1999), Bernissartia fagesii (Norell and Clark, 1990), Rugosuchus nonganensis (Wu et al., 2001a), and the Glen Rose form (Brochu, 1999). Isisfordia duncani (Salisbury et al., 2006) and most crocodylians instead have a well-developed external mandibular fenestra. The symphysis is only partly preserved and posterior to it, the mandibular rami diverge at an angle of approximately 40°.

The dentaries are almost complete in IGM 100/1195, having only the anterior symphyseal region missing. The anterior half of the dentaries is inset between the maxillae, whereas the posterior region of the dentaries is aligned with the lateral edge of the maxillae. Anteriorly, the height of the dentary is markedly low, forming a shallow mandibular symphysis (figs. 9, 10, 12). Posteriorly the mandibular rami diverge continuously and the dentaries gradually increase their dorsoventral depth. The ventral and lateral surfaces of the dentaries of IGM 100/1195 are deeply ornamented with the pitted pattern present on most skull bones, whereas the preserved portion of the dentaries of the holotype is smooth. The ventral end of the dentary is medially deflected, forming the lateral half of the ventral surface of the mandibular rami. This margin of the dentary is sutured to the ventrolateral margin of the splenial through a linear suture. The lateral surface of these elements is laterally convex, as in most neosuchian crocodyliforms (Ortega et al., 2000). The buccal margin of the dentaries is mostly occluded by the maxillae, except for their posterior region. At this point, specimen IGM 100/1195 shows a smooth laterodorsally exposed surface bordering the posteriormost region of the lower tooth row. In this smooth surface there are no signs of neurovascular foramina in IGM 100/1195.

The posterior ends of the dentaries are entirely sutured to the surangular and angular, thus forming the absence of an external mandibular fenestra (figs. 9, 10). The posterodorsal margin of the dentaries borders the acute anterior process of the surangular, along a straight suture directed posteroventrally. Ventral to this point, the posterior region of the dentary is sutured to the angular, although details on this contact cannot be determined in the two specimens studied here. The lower dentition cannot be observed in any of the specimens reported here due to the inset position of the mandibular rami between the maxillae.

The splenials form part of the mandibular symphysis (at least on its ventral surface). Along this region, the splenials form the medial third of the ventral surface of the mandibular symphysis. The participation of the splenials on the mandibular symphysis is also present in basal crocodylians (Brochu, 1999) and most noncrocodylian neosuchians (e.g., Rugosuchus nonganensis [Wu et al., 2001a], Terminonaris robusta [Wu et al., 2001b], Sarcosuchus imperator MNN 604, Hyposaurus rogersii [Denton et al., 1997], Theriosuchus pusillus BMNH R48330). Posteriorly, the splenials cover the medial surface of the mandibular rami and the medial half of their ventral surface (figs. 4, 9). The ventrolateral edge of the splenial is sutured to the dentary along a straight suture that extends on the anterior half of the mandibular ramus. The medial surfaces of the splenials are smooth and slightly convex. Although this surface is slightly damaged anteriorly in IGM 100/1195, there are no signs of a foramen intermandibularis oralis. Similarly, the foramina intermandibularis medius and caudalis seem to be absent in IGM 100/1195. The posteroventral margin of the medial lamina of the splenial is sutured to the anteromedial surface of the angular, through a slightly interdigitated suture running posterodorsally. The posterodorsal margin of the splenial lamina seems to be slightly broadened and deflected medially. Presumably, this surface formed a broad medial margin of the posterior region of the lower tooth row, a morphology also present in several basal globidontans (Brochu, 1999). The coronoids have not been preserved as isolated elements, although at the moment we cannot determine whether these elements were fused to the splenials or were lost during preservation (fig. 17).

Figure 17

Posterior mandible of referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in medial view. See appendix 5 for abbreviations. Asterisk indicates likely position of coronoid had it been present or preserved.

The angular forms more than half of the lateral surface of the posterior mandibular ramus (figs. 5, 9, 10, 17). In the two specimens studied here, the lateral surface of this element is heavily ornamented and bears a well-developed longitudinal ridge located at its dorsoventral midpoint. This ridge extends from its anterior margin to the posteroventral end of the mandibular ramus, where it is deflected ventrally (fig. 10). This ridge is absent in crocodylians, but a similar structure has been noted as present in Shamosuchus gradilifrons (Wu et al., 2001a) and is also recorded in some forms that lack an external mandibular fenestra, such as the neosuchians Rugosuchus nonganensis (Wu et al., 2001a), Theriosuchus pusillus (BMNH R48328), and the Glen Rose form. Additionally, an angular ridge is also present in some basal crocodyliforms (e.g., Zaraasuchus shepardi IGM 100/1321, Gobiosuchus kielanae ZPAL MgR-II/68) although their ridge is located at the ventral margin of the mandibular ramus, rather than on its lateral surface as in the above-mentioned neosuchians. The presence of this ridge, however, is not strictly correlated with the absence of an external mandibular fenestra since it is similarly present in forms that have this opening (e.g., Simosuchus clarki UA 8769) and absent in others that lack this fenestra (e.g., Bernissartia fagesii [Wu et al., 2001a], Geosaurus araucanensis MACN-N 95). Due to its position on the lower jaw, this ridge may have served as the posterior insertion point of the m. pterygoideus posterior.

The dorsal margin of the angular contacts the ventral margin of the surangular through an interdigitated suture oriented longitudinally along the lateral surface of the lower jaw (fig. 5). This suture is located slightly above the dorsoventral midpoint of the mandibular ramus and extends horizontally to the caudal end of the lower jaw, reaching the lateral margin of the retroarticular process (fig. 18). A horizontal and dorsally located angular-surangular suture is also present in Rugosuchus nonganensis (Wu et al., 2001a) and Bernissartia fagesii (Buscalioni and Sanz, 1990). The posteriormost region of the angular that borders the retroarticular process lacks the ornamentation present in the rest of the angular's lateral surface. The medial surface of the angular is smooth and seems to be dorsoventrally low, although it is not completely exposed in IGM 100/1195. Posteriorly, the ventral surface of the angular contacts the articular, covering the lateral region of the ventral surface of the retroarticular process (extending up to its posterior end; fig. 18). The ventral margin of the angular of Shamosuchus djadochtaensis is slightly deflected dorsally, in contrast to the highly curved angular of Rugosuchus nonganensis (Wu et al., 2001a) and Bernissartia fagesii (Buscalioni and Sanz, 1990; Norell and Clark, 1990) that form angles of 110–120° (see Wu et al., 2001a, for a discussion).

Figure 18

Cranioquadrate articulation of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in ventral view. See appendix 5 for abbreviations.

Figure 19

Coronal CT section through the braincase of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195. See appendix 5 for abbreviations.

The surangular extends anteriorly as an acute process that contacts the posterodorsal region of the dentary, reaching the level of the anteroposterior midpoint of the orbit. At this point, the lateral surface of the surangular is pierced by a moderately large surangular foramen (figs. 5, 10). Posterior to this point, the surangular increases its dorsoventral depth along its suture with the dentary and angular. The dorsal surface of the surangular of IGM 100/1195 is rather flat and not dorsally bowed as it seems to be in the poorly preserved holotype of Shamosuchus djadochtaensis. Except for the anterior process, the lateral surface of the surangular is heavily ornamented. This element bears a rounded ridge that extends along the posterior half of its dorsal margin. Posteriorly, this ridge curves ventrally following the outline of the caudal end of the mandibular ramus (figs. 7, 10). Like the angular ridge, this structure is also present in other forms that lack an external mandibular fenestra (e.g., Zaraasuchus shepardi IGM 100/1321, Gobiosuchus kielanae ZPAL MgR-II/68). The surangular ridge, however, is absent in other crocodyliforms that lacks such fenestra (e.g., Geosaurus araucanensis MACN-N 95). The posterior end of the lateral surface of the surangular reaches the posterolateral end of the retroarticular process, as in Rugosuchus nonganensis (Wu et al., 2001a) but in contrast to the condition of other advanced neosuchians (e.g., Bernissartia fagesii). This region is smooth and continuous with the smooth posterior end of the angular (fig. 10).

The articular is exposed only on its ventral, medial, and posterior surfaces. The lateral surface is completely covered by the angular and surangular and its dorsal surface occluded by the quadrate. The rostral end of the articular tapers anteriorly into a rod-shaped anterior process, as in most crocodyliforms. Posteriorly, the medial surface of the articular is concave and expands medially to support the articular facets for the quadrate (fig. 9). At its medial margin, the process that supports the internal articular facet for the quadrate is dorsoventrally thin. The caudal margins of these facets are bounded by a posterior buttress as in most crocodyliforms. Posteriorly, the articular continues into a broad and short retroarticular process (figs. 8, 17, 18). This process projects posteroventrally and its dorsal surface is slightly concave, smooth, and lacking the longitudinal ridge present in several crocodyliforms. Most of this surface faces posteriorly, but its posteromedial end is slightly deflected medially. The morphology of the retroarticular process of Shamosuchus djadochtaensis resembles that of some neosuchian taxa (e.g., Theriosuchus pusillus [Clark, 1986], Rugosuchus nonganensis [Wu et al., 2001a]). This condition, however, contrasts that of eusuchians (e.g., Alligator mississippiensis FMNH 8021, Paleosuchus palpebrosus FMNH 69867, Crocodylus niloticus FMNH 17157) and longirostrine forms (e.g., Terminonaris robusta [Wu et al., 2001b], Sarcosuchus imperator MNN 604), which have an elongated, dorsally facing, and dorsally recurved process.

The Axial Skeleton

A well-preserved cervical series was found associated with the skull of IGM 100/1195 and removed during preparation (figs. 20, 21). All eight cervical vertebrae are preserved with the exception of the neural arches of the atlas (fig. 21). The right side of the cervical series is obscured by three lateral cervical osteoderms, of which the two anteriormost are in overlapping articulation. The left side of the seventh cervical is slightly obscured by a partial lateral osteoderm. All other surfaces and views of the cervical column are well exposed. A proatlas is not present or was not preserved. In addition to this, the first three dorsal vertebrae with associated ribs were recovered, as well as five caudal vertebrae (figs. 22–Figure 2324). Prior to discovery of IGM 100/1195, the only reported vertebrae from species referred to Shamosuchus were amphicoelous (i.e., “Paralligator” spp.; Konzhukova, 1954). The cervical vertebrae and the first dorsal of IGM 100/1195, however, are procoelous. Unfortunately, the condition in the remaining dorsal vertebrae cannot be determined and all the preserved caudals are amphicoelous. Since the first caudal vertebra is not unequivocally identified among the preserved caudals, it cannot be ruled out that this vertebra is biconvex like in Bernissartia fagesii and eusuchians.

Figure 20

Anterior postcranial remains of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195. See appendix 5 for abbreviations.

Figure 21

Cervical vertebrae and osteoderms of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195, in A, dorsal view; B, right lateral view; C, ventral view; and D, left lateral view. See appendix 5 for abbreviations.

Figure 22

Cervicodorsal vertebrae of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195. A, eighth cervical vertebra in posterolateral view; B, first dorsal vertebra in anterolateral view; C, anterior dorsal vertebrae in posterolateral view.

Figure 23

Anterior dorsal ribs of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195. See appendix 5 for abbreviations.

Figure 24

Caudal vertebrae of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195. See appendix 5 for abbreviations.

The intercentrum of the atlas is a simple bone that is wedge shaped and broader than long, with a shallow concavity separating the two capitular facets (fig. 21). This morphology is fairly conserved among crocodylomorphs. The atlantal intercentra known in Hesperosuchus agilis AMNH FR 6758 (Clark et al., 2000), Mahajangasuchus insignis (Buckley and Brochu, 1999), Araripesuchus tsangatsangana (Turner, 2006), Sunosuchus junggarensis (Wu et al., 1996a), and Crocodylus porosus FMNH 15529 share similar shapes and proportions. This is in contrast to the relatively platelike atlas of alligatoroids.

The dorsal surface of the bone is weakly concave for the reception of the occipital condyle. The anteriormost portion of the dorsal surface is formed by the short knoblike anterior margin of the atlas. The neurapophyses were not preserved in the specimen.

The odontoid process is strongly sutured to the axis (fig. 21D). It is nearly 13 mm wide and 6 mm deep at its extremes, making it wider than the body of the axis. In dorsal view, the odontoid process is similar to that figured for Sunosuchus junggarensis (Wu et al., 1996a). There is a midline lingulate surface projecting from the otherwise gently curved anterior margin. The ventral surface of the odontoid is obscured by the atlas.

The centrum of the axis is approximately 11 mm long, similar in length to the postaxial cervicals (fig. 21). The axial centrum is constricted medially and depressions are present on either side of the vertebra. Most of the ventral surface of the centrum is covered by a small, partly preserved osteoderm. Nonetheless, an anterior hypapophysis appears not to be present. The suture between the neural arch and the centrum is closed. In lateral profile, the neural arch is as tall as the axial centrum. A small notch is located posteriorly between the neural arch and centrum. This notch is superficially similar to that of the eusuchian Diplocynodon hantoniensis (Brochu, 1999), but does not extend to the ventral margin of the postzygapophysis as in the later taxon. The right prezygapophysis is damaged. The left prezygapophysis is small and poorly developed, but appears to be complete. The postzygapophyses are robust, small, and curved laterally. The articular facets are approximately on level with the prezygapophyses and face ventrolaterally. The neural spine is anteroposteriorly broad, running the entire length of the neural arch. In lateral view, the dorsal surface of the spine has a horizontal profile. The axis neural arch lacks a lateral process (“diapophysis”). The facet for the tuberculum of the axial rib is located on the odontoid process. The capitular facets appear to be divided between the axial centrum and the ventrolateral surface of the odontoid process. Neither atlantal nor axial ribs are preserved.

All six postaxial cervical vertebrae are preserved in an articulated death posture (fig. 21). As a result the dorsal surface of these vertebrae are not well exposed. The centra are similar to those of most mesoeucrocodylians, being slightly longer than they are tall. As all the cervicals are in articulation, the nature of the articulation cannot be determined for most. The eighth cervical possesses a partly weathered condyle on its posterior articular surface. Therefore, at the very least, the eighth cervical was procoelous (fig. 22A). Because of the surface of the posterior condyle of the eight cervical is partially weathered, it cannot be determined at the moment whether the condyle was hemispherical (as in crocodylians) or slightly less developed.

The ventral surface of the cervical vertebra is constricted at the midpoint and bears a weakly developed keel (hypapophysis) anteriorly. This keel is very small on c3, with them becoming progressively larger moving posteriorly. The tip of the keel on c5 is damaged but the posterior aspect of the keel is marked by a shallow groove (fig. 21C). A very similar groove is present on the keel of c6 in Araripesuchus tsangatsangana (Turner, 2006) and Mahajangasuchus insignis UA 8654, and Wu et al. (1996a) noted a groove present anteriorly on the ventral ridges of c7 and c8 in Sunosuchus junggarensis. The anterior margins are notably more expanded than the posterior margins due largely to the well-developed parapophyses. Moving posteriorly through the cervical series, the centrum bodies become wider and the parapophyses broader with larger contact facets. The lateral surfaces of c3, c4, and c5 bear shallow depression between the parapophyses and diapophyses. The diapophysis of c3 is small and cylindrical. The remaining diapophyses are longer and flatter, with oval contact facets. The diapophyses present on c7 and c8 are roughly twice the length of the others and are located entirely on the surface of the neural arch.

The prezygapophyses are well developed. On c3, the prezygapophyses are smaller than succeeding ones and strongly curve medially. The remaining prezygapophyses are larger and project more laterally. The surface of the articular facets are not exposed, but are directed more dorsally than medially. The postzygapophyses are on level with, and roughly equal in size to, the prezygapophyses. These processes curve laterally, and the articular facets face ventrolaterally. A poorly developed ridge extends from the posterior surface of the postzygapophyses onto the posteroventral margin of the neural spines (fig. 21A). These ridges do not resemble the suprapostzygapophyseal laminae seen in Notosuchus terrestris (Pol, 2005), Araripesuchus gomesii AMNH FR 22450, or Araripesuchus tsangatsangana (Turner, 2006).

Neurocentral sutures are completely closed on the lateral surfaces of the cervical vertebrae (fig. 21D), although this suture is visible on the internal surface of the neural canal of the eigth cervical vertebra. The neural canal is large, square, and about as large as the condylar surface of the centrum. The neural spines are damaged on c3 through c6 and c8. The neural spine of c7 is tall and anteroposteriorly narrow, with the posterior distal surface slightly curved. Very small median laminae are preserved on the posterior surface of the spine.

Four cervical ribs are preserved (c5, c6, c7, and c8) (fig. 21C). These vertebrae exhibit a highly conserved crocodyliform morphology (Mook, 1921; Whetstone and Whybrow, 1983). In c5–c7, the tuberculum and capitulum are roughly equal in size and the shaft of the rib is perpendicular to these processes. The shaft consists of a tapered anterior process and a dorsoventrally broad posterior process, which narrows posteriorly. Ventrally, the shaft is distinct from the capitular process, marked by a slight ventral development. The c8 is transitional in form between typical cervical and dorsal rib morphologies, as in many crocodyliforms (fig. 23). The tubercular process is small and the capitulum is represented by a small, thin process. These processes are closely spaced as in cervical ribs. Unlike other cervical ribs, the shaft of rib c8 is not perpendicular to the rib process. The rib shaft is short and weakly curved if at all.

The three anteriormost dorsal vertebrae are preserved articulated with one another and the cervical series (fig. 22C). The first dorsal vertebra is procoelous (fig. 22B). The centra of the other two are not well enough preserved to determine their condition. In ventral view, the centrum of d1 is about as broad as long. All three dorsals possess a distinct hypapophysis. The hypapophysis becomes less robust and more keellike by d3.

Distinct parapophyses are present and located near the neurocentral suture in all three vertebrae, although the left parapophysis on d3 is unpreserved. In d1 the parapophysis is located just below the suture and is anteriorly displaced relative to the diapophysis. In d2 and d3, the parapophyses rest on or just above the suture and are aligned anteroposteriorly with the diapophyses. The parapophyses are shorter and stouter than those on the cervical centra. The diapophyses of d1 are nearly cylindrical and short, just extending beyond the margin delimited by the postzygapophyses. The diapophyses increase in size posteriorly, becoming more dorsoventrally flattened.

The first dorsal vertebra of IGM 100/1195 is very similar to Sunosuchus junggarensis (Wu et al., 1996a) with respect to a number of features. The neural canal is extremely large in both taxa, roughly equal in size to the anterior contact surface of the centrum. This is in contrast to the condition in Crocodylia where the neural canal is markedly smaller than the anterior surface of the centrum. Consequently, the neural arch is tall and slanted dorsolaterally. This causes wide separation of the prezygapophyses on the arch, lateral to the edge of the centrum and parapophysis (fig. 22B). They do not extend anteriorly beyond the margin of the centrum, however. The articular facets face medially, with the surface slanted at about a 45° angle. The postzygapophyses are also widely spaced and extend just slightly beyond the posterior margin of the centrum.

On d2 and d3, the prezygapophyses are broad, very weakly slanted medially, and again widely spaced on the neural arch and proximal diapophyses. This differs from the closely spaced and strongly slanted prezygapophyses in mesoeucrocodylians like Araripesuchus gomesii AMNH FR 22450, Araripesuchus tsangatsangana (Turner, 2006), Notosuchus terrestris (Pol, 2005), and Mahajangasuchus insignis UA 8654 (Buckley and Brochu, 1999), as well as in Crocodylia (Mook, 1921). The widely spaced prezygapophyses in Shamosuchus djadochtaensis correspond to the widely spaced postzygapophyses. The postzygapophseal facets are ovular and offset laterally from the main body of the postzygapophysis. Unlike crocodylians, the postzygapophyses are positioned at the same level as the prezygapophyses (fig. 22C). Viewed dorsally, a thin lamina connects the postzygapophysis with the posterior surface of the diapophysis. Located posteriorly between the postzygapophyses is a relatively deep median small triangular fossa.

Only d1 preserves a complete neural spine (fig. 22). The spine is tall and anteroposteriorly narrow. Similar to Notosuchus terrestris (Pol, 2005), Araripesuchus gomesii AMNH 22450, Araripesuchus tsangatsangana (Turner, 2006), and Sunosuchus junggarensis (Wu et al., 1996a), but unlike crocodylians and Mahajangasuchus insignis UA 8654, the spine does not slant posteriorly. A distinct posterior thin lamina appears lacking. The dorsal terminus of the spine flares slightly, but not to the extent seen in Crocodylia. Although the remaining spines are missing near the base, it is apparent that the spines become anteroposteriorly broader in the more posterior dorsal vertebrae.

Three right and four left dorsal ribs were preserved (figs. 22, 23). The anterior two right ribs correspond to d1 and d2, but are largely obscured by matrix (fig. 22C). The third right rib is nearly complete, with its posterior surface well exposed. The tuberculum is short and stout, forming a robust contact with the diapophysis. The capitulum is very long and cylindrical. The shaft of the rib curves away slightly from the head of the rib, with the remainder of the shaft straight. The proximal lateral edge of the rib appears as a thin flange. This is formed in part by a longitudinally running depression near this lateral edge.

The four left dorsal ribs are isolated from their complementary vertebrae, making their exact identification equivocal (fig. 23). The anteriormost of the four is interpreted as the left rib of d3. This is based on its relative position within the block containing the dorsal vertebrae and its close similarity to the right d3 rib. The remaining three left ribs corroborate this identification. The tubercula of these ribs are small to completely absent, which is what is predicted given the changes in rib morphology seen in Crocodylia (Mook, 1921). The shafts of these three ribs (d4–d6) are straight to very weakly flexed. No lateral flanges are preserved on the ribs' shafts.

The five caudal vertebrae are likely from the anterior part of the tail, although exact identification is not possible (fig. 24). This interpretation is based on their long transverse processes, anteroposteriorly broad neural spines, relatively short centrum bodies, and the presence of a partial ischium near the anteriormost of the five caudals. The anteriormost caudal vertebra is heavily damaged, missing most of the left side and both transverse processes. All of the caudal centra are amphicoelous. The centra of the anteriormost two caudals are stout, being anteroposteriorly short but still slightly longer than broad. The posterior three preserved caudal vertebrae are much more spool shaped. They are about twice as long as wide, and are broadly constricted along their length. The ventral surfaces of the centra are smooth like in those of Pachycheilosuchus trinquei (Rogers, 2003) and Sunosuchus junggarensis (Wu et al., 1996a), and they lack the parallel ridges seen in basal mesoeucrocodylians like Araripesuchus gomesii AMNH FR 22450 and in modern crocodylians (e.g., Alligator sinensis FMNH 197946).

Where preserved, the transverse processes are long (at least slightly longer than the vertebral body) and roughly 0.5 cm wide. The region of the neural arch connecting the transverse process to the prezygapophysis bears a semicircular depression. Prezygapophyses are long, narrow, and extend laterally at approximately a 45° angle. This angle decreases in the posteriormost preserved caudal vertebra, where the prezygapophyses are directed more anteriorly. The prezygapophyses extend beyond the anterior border of the centrum. Postzygapophyses are short, the contact facets ovular, and they sit more dorsally than the prezygapophyses. They project laterally in the anterior caudals (complementing the prezygapophyses), becoming more posteriorly directed in the more distal vertebrae. The neural spines are anteroposteriorly broad. They are not, however, particularly tall—less than the length of the transverse processes. Neither anterior nor posterior thin laminae can be discerned in any of the preserved caudal vertebrae. The distal ends of the neural spines are slightly expanded, as in the dorsal vertebrae.

Two chevrons are preserved with the caudal vertebrae (fig. 24). The first lies between the second and third preserved caudals. This chevron is heavily damaged. The second chevron is more complete and located between the third and fourth preserved caudals. This chevron is just over 3 cm long. The haemal canal is less than a centimeter tall. The anterior surface of the chevron below the haemal canal bears a pronounced ridge along its length.

The Shoulder and Forelimb

Appendicular remains are more scarce than axial or cranial elements. The left and right humerus and radius are preserved, as well as the left ulna. No identifiable carpal or manual elements are present among the preserved postcranial material. The distal end of the right coracoid is exposed in medial view (fig. 20). This bone is flat and noticeably expanded at its distal terminus. The shaft of the coracoid, however, is obscured by a series of articulated dorsal ribs.

The right humerus is complete, but fractured at the middiaphysis (fig. 25). Proximally, only the dorsal surface is exposed. The left humerus is less complete. Only the distal half is preserved, with the lateral and ventral surfaces exposed. The humerus of Shamosuchus djadochtaensis is expanded to a similar extent at both ends. The shaft is relatively straight like in Alligatorium meyeri (Wellnhofer, 1971), and not as sharply bent medially as in more derived neosuchians like “thoracosaurs” or species of Alligator. At the end of the shaft, the humeral head begins to gently curve dorsally. Medially on the head, a prominent internal tuberosity is present. The dorsal surface of the humeral head is broad, bearing a wide but shallow depression near its proximalmost edge. A fairly prominent angle delimits the dorsal surface from the lateral surface and the deltopectoral crest (fig. 25). Lateral to this angle, the attachment surface for the m. teres major is marked by a thin raised surface.

Figure 25

Left humerus and proximal left radius of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in dorsal view. See appendix 5 for abbreviations.

The distal articular surfaces are similar in size. The medial, ulnar hemicondyle, is slightly smaller than the capitulum (fig. 26). It is narrow and squarish, separated from the more rounded capitellum by a moderately deep trochlea. The ulnar hemicondyle extends distally farther than the capitellum forming a slight slant to the overall articular surface. Like other crocodyliforms, the articular surface of the capitellum curves ventrally, developing a short process that extends proximally. In dorsal view, the distal end is generally triangular in outline and prominent supracondylar ridges are absent. Laterally, an extensor ridge runs from near the dorsal edge ventrally to the terminus of the capitellum. Medially, a prominent flexor ridge is present on the ulnar hemicondyle near its ventral edge. This forms a triangular raised surface. In ventral view, the flexor ridge is expressed as a flat, lateral continuation of the ventral surface onto the edge of the hemicondyle.

Figure 26

Distal half of left humerus of referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in all views. See appendix 5 for abbreviations.

The left ulna is preserved (fig. 28). Only the lateral and part of the anterior surfaces are exposed. In general shape it is typical of crocodyliforms, expanded and triangular proximally while narrow and reduced distally. The ulna is bowed anteroposteriorly as in crocodylians, but not as pronounced as in the mesoeucrocodylian Pachycheilosuchus trinquei (Rogers, 2003). In lateral view, the proximal end bears a distinct process contacting the radius. The medially direct surface of this process may represent the dorsolateralmost extent of the m. pronator quadratus (Brochu, 1992; Meers, 2003), but this interpretation remains equivocal. Distal to this, the shaft begins to narrow gradually, narrowing more rapidly after midshaft and ending in an anteroposteriorly flat surface. Just proximal to the midpoint of the shaft, a ridge traverses the length of the ulna, distally becoming indistinct from the flattened end. In crocodylians, a similar ridge marks the division between the m. flexor ulnaris attachment posteriorly from the m. extensor carpi radialis brevis attachment anteriorly (Meers, 2003).

A complete left radius is preserved articulated to the ulna and humerus, while the right radius is represented only by the proximalmost portion (fig. 27). The radius is relatively slender. The proximal end expands abruptly from the shaft and is a long rectangle in cross-section. The distal end is also expanded, however this expansion is more gradual. Additionally, the expansion of the distal end is unequal as the medial portion is broader and thicker, while a thin lingular process comprises the lateral portion. The ventral articular surface is weakly concave. The radius shaft is narrow and the medial edge strongly concave while laterally it is nearly straight edged. Only anterior and lateral aspects of the left radius are visible, revealing little in the way of muscle scarring. The preserved portion of the right radius is better preserved and bears a protuberance medially corresponding the insertion of the m. humeroradialis (Fürbringer, 1876; Meers, 2003) (fig. 28).

Figure 27

Close-up of proximal portion of left radius of referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in lateral view. See appendix 5 for abbreviations.

Figure 28

Right distal half of humerus, right radius, and right ulna of referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in medial (left) and lateral (right) views. See appendix 5 for abbreviations.

The Pelvis and Hindlimb

The right ischium is largely complete (fig. 29). It is fractured in two places and lacks the anterior margin of the blade. The anterior iliac process is robust like other neosuchians (e.g., Theriosuchus sp. IVPP V10613 [Wu et al., 1996b], Sunosuchus sp. [Averianov, 2000], Sunosuchus junggarensis [Wu et al., 1996a], Alligator sinensis FMNH 197946). The process is nearly cylindrical, expanding slightly as it extends proximally. Although badly damaged and incomplete, it is clear that the blade of the ischium extended posteriorly and very slightly curved medially. The preserved portion of the anterior margin is more straight and thin than the posterior margin. Proximally below the posterior iliac process, the posterior margin of the ischiac blade is weakly concave, straightening distally on the blade. Little can be said regarding the lateral surface of the ischial blade as it is poorly preserved. What is discernible is a small, shallow longitudinal depression running ipsilateral to the anterior margin.

Figure 29

Partial right ischium of referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in lateral (left) and medial (right). See appendix 5 for abbreviations.

A partial, highly fragmentary element was recovered near the caudal vertebrae, ischium, and left tibia. It is interpreted here as the proximal shaft and distalmost end of the left femur. The bone was recovered in three pieces and is heavily effaced. A small, slightly rugose protuberance on the medial surface of the proximal portion likely corresponds to the fourth trochanter. The distal end of the femur consists of two large condyles separated by a shallow, triangular popliteal space (fig. 30). On the lateral hemicondyle a faintly developed crista tibiofibularis delimits a shallow groove.

Figure 30

Distal end of left femur of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in all views.

The left and right tibiae are well preserved with little distortion beyond cracking (fig. 31). The proximal ends are broken at approximately the same location on each one—just before the proximal shaft begins to expand. The right proximal end is missing, the left preserved. The proximal end of the tibia is composed of two depressions for reception of the distal femoral condyles. The buttresses for these depressions are unequally developed; the medial is larger and more posteriorly directed than the lateral. A triangular depression divides the two buttresses along the posteromedial surface of the tibia. Viewed anteromedially, the proximal end of the tibia is flat sided. A rugose tuberosity is present on this surface. Brochu (1992) interpreted a topographically equivalent structure in Alligator mississippiensis as the attachment for the internal lateral ligament. Laterally, just around the anterior face of the tibia, another protuberance is present and likely served as the insertion of m. tibialis anterior.

Figure 31

Left tibia of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in A, anterior; B, lateral; C, posterior; D, medial; E, distal; and F, proximal views. See appendix 5 for abbreviations.

The cross-section of the shaft is triangular both proximally and distally, becoming more circular throughout the diaphysis. The margin of the shaft is straight anteriorly and curves along the posterior margin. The medial face of the shaft is flat. Distal to the break surface, an angle distinguishes this surface from the more rounded posterior and lateral surfaces. This angle marks the origin of m. flexor digitorum longus (Brochu, 1992). At a similar level, but on the anterolateral surface, the tibia bears a low, centimeter-long ridge that is weakly fluted on the lateral aspect of the ridge. Topographically, this corresponds to the insertion of m. flexor tibialis internus (interior sensu Romer [1923] and Brochu [1992]).

The tibia widely flares distally. The posteromedial surface remains generally flat, though it becomes weakly depressed prior the lateral margin (fig. 31D). The anterolateral surface expands greatly into a round head when viewed laterally or distally. Normally, a depression for the attachment of the medial tibioastragalar ligament would be located on this surface (Brochu, 1992). However, its presence cannot be confirmed because the distal extent of this rounded surface is damaged. The distal tibia is divided into two contact surfaces for the tarsus. The larger, more developed facet contacts the astragalus. A small surface on the anteromedial aspect of the distal tibia articulates with the calcaneum. The astragalus contact extends distally, considerably farther than the contact with the calcaneum. Additionally, the facet for contact with the astragalus is far more developed distally than is the same contact surface in Crocodylia or the neosuchian Pachycheilosuchus trinquei (Rogers, 2003). The facet, in fact, terminates in an acute process similar to that of Araripesuchus gomesii AMNH FR 22450, Araripesuchus tsangatsangana (Turner, 2006), or Mahajangasuchus insignus UA 8654 (Buckley and Brochu, 1999).

The right fibula is complete but very heavily cracked and shattered. It was recovered loosely articulated to the right tibia. Nevertheless, some morphology is present and it can be determined that it bears a prominent proximodistal oriented ridgelike iliofibularis trochanter that slants anteroposteriorly and is located roughly one quarter of the way down the fibular shaft. As in Sunosuchus junggarensis (Wu et al., 1996a) and Goniopholis sp. AMNH FR 620, the trochanter is bordered proximally by a broad, but shallow depression leading into the head of the fibula. The anterior border of the depression bulges medially, marking the origin of m. flexor digitorum longus (Brochu, 1992). Distally, an additional ridge, separated from the trochanter by a small hiatus, extends down the shaft of the fibula. The fibularis longus muscle originates from this longer and more developed ridge (Reese, 1915; Brochu, 1992). The cross-section of the shaft is triangular distally. In axial view, the distal surface of the fibula is rectangular. The medial portion of the distal articular surface curves upward forming the “distal hook,” which contacts the lateral edge of the tibia.

The lateral surface of the left fibula is also exposed (fig. 32). It suggests that the fibular shaft was not completely straight (e.g., Alligator mississippiensis AMNH 1106CA, Araripesuchus gomesii AMNH FR 22450), but was bowed posterior slightly (e.g., Goniopholis sp. AMNH FR 620, Pachycheilosuchus trinquei [Rogers, 2003]). Immediately distal to the iliofibularis trochanter, an oval-shaped scute is present adjacent to the fibular shaft. Opposite to the left fibula in the block, the left tibia was removed. A small portion of the calcaneal contact facet of the left tibia remains in the block. In contact with this portion of the tibia, and lying less than a half of centimeter from the distal articular surface of the left fibula, a bone fragment is present that is interpreted here as the left calcaneum. This bone is poorly preserved and little can be said of its morphology.

Figure 32

Left fibula and accompanying appendicular osteoderms of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195. See appendix 5 for abbreviations.

The pes of IGM 100/1195 is only partially preserved. A single left metatarsal and phalanx were recovered near the left fibula (fig. 32). Also, three metatarsals and three phalanges are preserved with the right hindlimb elements. The metatarsals of this pes are interpreted as metatarsals III (specimen C15), IV (specimen C16), and V (specimen C11), respectively (fig. 33, 34). This identification is based on the short, hooked shape of the lateralmost metatarsal. This metatarsal is roughly 1.5 cm long. Metatarsal V is shorter in other neosuchians such as Alligatorium meyeri (Wellnhofer, 1971; pl. 11) and Alligator mississippiensis (AMNH 18707, AMNH 18697) (fig. 34). In this regard, Shamosuchus djadochtaensis bears closer resemblance to basal crocodyliforms (e.g., Protosuchus richardsoni AMNH FR 3024). The proximal articular surface is weakly convex, with a shallow midpoint depression. The shaft of the metatarsal is narrow distally, expanding and curving medial as one moves proximally. The medial surface bears a triangular contact surface for distal tarsal IV near the posterior border.

Figure 33

Metatarsal III (A) and metatarsal IV (B) of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in anterior and posterior views.

Figure 34

Partial proximal ends of metatarsal IV and V of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195 in lateral and posterior views.

The metatarsals are slender and long—slightly more than half the length of the tibia. The proximal articular surface is fan shaped. Laterally, the surface is nearly as thick as the main shaft of the metatarsal, while medially the proximal surface thins into a lamina underlying the proximoposterior surface of the preceding metatarsal. The cross-section of the metatarsal shafts is nearly circular. The distal ends are trochleated. Anteriorly, an extensor fossa is present above the trochlea. The preserved phalanges appear to pertain to digit three. The proximal articular surface is concave for the reception of the preceding pedal element. The distal articular surface is trochleated with the medial and lateral surfaces bearing ligament fossae.

Osteoderms

Numerous osteoderms displaying a wide range in morphology are present in partial articulation with the postcranial remains of IGM 100/1195. In the cervical region, several large osteoderms are present along the lateral margins of the specimen (fig. 21). The right side is best preserved, retaining at least six osteoderms laterally. The anteriormost two are adjacent to the axis and articulated to one another, likely representing the first two lateral cervical osteoderms. The posteriormost preserved osteoderms are located adjacent to the second and third dorsal vertebrae. These osteoderms are articulated via a cranially sloping imbrication. A discrete convexity is present along the anterior margin as in Bernissartia fagesii IRScNB n° R 46 (Buffetaut, 1975: fig. 4; Norell and Clark, 1990) and nonbrevirostran eusuchians (Brochu, 1999). The dorsal surface of these osteoderms is ornamented with a series of very small and shallow pits, and has a well-developed medial keel near their posterior margin. The dorsal ornamentation continues onto the surface of the keel. The anterior osteoderms, lying adjacent to the cervical vertebrae, bear the largest keels and are strongly curved to form a concave medial surface. The lateromedial curvature of the osteoderms lessens caudally until the level of the first dorsal vertebrae where curvature is extremely slight to nearly flat, and the medial keels become less developed. It is unclear whether a paired row of dorsal osteoderms were present between the lateral rows of osteoderms. However, a longitudinally running discontinuity zone is present near the dorsal margin of the lateral osteoderms. This zone may mark a suture with what might be interpreted as fragmentary dorsal osteoderms. Given that no unambiguous osteoderms in this location preserving a second distinctive morphology were recovered, it is unclear whether additional osteoderms were present.

Posteriorly, two osteoderm morphotypes are present along the right side of the caudal vertebrae as well as in two blocks corresponding to the left flank (fig. 35). This last set is located dorsal to the humerus but appears unassociated with it. The first type is rectangular in shape, being anteroposteriorly longer than it is wide and weakly bowed mediolaterally (fig. 35B). The second type is also rectangular, but these osteoderms are wider than they are long—the more pervasive condition among crocodyliforms (fig. 35A, C). These, like the other posterior osteoderms, are weakly bowed mediolaterally. Both posterior morphologies have the same ornamentation pattern as the anterior ones, although the medial keels diverge slightly to one side of the osteoderm. Laterally near the preserved left dorsal ribs, two osteoderms are preserved side by side articulated to one another. One of these two osteoderms displays the first posterior morphotype discussed above. The second one is too poorly preserved to determine its overall shape. Additionally, it is unclear whether these osteoderms are preserved in their life position.

Figure 35

Individual and systematic variation in the dorsal osteoderms of Shamosuchus djadochtaensis. Paramedian or near paramedian osteoderms (A and C) and more laterally located osteoderms (B). Note the posteriorly restricted location of median ridge (character 274.0).

Several ventral osteoderms are preserved between the two humeri, patches of which remain articulated (fig. 36). Additionally, fragments of ventral osteoderms are present along the posterior end of the cervical vertebrae and the proximal dorsal vertebrae. The ventral dermal armor is clearly distinct from the dorsal armor. The osteoderms are generally smaller and more square shaped in outline, except in some more laterally placed contacts where the osteoderms can take on less regular shapes. The ornamentation is similar to the dorsal elements in that it consists of a series of small, closely clustered pits. No keels are present on the preserved ventral elements. The anteriormost region lacks the smooth imbrication area present in the dorsal elements. This morphology is consistent with the fact that the ventral osteoderms, where articulated, are sutured to each other with no evidence of imbrication.

Figure 36

Ventral osteoderms of the referred specimen of Shamosuchus djadochtaensis IGM 100/1195. See appendix 5 for abbreviations.

The cervical osteoderms of Shamosuchus djadochtaensis are peculiar among crocodyliforms. They are most similar to the basal crocodyliform Zaraasuchus shepardi (Pol and Norell, 2004b), due to the presence of an extremely large lateral keel located along the posterior margin of each osteoderm. This similarity is largely superficial, however, as Shamosuchus djadochtaensis lacks the radiating anterior ridges and shallow grooved ornamentation present in the Zaraasuchus shepardi. Additionally, the cervical dermal armor of Shamosuchus is distinguished from Zaraasuchus shepardi and other crocodyliforms by being wider than long and more strongly flexed medially. Sunosuchus junggarensis also has dorsolateral neck osteoderms that bear a single medial keel (Wu et al., 1996a). These osteoderms, however, are largely dissimilar to Shamosuchus djadochtaensis. The osteoderms are small and leaf shaped, and the medial keel runs the entire length of the osteoderm in Sunosuchus junggarensis (Wu et al., 1996a). In this respect, the lateral dermal neck armor of Sunosuchus junggarensis is more similar to Alligator sinensis than Shamosuchus djadochtaensis.

Shamosuchus djadochtaensis shares with Bernissartia fagesii and nonbrevirostran euschians the derived presence of a discrete convexity on the anterior margin of each osteoderm, in addition to the lack of a peglike anterolateral process. The two osteoderm morphologies present in the trunk and caudal dermal series is inconclusive as to the number of rows of scutes that were present in Shamosuchus djadochtaensis. Eusuchians and some derived neosuchians closely related to this clade have more than two rows of dorsal osteoderms, flanked by accessory rows of osteoderms (Salisbury et al., 2006). Most other crocodylomorphs possess a single paired row (Clark, 1994), although some notosuchians are known to have more than two rows—e.g., Malawisuchus mwakayasyungutiensis (Gomani, 1997) and Simosuchus clarki UA 8679.

The dorsal armor of Shamosuchus djadochtaensis can be distinguished from most other crocodyliforms by its possession of a set of osteoderms that are longer than they are wide, although this condition also occurs in some basal crocodylians (e.g., Stangerochampsa maccabei; Wu et al., 1996c). The dorsal armor of Shamosuchus djadochtaensis is distinguished from all other mesoeucrocodylians by the presence of keels restricted to the posterior edge of each osteoderm.

In addition to the axial osteoderms, a number of small osteoderms were found adjacent to the radial and ulnar, tibial and fibular, and femoral shafts. Those associated with the radius and ulna are only partly preserved but appear subrectangular in outline. The osteoderms have the small-sized alveolar ornamentation present in the dorsal osteoderms. However, no keel is present on any of the preserved forelimb osteoderms. Only one femoral osteoderm is preserved and is partial and poorly exposed. Five osteoderms were recovered along the shafts of the left tibia and fibula. Opposed to the forelimb osteoderms, these hindlimb elements are narrow and elongated axially. The ornamentation consists of small pits, and at least one of them possesses a low longitudinal ridge near the distal edge of the osteoderm. The nature of the articulation in these elements is unclear, given that no appendicular osteoderm is preserved overlapping or even contacting another osteoderm. The phylogenetic significance of appendicular osteoderms is currently unclear because the distribution of this character within Crocodyliformes is poorly understood. They have been reported in widely disparate taxa with crocodyliformes, including the goniopholid Sunosuchus junggarensis (Wu et al., 1996a), alligatorids (Cong et al., 1998), and the basal crocodyliforms Gobiosuchus kielanae and Zaraasuchus shepardi (Osmólska et al., 1997, and Pol and Norell, 2004b, respectively).

Shamosuchus, Bernissartia, and the Glen Rose Form

The new information provided by IGM 100/1195 allows postulating a novel phylogenetic arrangement on the interrelationships of Shamosuchus djadochtaensis and two other advanced neosuchians: the Glen Rose form and Bernissartia fagesii.

Taxon Sampling and the Phylogeny of Advanced Neosuchians

The cladistic analysis presented here helps to clarify the phylogenetic position of Shamosuchus djadochtaensis as a closer relative to Eusuchia than to other advanced neosuchians, such as Bernissartia fagesii or the Glen Rose form. This analysis, however, is far from complete in terms of the taxonomic sampling. A large number of advanced neosuchians and basal eusuchians from Cretaceous beds of different regions of the world could not be included in this analysis for several reasons. Many of them are fragmentary forms, known for several decades. Others are recently discovered taxa, and though relatively complete and well preserved, they have been only briefly described or are still unpublished. In this section we briefly review the record of these forms according to the geographic regions where they had been found, pointing toward future directions that will help to understand more thoroughly the phylogeny of advanced neosuchians and the origin of Eusuchia. Furthermore, some of these taxa are preliminarily included in the phylogenetic analysis in order to conduct an exploratory test on their relationships as well as to assess the robustness of the topology presented above for advanced neosuchians. The results are discussed for each of these taxa and a summary is provided at the final remarks of this section.

Australia

Until recently, the evidence of advanced neosuchians from the Southern Hemisphere was restricted to fragmentary forms of uncertain affinities, leading some authors to treat the origin of Eusuchia as if it had taken place in the Northern Hemisphere (Sill, 1968; see also Salisbury et al., 2006). However, the Cretaceous record of Gondwana has recently offered important new information on advanced neosuchians.

Among these, one of the most recently described and probably one of the most relevant taxa for understanding the evolutionary origins of Eusuchia is Isisfordia duncani from the Cretaceous of Australia (Salisbury et al., 2006). Several specimens of this taxon, some of which are remarkably complete, have been found in the Winton Formation (late Albian–early Cenomanian). This taxon bears a unique combination of characters that suggest close affinities with basal eusuchians. In fact, Salisbury et al. (2006) interpreted this form as the most basal member of Eusuchia (see below). Among the derived features of Isisfordia duncani are the presence of weakly procoelous vertebrae in both the presacral and the caudal series. In contrast to some noneusuchians with procoelous vertebrae (e.g., Theriosuchus, Pachycheilosuchus), the vertebral condyle of Isisfordia duncani lacks a central depression on the articular facets (Salisbury et al., 2006). The morphology of the dermal armor also show a derived condition as it is composed by four parasagittal rows of dorsal osteoderms (two on each side of the trunk), in contrast to the two parasagittal rows representing the plesiomorphic neosuchian condition. Moreover, the central region of the paravertebral shield is flanked by a longitudinal row of accessory osteoderms, a derived feature of advanced neosuchians (e.g., Bernissartia fagesii, Susisuchus anatoceps) and Eusuchia. The skull of Isisfordia duncani also bears derived characters that suggest a close relationship with Eusuchia. The most remarkable one is the presence of a choanal opening completely enclosed by the pterygoids (Salisbury et al., 2006), one of the long-standing diagnostic characters of Eusuchia (Huxley, 1875). This opening, however, is not as posteriorly displaced as in the Hylaeochampsa vectiana and more derived forms, but is located close to the caudal margin of the suborbital fenestra (Salisbury et al., 2006), as in some noneusuchian neosuchians (i.e., Bernissartia fagesii and the Glen Rose form). Thus, Isisfordia duncani shows a combination of plesiomorphic and apomorphic characters previously unrecorded among advanced neosuchians (although see comments on the Las Hoyas neosuchian below).

Given the unique combination of characters present in Isisfordia duncani and its putative close relationships with Eusuchia (Salisbury et al., 2006), we have scored this taxon in our dataset using the published information in order to preliminarily test whether its inclusion has any impact on the phylogenetic hypothesis presented here (see appendix 4). Since we have not examined this material and its description is relatively brief (Salisbury et al., 2006), the scoring of this form was conducted using a conservative approach that assigned missing entries for those characters not explicitly described in the text or the data matrix published by Salisbury et al. (2006). Admittedly, many of the missing entries using this approach are likely to be scorable after a firsthand study of the specimen, but the risk of introducing spurious information into this exploratory analysis is reduced.

An exploratory phylogenetic analysis including Isisfordia duncani retrieves 60 most parsimonious trees (1044 steps) that show the same basic topology as the analysis presented above (excluding Isisfordia duncani). The only topological difference among these 60 trees regarding the relationships of advanced neosuchians consists of two alternative positions for Isisfordia duncani. Both of these corroborate the hypothesis of the close relationship of Isisfordia duncani to eusuchians recently proposed by Salisbury et al. (2006). Half of the most parsimonious trees depict Isisfordia duncani as the sister group of Hylaeochampsa vectiana and more derived eusuchians, identical to the analysis of Salisbury et al. (2006). The other half of the optimal trees, however, depict an alternative position for this taxon: as the sister group of the Asian clade (fig. 38). This alternative position is also compatible with the results of Salisbury et al. (2006), as they did not include the Asian forms in their study.

Figure 38

Reduced strict consensus of the 60 most parsimonious trees obtained in the exploratory analysis including Isisfordia duncani (based on the information provided by Salisbury et al., 2006), showing only the relationship among neosuchians. The two alternative positions retrieved for Isisfordia duncani are indicated with grey lines.

The former position is supported by two unambiguous synapomorphies. The first is the condition of the choanal opening of Isisfordia duncani: completely bounded by the pterygoids (character 43.2). As mentioned above, this condition is exclusively shared by Isisfordia duncani and eusuchians. The second synapomorphy is the absence of a shallow depression at the anteromedial corner of the supratemporal fossa (character 265.1), a condition shared by Isisfordia duncani, Hylaeochampsa vectiana, and crocodylians. Other advanced neosuchians (e.g., Shamosuchus djadochtaensis, the Glen Rose form, Goniopholis simus, Eutretauranosuchus delfsi, Theriosuchus pusillus) have a distinct rounded depression within the supratemporal fossa (Brochu, 1999). The absence of such fossa, however, seems to have a rather complex evolutionary history as it is optimized as convergently lost in some neosuchian taxa, such as Calsoyasuchus valliceps and the Sarcosuchus + Terminonaris clade. Furthermore, this condition is also present in some basal mesoeucrocodylians (e.g., Araripesuchus, Libycosuchus) and basal crocodyliforms (e.g., protosuchids).

In those trees in which Isisfordia duncani is depicted as the sister taxon of the Asian clade there are also two unambiguous synapomorphies supporting this arrangement. First, Isisfordia duncani and Shamosuchus djadochtaensis are the only advanced neosuchians that lack a notch at the ventral edge of the premaxilla-maxilla suture (character 9.0). As mentioned in the previous section the condition in Rugosuchus nonganensis is uncertain. Determining the condition of this character for this taxon is a critical issue, given that within the context of this dataset, if Rugosuchus nonganensis actually has a premaxillary-maxillary notch, Isisfordia duncani would be the sister group of Hylaeochampsa vectiana and crocodylians. The second synapomorphic feature is the presence of a well-developed sagittal ridge on the dorsal surface of the frontal (character 22.1). In contrast to the previous character, the frontal ridge is present in Isisfordia duncani, Shamosuchus djadochtaensis, and Rugosuchus nonganensis but is absent in other advanced neosuchians (including eusuchians). It must be noted, however, that other taxa that are not closely related also have a sagittal ridge on the dorsal surface of the frontal (e.g., Theriosuchus pusillus, some notosuchians).

In sum, the exploratory analysis shows that including the recently described Isisfordia duncani does not alter the pattern of relationships proposed in this study for advanced neosuchians. Furthermore, it underscores the relevance of the Australian taxon for understanding the early evolution of Eusuchia given that: (1) it undoubtedly represents a close relative of Hylaeochampsa vectiana and more derived forms and (2) its inclusion is critical for the character optimization of several important characters (e.g., choanal morphology). However, more information on this taxon (and others such as Rugosuchus nonganensis) is needed to resolve the uncertainties obtained in this exploratory analysis.

South America

The most interesting and informative advanced neosuchian described from the Early Cretaceous of South America is Susisuchus anatoceps (Salisbury et al., 2003). This taxon is known from a single specimen found in the Nova Olinda member of the Crato Formation (Aptian) of northeastern Brazil. The holotype consists of a skull and a partly articulated postcranium preserved in a limestone slab and only exposed in dorsal view, precluding the determination of several important characters (e.g., palate morphology). However, the available skull and postcranial information suggests this taxon represents another advanced neosuchian from the Southern Hemisphere. As noted by Salisbury et al. (2003, 2006), the dorsal dermal armor of Susisuchus anatoceps bears several derived features absent in other neosuchians, except for eusuchians and their closest relatives (e.g., Isisfordia duncani, Rugosuchus nonganensis). First, the dorsal dermal armor is composed of four longitudinal rows of parasagittal osteoderms (i.e., tetraserial paravertebral shield sensu Salisbury et al., 2006) flanked by two rows of accessory osteoderms. Second, the dorsal paravertebral osteoderms are square, rather than rectangular (lateromedially broader than anteroposteriorly long) as in more basal neosuchians. Third, the anterolateral process of these osteoderms is straight, lacking the distinct convexity present in some advanced neosuchians (e.g., Bernissartia fagesii) or the well-developed articular peg of other neosuchians (e.g., goniopholids, Theriosuchus pusillus). The skull of Susisuchus anatoceps also shows several neosuchian characters, such as the absence of an antorbital fenestra and quadrate condyles located at the level of the occipital condyle (Salisbury et al., 2003).

Susisuchus anatoceps, however, seems to be more distantly related to eusuchians than Isisfordia duncani because of the presence of some plesiomorphic characters (Salisbury et al., 2006). Most notably, the presacral vertebrae of Susisuchus anatoceps are amphicoelous (Salisbury et al., 2003), instead of the derived procoelous cervical and dorsal vertebrae of Isisfordia duncani, Shamosuchus djadochtaensis, and eusuchians. Susisuchus anatoceps also has a sagittal anterior projection of the frontals that separates the caudal end of the nasals (character 165.1), as in Bernissartia fagesii, goniopholids, and longirostrine crocodyliforms. In contrast, Shamosuchus djadochtaensis, Isisfordia duncani, and basal eusuchians (including Hylaeochampsa vectiana and the recently described Iharkutosuchus makadii; Ösi et al., 2007) have a blunt anterior end of the frontal that does not project between the nasals (character 165.0). More derived forms (e.g., crocodylians), however, also have an anterior process of the frontal wedging between the nasals (resembling the condition of Susisuchus anatoceps and basal neosuchians). This combination of plesiomorphic and apomorphic characters is reflected in the phylogenetic analysis of Salisbury et al. (2006), in which the Brazilian taxon was retrieved as more derived than Bernissartia fagesii but more basal than Isisfordia duncani. A preliminary analysis on the phylogenetic position of Susisuchus anatoceps in our dataset was conducted using a similar coding strategy as the one described above for Isisfordia duncani. In all of the most parsimonious trees Susisuchus anatoceps is more derived than Bernissartia fagesii (and the Glen Rose form) but more basal than Isisfordia duncani. In fact, Susisuchus anatoceps is placed as the sister taxon of a group composed by the Asian clade, Isisfordia duncani, and eusuchians (fig. 39). This arrangement is topologically compatible with the results of the phylogenetic study of Salisbury et al. (2006), given that Shamosuchus djadochtaensis and Rugosuchus nonganensis were not included in their study. The inclusion of Susisuchus anatoceps in this exploratory analysis does not affect the proposed relationships of advanced neosuchians presented previously, although it is decisive for determining the position of Isisfordia duncani as the sister taxon to the Asian clade (see Final Remarks at the end of this section). As interpreted by Salisbury et al. (2006), Susisuchus anatoceps indeed represents a previously unknown stage in the evolutionary history of advanced neosuchians, showing a lineage with a more plesiomorphic condition than Isisfordia duncani and the Asian clade but more advanced than Bernissartia fagesii.

Figure 39

Reduced strict consensus of the 39 most parsimonious trees obtained in the exploratory analysis with the addition of five taxa (highlighted in grey) that have been preliminary scored to test their positions as well as the robustness of the relationships proposed in figure 37. This analysis includes Isisfordia duncani, Susisuchus anatoceps, Pachycheilosuchus trinquei, Glichristosuchus palatinus, and Allodaposuchus precedens, in addition to the previously included taxa. Two of the alternative positions retrieved for Glichristosuchus palatinus are indicated with a dashed grey line. Other optimal positions depict this taxon nested within Crocodylia.

Another putatively relevant taxon from South America is Dolichochampsa minima (Gasparini and Buffetaut, 1980) from the Yacoraite Formation (Late Cretaceous), northwestern Argentina. This taxon is much more fragmentary than the two other forms mentioned above, and consequently its affinities are much harder to test. The type specimen consists of a small but extremely narrow and elongate dentary (MLP 73-II-28-16) with well-spaced everted alveoli, resembling of those of longirostrine crocodyliforms (e.g., dyrosaurids, gavialoids). Several isolated crocodyliform elements (cranial and postcranial) found at the same locality (or formation) have been referred to this taxon by Gasparini and Buffetaut (1980). The postcranial materials include procoelous vertebrae, a feature that prompted Gasparini and Buffetaut (1980) to include Dolichochampsa minima within Eusuchia. Unfortunately, the available remains do not provide much information for establishing the phylogenetic relationships of Dolichochampsa minima. The presence of procoely is now known to occur not only in Eusuchia, but also in closely related forms (e.g., Shamosuchus djadochtaensis, Isisfordia duncani) and other neosuchians (e.g., atoposaurids). Therefore, the affinities of this form must be considered dubious pending the discovery of more complete remains.

Africa

The Cretaceous record of continental Africa has provided remains of three taxa with valuable information concerning the diversity and distribution of advanced neosuchians. The first of these is Brillanceausuchus babouriensis (Michard et al., 1990) from the Lower Cretaceous beds of Babouri-Figuli Basin of northern Cameroon. Although some authors have regarded this taxon as closely related to atoposaurids (Michard et al., 1990; Rogers, 2003; Salisbury et al., 2006), its remains display not only procoelous vertebrae (as Theriosuchus) but also lack the anterolateral articular peg on the parasagittal row of osteoderms (Salisbury and Frey, 2001) and have been interpreted as having a pterygoid bounded choanal opening (Michard et al., 1990; Brochu, 1999; Buscalioni et al., 2001). Despite the simultaneous presence of these derived characters (the first two of which have traditionally been considered eusuchian synapomorphies), this taxon has never been thoroughly described or included in extensive phylogenetic analyses. Further study of this material is needed to understand its anatomy and thoroughly test its phylogenetic affinities.

The second taxon from the Cretaceous of continental Africa is Stomatosuchus inermis (Stromer, 1925) from the Bahariya Formation (Cenomanian) of Egypt. This large animal represents one of the most bizarre crocodyliforms, and is characterized by an elongate, flat, and parallel-sided rostrum and a possibly edentulous lower jaw. Most authors have considered this form as related to Eusuchia based on the presence of procoelous vertebrae and a choanal opening enclosed by the pterygoids (Stromer, 1925, 1933; Steel, 1973). The only phylogenetic analysis that has included this form supported the eusuchian affinities of Stomatosuchus inermis (Benton and Clark, 1988), depicting it as a noncrocodylian eusuchian (forming a polytomy with Hylaeochampsa vectiana and Borealosuchus). Unfortunately, only the original descriptions remain, as the only known specimen was destroyed during World War II. New material is needed to further understand the anatomy and relationships of this enigmatic taxon, although available information suggests affinities with advanced neosuchians (or basal eusuchians). This suggests that this “grade” was morphologically diverse and widely distributed in the Cretaceous of Gondwana. The third taxon is Aegyptosuchus payeri (Stromer, 1933), also found in the Bahariya Formation of Egypt. The only known specimen is fragmentary (a skull roof with part of the occipital table; BSP 1912.VIII.177) but has fortunately been preserved in the BSP collections. This is a poor specimen and it cannot be determined whether this taxon is closely related to eusuchians or not. Several characters of the skull roof, however, resemble the condition described for Stomatosuchus inermis. These include the extremely reduced supratemporal fossa located anteriorly on the skull roof leaving an anteroposteriorly extensive surface of the squamosal that occupies the posterior half of the skull table, orbital openings facing dorsally and separated by a narrow bar of the frontal, dorsal surface of skull roof with a slightly developed sculpture (rather than the well-developed subcircular pits of other neosuchians), and the elongate posterolateral process of the squamosal. This combination of characters is unusual for a neosuchian crocodyliform and suggests a close affinity of Aegyptosuchus payeri and Stomatosuchus inermis, in agreement with Romer's (1956) inclusion of both taxa in the family Stomatosuchidae. The fragmentary nature of this taxon and the current lack of material of Stomatosuchus inermis, however, preclude resolving their affinities with certainty. Further materials from the Bahariya Formation will undoubtedly help to understand these unusual crocodyliforms.

North America

The most relevant advanced neosuchian from North America is the Glen Rose form, which still needs to be described in detail, including a thorough evaluation of the postcranial material previously referred to this taxon (see above). The Cretaceous record of North America has provided other taxa that need to be considered to achieve a more complete picture on the evolution of advanced neosuchians. The first of these is Pachycheilosuchus trinquei described by Rogers (2003) from the Early Cretaceous Glen Rose Formation. This taxon is based on several hundred isolated crocodyliform elements collected from a single locality of the Glen Rose Formation in Texas (SMU 331; Rogers, 2003). Although these elements were not found in articulation, their referral to a single taxon seems a reasonable hypothesis based on anatomical and taphonomic bases (see Rogers, 2003). The craniomandibular remains distinguish Pachycheilosuchus from the Glen Rose form based on the presence of a more dorsoventrally compressed maxilla with an expanded and unfestooned buccal margin and a more gracile dentary (Rogers, 2003). Pachycheilosuchus trinquei is also a neosuchian crocodyliform but has been interpreted as closely related to atoposaurids, because of the presence of a jugal anterior ramus subequal in depth to posterior ramus and the presence of procoelous presacral vertebrae (Rogers, 2003). Furthermore, Pachycheilosuchus trinquei shares with Theriosuchus the presence of a relatively large central pit (dimple) on the condyle of most procoelous vertebra (Rogers, 2003; Clark, 1986). The phylogenetic position of this taxon, however, is weakly supported and Pachycheilosuchus trinquei shares some derived characters with neosuchians more advanced than Theriosuchus, such as the presence of procoelous dorsal vertebrae, a biconvex first caudal vertebra, and the absence of a well-developed articular peg in the anterolateral corner of each dorsal osteoderm. Despite the evident conflict among these characters, a preliminary scoring of Pachycheilosuchus trinquei in our dataset retrieves this taxon as the sister group of Atoposauridae (fig. 39), as originally proposed by Rogers (2003). As we have not examined the specimens, these scorings have been taken from the original publication with the same coding strategy described above for Isisfordia duncani and Susisuchus anatoceps (and therefore the results should be taken with similar caveats; see appendix 4 for these scorings).

Another relevant neosuchian from North America is Glichristosuchus palatinus (Wu and Brinkman, 1993) from the Milk River Formation (Santonian-Campanian) of Canada. Although this taxon is known from an incomplete small specimen (partial skull and cervical vertebra) it has been interpreted as an advanced neosuchian, purportedly more closely related to Eusuchia than Bernissartia fagesii (Wu and Brinkman, 1993). This interpretation is based on the presence of a unique combination of derived and plesiomorphic characters. The single cervical vertebra known for Glichristosuchus palatinus is procoelous, a feature known to occur (among neosuchians) in only two groups: eusuchians and related forms (e.g., Isisfordia duncani, Shamosuchus djadochtaensis) and Theriosuchus pusillus and related forms (i.e., Pachycheilosuchus trinquei; Clark, 1986; Salisbury and Frey, 2001; Rogers, 2003). Among the most advanced cranial features of Glichristosuchus palatinus is the postorbital-parietal suture extending slightly onto the skull roof, with the frontal almost reaching the anterior margin of the supratemporal fossae. This feature resembles the derived condition of some crocodylian clades (e.g., crocodylids, alligatorids) but differs from the morphology present in the most basal eusuchians and outgroups. The choanal opening of Glichristosuchus palatinus is unique among Crocodyliformes because of the simultaneous presence of two characters. First, it is located posterior to the caudal margin of the suborbital opening, an advanced feature present in eusuchians but not in other advanced neosuchians (including Isisfordia duncani, Shamosuchus djadochtaensis, and the Glen Rose form). Second, in contrast to the condition of eusuchians (and Isisfordia duncani) the pterygoids do not enclose the choana, but they form the entire lateral margin of this opening, resembling the condition in some advanced neosuchians (e.g., Shamosuchus djadochtaensis, Bernissartia fagesii [Buscalioni and Sanz, 1990], the Glen Rose form).

The simultaneous presence of these skull features in Glichristosuchus palatinus generates some conflict in the character-state distribution among advanced neosuchians. Therefore the proposed phylogenetic position of this taxon (Wu and Brinkman, 1993) needs to be tested within a comprehensive phylogenetic context. Although the available information on this taxon is limited because of the incompleteness of the type specimen (Wu and Brinkman, 1993) and we have not examined the specimens, a preliminary scoring of this taxon in our dataset was performed following the same coding strategy described above (see appendix 4). The exploratory analysis on the position of Glichristosuchus palatinus retrieves this taxon in multiple equally parsimonious positions (fig. 39). The most basal of them depicts this taxon as the sister group of Hylaeochampsa vectiana and more derived crocodyliforms (e.g., Eusuchia sensu Brochu, 1999), being more derived than any other noneusuchian neosuchian (including Isisfordia duncani). Other most parsimonious trees of this exploratory analysis invariably depict Glichristosuchus palatinus clustered with Crocodylia (either as its sister taxon or nested within this clade; see fig. 39). The position of Glichristosuchus palatinus allied with eusuchians is supported (in all most parsimonious trees) by the posterior location of the rostral margin of the choanal opening with respect to the caudal border of the suborbital opening (character 44.2).

The unstable behavior of Glichristosuchus palatinus in the phylogenetic analysis is undoubtedly influenced by the large proportion of missing data in the scorings of this taxon (approximately 87%). However, the alternative positions retrieved for this taxon are supported by conflicting character-state distributions. For instance, the participation of the palatine in the choanal opening (character 43) supports the exclusion of Glichristosuchus palatinus from the clade formed by Hylaeochampsa vectiana and more derived forms in some trees. Alternatively, Glichristosuchus palatinus is nested within Crocodylia in other trees because of the presence of a postorbital-parietal contact in the supratemporal fossa (character 23). The most basal eusuchians included in this analysis (e.g., Hylaeochampsa vectiana, Borealosuchus formidabilis) and their outgroups (e.g., Shamosuchus djadochtaensis, the Glen Rose form) have an extensive participation of the frontal in this opening, precluding the postorbital to contact the parietal. As has been noted for many other cases dealing with incompletely known taxa, character conflict produces alternative positions for an incomplete taxon and the lack of information (i.e., missing entries) precludes resolving such conflict (Kearney, 2002).

The derived position of Glichristosuchus palatinus in this exploratory analysis suggests this form may play a critical role in understanding the sequence of evolutionary transformations toward the derived choanal condition of Eusuchia. Salisbury et al. (2006) had rightfully suggested, based on their phylogenetic results, that the choanal morphology of Isisfordia duncani (enclosed by the pterygoids but located relatively anteriorly on the palate) represents an intermediate stage in the choanal evolution of eusuchians (pterygoid bounded and posteriorly located). The choanal condition of Glichristosuchus palatinus (posteriorly located but with a palatine participation) and its sister group relationship with Eusuchia (sensu Brochu, 1999) in some of the most parsimonious trees suggests a possible alternative scenario: a posteriorly positioned choana may have appeared before the complete enclosure of this opening by the pterygoids in the early evolution of the eusuchian lineage. Under such a scenario, the pterygoid-bounded choana of Isisfordia duncani must be interpreted as convergent with that of eusuchians. A detailed firsthand study of the available remains, a denser taxon sampling within Eusuchia, and more complete material of Glichristosuchus palatinus are needed to clarify the affinities of this taxon and to achieve a more complete understanding on the evolutionary origins of the eusuchian palate.

Europe

As discussed above, Hylaeochampsa vectiana and Bernissartia fagesii are the most critical taxa for understanding the origins of Eusuchia in most studies conducted during the last century. The Cretaceous record of Europe, however, has provided remains of other forms that should play an equally important role in future studies.

The first of these is Allodaposuchus precedens, originally known from relatively fragmentary material found in the Late Cretaceous beds of the Hateg Basin, Romania (Nopsca, 1928). A recent revision and redescription of the type material was conducted by Buscalioni et al. (2001), who also referred several more complete specimens to this taxon from the Late Cretaceous of France and Spain. The type material of Allodaposuchus precedens is currently restricted to a skull roof and occipital region (Buscalioni et al., 2001), whereas the rest of the original material from Romania consists of isolated postcranial remains that include several procoelous vertebrae. The material from Spain and France referred to Allodaposuchus precedens by Buscalioni et al. (2001) consists of five fragmentary skulls that provide a more complete understanding of the anatomy of this taxon, including critical information to test its phylogenetic relationships (e.g., pterygoid-bounded choana located posteriorly on the palate, interdigitating occlusion of dentary teeth). Although Salisbury et al. (2006) question the taxonomic validity of this taxon as defined by Buscalioni et al. (2001), these authors noted the presence of two autapomorphic characters in the type material and the referred specimens (i.e., posterior bottom of supratemporal fossa formed by the quadrate without contribution of squamosal or parietal and extensive prefrontal contribution of the orbital margin, with the frontal restricted to the caudomedial edge of orbit).

The phylogenetic position of Allodaposuchus precedens was tested by Buscalioni et al. (2001) through a cladistic analysis focusing on the relationships of eusuchians, expanding the dataset published by 9Brochu (1997, 1999). This study retrieved Allodaposuchus as one of the most basal eusuchians, specifically as the sister taxon of the Crocodylia crown group (i.e., more derived than Hylaeochampsa vectiana). A similar result has been obtained by Delfino et al. (2005) in a slightly modified version of the same dataset and has been accepted by Brochu (2003) in a recent review of crocodylian phylogeny. Salisbury et al. (2006) also included Allodaposuchus precedens in their phylogenetic analysis but obtained this taxon as the sister group of Hylaeochampsa vectiana, whereas Ösi et al. (2007) retrieved this taxon in a trichotomy together with Crocodylia and the clade formed by Hylaeochampsa vectiana and Iharkutosuchus makadii (see below). Irrespective of these minor disagreements, the combination of characters present in the Allodaposuchus precedens material described by Buscalioni et al. (2001) clearly suggests a basal position within Eusuchia for this taxon. The published information on Allodaposuchus precedens was used to score this taxon in our dataset (following the same coding strategy described above; see appendix 4), as we have not examined these specimens. The inclusion of Allodaposuchus precedens in an exploratory analysis along with the other preliminary scored taxa unequivocally places this taxon as more derived than Hylaeochampsa vectiana, being the sister group of Borealosuchus and crocodylians (fig. 39). This position is equivalent to the results obtained by some previous authors (Buscalioni et al., 2001; Delfino et al., 2005). In addition to the evidence cited by Buscalioni et al. (2001), the derived position of Allodaposuchus precedens as closer to Crocodylia than Hylaeochampsa vectiana is supported in our exploratory analysis by the presence of lateral margins of the frontal flush with the dorsal skull surface (character 266.0). Most advanced neosuchians (e.g., the Glen Rose form, Susisuchus anatoceps, Shamosuchus djadochtaensis) and Hylaeochampsa vectiana, in contrast, have elevated orbital ridges on the lateral margins of the frontal. Allodaposuchus precedens is however excluded from the clade formed by Borealosuchus and more derived eusuchians because of the short extension of the paroccipital process lateral to the cranioquadrate opening (character 268.0), thus resembling the short paroccipital process of Hylaeochampsa vectiana, Shamosuchus djadochtaensis, Isisfordia duncani, and the Glen Rose form.

Another basal eusuchian recently described from Europe is Iharkutosuchus makadii (Ösi et al., 2007) found in the Late Cretaceous Csehbánya Formation of Hungary (Ösi, 2004). This taxon is highly autapomorphic, including unusual features such as the complete closure of the supratemporal fenestra, elongate posterior process of the pterygoid, and complex multicusped posterior teeth with wear facets indicating buccolingual jaw movement during occlusion. Although complex teeth and extensive wear facets have been reported in some non-neosuchian crocodyliforms (e.g., Clark et al., 1989; Carvalho, 1994; Wu and Sues, 1996; Pol, 2003; Pol et al., 2004), the presence of such features was unknown within Neosuchia prior to the description of Iharkutosuchus makadii. Moreover, the multicusped teeth of this taxon differ markedly from other multicusped crocodyliform teeth (e.g., the posteriormost teeth are flat and bear radial rows of small cusps surrounding a row of three central cusps; Ösi et al., 2007). The phylogenetic position of this taxon has been tested by Ösi et al. (2007) expanding the dataset previously used by Buscalioni et al. (2001), which was itself based on Brochu (1997a, 1999). The results of this analysis depicted Iharkutosuchus makadii as the sister taxon of Hylaeochampsa vectiana, forming the clade Hylaeochampsidae. According to these authors several characters, including the presence of a linear frontoparietal suture, a prefrontal longer than the lacrimal, and a vertical ridge on the paroccipital process, support the monophyly of this group. We have tested the inclusion of this taxon in our dataset, performing a preliminary scoring based on the information published by Ösi et al. (2007). Unfortunately, the published description is brief and several key features cannot be scored and this taxon is depicted as unstable collapsing most nodes of advanced neosuchians and basal eusuchians (although in many of the MPTs it clusters with Hylaeochampsa vectiana as proposed by Ösi et al. [2007]). The uncertainty in these results will likely be solved when a more extensive description of this taxon becomes available.

Although the precise position of Iharkutosuchus makadii within Eusuchia should be further tested using more extensive taxon sampling, this taxon will probably become highly relevant in future studies on the relationships of advanced neosuchians and the origin of Eusuchia. First, the combination of characters described for this taxon clearly suggests strong affinities with basal eusuchians adding a new taxon known from relatively complete and abundant materials (see Ösi et al. 2007). Second, if its inclusion in Hylaeochampsidae is corroborated in future studies, the multiple skulls of Iharkutosuchus makadii reported by Ösi et al. (2007) would significantly increase our anatomical knowledge on this peculiar group, which has been historically pivotal for understanding the origins of eusuchians despite being known from a single fragmentary skull described by Owen (1874; the holotype of Hylaeochampsa vectiana). Third, the unusual craniodental features of Iharkutosuchus makadii broaden the known morphological and ecological diversity of advanced neosuchians.

Another neosuchian taxon from the Cretaceous of Europe that may play a critical role in future studies is the “Las Hoyas neosuchian,” found in the Calizas de la Huéguina Formation (Barremian) at the Las Hoyas locality in Spain (Ortega and Buscalioni, 1995). Although it remains undescribed, it has been discussed and included in a number of recent phylogenetic analyses. One of the most remarkable features is its derived osteoderm morphology and arrangement, including the presence of double-keeled osteoderms (Ortega and Buscalioni, 1995; Salisbury and Frey, 2001). Other skull characters suggest this form is closely related to advanced neosuchians and basal eusuchians. For instance, the choanal opening of this taxon has been scored as completely enclosed by the pterygoids but located relatively far from the posterior margin of the skull (Ortega et al., 2000; Buscalioni et al., 2001), resembling the choanal condition of Isisfordia duncani (Salisbury et al., 2006).

The phylogenetic position of this form has been tested in several phylogenetic analyses (Ortega et al., 2000; Buscalioni et al., 2001; Hua and Jouve, 2004; Company et al., 2005; Delfino et al., 2005). In most of these analyses the Las Hoyas neosuchian was retrieved as the sister taxon of Hylaeochampsa vectiana and more derived forms (i.e., Eusuchia sensu Brochu, 1999). Alternatively, the analysis of Buscalioni et al. (2001) depicted this taxon as the sister group of Hylaeochampsa vectiana. Despite this difference in results, a close relationship with the most basal nodes of Eusuchia has been consistently retrieved in previous studies, even in analyses based on a markedly different taxon and character-sampling schemes (e.g., Ortega et al., 2000; Company et al., 2005). Ever since its discovery the Las Hoyas neosuchian has been recognized as an important form (e.g., Ortega and Buscalioni, 1995) and a detailed description and revision of this taxon is needed. This information, added to that of the other advanced neosuchians commented above, needs to be gathered before a full understanding on the evolutionary origins of Eusuchia can be achieved.

The Cretaceous record of Europe has also provided remains of basal crocodylians that bear relevant information for understanding the origins of Eusuchia in future studies. One of these forms is Acynodon iberoccitanus, formerly known from an isolated maxilla (Buscalioni et al., 1997) and recently redescribed based on a series of six individuals (Martin, 2007). Acynodon iberoccitanus was originally referred to Alligatoridae (Buscalioni et al., 1997) and later considered a basal alligatoroid based on a phylogenetic analysis (Martin, 2007). As noted by this author, this taxon resembles in many features other short- and blunt-snouted basal globidontans (e.g., Stangeorchampsa, Brachychampsa, Albertochampsa), but it also bears several plesiomorphic characters that may be relevant for understanding the evolution of some features among basal eusuchians. For instance, in Acynodon iberoccitanus the external mandibular fenestra is absent, the frontal enters the supratemporal fossa, and the frontal has elevated orbital ridges. These conditions occur in basal eusuchians and advanced neosuchians (e.g., Hylaeochampsa vectiana, Iharkutosuchus makadii, Shamosuchus djadochtaensis, Rugosuchus nonganensis, Bernissartia fagesii, the Glen Rose form) but also are present in some basal gavialoids and/or alligatoroids (as well as some species of Borelaosuchus). Additionally, the absence of premaxilla-maxillary notch and the flared anterior half of the palatine bar is present in A. iberoccitanus (and alligatorids) and in basal eusuchians and some advanced neosuchians (e.g., Iharkutosuuchus makadii, Isisfordia duncani, Shamosuchus djadochtaensis).

The presence of these plesiomorphic characters in Acynodon iberoccitanus suggests it may play an important role in future studies of eusuchian origins. At the very least, this and other basal crocodylians are relevant because they can affect the character optimizations and influence evaluations of alternative hypotheses of relationships among basal eusuchians and advanced neosuchians. On the other hand, the numerous plesiomorphies and the evident character conflict in the distribution of these characters suggest the need of a more extensive evaluation on the relationships of this taxon and other basal forms of Crocodylia.

Asia

Rugosuchus nonganensis and Shamosuchus djadochtaensis are, as shown in this contribution, critical forms for understanding the relationships of advanced neosuchians and the origins of Eusuchia. Current knowledge on the cranial anatomy of these two forms is relatively complete, although more complete postcranial remains of Shamosuchus djadochtaensis are needed to define some relevant characters related to vertebral and osteoderm morphology. The postcranial anatomy of Rugosuchus nonganensis will also offer important information, as noted by Wu et al. (2001a: 1660). Other forms from the Cretaceous of Asia, however, are still less studied and often ignored in recent phylogenetic approaches to the relationships of advanced neosuchians, despite their importance in understanding the relationships of neosuchians. Among these, the most significant materials are the abundant remains referred to different species of Shamosuchus found in Late Cretaceous beds of Mongolia and Uzbekistan (Efimov, 1988). Storrs and Efimov (2000) noted that some of the 10 species described for Shamosuchus might be synonymous, although some of them differ markedly with respect to the remains of Shamosuchus djadochtaensis described here. In particular, Shamosuchus gradilifroms (Konzhukova, 1954) and Shamosuchus tersus (Efimov, 1983) are known from almost complete skulls (holotypes) that have a much longer and platyrostral snout compared to the specimen described here. Other remains can provide useful phylogenetic information because of the association of almost complete skulls with articulated postcranial remains, such as Shamosuchus ulgicus (Efimov, 1988; Storrs and Efimov, 2000), that shows a posteriorly placed choanal opening in combination with an osteoderm shield arranged in more than two parasagittal rows. A detailed revision of these materials is needed to achieve a solid alpha taxonomy. This will not only provide a more complete understanding of the diversity of advanced neosuchians from the Late Cretaceous of Central Asia, but will also form the basis for evaluating their evolutionary relationships and impact on the phylogeny of advanced neosuchians.

Final Remarks

The overview presented in this section aims to acknowledge a number of taxa that bear relevant information for assessing the relationships of advanced neosuchians and the origin of Eusuchia. Several points are made here. First, many advanced neosuchians with eusuchian affinities have been published during the last six years (e.g., Rugosuchus nonganensis, Susisuchus anatoceps, Pachycheilosuchus trinquei, Isisfordia duncani, Iharkutosuchus makadii). These taxa are represented by well-preserved material and, along with some forms yet to be formally described (e.g., Las Hoyas neosuchian, the Glen Rose form), increase our knowledge on the diversity of advanced neosuchians. Their information undoubtly helps to broaden previous discussions about the evolutionary origins of Eusuchia, which have traditionally been centered on the information provided by a few key taxa (e.g., Hylaeochampsa vectiana, Bernissartia fagesii, Theriosuchus pusillus). Second, adding the new records to previously known specimens, even if represented by fragmentary remains (e.g., Aegyptosuchus payeri, Dolichochampsa minima), results in a remarkably widespread distribution of advanced neosuchians during the Cretaceous. Except for Antarctica, advanced neosuchians have been found on all major continental landmasses. As previously noted by Salisbury et al. (2006) such a broad distribution of advanced neosuchians argues against the hypothesis of Laurasian origins of Eusuchia, suggesting that hypotheses of ancestral areas for this group should be taken with caution. Third, although most advanced neosuchians are relatively similar in terms of body size and general morphological features, there are cases of extensive morphological disparity. These include the recently described Iharkutosuchus makadii with multicusped teeth and presumably herbivorous diet (Ösi et al., 2007) and the gigantic Stomatosuchus inermis with a toothless mandible (Stromer, 1925). Such cases suggest advanced neosuchians (and early eusuchians) could have been among the most ecologically diverse groups of crocodyliformes (probably rivaled only by notosuchians in their morphological disparity).

In relation to the extended exploratory phylogenetic analysis presented in this section, several conclusions can be drawn. Five of the taxa listed above were preliminarily scored in our datasets: Isisfordia duncani, Susisuchus anatoceps, Glichristosuchus palatinus, Pachycheilosuchus trinquei, and Allodaposuchus precedens. The first objective of this analysis was to test to determine whether their inclusion affects the relative position of other advanced neosuchians. The results show that the basic topology presented in the previous section (fig. 37) is maintained in the exploratory analysis, depicting the Asian clade (Shamosuchus + Rugosuchus nonganensis) as closer to Eusuchia (Hylaeochampsa vectiana and more derived forms) than the clade composed by Bernissartia fagesii and the Glen Rose form. None of the five additional included taxa alter this basic result. This indicates that, within the context of this dataset, this topology is robust despite the relatively low values of branch support.

The second objective was to preliminarily test the phylogenetic relationships of these five neosuchian taxa. In general terms, the results of the exploratory analysis retrieved topologies that are highly congruent with previous hypotheses on the position of these five taxa (fig. 39). Pachycheilosuchus trinquei was retrieved as the sister group of atoposaurids, as interpreted by Rogers (2003). Susisuchus anatoceps and Isisfordia duncani are depicted as successively more closely related to Hylaeochampsa vectiana and eusuchians than Bernissartia fagesii, as interpreted by Salisbury et al. (2006). Glichristosuchus palatinus is similarly depicted as closer to eusuchians than Bernissartia fagesii, in agreement with the hypothesis presented by Wu and Brinkman (1993) and Allodaposuchus precedens is retrieved as closer to crocodylians than Hylaeochampsa vectiana as in Buscalioni et al. (2001). This consistency of results was not expected given that our dataset differs in terms of taxon and character sampling with respect to the various datasets used in previous studies of the relationships of these five taxa. However, it should be noted that the scorings of these five taxa in our dataset were based on the descriptions and datasets of previous authors, hence following their interpretations.

Within the context of our dataset, the phylogenetic placement of the five added neosuchians is relatively stable to variations in the taxon sampling scheme: they are retrieved in the same position irrespective of the inclusion or exclusion of the other added taxa. The only exception occurs for Susisuchus anatoceps and Isisfordia duncani, which vary their position when they are not simultaneously included. When Susisuchus anatoceps is not included in the analysis Isisfordia duncani is retrieved in two alternative positions (see fig. 38), whereas only one of these positions is obtained for the Australian taxon when Susisuchus anatoceps is included in the dataset (see fig. 39). Conversely, Susisuchus anatoceps is allied with Hylaeochampsa vectiana and crocodylians when Isisfordia duncani is excluded from the analysis but occupies a more basal position if the latter is included (fig. 39). Irrespective of these caveats, the results of the exploratory analysis provides novel phylogenetic information, such as the relative position of these five taxa and other neosuchian crocodyliforms, most of which had never been simultaneously included in a cladistic study. For instance, the position of Isisfordia duncani and Susisuchus anatoceps relative to the Asian clade or Glichristosuchus palatinus (see fig. 39) had not been tested in previous studies (e.g., Wu and Brinkman, 1993; Salisbury et al., 2006).

In addition to a more thorough study of the relationships of these advanced neosuchians and basal eusuchians, future analyses should aim to expand the taxon sampling to include other taxa that bear critical information for understanding the origins of Eusuchia. A critical point will be the addition of more basal representatives of the major crocodylians clades, as our analysis has a reduced sampling on these taxa in comparison with other studies (e.g., Brochu 1997a, 1999, 2004). As noted above, several conflictive character-state distributions are present among these forms (see discussion of Acynodon iberoccitanus), which need to be tested by character congruence in a phylogenetic analysis within an extensive character and taxon sampling regime. Progress here will help to clarify critical issues concerning the relationships of advanced neosuchians and to achieve a more complete picture on the evolution of important features among advanced neosuchians and basal eusuchians.

Shamosuchus and the Evolutionary Origins of Eusuchia

Eusuchia has likely been the most stable and widely recognized group of crocodyliforms since the original classification scheme provided by Huxley (1875). Unlike the other major groups of Crocodyliformes (e.g., Protosuchia, Mesosuchia, Metamesosuchia, Metasuchia), eusuchian monophyly has never been seriously questioned, either in traditional or cladistic analyses. As originally conceived, this group was diagnosed by the presence of two key characters: the procoelous vertebrae and the choanal opening completely enclosed by the pterygoids. Although the origins of this clade and its relationships with other crocodyliforms have never been thoroughly understood, a clear morphological gap existed between eusuchians and other fossil forms (i.e., “mesosuchians”). Two taxa from the Early Cretaceous of Europe have long been considered key for understanding the origin of Eusuchia, representing the condition at both ends of this gap: Bernissartia fagesii (Dollo, 1883) and Hylaeochampsa vectiana (Owen, 1874). As noted above most authors consider Bernissartia fagesii as an advanced neosuchian more closely related to Eusuchia than to other neosuchians (e.g., goniopholids, atoposaurids), whereas Hylaeochampsa vectiana has been considered as the earliest and most basal eusuchian (Steel, 1973; Buffetaut, 1975; Clark, 1986; Benton and Clark, 1988; Clark and Norell, 1992). This morphological gap between Eusuchia and other neosuchians has been partly filled through the discovery of several Cretaceous fossils during the last few decades (see above). The new specimen of Shamosuchus djadochtaensis provides novel anatomical information that further contributes to our knowledge of this gap.

Although we consider premature the postulation of a robust and comprehensive scenario for the evolutionary origins of Eusuchia given the incomplete taxon sampling of our analysis (see previous section), the new information and the phylogenetic analyses presented here highlight two relevant issues that deserve further comments: (1) choanal and procoely evolution and (2) ghost lineages among advanced neosuchians.

Choanal and Procoely Evolution

The new material of Shamosuchus djadochtaensis has a previously unknown combination of characters for a eusuchian close relative: the presence of derived procoelous presacral vertebrae in combination with two plesiomorphic characters—amphicoelous caudal vertebrae and a secondary palate anteriorly located and bounded by the palatines and pterygoids. This combination of characters, added to the phylogenetic analysis presented above, provides new insights into the evolution of the two characters that have traditionally characterized Eusuchia.

The evolution of the choana in Crocodyliformes had traditionally been postulated as one of the most remarkable examples of gradual change across the evolutionary history of a major group. This classic scenario portrays a gradual shift of the location of the choanal opening from an anterior position in basal forms to a posterior placement in modern crocodyliforms. Within this hypothesis the progressive shift in the position of the choana is reflected in the architecture of different palatal bones forming the anterior margin of this opening, which defined three evolutionary grades: the protosuchian condition (anterior margin bordered by maxillae), the mesosuchian condition (anterior margin formed by palatines), and the eusuchian condition (anterior margin formed by pterygoids).

Several authors have pointed to the presence of parallel trends toward a posteriorly located choana in addition to the one present in the lineage leading to extant crocodyliforms. Such cases have been noted to occur among basal crocodyliforms (e.g., Zosuchus; Pol and Norell, 2004a), basal mesoeucrocodylians (e.g., Iberosuchus; Ortega et al., 2000, and Mahajangasuchus; Turner and Buckley, 2008), and dyrosaurids (Jouve et al., 2006). Thus, the progressive nature of choanal evolution within the entire diversity of Crocodyliformes had already been questioned. The evolutionary pattern within advanced neosuchians was however, up to now, free of homoplasy. The new information provided by Shamosuchus djadochtaensis (and Rugosuchus nonganensis) combined with the derived position of the Asian clade as closer to Hylaeochampsa vectiana and more derived forms than Bernissartia fagesii and the Glen Rose form introduces some homoplasy in the optimization of choanal morphology among advanced neosuchians.

Early changes are inferred to occur in the choana of advanced neosuchians reflected in the relatively posterior position of this opening (i.e., at the posterior margin of the suborbital fenestra; character 44.1) and the pterygoids forming the entire lateral margin of the choana (character 43.1; see fig. 40). Examples of this primitive condition are present in Bernissartia fagesii (Buscalioni and Sanz, 1990) and the Glen Rose form. Although eusuchians extend this trend by having an even more posteriorly positioned choana completely enclosed by the pterygoids (fig. 40), the choanal morphology of the Asian clade implies a reversal to the plesiomorphic position of the anterior margin of the choanal opening present in most other mesoeucrocodylians (fig. 40). The degree of pterygoid participation from the lateral margin of the choana, however, does not show such a reversal in the Asian clade, maintaining its extensive participation from this edge of the opening. Thus, the choanal morphology of both Shamosuchus djadochtaensis and Rugosuchus nonganensis shows the phylogenetic history of the neosuchian choana was more complicated than ordinarily supposed, as anticipated by Langston (1973). The eusuchian choana (both posteriorly located and pterygoid bounded), however, still seems to be diagnostic of Hylaeochampsa and more derived neosuchians, although the pterygoid bounded choanal opening of Isisfordia duncani may have been acquired independently. The palatine participation in the choanal opening of Glichristosuchus palatinus and its derived position in the tree (see fig. 39) add additional homoplasy to the choanal evolution in Neosuchia (fig. 40).

Figure 40

Simplified cladogram of advanced neosuchians showing the decoupled nature of the evolution of choanal and vertebral condition previously considered diagnostic of Eusuchia (based on the extended exploratory phylogenetic analysis discussed in the text). Only six selected taxa are shown in this tree representing the different combinations of character states among advanced neosuchians. The most parsimonious optimization of presacral vertebral morphology is shown in the cladogram. Grey lines indicate the plesiomorphic amphicoelous vertebrae (denoted with an A) and black lines represent the derived procoelous condition (denoted with a P). The vertebrae of Hylaeochampsa are unknown but the procoelous condition is marked for this figure representing the morphology known in all other eusuchians. Schematic drawings for the choanal morphology of selected taxa are shown to the right of the cladogram, with arrows indicating the condition of the two choanal characters discussed in the text (characters 43 and 44; see appendix 1 for character-state definitions). Goniopholis is based on G. simus (BMNH 41098), Bernissartia modified from Buscalioni and Sanz (1990), Shamosuchus based on S. djadochtaensis (IGM 100/1195), Isisfordia modified from Salisbury et al. (2006), Glichristosuchus modified from Wu and Brinkman (1993), Hylaeochampsa modified from Clark and Norell (1992), with pterygoid process reconstructed based on Iharkutosuchus makadii (Ösi et al., 2007). Images not to scale.

The presence of procoelous vertebrae has also traditionally characterized Eusuchia, although this feature had been noted to recurrently appear in other crocodyliforms, such as atoposaurids (Buffetaut, 1975; Clark, 1986; Salisbury and Frey, 2001; Rogers, 2003), basal mesoeucrocodylians (Clark, 1985), and basal crocodylomorphs (Clark et al., 2004). The presence of this character among advanced neosuchians has been considered exclusive of Eusuchia, despite the debated presence of an incipient procoely in some vertebrae of Bernissartia fagesii (e.g., Buffetaut, 1975; Buscalioni and Sanz, 1990; Norell and Clark, 1990; Salisbury and Frey, 2001). The new information provided by Shamosuchus djadochtaensis suggests procoelous presacral vertebrae had appeared before the origin of Eusuchia in the evolution of advanced neosuchians (fig. 40). This feature may have an even earlier appearance than currently optimized if the debated procoelous condition seen in both Bernissartia fagesii and the Glen Rose form are confirmed in future studies. The simultaneous presence of procoelous presacral vertebrae and amphicoelous caudal vertebrae in Shamosuchus djadochtaensis also suggests that the evolutionary origin of procoely had progressively appeared craniocaudally, which is reflected in the different degrees of procoely recorded among extant crocodylians (Salisbury and Frey, 2001). The presence of such craniocaudal trend may not be exclusive of Eusuchia and its closest relative, as a differential degree of procoely has also been noted in Pachycheilosuchus trinquei (Rogers, 2003), which presumably has independently acquired the procoelous condition.

The new information and the phylogenetic results imply the traditional diagnostic choanal and vertebral conditions of Eusuchia appeared at different times during the evolutionary history of advanced neosuchians (fig. 40). The eusuchian-type palate is still a synapomorphy of this clade, but the procoelous presacral vertebrae diagnose a broader group. The decoupling of the evolutionary history of these two features is hardly unexpected, as has been proposed in early cladistic studies of this group (Benton and Clark, 1988; Norell and Clark, 1990; Wu and Brinkman, 1993) and has been predicted by previous authors (Lyddeker, 1887; Buffetaut, 1975). The decoupled evolution of these two conditions, however, contrasts with the results of the most recent review of the evolutionary history of Eusuchia (Salisbury et al., 2006), which postulated the appearance of both conditions at the same node in the tree.

This new interpretation on the evolution of the choanal and vertebral morphology has bearings on the three alternative definitions of Eusuchia that have been defended in recent years. The first of them restricted the name to those forms with a pterygoid bounded choanal opening (Benton and Clark, 1988; Clark and Norell, 1992), given that the Glen Rose form was then considered to have procoelous vertebrae. The second proposal was offered by Brochu (1999), who argued for a node-based definition of Eusuchia, as the clade that includes the last common ancestor of Hylaeochampsa vectiana and Crocodylia. The choice of Hylaeochampsa vectiana as the anchor or specifier taxon made by Brochu (1999) was based on the fact that this form has been considered the most primitive eusuchian by most authors (Owen, 1874; Huene, 1933; Clark and Norell, 1992). Brochu (1999) discussed the convenience of using a node-based definition given that apomorphy-based definitions can be problematic when further studies reveal homoplastic instances of the chosen character within the group of interest (Rowe and Gauthier, 1992; Bryant, 1994; Holtz, 1996). This position was echoed by Buscalioni et al. (2001), given the presence of conflicting characters and missing data throughout the evolution of neosuchian taxa that conform the stem group of Crocodylia, including the pterygoidean choana, procoely, and dorsal armor. More recently, Salisbury et al. (2006) obtained in their phylogenetic analysis both procoely and the pterygoid-bounded choana as synapomorphic features of the same node. This resulted, in part, because of their simultaneous presence in Isisfordia duncani (and because the procoely of the Glen Rose form was no longer accepted). This result led these authors to argue for an apomorphy-based definition based on the simultaneous presence of the two characters, based on the historical (Huxley, 1875) and biomechanical (Salisbury and Frey, 2001) significance attached to these features.

The disparity of the choanal and vertebral optimization in our phylogenetic analysis poses some problems to the apomorphy-based definitions. The decoupled evolution of both diagnostic features is a minor part of the problem and only applies to the definition proposed by Salisbury et al. (2006). The most important problem is the homoplasy involved in the posterior migration of the choanal opening and the participation of the pterygoids from the anterior margin of this opening. These homoplastic patterns coupled with the alternative positions of some taxa in our exploratory analyses suggest the optimization of these features is still unstable. For instance, the pterygoid-bounded choana of Isisfordia duncani is depicted as a convergence in some trees and the participation of the palatines from the anterior margin of the choana in Glichristosuchus palatinus requires a reversal to the plesiomorphic state in trees that depict this taxon as more derived than Hylaeochampsa vectiana. These issues are not likely to be completely solved in the near future given the fragmentary nature of some specimens and the presence of so much character conflict among advanced neosuchians. Therefore, we consider the use of an apomorphy-based definition of Eusuchia premature, as the relationships of its most basal members and outgroups are not sufficiently well established to produce relatively stable optimizations in the relevant characters.

Ghost Lineages among Advanced Neosuchians

In addition to the changes in the traditional diagnosis of Eusuchia, the Laurasian origin of this clade (Salisbury et al., 2006) has also been challenged. Historically, the best-known advanced neosuchians (atoposaurids, goniopholids, Bernissartia fagesii) and basal eusuchians (Hylaeochampsa vectiana) have been found in Early Cretaceous (or Late Jurassic) beds of Europe and North America. This distribution suggested a Laurasian diversification of advanced neosuchians and basal eusuchians (Sill, 1968). Salisbury et al. (2006) have recently argued that the biogeographic origins of Eusuchia are ambiguous, postulating either eastern Gondwana (Australia) or Laurasia (western Europe) as equally likely candidates for an ancestral area. They noted, however, that the close phylogenetic relationship of Susisuchus anatoceps and Isisfordia duncani with eusuchians might be cited as evidence supporting a Gondwanan placement for the transition from Neosuchia to Eusuchia.

The increased taxon sampling of our phylogenetic analysis and the review of the distribution and diversity of advanced neosuchians (see above) allow a critical revision of this issue. In particular, combining the geographic distribution of advanced neosuchians with their chronostratigraphic information provides interesting insight into this problem.

Advanced neosuchians first appear in the fossil record by the Aptian (approximately 112–125 mya; Early Cretaceous), although this date may extend back to the Barremian (approximately 125–130 mya) based on the age of Las Hoyas neosuchian as well as on the new palynological data of the Bernissart beds in which Bernissartia fagesii was found (Yans et al., 2005). By the end of the Early Cretaceous, advanced neosuchians are present in virtually all major landmasses. This widespread distribution coupled with their striking morphological disparity suggests a significant part of their evolution may have occurred well before their first appearance in the fossil record. A calibration of our phylogenetic tree with the chronostratigraphic information of the fossil taxa lends support to this interpretation (fig. 41). In particular, the recent discovery of the goniopholid Calsoyasuchus valliceps in the Early Jurassic Kayenta Formation (Tykoski et al., 2002) has important consequences for this issue. Although the ages of most previously known neosuchians are either Late Jurassic or Early Cretaceous (e.g., Theriosuchus, Goniopholis, Eutretauranosuchus), the early occurrence of Calsoyasuchus valliceps (Sinemurian-Pliensbachian; approximately 183–190 mya) places a minimum age of origin for the lineage leading to advanced neosuchians more than 50 million years before their first appearance in the fossil record. This yet unknown significant part of the early evolutionary history of advanced neosuchians may explain their morphological disparity and worldwide distribution by the Early Cretaceous (as some of their major lineages could have diversified before the effective isolation of some regions caused by the breakup of Pangea).

Figure 41

Phylogeny of Crocodyliformes calibrated against geological time. The solid grey line highlights the presence of an extensive ghost lineage that trace the origins of advanced neosuchians back to the Early Jurassic. This extension implies a long unknown period in their history that predates their first appearance by more than 50 million years. Such a remote origin, coupled with the worldwide distribution of the group can be indicative of an initial diversification prior to the effective separation of the major continental landmasses.

An Early Jurassic origin of the advanced neosuchian lineage, however, does not necessarily mean all their groups had diversified by that time. Therefore, the origin and diversification of Eusuchia may still have taken place in the Early Cretaceous, as traditionally thought. It is interesting to note that all the noncrocodylian eusuchians included in our phylogenetic analysis are forms from the northern hemisphere, particularly in the Early Cretaceous of Europe. Even in our extended phylogenetic analysis the added taxa that fall within Eusuchia are Laurasian forms (e.g., Glichristosuchus palatinus, Allodaposuchus precedens), as well as other taxa proposed as basal eusuchians (e.g., Iharkutosuchus makadii, Las Hoyas neosuchian). Thus, the analyzed data still seems to indicate the origin and diversification of Eusuchia was a strictly Laurasian event. Despite this result, we agree with Buscalioni et al. (2001) and Salisbury et al. (2006) in that the Laurasian origin of Eusuchia can only be seriously proposed after the phylogenetic relationships of several southern taxa are better understood (e.g., Stomatosuchus inermis, Brillanceosuchus babouriensis, Dolichochampsa minima). At any rate, even if Eusuchia turns out to be distributed worldwide, its diversity and abundance in Cretaceous beds of the Northern Hemisphere seem to be much higher than in Gondwana.