Chatterjee, 1991. Cranial anatomy and relationships of a new Triassic bird from Texas. Philosophical Transactions of the Royal Society of London Series B. 332(1265), 277-342.

This paper described the skull of Chatterjee's new supposed bird, Protoavis. It featured a small cladistic analysis of cranial and mandibular characters.


Chatterjee's phylogeny was


Taxon Issues

Sphenosuchids- Chatterjee primarily uses Dibothrosuchus, though he also mentions Sphenosuchus itself. Recent analyses (e.g. Clark et al., 2004) suggest Sphenosuchidae is paraphyletic, with Dibothrosuchus closer to Crocodyliformes than Sphenosuchus.
Theropods- This is of course paraphyletic to the ingroup. Chatterjee mentions several theropods in his comparison of Protoavis to that group- Megapnosaurus, Dilophosaurus, Allosaurus, Tyrannosaurus, Ornithomimidae, Troodon and Dromaeosaurus. These will be used for coding.
Protoavis- This taxon is now recognized as a chimaera, consisting of drepanosaurid, pterosauromorph and theropod (including coelophysoid) elements. The cranial material coded here is partially theropod (holotype basioccipital, prootic, exoccipital, epiotic, supraoccipital), while the other (holotype premaxilla, lacrimal(?), squamosal, quadrate(?), frontal, parietal, posterior mandible(?); paratype maxilla, nasal(?), lacrimal, jugal, quadratojugal, quadrate(?), parietal(?), basioccipital(?), vomer, palatine, pterygoid, partial dentary, posterior dentary(?), posterior mandible) may include drepanosaurid or misidentified elements.

Coding Issues

1. This character (metotic structure) deserves comment. Chatterjee identifies a "metotic ossification" in Protoavis which spans the area between the paroccipital process and basal tubera, separating the fenestra pseudorotunda from the vagus (X) nerve. This structure is seemingly part of the exoccipital-opsthotic (there is no suture) and causes the vagus nerve to exit on the occipital surface. This is the same as what's usually called the metotic strut in theropods, but Chatterjee codes theropods as lacking the character. The text makes it clear that he believes the theropod structure (which he terms a subcapsular process) is non-homologous with the avian structure because Rieppel (1985) found that the subcapsular process in crocodilians develops from the exoccipital while the metotic strut of birds arises from a separate condensation. Yet developmental processes also evolve, and the condition in theropod development is unknown in any case. Even worse, the developmental origin of the structure in Protoavis, Archaeopteryx or Hesperornis is also unknown, making Chatterjee's coding for them as having a metotic and not a subcapsular process an a priori assumption based on his "knowing" they are birds. As derived theropods (e.g. Dilophosaurus- Welles, 1984; Tyrannosaurus- Brochu, 2003; Troodon- Currie and Zhao, 1994; Dromaeosaurus- Currie, 1995) have a metotic strut identical to the condition in Protoavis and Archaeopteryx, theropods are coded as polymorphic. Avimimus (Kurzanov, 1986) and Ichthyornis (Clarke, 2004) also have metotic struts. The condition in Protoavis is illustrated by Chatterjee as having a metotic strut, but comparing it to "Megapnosaurus" kayentakatae suggests it's possible the supposed anterior opening of the posterior tympanic recess is the actual vagus nerve, while the supposed vagus foramen could be the larger of two hypoglossal (XII) foramina. Witmer (2001) had no problem with the metotic strut interpretation however. In any case, Protoavis is coded unknown pending further study.

2. This character is the unquantified "small olfactory lobes", which is coded as present in all bird taxa and absent in the outgroups. Comparing olfactory bulb width to cerebrum width does indeed suggest Archaeopteryx has narrower bulbs (27%) than Troodon (32%) or Dromaeosaurus (48%). However, Avimimus is described as having "relatively large" olfactory bulbs by Kurzanov (1985) and the skull seems broken posterior to where the bulbs were placed anyway, with the small triangular olfactory (I) nerve foramen visible in anterior view. Avimimus is thus recoded as unknown. Edinger (1951) indicates that Marsh's (1888) endocast of Hesperornis was largely imaginary and notes no data concerning olfactory bulb size can be determined from the specimens. Whether newer specimen KUVP 71012 can shed light on its morphology is unknown, as this area has not been described. Hesperornis is recoded unknown. While no ventral illustration of the frontals or dorsal/ventral illustration of Protoavis' endocast is shown, the drawing of the brain (figure 16d) by Chatterjee indicates its olfactory bulbs were ~42% as wide as the cerebra. So assuming the composite endocast is even trustworthy, Protoavis would seem to have large olfactory bulbs. It is here provisionally coded as unknown.

3. Character 3 is cerebellar fossa extended on to supraoccipital, but contra Chatterjee this is seen in even basal theropods (e.g. Ceratosaurus- Sanders and Smith, 2005; Tyrannosaurus- Osborn, 1913). The Ichthyornis material is far too crushed to determine this character's presence.

4. This character involves the presence of a cerebral vein ("sinus canal") penetrating the supraoccipital-epiotic suture. This is present in theropods as well ('caudal middle cerebral vein' of "Megapnosaurus" katentakatae in Tykoski, 1998; 'dorsal cerebral vein' of Tyrannosaurus in Brochu, 2003; 'posterior canal of the middle cerebral vein' of Dromaeosaurus in Currie, 1995).

5. Troodon lacks a prefrontal (Russell, 1969), making theropods polymorphic. The condition in Gobipteryx is uncertain, as no specimen preserves that area. The presence of this bone is uncertain in Protoavis, as the lacrimal is preserved attached to the maxilla, while the frontal is not just preserved separately, but comes from a different larger individual. Thus they cannot be articulated, and the frontal itself only seems to bear an anterior (nasal?) articular surface, with no apparent lateral articular surface for a lacrimal and/or prefrontal.

6. Chatterjee's next character is confusing. It is "braincase: (basioccipital-basisphenoid-parasphenoid complex): (0) vertical; (1) horizontal." Chatterjee elaborates by discussing the outgroups' braincase as deep, but birds' as flat. Yet the braincase of Archaeopteryx does not differ much from Troodon's in overall proportions. The ventral braincase of most small theropods (Megapnosaurus, Troodon, etc.) is basically horizontal, with an anteriorly directed parasphenoid cultriform process. Most theropods' ventral braincase is deeper than birds' due to a vertically deep basisphenoid sinus, and even taxa like Troodon which lack an external sinus have deep internal sinuses there. Unfortunately, sinus depth is difficult to measure in most specimens, though coelophysids have shallower sinuses than other theropods (Raath, 1989). Protoavis has a sinus on the ventral surface of its supposed basioccipital, though the depth of the element is never illustrated. It would resemble coelophysids if not for the supposed basisphenoid (Witmer, 2001 doubted its identity) from a different individual with its small basipterygoid processes. The reason for coding Archaeopteryx as derived is uncertain, as the Munich specimen is the only one which preserves this area and was not known until after Chatterjee's paper was published. In any case, its deep basipterygoid processes suggest a deep sinus if anything (Elzanoski and Wellnhofer, 1996). Avimimus and Ichthyornis are also unknown, as their articulated braincases do not allow ventral depth to be measured. Dibothrosuchus has a deep basisphenoid like theropods, though due to the ventrally projecting basipterygoid processes, as it lacks a ventral basisphenoid sinus. The question remains how to code this character. The condition in most theropods is non-homologous with sphenosuchids, Protoavis may resemble coelophysids, and Hesperornis has a flat morphology due to lacking promiment basipterygoid processes or a ventral sinus. It is here deleted (so all eight taxa are recoded as unknown), though basiphenoid sinus depth or basipterygoid process size might be good substitutes.

7. This character involves the external naris being posteriorly placed, and indeed Archaeopteryx, Gobipteryx and Hesperornis have more posteriorly placed nares than the included avetheropods (distance from snout tip to front of nares 20% or more of height of orbit+jugal). But Megapnosaurus and Dilophosaurus have nares placed more posteriorly than Archaeopteryx, making theropods polymorphic. The condition in Ichthyornis is unknown as the described premaxillary fragment does not indicate the nares' anterior edge. Protoavis actually has a rostrally placed naris based on the illustrated premaxilla and posterior skull of the large individual, though Chatterjee changed this for his cranial reconstruction to more closely resemble birds.

8. Archaeopteryx has a complete postorbital, which probably contacted the jugal (Tischlinger, 2005). The condition in Gobipteryx is unpreserved. Protoavis is supposed to lack a postorbital based on a lack of facets on the frontal and squamosal, which were confirmed by Witmer (2001). Yet the laterosphenoid has a large capitate process (termed the postorbital process by Chatterjee) with what appears to be an articular surface, which fits into a socket on the postorbital in most theropods. The jugal is described as having only a rudimentary dorsal process, though the original material is only illustrated in ventral view. Thus there is not a complete lack of postorbital articular surfaces, and the evidence is contradictory. Perhaps being such a small or juvenile individual, Protoavis did not have a strongly attached postorbital. For now, it is tentatively coded as unknown.

9. This character (maxillary tooth row ends anterior to antorbital fenestra) is coded as an autapomorphy of Archaeopteryx, making it useless for the analysis and artificially increasing the CI. However, Archaeopteryx actually lacks this character as shown by the Eichstatt specimen (Paul, 2002). Even worse, Protoavis and Gobipteryx are coded as lacking it, despite not having maxillary dentition at all! Similarly, even if the premaxillary denticles of Avimimus are interpreted as teeth (contra Vickers-Rich et al., 2002), the maxilla is unpreserved. Ichthyornis only preserves a small maxillary fragment of uncertain position (Clarke, 2004). All of these taxa should be coded as unknown. The condition in Hesperornis is difficult to code, as a medial antorbital wall which defines the antorbiral fenestra anteriorly in more basal theropods is generally thought to be absent (Witmer, 1990). Even if the palatine process is homologous, without it extending dorsally, there is no way to judge an anterior boundary to the antorbital fenestra. Hesperornis is thus recoded as unknown.

10. The fused quadratojugal and quadrate in Avimimus make judging the presence of a dorsal quadratojugal process impossible. Neither Gobipteryx or Ichthyornis preserve the quadratojugal, so are both recoded as unknown. The quadratojugal of Protoavis is posterodorsally curved in the material, unlike Chatterjee's reconstruction. This posterodorsal section would seem to be a valid dorsal process, so Protoavis is recoded.

11. This character (premaxillary teeth present but anterior dentary teeth absent) is another supposed autapomorphy, and is problematic for being a composite as well. It combines a primitive state for one element (premaxillary teeth present) with a derived one from another element (anterior dentary teeth absent). The latter is also present in some theropods (e.g. ornithomimids) and Gobipteryx, so those are changed to polymorphic and derived respectively. The condition in Protoavis is uncertain, as the one anterior dentary fragment seems broken at both ends, making its position unknown.

12. This character (temporal configuration fully modified) is correlated with character 8 (diapsid temporal configuration partially modified). Thus all taxa coded 0 for character 8 (sphenosuchids, theropods, Archaeopteryx) are coded unknown. Gobipteryx and Protoavis are again coded as unknown.

13. Chatterjee's 13th character is "alaparasphenoid covers anterior tympanic recess." The alaparasphenoid of birds is homologous to the structure variously called the crista prootica, otosphenoidal crest, ala basisphenoid and preotic pendent in theropods. In some theropods (e.g. Allosaurus- Chure, 2000; Tyrannosaurus- Brochu, 2003), this covers the anterior tympanic recess. Theropods are thus recoded as polymorphic. Archaeopteryx has only a small ridge, so is coded as plesiomorphic. While the supposed alaparasphenoid of Protoavis (on the possible basisphenoid) does not seem to have much projection posterodorsally to cover the anterior tympanic recess, the prootic fragment has a dorsal overhang (figure 12f) that is similar to that in Megapnosaurus.

14. Chatterjee codes the zygomatic process of the squamosal as lacking in theropods, but it is homologous to the quadratojugal process, which is present in almost all theropods. Theropods are recoded as having the process. Fusion makes its presence in Avimimus uncertain (Kurzanov, 1987), while Archaeopteryx has the process (Elzanowski and Wellnhofer, 1996). The squamosal of Gobipteryx is unknown. The squamosal of Protoavis is described as having a zygomatic process, but this is clearly the anterior process instead, as it is the one Chatterjee notes is peculiar in lacking a postorbital facet. Thus assuming it is complete, Protoavis actually lacks a significant ventral process.

15. Witmer (2001) notes the supposed quadratojugal cotyla of Protoavis' quadrate is subtle and "not all that different" from other theropods such as Deinonychus or juvenile tyrannosaurids. It is thus recoded as absent. The condition in Avimimus is uncertain, as the quadratojugal is fused to the quadrate. The area of contact is unpreserved in Gobipteryx's quadrate (Elzanowski, 1977).

16. This character (orbital process of quadrate present) involves the pterygoid's articulation with the quadrate more than the quadrate's structure itself, as sphenosuchians and theropods both have the homolog to an orbital process (the pterygoid process). Yet in derived birds, the pterygoid only articulates with the quadrate ventrally, leaving the former pterygoid process free. The condition in Protoavis is difficult to determine, as Witmer (2001) noted the supposed pterygoid condyle of the quadrate was not present. The pterygoid's quadrate ramus is described as a narrow bar though, which would indeed have trouble articulating in a broad overlapping suture typical of theropods. Protoavis' derived coding is tentatively retained.

17. This character (pterygoid condyle present on quadrate) is correlated with the previous one, as only a taxon with a free orbital process will have a pterygoid condyle. Thus sphenosuchids, theropods, Archaeopteryx and Avimimus are all recoded as unknown. As noted above, the pterygoid condyle on Protoavis' quadrate is not there according to Witmer (2001), so Protoavis is changed to plesiomorphic.

18. Gobipteryx (Chiappe et al., 2001) and Hesperornis (Witmer, 1990) have large dorsal maxillary processes, so are recoded. That of Protoavis was originally illustrated as broken off (although described as absent), but Chatterjee (1999) later described a new element as the maxilla, this time with a large dorsal process (illustrated in Chatterjee, 1998). Assuming this is really a maxilla, it does indeed have a large process. Protoavis is recoded.

19. This character involves cranial kinesis, which is especially difficult to demonstrate in fossils (Holliday and Witmer, 2008). Specifically, it involves prokinesis, which needs several components to function- mobile joints at the nasofrontal, quadrato-otic, quadrato-quadratojugal, maxillojugal and anterior palate, and loss of a connecting postorbital, epipterygoid and ectopterygoid. Based on their criteria, Gobipteryx probably lacked prokinesis due to its ectopterygoid so would need to be recoded. The condition in Protoavis is complicated by incompleteness and uncertain identification. The nasofrontal joint does indeed seem to be reduced, and like most theropods the quadrate has a synovial joint with the sqmousal. While perhaps not a true ball and socket joint, the quadrate-quadratojugal articulation is loose and convexoconcave. However, the maxilla appears to have a broad, forked articulation with the jugal. The evidence it lacked a postorbital is ambiguous (see above), and the absence of small elements such as epipterygoids and ectopterygoids even more so. Note this character is correlated with characters 8, 10, 12, 15, 16 and 20, so is deleted and all seven taxa recoded unknown.

20. The ectopterygoid is present in Gobipteryx (Elzanowski, 1995), while its absence in Protoavis cannot be demonstrated given the disarticulation of the material.

21. Clarke (2004) notes both Hesperornis and Ichthyornis have a bicondylar quadrate, with the supposed caudal condyle being a posterior continuation of a single medial condylar surface. The tricondylar quadrate of Protoavis is supposedly illustrated in figure 11 of Chatterjee's 1991 paper, but that figure is actually missing. Still, the paratype quadrate is illustrated in his 1998 paper as being tricondylar, so is provisionally still coded that way. Witmer (2001) notes the holotype's quadrate seems to lack mandibular condyles at all.

22. Like character 11, this is both a useless autapomorphy, and also a composite character. Supposedly an apomorphy of Protoavis, it is "teeth restricted to tips of premaxilla and dentary." The two illustrated premaxillary alveoli are in the middle of the bone, while Chatterjee describes a total of three alveoli which are "in front." Chatterjee claims there were only two anteriorly placed dentary teeth, but the only section of dentary preserved which would be expected to have teeth is from an unknown position and is associated with three disarticulated teeth (one supposedly a replacement tooth). Assuming it is even a dentary and the teeth belong to that bone, there is no way of knowing how they were distributed. The potentially toothless posterior portion of the premaxilla is matched by some theropods such as coelophysoids, while Avimimus, Gobipteryx, Hesperornis and Ichthyornis lack premaxillary teeth entirely. Thus Protoavis is tentatively coded as lacking posterior premaxillary teeth, theropods are coded as polymorphic, and Gobipteryx, Hesperornis and Ichthyornis are coded as apomorphic (Avimimus' posterior premaxillae are unpreserved).

23. The nasals do not contact on the median in Ichthyornis (Clarke, 2004).

24. This character (nasal process of premaxillae extend posteriorly to lacrimals) is directly correlated with the previous one (nasals do not contact on median), as the premaxillae are what separates the nasals from one another. It is therefore deleted and the eight coded taxa recoded as unknown.

25. The mesethmoid is ossified in many theropods (e.g. Tyrannosaurus- Hurum and Sabath, 2003; Troodon- Ali et al., 2008; Dromaeosaurus- Currie, 1995), so Theropoda is recoded as polymorphic. Avimimus also has this element (Kurzanov, 1981), as does Ichthyornis (Clarke, 2004). Its presence in Archaeopteryx and Protoavis is uncertain due to the lack of articulated material in the latter and easy separation from the laterosphenoids and frontals.

26. This is another useless autapomorphy, and yet again the taxon in question (Hesperornis) does not exhibit the feature (pterygoid process fits into basisphenoid facet), as Hesperornis has low but convex basipterygoid processes (Elzanowski, 1991). The condition in Avimimus is uncertain due to fusion (Kurzanov, 1987), while the condition in Ichthyornis is uncertain as the basipterygoid area is unpreserved (Clarke, 2004).

27. Another autapomorphy that Hesperornis does not in fact possess, and a composite character to boot- "combination of short, complex pterygoid with narrow, elongate palatine." Chatterjee misidentifies the palatal elements of Hesperornis (Chatterjee, 1991), which only appears to have a short pterygoid because of the unfused anterior portion (hemipterygoid of Elzanowski). The real palatine of Hesperornis is indeed narrow and elongate, but those of Avimimus, Gobipteryx and Ichthyornis are unpreserved. The palate of Protoavis is very difficult to intrepret, but the supposed choana makes more sense as a suborbital fenestra given its position behind the maxillojugal suture. Any coding regarding the shape of its palatines should be changed to unknown.

28. This character (pterygoid condyle of quadrate large) is correlated with character 17 (pterygoid condyle present), so any taxa scored as 0 for 17 (sphenosuchids, theropods, Protoavis, Avimimus, Archaeopteryx) should be coded unknown for this one. However, the character is still a useless autapomorphy, being found only in Ichthyornis.

29. This character (toothless jaws) is correlated with parts of 11 (dentary teeth absent) and 22 (posterior premaxillary teeth, maxillary teeth, and posterior dentary teeth absent). It is intended as an apomorphy of Gobipteryx, making it useless. However, as Gobipteryx differs from most included taxa by lacking maxillary teeth specifically, that portion of the character is retained for this analysis. Protoavis apparently also lacks maxillary teeth, so is recoded. Some theropods (e.g. ornithomimids) lack them as well, and Avimimus does not preserve the maxilla.

30. The final character is another useless apomorphy of Gobipteryx- dentary symphysis fused. Protoavis cannot be scored for it, as the anteriormost dentary portion is unpreserved.

General analysis conclusions- This is a highly flawed analysis. It only uses characters from one area of the body, which makes it less reliable. Also, it only uses characters which are known in Protoavis, a trend which is common when authors are trying to phylogenetically place their taxon under study. Yet characters unpreserved in Protoavis could nonetheless strongly support the topology of the tree, which will in turn affect character distribution and thus Protoavis' placement. The characters are all designed to support certain clades, so this is another cladistic demonstration instead of a cladistic analysis. Having an all-zero outgroup is generally flawed, and in this case theropods should be scored as 0/1 or for 11 (37%) of the characters, though sphenosuchids do indeed seem to be coded correctly as all zero. Nine of the characters (30%!) are designed to be autapomorphies, which do not influence the analysis at all, making the already small matrix even smaller. 6 (20%) of the characters are correlated with other characters. Characters 11, 22 and 27 are especially poorly formed, as they involve the morphology of more than one element each. Several times Chatterjee assumes non-homology between avian and theropod features (metotic strut, alaparasphenoid, zygomatic process) and codes as if this assumption were necessarily true. Taxa are commonly coded for elements they do not preserve (especially Ichthyornis and Gobipteryx) or for characters which are inapplicable (tooth characters in toothless taxa, articulation characters in taxa with fusion). In total, 110/240 (46%) of the characters are miscoded. When recoded, the new result is-

   |--other theropods
|*-Gobipteryx |--Protoavis `--Ornithuromorpha |--Hesperornis `--Ichthyornis

The resolution is decreased, which isn't surprising since there are only 11 parsimony-informative characters left.

Phylogenetic conclusions- The table shows the number of extra steps needed to accomodate each rearrangement using Chatterjee's original matrix, and his recoded matrix. A negative number means the arrangement is already most parsimonious, but that many steps are needed to undo it.

rearrangement original recoded
(sphenosuchids,theropods,Avimimus(Archaeopteryx,Hesperornis)) 1 1
(sphenosuchids,theropods,Protoavis(Archaeopteryx,Hesperornis)) 9 1
(sphenosuchids,theropods((Archaeopteryx,Gobipteryx)(Hesperornis,Ichthyornis))) 10 3

Neither the original or recoded analysis provides much support at all for placing Avimimus closer to modern birds than Archaeopteryx. The original analysis did provide moderate support for placing Protoavis in this position, but once recoded its position relative to Archaeopteryx is basically ambiguous. Sauriurae was rejected with moderate confidence originally, but recoding has made this support weak. Basically, Chatterjee's matrix is too small to securely reject any plausible hypothesis.

Experiments with controversial taxa- It's somewhat pointless to add more taxa to a matrix with such little power to resolve relationships. One question might be if resolution is improved by deleting the chimaerical Protoavis. Another is if more basal oviraptorosaur such as Incisivosaurus would cause Avimimus to be excluded from birds. If Protoavis includes some drepanosaurid elements, adding Megalancosaurus to the matrix might be informative. Finally, it is odd that Chatterjee excluded modern Aves. Deleting Protoavis has no effect on tree topology or resolution. Megalancosaurus, Incisivosaurus and Aves were all added as OTUs. The resulting tree is-

   |--other theropods

Avimimus still falls out as a bird, though Gobipteryx has a more stable position more derived than Protoavis. The fact Megalancosaurus falls out by Protoavis within Theropoda is yet more indication this matrix is insufficient to test phylogenies. Megalancosaurus is excluded from Theropoda with 1 more step.