Movatterモバイル変換

Variation in Size

There is a considerable range in the size of individuals represented by specimens referred toIchthyornis, but the different sizes do not occur with the same frequency across the YPM material. Only two specimens (i.e., YPM 1738 and YPM 1765) are significantly smaller than theIchthyornis dispar holotype (YPM 1450). And only four other specimens (i.e., YPM 1460, YPM 1462, YPM 1766, and YPM 9148) are approximately the same size as or just slightly smaller than theIchthyornis dispar holotype. That is, of 77 specimens referred toIchthyornis, only two specimens, or 2.6%, are notably smaller than theIchthyornis dispar holotype, four specimens, or 5.2%, are the same size, and 92.2% are larger.

Those YPM specimens slightly larger than theIchthyornis dispar holotype include, for example, YPM specimens 1730, 1733, 1749, 1756, and 1764. However, the majority (85.7%) of the YPM material referred toIchthyornis is significantly larger than theIchthyornis dispar holotype. Most of these specimens were previously referred to “Ichthyornis victor” and include, for example, YPM specimens 1447, 1452, 1457, 1461, 1720, 1721, 1722, 1725, 1729, 1737, 1741, 1742, 1747, 1748, 1750, 1755, 1757, 1762, 1763, 1773, and 1775.

The data presented intable 3 as well asfigures 5–9 are not suggestive of the presence of distinct size classes, but rather a near continuum of differently sized individuals. A more detailed morphometric analysis would rigorously explore whether there are distinct size classes to be discriminated.

The width and depth of the distal humerus were compared for the only four undistorted YPM specimens (fig. 5). The smallest undistorted distal humerus (YPM 1738) measured (table 3) is one of the two YPM specimens significantly smaller than theIchthyornis dispar holotype (YPM 1450). The undistorted humerus of YPM 1764 is slightly larger than theIchthyornis dispar holotype; it plots between the holotype ofIchthyornis dispar (YPM 1450) and the largest undistorted specimen, YPM 9685 (table 3;fig. 5). YPM 9685 is approximately the same size as the majority of preserved YPM specimens. Based on comparison of the distal humeral dimensions of the two smallest undistorted humeri to the largest (YPM 9685), one of the two smallest specimens in the YPM material (YPM 1738) is 63–65% of the largest depending on what measurement is compared (table 3). By contrast, theIchthyornis dispar holotype (YPM 1450) is 76–82% of the largest specimen compared (YPM 9685, or YPM 1742 in the case of total humeral length;table 3).

Placing the size of specimens in the context of time allows the investigation of anagenetic change as an explanation for some or all of the size differences among the material. Specimens referred toIchthyornis have been described from the late Cenomanian through the early Campanian, an interval of more than 10 million years. The specimens referred from the late Cenomanian (Tokaryk et al., 1997) are all significantly larger than theIchthyornis dispar holotype. However, the three next oldest (Turonian) specimens referred toIchthyornis, one from the early Turonian of Alberta (Fox, 1984) and Kansas (Martin and Stewart, 1982), as well as from the late Turonian of Texas (Lucas and Sullivan, 1982), are all notably smaller than or approximately the same size (figs. 6,7) as the holotype ofIchthyornis dispar (YPM 1450). The oldest specimen is the smallest of the three (figs. 6,7). Two slightly later specimens (TMM 31051–24, TMM 31051–25) from the early Coniacian of Texas (Parris and Echols, 1992;table 3) are slightly larger than YPM 1450. All of this material is smaller than the majority of the YPM material, for which more precise age estimates are generally not possible.

Unfortunately, there is almost no detailed stratigraphic information for the YPMIchthyornis material collected from the Smoky Hill Chalk Member. Only a few YPM specimens referred toIchthyornis have locality information more detailed than, for example, “Wallace County, Kansas” (YPM VP Catalogue). The entire Smoky Hill Chalk Member of the Niobrara Formation is upper Coniacian–lower Campanian in age (Stewart et al., 1990), and specimens ofIchthyornis have been estimated to be present in all recognized intervals of the Chalk (Stewart, 1990). However,Bennett (1990) concluded that, historically, nearly all of the fossil vertebrates, including the YPM material, were collected in the upper part of Smoky Hill Chalk Member from between Marker Unit 15 and 20 ofHattin (1982).Stewart (1990) estimated the interval containingHattin's (1982) Marker Units 8–10 to be upper Santonian in age. Marker Units 15–20 must therefore be, at the earliest, from later in the upper Santonian through the early Campanian. Thus, the majority of the YPM material would appear to be from the late Santonian–early Campanian.

By contrast, the holotype ofIchthyornis dispar (YPM 1450) appears to be early Santonian in age:Bardack (1965) specified the locality of the holotype within Rooks County (seeIchthyornis dispar below), and this locality was mapped in theCladoceramus undulatoplicatus byStewart (1988). The “Zone ofCladoceramus undulatoplicatus” was given as early Santonian in age byStewart (1990: 22). The top of this zone was dated more precisely by⁴⁰Ar/³⁹Ar as 84.88 ± 0.28 Ma, which is in the late Santonian (Obradovich, 1993). Thus, more than 80% of the YPMIchthyornis material is significantly larger, and inferred as younger, than the holotype ofIchthyornis dispar.

Figures 5–7 summarize the data discussed above. Shading infigure 6 illustrates an explanation of the size-through-time data as anagenetic change, andfigure 7 illustrates this data as differential sampling through time of a consistent size range. Measurements of the six complete humeri in the YPM collection were taken (table 3). These measurements, taken from the material itself, differ from those given inMarsh (1880) andLucas and Sullivan (1982). Other published measurements of complete humeri referred toIchthyornis from the other localities mentioned above were also included (i.e.,Fox, 1984;Parris and Echols, 1992). Further, qualitative estimates of the size of several additional specimens referred toIchthyornis that do not include complete humeri but have comparatively good constraint on their age were also included. For example, a proximal carpometacarpus (SMM 2139) referred toIchthyornis (although it preserves no apomorphies) from the early Turonian of Kansas (Martin and Stewart, 1982) was included. It is approximately the size of theIchthyornis dispar holotype (YPM 1450; Clarke, personal obs.). The comparative sizes of the holotype ofIchthyornis antecessor from the Campanian of Alabama (Olson, 1975) and two specimens referred to that taxon from the Campanian of Texas (Parris and Echols, 1992) plotted infigures 6 and7 were based on comparisons of published measurements for the distal humeri that comprise these specimens to comparable measurements of complete YPM humeri.

There is also slightly more precise locality data for one of the four specimens the size of theIchthyornis dispar holotype. This specimen, YPM 1460, was collected at Twin Butte Creek, a well-known locality representative of the upper part of the Smoky Hill Chalk Member (Stewart et al., 1990).Bennett (1990) measured seven stratigraphic sections along Twin Butte Creek and identified the interval represented to be between Marker Unit 15 and 19 ofHattin (1982); this is the same interval as the majority of vertebrates from the Chalk (Bennett, 1990) and, presumably, the majority of the YPMIchthyornis material. Indeed, YPM 1209 is approximately the size of 80% of the YPM material and also was also collected at Twin Butte Creek (e.g.,Marsh, 1880).

Stewart (1990: 22) reportedIchthyornis “c.f.Ichthyornis anceps” and a “smaller species” as present in his late Coniacian “Spinaptychus n. sp.” zone. This reference is cryptic, as the holotype of “Ichthyornis anceps” is a specimen from a large individual comparable in size to most of the YPM material. However, the one specimen referred to “Ichthyornis anceps” is from an individual midsize between theIchthyornis dispar holotype and the size of most of the YPM material. All that can be gleaned from this information is that individuals of different sizes are present at this earlier time as well (i.e., lower in the Smoky Hill Chalk Member). However, measurements of these individuals were not reported inStewart (1990).

In sum, there appears to be no evidence of individuals as large as more than 80% of the YPMIchthyornis material from older deposits from the Smoky Hill Chalk Member (figs. 6,7). There is some evidence that smaller individuals (i.e., the size of theIchthyornis dispar holotype), in low abundance, are present in the upper part of the Smoky Hill Chalk Member where most of the material is large. The only known Cenomanian specimens (Tokaryk et al., 1997) are significantly larger than theIchthyornis dispar holotype, while all of the Turonian specimens (Lucas and Sullivan, 1982;Fox, 1984; Clarke, personal obs.) are smaller than or the same size as the early SantonianIchthyornis dispar holotype.

Variation in Morphology

YPM specimens showing morphological differences from the holotype ofIchthyornis dispar were divided into two kinds: (1) difference in specimens otherwise supported by other evidence as part ofIchthyornis and (2) difference in specimens not supported by other evidence as part ofIchthyornis. For the four YPM specimens not supported as part ofIchthyornis (table 2), noted differences were used to identify specimens for inclusion as separate terminal taxa in the phylogenetic analyses (Part II). A description of these four specimens (listed intable 2) and evaluation of their taxonomic status is provided below in the Taxonomic Revision.

Subtle variation that appears to represent intraspecific variation is seen among specimens referred toIchthyornis. A large ovoid scar on the distal end of the anterior surface of the deltopectoral crest (e.g., YPM 1461, YPM 1720, YPM 1742) and a smaller scar just anterodorsal to the humeral head appear to be more pronounced in larger specimens (e.g., YPM 1742) as does a scar on the distal radius, character 7 from the diagnosis ofIchthyornis. These scars are commented on in more detail in the Anatomical Description. In addition,Olson (1975) reported a nutrient foramen just proximal to the dorsal condyle of the humerus inIchthyornis dispar andIchthyornis antecessor. It is also developed in other YPM humeri (e.g., YPM 1748, YPM 1764) but may not be present in all (e.g., YPM 9685) and can vary slightly in position and development. In theIchthyornis dispar holotype, the foramen is conspicuous on the right but not on the left humerus. These subtle differences were observed in multiple YPM specimens.

The only other morphological variation observed in YPM material referred toIchthyornis was observed only in single specimens. Nine YPM specimens show such differences. These specimens, the morphological differences noted, and the basis for referral toIchthyornis dispar are listed intable 4. Unfinished bone at the ends of the ulna that comprises YPM 1740 suggests that it may be an immature individual. It is intermediate in size between theIchthyornis dispar holotype and the large specimens that are the majority of the YPM material. YPM 1774 is a large coracoid (i.e., the size of the majority of the material) with what appears to be a pathology; the glenoid facet has an irregular boss of bone on its posterolateral edge and slopes smoothly into the corpus.

Evidence for the Presence of Distinct Ontogenetic Stages

There is some evidence for the presence of individuals of different ontogenetic stages in the YPM material referred toIchthyornis. In theIchthyornis dispar holotype, septa between most of the posterior alveoli of the mandible are thickened and completely formed. By contrast, in several specimens larger than theIchthyornis dispar holotype, these septa are thin or apparently absent (e.g., YPM 1735, YPM 1749, SMM 2503). However, in both theIchthyornis dispar holotype and in these larger specimens, most of the bones of the mandible appear to be fused.

In YPM 1733, a midsized specimen larger than theIchthyornis dispar holotype but smaller than the majority of the YPM material, the intercentrum of the atlas is completely fused to the body of the axis: No suture is visible. However, in the larger YPM 1775, which is the same size as the majority of the YPMIchthyornis material, this suture is incompletely obliterated. Finally, as mentioned above, YPM 1740 is an ulna intermediate in size between theIchthyornis dispar holotype and larger individuals, and is possibly subadult, based on the bone surface texture of its ends.

All specimens (with the exception of YPM 1740) appear to be adult. Muscle scars are well developed on all bones and some, as noted, appear more pronounced in the material larger than theIchthyornis dispar holotype. All known carpometacarpi, tibiotarsi, and tarsometatarsi are completely fused. All skull bones, with the possible exception of the frontoparietal contact, are fused in both theIchthyornis dispar holotype and the apparently just slightly larger YPM 1728.

Conclusions

Several possible conclusions can be drawn about the number of species represented in theIchthyornis material from the variation in size, morphology, and ontogenetic stage. First, it must be emphasized again that the specimens referred toIchthyornis do not sample contemporaneous individuals. Specimens referred toIchthyornis have been described from the late Cenomanian–lower Campanian, a period of approximately 15 million years. Furthermore, even if the majority of the YPM specimens are only from the upper Santonian–lower Campanian, this interval still represents several million years. The individual represented in the holotype ofIchthyornis dispar, collected from the lower Santonian, may have existed as many as a million years earlier than the majority of YPM specimens.

Explanation of some of the observed variation in size by anagenetic change may be supported by apparent size variation across time (fig. 6; although see an alternative explanation of the data as differential sampling through time [fig. 7]). While the distribution of referred material from the Cenomanian is all significantly larger than the early Santonian holotype ofIchthyornis dispar, all Turonian referred specimens are smaller than or the same size as theIchthyornis dispar holotype. The latest specimens (late Santonian–early Campanian) include the largest specimens, which are significantly larger than the holotype ofIchthyornis dispar. If the Cenomanian-referred material is borne out asIchthyornis, there may be a sharp decrease in size across the Cenomanian–Turonian boundary. This may be consistent with anagenetic change tracking environmental degradation associated with a global biotic event that has been recognized in marine environments (Raup and Stepkoski, 1986) and isotopic (δ¹³C and δ¹⁸O) fluctuations (Kyser et al., 1993).

However, within the comparatively short period of time represented in the Smoky Hill Chalk Member, there is variation in size: A small individual, approximately the size of theIchthyornis dispar holotype, and a larger one the size of the majority of referred specimens are copresent in the upper part of the Smoky Hill Chalk Member, and two specimens in the YPM Smoky Hill material are smaller than the earlier (Turonian) material. Anagenetic change within a single lineage, thus, does not appear sufficient to explain all observed variation in size.

This “residual” variation may be explained by the presence of more than one species ofIchthyornis or as intraspecific variation. At this time, the latter explanation cannot be rejected. Indeed, because of the distribution of morphological differences among the YPM specimens referred toIchthyornis, and because other Mesozoic avialan species show gross differences in size among referred individuals (Houck et al., 1990;Chiappe et al., 1999), this hypothesis is currently considered the more strongly supported.

Two forms of intraspecific variation are specifically considered here: difference due to ontogenetic stage and due to sexual dimorphism. Generally, there is little information on intraspecific variation in size or morphology (polymorphism) for Mesozoic avialan species to contextualize the variation seen inIchthyornis because most of these species are known from single specimens. However, for the few species that are known from multiple specimens, considerable variation in size among individuals has been described. For example, bothArchaeopteryx lithographica (Houck et al., 1990) andConfuciusornis sanctus (Chiappe et al., 1999) are known from multiple specimens that show a considerable range in size. InArchaeopteryx lithographica, the largest (Solnhofen) exemplar is roughly twice the size of the smallest (Eichstätt;Houck et al., 1990). InConfuciusornis sanctus,Chiappe et al. (1999) noted that the length of the humerus from the smallest specimen sampled was approximately 60% the length of the humerus of the largest specimen considered. The smallest humerus of the YPM specimens (table 3) is approximately 63% percent of the largest while that of theIchthyornis dispar holotype is approximately 80% of the largest individual measured (table 3). However, only 2 specimens of the total 77 referred are this small, while most of the variation is between 80% and 100% of the largest specimen. Thus, while size variation as extreme as that observed inConfuciusornis sanctus is represented in the YPM material, the majority of the variation is less extreme than in that taxon or inArchaeopteryx.

Houck et al. (1990) explained the variation inArchaeopteryx as a growth series by presenting morphometric data, indicating that the known specimens scale allometrically. These authors further suggested that allArchaeopteryx specimens might represent subadult individuals, as all lack the compliment of fusions seen in adult coelurosaurs (Houck et al., 1990).Chiappe et al. (1999) also considered the size variation inConfuciusornis sanctus to represent a growth series. Other authors, with reference to theIchthyornis material, have considered individuals even slightly different in size to be parts of distinct species (e.g.,Olson, 1975). Apparently, this is because a comparatively minute difference in size is all that appears to distinguish some crown clade species, and size within species is often assumed to be near invariant for adults of crown clade species. It has also been generalized that birds reach adult size in weeks to months after hatching (de Ricqlés et al., 2001). However, becauseIchthyornis lies outside the avian crown clade it cannot, without additional evidence, be justifiably inferred to have grown like a living bird (Gauthier and de Queiroz, 2001). That certain muscle impressions in the YPMIchthyornis material appear more strongly developed in larger specimens appears consistent with these larger individuals being more mature. Furthermore, plots (figs. 5,8,9) of the differently sized individuals inIchthyornis appear closer to a continuum than to support the presence of distinct size classes.

As noted above, nearly all of theIchthyornis specimens appear adult or nearly adult as indicated by the fused elements, muscle impressions, and bone texture that subadult Aves often lack.Wang et al. (2000) conducted a statistical analysis of several hundred specimens ofConfuciusornis sanctus and concluded that all of these specimens were of adult individuals. These observations would appear inconsistent with the interpretation of the variation in size in both taxa as representing a growth series.

Some histological studies have been taken to indicate that a shift to an avian pattern of rapid sustained growth occurred phylogenetically afterPatagopteryx deferrariisi (as well as Enantiornithes andConfuciusornis sanctus), and beforeHesperornis regalis andIchthyornis (Chinsamy et al., 1998). This conclusion was based on the apparent absence inHesperornis regalis andIchthyornis of lines of arrested development (LAGs) and on the presence of highly vascular primary bone (Chinsamy et al., 1998). At least one LAG was interpreted as present inPatagopteryx deferrariisi and a sampled enantiornithine (Chinsamy et al., 1995). However, this hypothesis has been critiqued more recently, as the significance of LAGs, as well as their known distribution, has been reappraised (de Ricqlés et al., 2001). Furthermore, the notion that an avian growth pattern is seen inHesperornis regalis andIchthyornis is considered speculative (Castanet et al., 2000). Complicating the interpretation of the data from histological studies, the single humerus ofIchthyornis so far sampled has been considered possibly subadult (Chinsamy et al., 1998;de Ricqlés et al., 2001). Indeed, that this element was apparently not assessed to be subadult (which would make it less than ideal for sampling) based on morphology alone is suggestive that otherIchthyornis specimens, previously assumed to be fully mature individuals, may also be subadult.

Rapid rates of growth during some part of ontogeny have been associated with highly vascularized primary bone that is seen not only in basal avian taxa (Castanet et al., 1996,2000) but in nonavialan dinosaurs (Horner et al., 2001). However, it has been postulated that histological evidence from basal avialans suggests that some part of their ontogeny involved a slow growth phase and that they may have taken longer to reach adult size (de Ricqlés et al., 2001). This conclusion is consistent with the size ranges described forArchaeopteryx andConfuciusornis sanctus (de Ricqlés et al., 2000,2001) and might explain the size variation observed inIchthyornis. We appear to be early in developing understanding of the evolution of dinosaur ontogeny, and further study of the histology of noncrown clade avialans is necessary to document the changes that occur in growth pattern (de Ricqlés et al., 2001). Planned future histological work investigating samples fromIchthyornis should provide key insights into whether small and large individuals are, indeed, all adult. This data is essential to resolving the number of species represented in the YPMIchthyornis material and to determining when, phylogenetically, the growth pattern seen in Aves (in which adult size is achieved in a matter of weeks to months) arose. Currently, the explanation of the variation in size present among the YPMIchthyornis material as distinct growth stages, perhaps within a protracted slow growth phase, fits histological data and is consistent with data from more basal Mesozoic avialans. If minute variation in size is considered to distinguish distinct species, it is not currently clear where the lines distinguishing these size classes could be meaningfully drawn.

While the variation in size among the YPM specimens appears largely explained by some anagenetic change and by individuals at different growth stages, it may not explain the three cases described above: (1) the axis is preserved in two specimens, and the larger of the two (YPM 1775) has an open suture that is closed in a slightly smaller specimen (YPM 1733); and (2) the septa separating the posterior alveoli of the dentary tooth row in theIchthyornis dispar holotype are fully formed while in larger specimens (e.g., YPM 1735), they appear slightly more weakly developed or absent. Finally, unfinished bone at the ends of an ulna (YPM 1740) suggest it is from a very young individual, but the ulna is larger than that of theIchthyornis dispar holotype. These data may suggest the presence of at least two species. However, we do not have data that any of the mentioned specimens represent nearly contemporaneous individuals. Furthermore, to compare the case of the axis and the ulna: The first case involves a specimen that falls nearly exactly intermediate in size between theIchthyornis dispar holotype and the majority of the YPMIchthyornis material and that appears adult relative to a larger individual. In the case of the ulna, a specimen that is, again, squarely intermediate in size appears to be from a very young individual. Thus, the signal from these cases is not clear. Even if two species were present, intermediary-sized material could potentially belong to either, and for other specimens, ontogenetic clues are not available. It is also possible that this variation is anagenetic at a finer scale than that discussed, with small scale climatic variation causing fluctuations in adult size or growth rates. On the basis of geochemical (%CaCO₃, %Al, %organic carbon) logs, gamma-ray logs, and bedding couplets,Pratt et al. (1993) identified four cycles in Niobrara Formation (Fort Hays Limestone and Smoky Hill Chalk Members) with estimated periodicities of 1.7 m.y. (30-m cycles), 280 k.y. (5-m cycles), and 100 k.y. (2-m cycles) and 41 k.y. for the bedding couplets. The 41 k.y. and the 100 k.y. periodicities are close to expected periodicities of the precession and eccentricity of the earth's orbit (Pratt et al., 1993) and may reflect cyclic variation in the depositional environment.Pratt et al. (1993) concluded that proximate causes of such cycles transcribed in the sediment record were changes in regional climate and paleo-oceanographic conditions, both of which would be expected to impact the local fauna; food supply and surface water temperatures have been shown to relate to intraspecific differences in adult body size in an array of avian taxa (e.g.,Graves, 1991;Leafloor et al., 1998). However, even if the YPM specimens are assumed to represent contemporaneous individuals (contra limited stratigraphic information available), an explanation other than the presence of more than one species lineage is plausible.

Sexual dimorphism in size is a widespread phenomenon in Aves, having arisen many times and ranging from minor to extreme difference in size. Although sexual size dimorphism appears highly homoplastic across Aves, the presence of dimorphism in size in palaeognaths and galloanserines (del Hoyo et al., 1992–1999) suggests it may be ancestral to Aves. Furthermore, although dimorphism in size has not been investigated forConfuciusornis sanctus, there is evidence of sexual dimorphism in plumage (Chiappe et al., 1999). A variety of nonavialan dinosaurs has also been described as possibly sexually dimorphic in the development of particular morphologies or as having gracile and robust morphs (e.g.,Chapman et al., 1997). Sexual dimorphism in growth rates and growth duration is frequently documented in birds (reviewed inTeather and Weatherhead, 1994). Recent studies have found that in avian species in which males are larger than females, they may reach adult size after females (Badyaev et al., 2001). The growth rates of different traits and the duration of different growth phases varied in their relative time of onset between males and females (Badyaev et al., 2001).

The difference in size between theIchthyornis individuals discussed is comparatively minimal. In the more extreme of the two cases discussed, the axis with the closed suture is approximately 85% the size of that of the larger individual. This would yield a difference between these individuals in a ratio of larger/smaller of 1.18. This is well below the most extreme instances of dimorphism in Aves where, for example, in one icterid (Passeriformes) the male/female ratio for wing measurements is 1.33 (Webster, 1997), a value approached by a variety of other avian taxa (e.g.,Tetrao urogallus, Galliformes: 1.30 for the wing andOtis tarda: Gruiformes: 1.27 for the wing; K. Zyskowski, personal commun.). Sexual size dimorphism in the extinct dodo (Raphus cucullatus) and solitaire (Pezophaps solitaria;Livezey, 1993: table V) is also close to or exceeds size difference amongIchthyornis specimens (e.g., ratio of humerus length in “males”/“females”: 1.10 and 1.30, respectively) and these differences were similarly considered by some early workers to be evidence of distinct species (Livezey, 1993).

The morphological variation noted in single specimens (table 4) is not considered sufficient to merit the recognition of distinct species. Discovery in additional specimens of any of the morphological variants described, especially if associated with additional differences from the morphologies of the holotype ofIchthyornis dispar, may form the basis for recognizing distinct species ofIchthyornis. Variation in a variety of morphological characters has been noted acrossConfuciusornis sanctus specimens (e.g., in the development of the sternal midline ridge, fusion of the dentaries, and possible presence of uncinate processes;Chiappe et al., 1999). These differences, like those observed in theIchthyornis material, may be related to the presence of individuals of differing ontogenetic stages or sexes among the material. To assume that in a sample of specimens representing individuals from across millions of years there would be no variation, even as minor as that described, would seem unrealistic.

Two previously named species referred to the genusIchthyornis (Marsh, 1880) are not part of the cladeIchthyornis (see Part II, Results).Apatornis celerMarsh, 1873a (Marsh, 1873b) is recognized as a valid taxon. The holotypes ofIchthyornis tenerMarsh, 1880,Ichthyornis lentusMarsh, 1877b (Marsh, 1880), andApatornis celerMarsh, 1873a (Marsh, 1873b), are differentiated fromIchthyornis dispar and were discovered more closely related to or part of Aves or placed as part of Aves in the phylogenetic analyses (Part II, results). The species epithets “tener”, “lentus”, and “celer” as well as the name “Apatornis” are converted in the context of a system of phylogenetic taxonomy (de Queiroz and Gauthier, 1990,1992;Cantino and de Queiroz, 2000). Two new clades are named to whichtener andlentus are designated internal specifiers. The one specimen previously referred toApatornis celer (YPM 1734) cannot be compared to the holotype ofApatornis celer (YPM 1451) and is designated as the holotype of a new species. The five remaining previously named species ofIchthyornis are junior synonyms ofIchthyornis dispar, as mentioned above. Descriptions are provided of the holotypes of all previously named and newly identified species. Commentary is provided on the prior referral of specimens to these species. With reference to previously-named species that are recognized as junior synonyms ofIchthyornis dispar, the descriptions and commentary given support the synonymy of these species and serve as a resource for future work.

Most specimens previously referred to species ofIchthyornis andApatornis cannot be compared to the holotypes of these species. Indeed, many of the referrals appear to have been largely arbitrary. Even if some of the species names synonymized withIchthyornis dispar are found to be valid, almost without exception, the previous referral of specimens to these taxa will not. Size appears to have been the basis for the referral of the majority of specimens toIchthyornis victor; however, even this criterion appears to have been inconsistently applied. For example,Marsh (1880) referred YPM 1733 toIchthyornis victor, while a specimen of approximately the same size (YPM 1764) was referred toIchthyornis dispar (Olson, 1975).

Ichthyornis anceps

Graculavus anceps (Marsh, 1872a) was named several months beforeIchthyornis dispar (1872b) and was later referred toIchthyornis (Marsh, 1880).Graculavus anceps is not the name-bearer of the taxon “Graculavus” and, thus, the name of the species was changed toIchthyornis anceps (Marsh, 1880). The specimen number of the holotype was not given in the original publication but was later specified (Marsh, 1880).

Ichthyornis agilis

Graculavus agilis (Marsh, 1873c) was named shortly afterIchthyornis dispar (Marsh, 1872b) and was subsequently referred toIchthyornis (Marsh, 1880). It was described in two short sentences, and a holotype specimen was not named in the publication. The holotype specimen was not mentioned to be a proximal carpometacarpus, and the differentiation provided ofGraculavus agilis fromGraculavus anceps (Marsh, 1873c) is unsupported, primarily because the two holotypes are nonoverlapping parts of a left carpometacarpus that cannot be directly compared.

Ichthyornis victor

Ichthyornis victor was named byMarsh (1876) and differentiated from the holotype ofIchthyornis dispar (YPM 1450) as being one-third larger than that species (Marsh, 1876).

Ichthyornis validus

Ichthyornis validus was named inOdontornithes. The holotype was not described, differentiated, or diagnosed, although it was figured (Marsh, 1880: pl. XXX, figs. 11–14).

Ichthyornis antecessor

Plegadornis antecessor was named byWetmore (1962) based on a distal humerus.Kashin (1972) noted thatPlegadornis was preoccupied, and changed the name toAngelinornis antecessor.Olson (1975) used characters shared byIchthyornis andAngelinornis antecessor to synonymize the latter genus and refer the specimen toIchthyornis, but still recognized it as a valid species,Ichthyornis antecessor.

Guildavis (new clade name)

tenerMarsh, 1880 (converted species name)

ApatornisMarsh, 1873b (converted clade name)

celerMarsh, 1873a (converted species name)

marshi (new species name)

AVESLINNAEUS, 1758 PANGALLIFORMES (NEW PROVISIONAL CLADE NAME)Austinornis (New provisional clade name)

lentusMarsh, 1877b (converted species name)

Cranial Elements

The description of the cranium (exclusive of the mandible) inOdontornithes (Marsh, 1880) is limited. It included measurements (Marsh, 1880: 125) of the holotype ofIchthyornis dispar (YPM 1450) and of a second specimen, YPM 1459.Marsh (1880: 124) also commented that two specimens were used to estimate the size of theIchthyornis brain, but did not indicate which specimens these two were.

Cranial material incorporated in theIchthyornis dispar panel mount (YPM 1450;fig. 1) is comprised of one partial cranium and one fragment of the tooth-bearing portion of the left maxilla (fig. 18). However, we can only surmise the specimen numbers of three cranial fragments in theIchthyornis victor mount (fig. 19A). Henry Gibb's notes (fig. 3) placed a single arrow applying the number “[YPM] 1728” to the illustration of the cranium on the annotated copy of the reconstruction ofIchthyornis victor (Marsh, 1880: pl. XXXVI) that he used to indicate the contents of theIchthyornis victor mount. However, the single, large portion of the cranium illustrated (Marsh, 1880: pl. XXXVI, and copied in Gibb's notes,fig. 3) is simply an enlargement of theIchthyornis dispar holotype skull (compareMarsh, 1880: pl. XXI), and not based on material referred toIchthyornis victor. Further, Gibb (fig. 3) applied only one number (YPM 1728) to theIchthyornis victor reconstruction that he annotated, while three separate cranial fragments were included in theIchthyornis victor mount (fig. 19A).

In this section, the identity of the specimen(s) represented by the three skull parts in theIchthyornis victor panel mount is discussed. Next, the basis for the referral of specimens known only from cranial material toIchthyornis dispar is appraised. Finally, the morphology of the cranium is detailed, as inferred from the partial cranium of theIchthyornis dispar holotype and specimens assessed to be reliably referred to this species.

The YPM 1728 specimen number is here applied to all of the cranial material from theIchthyornis victor mount consistent with the label in Gibb's notes. However, it is possible (but considered unlikely) that one of these fragments, which includes the occipital condyle (figs. 20–22), is part of specimen YPM 1459. Because the fragment in the mount is the only YPM specimen to preserve basicranial morphology other than the holotype ofIchthyornis dispar (YPM 1450), the specimen number of this fragment is nontrivial. Its identity determines which other material is associated with it, either as part of YPM 1728 or YPM 1459.

Measurements were given (Marsh, 1880: 129) for a supposedly “cordate” (Marsh, 1880: 121) occipital condyle in YPM 1459 (i.e., transverse diameter: 3.0 mm; median vertical diameter: 2.2 mm). These dimensions are approximately the same as those of the basicranial fragment in theIchthyornis victor mount identified by Gibb as YPM 1728 (i.e., 3.0 mm and 2.5 mm, respectively). Does this suggest that the mount specimen is YPM 1459? The dimensions of the frontals in YPM 1459 and those in a second part of the cranium in theIchthyornis victor mount (presumably YPM 1728) suggest they are from individuals of approximately the same size. Thus, the occipital condyle in the mount could be from the similarly sized YPM 1728 and, for this reason, closely matchesMarsh's (1880) dimensions for YPM 1459. However, no occipital condyle appears currently among the YPM 1459 material; what didMarsh (1880) measure? It is possible that an occipital condyle was once part of YPM 1459 and is now lost. Alternatively,Marsh (1880) could have identified a small, extremely crushed fragment from YPM 1459 as an occipital condyle; he identified such a condyle in the holotype ofIchthyornis dispar (Marsh, 1880: pl. XXI, “bo”), where no morphology resembling a condyle is present. In 1983, Whetstone identified anIchthyornis braincase as YPM 1459 when describing anatomical details of thefenestra pseudorotunda and pneumatic features of the braincase. These morphologies are also not represented in material that is currently part of YPM 1459, although they are represented in the fragment included in the mount with the well-preserved occipital condyle.

The question of whether the partial basicranium in theIchthyornis victor mount with the well-preserved occipital condyle is part of YPM 1728 (per Gibb's notes) or a missing and measured part of YPM 1459 may never be known with complete certainty. In this paper, however, the decision was made to apply YPM 1728 to this specimen for three reasons. First, the positive identification of the cranial material in the mount as YPM 1728 in Gibb's notes is considered stronger evidence than the apparent lack of basicranial material in YPM 1459. Second, it seems thatWhetstone (1983) simply assumed that the measurements given inMarsh (1880) pertained to the mount fragment and applied YPM 1459 to that specimen. Gibb's notes and other information on the composition of the mount were unknown at that time and would, thus, have been unavailable toWhetstone (1983). Finally, the pale yellow-orange color of the element removed from the mount matches the color of the other two cranial fragments removed from theIchthyornis victor mount, while the YPM 1459 material has a slightly darker, yellow-gray tone.

Neither of the two YPM specimens referred toIchthyornis dispar to include cranial material (i.e., YPM 1459 and YPM 1728) has associated postcranial material. Their identification asIchthyornis dispar is only supported by morphological correspondence to the cranium of theIchthyornis dispar holotype (YPM 1450).

YPM 1728 consists solely of parts of the cranium (which accords with the original inventory of the material) and two fish teeth. YPM 1459 includes a fragment mounted as part of the premaxillae in theIchthyornis victor mount (per Gibb's notes), an identification with which this reevaluation agrees. The rest of YPM 1459 includes a fragment (marked “YPM 1459A”) labeled “fish”, which is actually fused avian frontals (fig. 23B) for which the measurement of the “distance between upper margins of orbits” ofIchthyornis victor is given inOdontornithes (Marsh, 1880: 125). A final vial contains a proximal coracoid collected a year following the cranial material and not considered associated with it (seetable 1). Thus, the morphology of the partial coracoid cannot be used to refer YPM 1459 toIchthyornis dispar.

The frontals in YPM 1728 (fig. 23B), YPM 1459, and the holotype ofIchthyornis dispar (YPM 1450) correspond in the following details: (1) frontals fused; (2) paired shallow glandular depressions (i.e., fossae glandulae nasalis;Baumel and Witmer, 1993) on the dorsal supraorbital margins; (3) glandular depressions do not reach midline; (4) anteroventral midline of frontals fused with mesethmoid (only barely discernable in YPM 1450); (5) the position of the shallow groove for the olfactory nerve and ethmoid artery (i.e., sulcus n. olfactorii;Baumel and Witmer, 1993), which is developed in the angle between the interorbital septum and the dorsal lamina (Baumel and Witmer, 1993) of the mesethmoid (which is fused to the frontals). Morphology “5” is visible on the right side of theIchthyornis dispar holotype (YPM 1450;fig. 24A) in ventral view and on both the right and left sides in YPM 1459 and YPM 1728. The correspondence of these morphologies is consistent with identification of the material to the same species, a conclusion supported by other similarities discussed below. These characters, however, are not considered demonstrated apomorphies ofIchthyornis dispar.

The nearly complete frontals from YPM 1728 preserve contacts with the nasals, premaxillae, and lacrimals (fig. 23B). On the dorsal midline of the anterior tip of the fragment, a tongue-shaped depression, underlain by the mesethmoid, is interpreted as a surface for the articulation of the premaxillae (fig. 23B). There is no evidence (such as a midline ridge on the dorsal surface of the mesethmoid) to indicate that the premaxillae were unfused to each other posteriorly, althoughMarsh (1880: 121) described them as unfused.

Frontal processes of the premaxillae are known from YPM 1459 (previously incorporated in theIchthyornis victor panel mount). The premaxillary portion of the facial margin has been reported to lack teeth (Lamb, 1997) but is not preserved in any of the YPM specimens. The section of the frontal processes preserved in YPM 1459 is incomplete proximally and distally (fig. 23A). The premaxillae are fused for the length of this section such that a midline ridge is seen in ventral view (YPM 1459;fig. 23A). The ventral surface is concave on either side of the midline ridge at the closed suture. The fragment is slightly broader and flatter at one end, interpreted as posterior. Dorsally, a styloid midline depression, not an open suture, is visible for approximately the posterior half of the fragment (fig. 23A). This line deepens into a pit at approximately the midpoint of the dorsal surface (fig. 23A).

Based on the following two morphologies, it is inferred that the fragment would have formed most of the dorsal margin of the external nares (its approximate position in theIchthyornis victor mount): (1) the presence of a ventrally projected midline flange anteriorly, which in Aves extends toward the body of the premaxillae and sits just anterior to the external nares (fig. 23A); and (2) the slight posterior flattening that marks the nasofrontal hinge in Aves.

Based on the position of the nasals inHesperornis regalis (Bühler et al., 1988) and the anteroventral midline flange, it may be that the preserved part of the premaxillae represents all or nearly all of the supranarial bar. If so, the supranarial bar would have been relatively thick dorsoventrally, as has been reported forHesperornis regalis (Bühler et al., 1988). The nature of the premaxilla/nasal contact, and specifically the anterior extent of the nasals, is somewhat unclear. Two lateral flanges, which extend one-third of the length of the ventral surface of the fragment, are interpreted as representing the anterior tips of the nasals (fig. 23A). The nasals do not appear to have met or approached each other on the ventral midline, but rather to have only underlain posterolateral edges of the premaxillae (fig. 23A).

The posterior tips of the nasals do not contact the premaxillae. Narrow, tonguelike processes of the frontals lie between these bones and the nasals in dorsal view (YPM 1728;fig. 23B). Processes of the frontals also underlie and pass lateral to the nasals. These lateral processes bear well-developed facets for the lacrimals, the articular surfaces of which are perpendicular to the dorsal surface of the frontals (fig. 23B). The lacrimals may have also articulated with the nasals; the facets are incomplete anteriorly.

The presence of facets indicate that sutures between the lacrimals and frontals were not closed inIchthyornis dispar. Lateral facets on the frontals inHesperornis regalis (e.g., YPM 1206) indicate that, in this taxon, the lacrimals were also not fused to the frontals. By contrast, the two bones are often coosified in Aves (Cracraft, 1968), a feature that may be synapomorphic of that clade or derived within it.

In Aves, the lacrimal can articulate with (or coosify to) both the nasal and frontal, but may articulate exclusively with either of these two bones (Cracraft, 1968). InStruthio camelus andRhea americana, for example, the lacrimal usually contacts only the nasal (Cracraft, 1968), unlike the condition inIchthyornis dispar, while in tinamous,Apteryx, andGallus gallus, it contacts the frontal and nasal (Cracraft, 1968). In Anhimidae, it also articulates exclusively with the nasal (Cracraft, 1968).Hesperornis regalis has the lacrimal primarily contacting the nasal, although also contacting the anterolateral portion of the frontal (Bühler et al., 1988).

As noted above, the nasals (from YPM 1459;fig. 23A) do not appear to have met on the ventral midline or to have extended under the premaxillae to approach the anterior margin of the external nares. These two morphologies appear also to be potential synapomorphies of Aves relative toIchthyornis dispar, as theIchthyornis condition is also present inHesperornis regalis (e.g.,Bühler et al., 1988).

While the mesethmoid inIchthyornis dispar (YPM 1450, YPM 1459, YPM 1728) is fused to the frontals (fig. 23B), it did not form the complete or nearly complete interorbital septum seen in many adult avians. Posterodorsally, the unbroken ventral edges of the frontal indicate they did not contact the mesethmoid (fig. 23B). Thus, the ventral processes of the frontals did not meet to completely close the anterodorsal part of the cranial cavity; the mesethmoid contacts and fuses to the frontals only anteriorly. In Aves, this morphology would appear to be described as a broadly open, single foramen opticum (Baumel and Witmer, 1993). The mesethmoid may have extended posteriorly to contact the ventral processes of the frontals and/or part of the basicranium only ventrally, a condition that appears to be developed inHesperornis regalis (seeBühler et al., 1988).

The mesethmoid is visible in YPM 1728 extending from anterior to the anterior edges of the ventrally projecting processes of the frontal to the anteriormost tip of the fragment (fig. 23B). It is fused indistinguishably with the frontals for the length of its preserved contact with them. The mesethmoid inHesperornis regalis was not fused to the frontals (Bühler et al., 1988), unlike the condition inIchthyornis dispar (YPM 1450, YPM 1459, YPM 1728).

The mesethmoid is interpreted as having been overlain by the premaxillae; its apparent dorsal exposure (fig. 23B) is here considered an artifact of preservation. Where the mesethmoid is exposed dorsally (e.g., Ratitae;Cracraft, 1986), its dorsal surface is coplanar with the frontals, which is not the condition in YPM 1728 (fig. 23B). The thin and shallow sulcus for the olfactory nerve developed on the mesethmoid angles laterally at approximately the premaxilla/frontal contact, where the mesethmoid also broadens (fig. 23B). The anterior extent of the mesethmoid unfortunately cannot be ascertained from YPM 1728, but it may thin slightly (to taper out?) at the anteriormost tip of the fragment (fig. 23B).

The maxilla is represented only by a small, tooth-bearing fragment that is part of the holotype ofIchthyornis dispar (YPM 1450;fig. 18). It was described and figured inOdontornithes (Marsh, 1880: pl. XXI, fig. 1); however, this fragment differs fromMarsh's (1880) illustration.Marsh (1880: pl. XXI, fig. 1b) indicated a thin sheet of bone covering the alveoli dorsally, but the alveoli in the preserved fragment are completely perforate (fig. 18). Further,Marsh (1880: pl. XXI, fig. 1a) illustrated a curved edge he interpreted as the medial edge of the left maxilla. However, the fragment is a mirror image of its representation inOdontornithes (Marsh, 1880: pl. XXI, fig. 1), the curved edge reported seen on the opposite side of the fragment (fig. 18), and is here interpreted as a dorsally projected flange (fig. 18B), laterally placed, and increasing in height posteriorly. That is to say, the fragment would be from the left maxilla where Marsh placed it in his reconstructed outline of the skull (Marsh, 1880: pl. XXI, fig. 1), but a dorsal flange that was not illustrated would have been visible in lateral view.

The straight medial edge of the maxilla has a slightly rounded edge, which does not bear any indication of articular surfaces for any palatal elements (fig. 18B). On this side of the alveolar surface, interpreted as medial (labial), minute foramina lie close and slightly posterior to the alveoli themselves. These foramina are not present in the dentaries (YPM 1450). Similar features are observed on the medial surface of the maxilla in crocodilians (J. Gauthier, personal obs.).

Marsh (1880: 124) interpreted the size of the maxillary teeth as larger than the dentary teeth that would have opposed them. However, the exact position of the maxillary fragment in the jaw is uncertain and the alveoli are not significantly larger than those in the lower jaw. Further, the posterior teeth that Marsh interpreted as smaller than those in the middle of the dentary tooth row may be simply in a different stage of replacement.Hesperornis regalis has been proposed to have a groove in the maxilla into which subnarial processes of the premaxillae/nasals contacted in a unique tongue and groove articulation (Bühler et al., 1988). This morphology is unfortunately missing data for the alveolar portion of the maxilla preserved in YPM 1450.

Severely crushed portions of the posterior part of the skull are preserved in the holotype ofIchthyornis dispar (YPM 1450;figs. 24,25), and in YPM 1728 (figs. 20–22). Two relatively large portions of the skull roof were removed from theIchthyornis victor mount. These include the fragment with a well-preserved occipital condyle discussed above and identified as belonging to YPM 1728 (figs. 20–22) and one fragment explicitly marked with the specimen number “1728” (fig. 26). The anatomy of the posterior part of the skull inIchthyornis dispar is described from these two specimens. It is inferred from comparison of features in the auditory region and frontals that there was no conspicuous difference in size between these two specimens (although YPM 1450 is the holotype ofIchthyornis dispar, and YPM 1728 was referred toIchthyornis victor [Marsh, 1880]).

The larger fragment removed from theIchthyornis victor mount and marked with “1728” is interpreted as part of the frontals (fig. 26). It is extremely thin, dorsally convex, and its arch is interpreted as preserving the contour of the right cerebral hemisphere (fig. 26). On the dorsal midline (fig. 26A), there is a moderately developed frontal depression (depressio frontalis;Baumel and Witmer, 1993). A faint ridge (crista temporalis;Baumel and Witmer, 1993) appears to border a shallowly depressed portion of the temporal fossa (fossa temporalis;Baumel and Witmer, 1993) surrounding the upper temporal fenestra (fig. 26A).

In ventral view, there is a ridge (crista frontalis interna,Baumel and Witmer, 1993) demarcating the virtually complete impression of the right cerebral hemisphere from the preserved portion of the left (fig. 26B). At the anteriormost edge of the fragment close to and subparallel with this ventral midline ridge is a short, low ridge (fig. 26B).

The frontal/parietal contact is not discernable in YPM 1459 or YPM 1728 but may be in YPM 1450 (fig. 24). If present in YPM 1450, the frontal/parietal suture appears obliterated or closed (fig. 24B). A transverse ridge possibly associated with the back of the skull at or near the supraoccipital/parietal contact (YPM 1450;fig. 24A,B). Anterior to this contact, the bone surface in left lateral view is smooth, and the faint contour of the cerebral hemisphere appears visible. Posterior to this contact there is a change in the slope of the dorsolateral surface of the skull (fig. 24B). This change in slope is also seen in many avians and can be observed in juveniles when the sutures are still open (e.g.,Gallus gallus).Marsh (1880) described a sagittal crest in the holotype; however, this is an artifact of crushing (fig. 24B).

Marsh's (1880) interpretations of the size and shape of the brain inIchthyornis must have been made from YPM 1728. Though Marsh comments that his reconstruction of the brain is based on two specimens (Marsh, 1880: 122), no other specimen was found that would appear informative about brain morphology. One possible exception is part of YPM 1773, which could be interpreted as an endocast, but is not mentioned by number inOdontornithes (Marsh, 1880). While the estimate of the size of the cerebral hemispheres is partially discernable in YPM 1728, the rest of the reconstruction (e.g., of the cerebellum) must be considered highly, if not entirely, speculative.

The occipital condyle is well preserved in the second fragment removed from theIchthyornis victor mount identified as belonging to YPM 1728 (see above) but not in theIchthyornis dispar holotype (YPM 1450). The basioccipital and exoccipital components of the occipital condyle are fused (figs. 20,22). The condyle appears to have been large relative to the size of the foramen magnum (fig. 22) compared to the condition in Aves. However, the size of the foramen magnum can only be roughly interpreted, as it is crushed in both YPM 1450 (fig. 24A) and YPM 1728 (fig. 22). The dorsal edge of the foramen magnum in YPM 1728 is relatively well preserved, although its lateral edges are fractured and crushed (figs. 20,22). The occipital condyle in YPM 1728 is artifactually appressed to the left basal tuber. These tubera in both YPM 1450 (figs. 24A,25) and YPM 1728 (figs. 20,22) are conspicuously robust (figs. 20,22).

The anterior portion of the auditory recess is clearly visible, as the ala parasphenoidalis (Baumel and Witmer, 1993) is apparently broken away from this region in both YPM 1450 (figs. 24A,25) and YPM 1728 (fig. 20). In YPM 1728, the metotic strut (sensuWitmer, 1990) is visible in left ventrolateral view (fig. 20). Portions of the fenestra ovalis, fenestra pseudorotunda, and a depression interpreted as a part of the posterior tympanic recess are visible anterior to this strut (fig. 20).Whetstone (1983) identified a caudal (posterior) tympanic recess in a specimen he identified as YPM 1459; he appears, however, to have been referring to the specimen from theIchthyornis victor mount, which is YPM 1728. A conspicuous foramen lies anterior to and at about the same dorsal height as the fenestra ovalis (fig. 20); this feature is interpreted as the exit of the fifth cranial nerve (trigeminal; or foramen n. maxillomandibularis;Baumel and Witmer, 1993).

Paired large openings on the ventral surface of the basisphenoid plate visible in YPM 1450 (fig. 25) and to a lesser extent in YPM 1728 (figs. 20,22) may be the openings of ossified eustachian tubes.Witmer (1990) describes paired foramina inEnaliornis barretti as such openings. On the right side of the preserved section of the basisphenoid plate in YPM 1728 and just ventral to the right basal tuber, a flat fragment that does not appear grossly displaced may be a part of the parasphenoid. A conspicuous foramen in this fragment is possibly the anterior exit of an external carotid (fig. 21). Below this relatively large opening is a straight fracture. On the ventrolateral surface of the left basal tuber, two oblong grooves are developed which appear to be artifactually exposed pneumatic cavities (fig. 20).

Most of the brief description of the skull of YPM 1450 (Marsh, 1880: 120) is inaccurate. However, it can be inferred, asMarsh (1880) suggested, that the rostrum inIchthyornis dispar is extremely long, based on the preserved length of the mandibles. Although,Marsh (1880) described the occipital condyle as “very small, and directed backward” in YPM 1450, it is not preserved. What was indicated as the occipital condyle (Marsh, 1880: pl. XXI) is a part of a flat piece of bone perforated by two foramina discussed below (figs. 24A,25).

In YPM 1450, the fragment identified as bearing the occipital condyle byMarsh (1880: pl. XXI) is visible in right lateral view (fig. 25). This flat fragment appears to be a portion of the exoccipital. It is perforated by two foramina, the more dorsal of which is the larger (fig. 25). These foramina may represent two exits of the 12th cranial nerve (hypoglossal), but this identification, given the obvious displacement of the fragment, is tentative. Slightly anterodorsal to this fragment lies another that appears to preserve the margin of a third and larger foramen (fig. 25) that may represent the exit of the 10th cranial nerve (vagus). Anterior and slightly ventral to the above-described fragment lies a third that ventrally contacts the right basal tuber. It appears in approximately its natural position relative to the right basal tuber and is tentatively identified as the metotic strut (fig. 25).

Anterior to the metotic strut (sensuWitmer, 1990) an elongate, crushed fossa interpreted as representing both the fenestra pseudorotunda and fenestra ovalis is present (fig. 25). A posterior projection is interpreted as an incomplete portion of the bar that separated these two fenestrae (crista interfenestralis). Dorsal to these depressions and close to the preserved fragments of the paraoccipital process is a depression possibly representing the posterior (caudal) tympanic recess (fig. 25). Anterior to the area of the fenestra pseudorotunda and fenestra ovalis is a small foramen that may represent the exit of the seventh cranial nerve (facial), but it appears slightly ventral to its typical position at approximately the same level as the fenestra ovalis (fig. 25). The auditory region lacks the extreme pneumatization seen, for example, in Palaeognathae and Galliformes (Witmer, 1990).

Anterior to this foramen is a much larger, roughly ovoid, depression interpreted as the exit of the fifth cranial nerve (trigeminal; foramen n. maxillomandibularis;Baumel and Witmer, 1993). Ventrally adjacent and just slightly anterior to this structure is an opening to an anteroventrally directed passage or pocket that may be the rostral tympanic recess (fig. 25). Ventral and slightly posterior to this structure is a depression (fig. 25) that could only be interpreted as a portion of the ossified eustachian tube entrance (comparing favorably withGavia stellata, e.g.). The remains of the ala parasphenoidalis (seeBaumel and Witmer, 1993, for a discussion of this structure, which has its own ossification center in Aves) are interpreted as present anteroventral to this structure. The ala parasphenoidalis shows no sign of being particularly large or thick, morphologies present in Hesperornithes (Witmer, 1990). No further morphologies of the parasphenoid could be discerned; basipterygoid processes, for example, are not preserved.

The articular cotylae for the quadrate are best preserved in the holotype ofIchthyornis dispar, YPM 1450 (fig. 25). The otic and squamosal cotylae for the quadrate appear separated by a very narrow incisure marking the entrance to the dorsal tympanic recess visible on the right side of YPM 1450 (fig. 25) and on the left in YPM 1728 (fig. 20). A diminutive entrance to the dorsal tympanic recess between the capituli is consistent with the slight separation of the two articular surfaces on the head of the quadrate (YPM 1775;fig. 27D). In YPM 1450, there is no evidence of an elongate ventral, or “zygomatic”, process of the squamosal, but the lateral surface of this element is poorly preserved (fig. 25).

The only quadrates known forIchthyornis dispar belong to YPM 1775 (fig. 27), a specimen referred toIchthyornis dispar on the basis of apomorphy (seetable 1). The left quadrate was figured in relative detail inOdontornithes (Marsh, 1880); however,Marsh (1880) commented only on the braincase articulation, which he considered “single headed”.Witmer (1990) figured in stereo photographs and in a line drawing the complete left quadrate of YPM 1775. The otic process of the right quadrate is also preserved and has not been figured; it contains some additional information about the morphology of the head and the shape of the orbital process.

As described byWitmer (1990), the quadrate has a large pneumatic foramen lying close to the pterygoid condyle on the ventromedial surface of the shaft (fig. 27A,C).Witmer (1990) also noted that the position of the foramen is seen in some birds that were considered at one time “primitive” or basal within crown clade Aves (e.g., in the neoavianFregata minor). The foramen inIchthyornis dispar and these taxa, located close to the pterygoid articulation, differs, however, from the condition in the basal neognaths (i.e., galloanserines) surveyed for the phylogenetic analyses (see Part II, Materials and Methods), which have a foramen on the posterior surface of the shaft (seeappendix 1). Further, the extremely dorsally located foramina in Palaeognathae are found to be nonhomologous with the more basal foramina in the other included Aves because the two foramina co-occur at least inCrypturellus undulatus (appendix 1). They are only observed in the one crown clade palaeognath exemplar included and are, thus, optimized as an autapomorphy of that taxon (Part II, Results). They appear to be synapomorphies of crown clade Palaeognathae asWitmer (1990) suggested. The plesiomorphic position of a quadrate pneumatic foramen is ambiguously optimized (Part II andappendix 1).

The head of the quadrate is well preserved in both the right and left quadrates of YPM 1775 (fig. 27). The otic capitulum (articulating with an “otic cotyla”, or facet, variably of prootic or opisthotic;Baumel and Witmer, 1993) is more globose and projects farther dorsally than the squamosal capitulum. As noted byWitmer (1990) forHesperornis regalis and figured forIchthyornis (Witmer, 1990), the otic and squamosal capituli are distinct, separated by a nonarticulating shallow incisure. As also noted byWitmer (1990), this condition is distinct from that seen in extant Palaeognathae, in which there is no such incisure and the capituli are confluent. Both of these conditions are distinguished from the lack of an otic articulation, or “single headed” condition (Witmer, 1990), primitive to theropod dinosaurs.

The squamosal capitulum extends ventrally in a narrow ridge down the lateral edge of the otic process of the quadrate (fig. 27B). This feature is slightly broken on the left quadrate but is clearly seen on the right. The capitulum and ridge project slightly anteriorly to overhang the body of the otic process (fig. 27A,G). The otic capitulum has a slight anterodorsal indentation or pit (fig. 27G). A similar pit was described on the head of the quadrate ofPotamornis skutchi near the medial edge of its dorsal surface (Elzanowski et al., 2000). It was interpreted as an autapomorphy of that taxon and was considered to possibly delineate edges of the capituli that otherwise appear confluent in that taxon. A similar feature inIchthyornis dispar in approximately the middle of the prootic capitulum suggests that this feature should not be taken to indicate the edge of a capitulum or to be homologous with the intercapitular incisure, because this feature co-occurs with an indentation inIchthyornis.

The body of the otic process is only slightly narrower mediolaterally than the width of the head (fig. 27A). The process is virtually straight, although it bows slightly anteriorly close to its base. It is virtually uncontoured, lacking any muscular impressions (e.g., impressio medialis;Elzanowski et al., 2000) or processes (e.g., eminentia articularis;Lowe, 1926). The orbital process begins to rise below the head as a narrow ridge (fig. 27G). The dorsal portion of this low ridge of the process is better preserved on the right quadrate, which can be compared with the left to observe its relative extent down the body of the quadrate. The ventral portion of the process is missing. Posteromedial to the process, a slight depression is developed that deepens ventrally (fossa basiorbitalis;Elzanowski et al., 2000). At the broken base of the process is the large pneumatic foramen commented on above; a flat, dorsally facing surface lies between it and the pterygoid condyle (fig. 27A). The edges of the pterygoid articular surface are indistinct.

The pterygoid condyle is round, with a slightly constricted base (fig. 27A). It projects anteromedially well past the medial condyle and there is a cleft between it and the medial condyle. InIchthyornis dispar, the pterygoid condyle is not an extension of a facet with a flat to concave dorsal surface and a projected edge (also a “pterygoid condyle” sensuElzanowski et al., 2000), but rather it is a hemispherical, knoblike process (seeappendix 1, character 32).

The maximum width of the ventral quadrate, with its condyles for articulation with the mandible, is more than half the maximum dorsal/ventral height. This finding is contraElzanowski et al. (2000), who claimed the width of the mandibular articulation was less than half the height of the quadrate, and considered that condition to be synapomorphic of a monophyletic “Odontognathae”, includingIchthyornis and Hesperornithes. This condition, by contrast, appears to be primitive to at least Avialae (seeConfuciusornis sanctus,Chiappe et al., 1999).

Ichthyornis dispar (fig. 27E) andHesperornis regalis have a medial condyle, with its long axis extending posterolaterally. In Palaeognathae and Galloanserae, it is the lateral condyle rather than the medial that angles posteriorly. The gain of a posteriorly projected lateral versus medial condyle may be a synapomorphy of Aves relative toIchthyornis.Elzanowski et al. (2000) considered the condition inHesperornis regalis andIchthyornis to be a state in which the medial condyle is continuous with the “caudal condyle” and considered this morphology synapomorphic of a monophyletic Odontognathae. However, a posterior projection of the medial condyle is also present inConfuciusornis sanctus. Thus, this morphology does not appear synapomorphic but symplesiomorphic within Avialae, and it may be present in taxa still deeper within Theropoda.

It seems inappropriate to call the posterior continuation of single medial condylar surface a “caudal condyle”. The presence of a caudal condyle has been considered a synapomorphy of Aves relative toIchthyornis (e.g.,Cracraft, 1986). In palaeognaths and most neognaths (but not Galloanserae; reviewed inCracraft and Clarke, 2001), the mandibular articulation of the quadrate has a triangular aspect in ventral view, with a posterior articular surface developed that is variably distinct.Elzanowski et al. (2000) describeIchthyornis andHesperornis regalis as having such an apomorphic posterior articulation and triangulate articular surface.

The highly recurved and large medial condyle inHesperornis regalis is strikingly different from the flat and relatively shorter condyle inIchthyornis dispar (fig. 27E). The ventral surface of the quadrate inHesperornis regalis andConfuciusornis sanctus (Chiappe et al., 1999) could be described as “triangulate” (Elzanowski et al., 2000; with closely placed condyles and a posteriorly directed medial condyle). By contrast, the broad, shallow separation between the medial and lateral condyles (sulcus intercondylaris;Baumel and Witmer, 1993), and the preserved morphology of the medial condyle (fig. 27A,E) do not confer a triangulate aspect on the ventral surface of theIchthyornis dispar quadrate (fig. 27A,E).

Given the extensive range of conditions in Aves, which includes many cases in which the caudal condyle is virtually indistinguishable from the medial or, alternatively, the lateral condyle, or in which the caudal condyle is apparently secondarily lost, it is reasonable to presume thatIchthyornis andHesperornis regalis might represent such a case outside of Aves. However, it is a less supported explanation given that the medial condyle inConfuciusornis sanctus,Ichthyornis, andHesperornis regalis is a single surface with such a posterior component and no nonavian taxa are known to have a distinct caudal condyle. Additionally, none of the preserved mandibular fragments that are a part of the YPM material (i.e., 1450, 1761, 6264) indicate the presence of a distinct posterior facet (fig. 28). Neither in theseIchthyornis dispar exemplars nor inHesperornis regalis (contraElzanowski et al., 2000) is there any indication that the slight posterior component of the medial articulation is a separate articulation or condyle.

The quadratojugal articulation is shallow and dorsal to the lateral condyle (fig. 27B). This articulation has also been described as shallow in Hesperornithes andPotamornis (Elzanowski et al., 2000), but it is deeper in those taxa than inIchthyornis dispar orConfuciusornis sanctus (Chiappe et al., 1999). The pit for the quadratojugal articulation also appears deeper inPatagopteryx deferrariisi (Chiappe, 1996) than inIchthyornis dispar. Furthermore, inHesperornis regalis andPotamornis, this pit lies close to the lateral condyle (Elzanowski et al., 2000), while it is distinctly dorsal to this condyle inIchthyornis dispar.

Mandible

Portions of the mandible are represented in the holotype ofIchthyornis dispar (YPM 1450) and in referred specimens YPM 1735, YPM 1749, YPM 1761, YPM 1775, and YPM 6264, which are referred on the basis of apomorphy or morphological correspondence with the holotype (seetable 1). The right mandible of YPM 1450 was mounted in theIchthyornis dispar panel mount as the left (fig. 19A; as noted inGregory, 1952, and inGingerich, 1972). The partial left mandible from YPM 1749 was mounted in theIchthyornis victor mount as the left maxilla (fig. 19A). It was identified as part of YPM 1749 with reference to Gibb's notes (fig. 3) and description of the material inOdontornithes (Marsh, 1880: 124–125). The mandible of YPM 1761 (fig. 28A) is larger than that of the holotype (YPM 1450) or YPM 6264 (fig. 28B).

Marsh (1880) commented on very few aspects of mandibular morphology. He considered there to be no ossified mandibular symphysis inIchthyornis and no “distinct symphysial surface” (Marsh, 1880: 123). He described the rami of the jaws as nearly straight, large and massive, and compressed mediolaterally, noting that nearly all sutures were closed with the exception of the splenial-angular intramandibular joint (Marsh, 1880: 123). Finally, he commented that the upper margin of the dentary was straight and that there was no mandibular foramen and no retroarticular process (i.e., that the jaw was “truncated” directly behind the articulation of the quadrate;Marsh, 1880: 123). As is further elaborated below, this analysis confirmed the lack of an ossified symphysis and several characters of the jawMarsh (1880) described. Visible sutures in the jaw and a small mandibular fossa (contraMarsh, 1880) and previously undescribed morphologies are treated below.

Gregory (1952) was the only other worker to figure the jaw, hypothesizing the identity and topological relations among the bones that comprise it.Gingerich (1972) commented on the reconstruction byGregory (1952) and discussed additional details of the articulation with the quadrate based particularly on a newly uncovered specimen (YPM 6264) that, among the YPM material, best preserves the posteriormost portion of the jaw (Gingerich, 1972).Figure 19B,C, andfigures 28–32 compare representations and reconstructions of the jaw inOdontornithes (Marsh, 1880;fig. 19B), byGregory (1952;fig. 19C), and by this analysis (figs. 28–32).

The posterior part of the mandible and the cotylae for the quadrate are represented in YPM 1450 (right side), YPM 1761 (fig. 28A), and YPM 6264 (fig. 28B).Gingerich (1972) described this portion of the jaw in detail, and the present analysis largely agrees with his description. What was named the “anterior cotylus” byGingerich (1972) corresponds to the lateral cotyla of other authors (cotyla lateralis;Baumel and Witmer, 1993) and the “posterior cotylus”, to the medial (cotyla medialis;Baumel and Witmer, 1993). The quadrate cotylae appear as two posterolaterally angled grooves and are best preserved in YPM 6264. Directly posterior to the medial cotyla is the pneumatic foramen (fig. 28) that was commented on byGingerich (1972) andWitmer (1990). The form of this feature can also be seen, undistorted, in YPM 1450; it is ovate and sits in a triangular depression. In YPM 1761, the morphology of the fossa is distorted by crushing (fig. 28A), which accounts for the interpretation of it as “oblong” inElzanowski et al. (2000: 718).

Medial to this foramen and directly posterior to the edge of the medial cotyla in YPM 6264, the entrance and exit of the n. corda tympani (Oelrich, 1956) are visible in what would be part of the prearticular, although sutures between this element and the articular are not visible. The entrance is located on a slight dorsal prominence medial to the foramen and the exit (anteroventral to it) is visible in medial view close to the posterior edge of the cotyla. This feature is infrequently preserved and highly variable in Aves. While the passage of the n. corda tympani was demarcated in one specimen ofChauna torquata (Anhimidae), it was not visible in a specimen of the closely relatedAnhima cornuta (also Anhimidae). It is much more conspicuous inGavia stellata, where the anterior passage of the nerve could be traced for much of the length of the posterior mandible. By contrast, it was not demarcated in the exemplars of palaeognaths and galliforms surveyed (see Comparative Materials and Methods). The groove demarcating the passage for the nerve does not appear developed in the holotype ofIchthyornis dispar (YPM 1450) or in YPM 1761. In YPM 1450, however, the relevant area of the posterior mandible is not well preserved.

The morphology of the lateral cotyla is partially obscured by crushing in all three specimens. This cotyla is best preserved in YPM 6264 (fig. 28B), where its depth and medial extent are visible. Both of these dimensions of the lateral cotyla inIchthyornis dispar are greater than previously illustrated. YPM 1761 was illustrated (Elzanowski et al., 2000) as virtually uncrushed; it is, however, strongly crushed dorsoventrally (fig. 28A). The lateral cotyla is depicted as a shallow, round facet restricted to the lateral edge; however, there is a fracture that runs anteroposteriorly through the specimen from the posterolateral edge toward the midline (fig. 28A). It is this fracture and crushed bone that form the artifactual medial terminus of the cotyla illustrated inElzanowski et al. (2000). The area of the anterior edge of the medial cotyla and what is visibly part of the lateral cotyla is also obliterated by damage, explaining the illustration of YPM 1761 inElzanowski et al. (2000) in which the medial condyle appears smaller and shallower than in the condition visible in YPM 6264 (fig. 28B) and to some degree in YPM 1450.

A posterior prominence at the lateral terminus of the medial cotyla has been dubbed the articulation of a third condyle of the quadrate (Elzanowski et al., 2000). As discussed above with reference to the morphology of the quadrate,Ichthyornis dispar lacks a distinct third or “caudal” condyle, as is plesiomorphic for Avialae. There is a slight, posteriorly facing extension of the medial condyle, and it would have articulated with this prominence. However, as discussed above, to consider this feature the articulation of a distinct “caudal condyle” seems unjustified. This feature is considered simply the posterolateral edge of the medial cotyla. The posterior mandibular midline is broken in YPM 1450 and YPM 6264. In both of these specimens crushing was mediolateral, creating a central break and/or ridge.

As previously noted (Marsh, 1880;Gregory, 1952;Gingerich, 1972),Ichthyornis dispar, unlikeHesperornis regalis, lacks a retroarticular process. A short medial projection, or medial process (processus medialis mandibulae,Baumel and Witmer, 1993), is developed on the posterior end of the articular (Gingerich, 1972;fig. 28). The posterior surface of the mandible is slightly concave (fig. 28; shallow fossa caudalis,Baumel and Witmer, 1993).

Most of the contacts between the elements of the posterior mandible (i.e., posterior to the last alveoli) are best represented in YPM 1450 (figs. 29,30) and YPM 6264, with a portion of the angular/surangular and angular/prearticular contacts also visible in YPM 1775. Portions of the angular, surangular, and prearticular are preserved in YPM 1761, but the sutures are all but completely closed. YPM 1775 preserved the contour of the posterodorsal margin of the dentary. YPM 1761 is larger than YPM 1450, YPM 1775, and YPM 6264. In YPM 1450, the morphology of both the right and left jaws is comparable; the left jaw, however, was depressed by crushing, especially at the posterior terminus of the preserved fragment (figs. 29,30). Similarly, compression slightly exaggerated the height of the right jaw.

Gingerich (1972) noted that the angular forms the ventral edge of the mandible posterior to the angular/splenial contact (intramandibular joint).Gingerich (1972) further described that at this contact, the angular has an articular facet for the splenial; the surangular forms more than one-half the height of the lateral mandibular surface posterior to its splenial contact; the prearticular extends between the splenial and the dentary on the dorsal edge of the jaw. These morphologies were confirmed in YPM 6264 and observed in the holotype ofIchthyornis dispar (YPM 1450;figs. 29,30).

The posterodorsal splenial margin forms a smooth arc from close to the last tooth to the intramandibular joint (figs. 29,30). Just anterior to this joint, the splenial passes laterally, making up the ventral margin of the jaw. On the lateral surface of the left jaw, 1 mm from the ventral edge, approximately 3 mm of the splenial/dentary suture is visible passing anteriorly, roughly subparallel to this ventral margin. The anterior angular margin is slightly cupped ventrally to receive the splenial (fig. 30).

The posterior contacts of the dentary are best preserved on the right jaw of YPM 1450. In lateral view, the dorsal surface of the dentary arcs from its ventral contact with the splenial to the tooth row. Just ventrolateral to the alveolar margin, the dentary receives a short, narrow tongue of the surangular (fig. 30). In Aves, this process of the surangular can be much more extensive, often extending anteriorly for approximately one-half of the rostrum (seeSanz et al., 1997, for a discussion of the distribution of this condition). On the left jaw, the anterior process of the surangular appears slightly longer than on the right jaw; however, this appears artifactual. A portion of this contact is more visible in YPM 1775; in medial view, the dorsal edge of the dentary in both jaws visibly curves posteroventrally. The exposed medial surface of this portion of the dentary bears the impressions of several incompletely formed alveoli: Two faint ridges separate three indentations that are approximately the dimensions of the alveoli that are completely formed in the anterior part of the tooth row (fig. 30C). The more posterior of these is most weakly developed. In the left jaw (that is not compressed significantly mediolaterally) two ridges are also visible with three alveolar impressions, and three corresponding teeth are preserved. The most posterior of these teeth is least erupted, and the posterior terminus of its incompletely formed alveolus and the apparent terminus of the dentary itself do not appear artifactual.Gregory (1952) considered the dentary to extend significantly posterior to the preserved portion of the prearticular; however, the dentary terminates anterior to the splenial/angular contact as described below.

A large coronoid was described for the holotype ofIchthyornis dispar (YPM 1450) byGregory (1952).Gingerich (1972), however, saw no sign of a coronoid bone in the holotype (YPM 1450). This analysis confirms the presence of a coronoid inIchthyornis dispar (e.g., YPM 1450), although it differs considerably fromGregory's (1952) illustrations (comparefig. 19C withfigs. 29 and30). In the left mandible of YPM 1450, a suture is visible between the posterior end of the dentary and part of element dorsal/posterodorsal to it (fig. 30C). Between this element, identified as the coronoid, and the surangular laterally, another suture is visible (in dorsal view;fig. 30B). The posterodorsal contact between the dentary and the coronoid is curved dorsoventrally. The shape of this contact can also be seen in both the right mandible of YPM 1450 and in YPM 1749. Posterior to the end of the dentary, the coronoid has an open sutural contact with the prearticular ventrally and the continuation of its contact with the surangular laterally (fig. 30). The coronoid is also visible continuing just posterior to a contact between the prearticular and the surangular, which is an artifact of the compression of this mandible.

The position and contacts of the coronoid in the right jaw of YPM 1450 are the same as those preserved in the left. In the right jaw, a small, dorsoventrally flat facet that would have been occupied by the coronoid is preserved just dorsal to the posterior end of the dentary (fig. 29). The facet lies in the same position occupied by the anterior part of the coronoid in the left jaw and is formed of an artifactually exposed surface of the surangular contact. This exposed surface is continued posteriorly as an open suture (visible in dorsal view) between the surangular and coronoid, which is exposed as a diminutive tongue of bone (fig. 29B). A brief, tapering continuation of the coronoid is faintly visible continuing on the other side of a conspicuous fracture crossing the mandible (fig. 29). The coronoid/prearticular suture appears closed at the coronoid's posterior terminus.

Gregory (1952) described a massive coronoid with some of the same contacts described above. He described a large fragment of bone supposed to lie on the medial surface of the right jaw of YPM 1450 and considered this to be a part of the coronoid (fig. 19C). No such large fragment of bone was observed in the specimen removed from the mount, and it seems that what was represented was merely a portion of the prearticular anterior to the shallow mandibular fossa (fig. 29).

Ventral to the posterior end of the coronoid in YPM 1450, the surangular/prearticular contact is visible. Just posterior to this contact is the anterodorsal margin of a mandibular fossa (i.e., fossa aditus canalis mandibulae;Baumel and Witmer, 1993;fig. 29). A portion of the anteroventral margin of this fossa can be discerned in YPM 1761, YPM 1775, and YPM 6264. It appears to be formed of the prearticular (compare YPM 1775, where some of the prearticular/angular/surangular contacts remain open, and YPM 1761 in which the sutures appear closed; YPM 1761 is also the largest mandible of the YPM specimens). The lateral and dorsal edges of this fossa appear to have been formed of the surangular.

In YPM 6264, a large foramen is visible in medial view near the ventral edge of the preserved portion of the surangular and at close to the midpoint of the mandibular fossa. This foramen was not described byGingerich (1972); it is directed posteriorly with no visible exit on the preserved portion of the lateral surface. A somewhat similar configuration was noted inCrypturellus undulatus, where foramina in the surangular are directed both anteriorly and posteriorly, and only the anterior of which appears to penetrate to the lateral surface. The feature probably represents a branch of fifth cranial nerve (trigeminal), possibly the n. intramandibularis (Dubbeldam et al., 1993). This foramen could not be seen in YPM 1450, where there is considerable breakage in this region. It is also possible that the prearticular could have covered the foramen medially, as it is posterodorsally incomplete in all specimens.

The posterior extent and contacts of the surangular, prearticular, angular, and articular are not well preserved in any of the specimens. However, the sutures between the angular/surangular, angular/prearticular, and angular/articular appear closed posteriorly (YPM 1450, YPM 1761, YPM 1775, YPM 6264). The anterior portion of the angular/ surangular contact appears to be open in YPM 1775, and it may also be open in the smaller holotype ofIchthyornis dispar (YPM 1450). The angular extended dorsally approximately a third of the height of the posterior mandible in lateral view. Medially, the prearticular angles toward the ventral margin. There may be a slight indication of this contact in the right jaw of YPM 1450, anterior to the medial cotyla close to the ventral margin of the jaw (fig. 29). No suture between the prearticular and angular was discernable in YPM 6264, thoughGingerich (1972) described the angular tapering to a point ventral to the articular cotylae for the quadrate.

The dorsal surface of the surangular is unbroken in YPM 1450, and it suggests that no coronoid process (processus coronoideus;Baumel and Witmer, 1993) was developed. The dorsal edge of the posterior mandible rises slowly from the level of the articular fossa and is highest just posterior to the level of the angular/splenial contact. No pseudotemporal tubercle (tuberculum pseudotemporale;Baumel and Witmer, 1993) is visible.

The dentigerous portions of both dentaries are well preserved in YPM 1450. Both right and left are complete with the exception of the anteriormost tip of the left, which is slightly abraded. The total length of the right jaw in YPM 1450 is 67 mm. The socketed tooth row extends 41 mm of this length (from posterior edge of last tooth to tip of dentary) confirming Marsh's measurements (Marsh, 1880: 125). Portions of the anterior jaw are additionally represented in YPM 1735 (fig. 31A), YPM 1749, and YPM 1775 (fig. 32). Anterior to the intramandibular joint, the dentary and splenial are completely fused in all specimens except for the short open suture on the lateral surface of left jaw of YPM 1450.

The splenial inIchthyornis was described as making up the medial margin of the last four alveoli and then angling ventrally (Martin and Stewart, 1977; SMM “13520” = SMM 2503, YPM 1450). However, the location of the dorsal dentary/splenial contact is not conspicuous in any YPM material. A series of small foramina on the left jaw of YPM 1450 (approximately one per alveoli) lie close to the posteriormost complete socket and move slightly ventrally relative to the alveoli for each of the three anteriorly adjacent sockets. These foramina could demarcate the dorsal edge of the splenial in this area. In YPM 1735, five small foramina are aligned subparallel to the posterodorsal margin of the preserved fragment (which does not include the posterior terminus of the dentary;fig. 31A). The anterior two are positioned slightly ventral to the other three (fig. 31A). The line of foramina does not appear to extend further forward. It is also possible that these foramina are the same as those in YPM 1450, and demarcate a portion the dorsal splenial/dentary contact. That they are only developed posteriorly, and not along the entire tooth row, suggests that they may not be related to dental succession or replacement.Martin and Stewart (1977) reported that in a large specimen ofIchthyornis they referred to as SMM “13520” (which is here found to be SMM 2503 and referred to as such in this document; see Taxonomic Revision), the ventral splenial suture was obliterated but marked by microscopic foramina. No such foramina were observed in the YPM material. Further, there is no apparent reason that the splenial/dentary contact should be indicated by foramina, as this is not its typical development in Aves, or elsewhere.

The position of the anterior terminus of the splenial is unclear.Norell and Clarke (2001) considered the splenial to extend to the tip of the mandible and to be involved in the symphysis. The smooth anteromedial surface of the jaw and the absence of a conspicuous Meckel's groove (fig. 32A,C) suggested this scoring forIchthyornis.Gregory (1952) inferred that Meckel's groove was manifest as a short groove passing from a large foramen close to the tip of the jaw in YPM 1450 (fig. 19C) and what he interpreted as the “symphysial area” (Gregory, 1952: 78). He inferred that the splenial terminated at this foramen, though he could discriminate no suture (Gregory, 1952).Martin and Stewart (1977) also considered the splenial to end at a large foramen, which they described as 19 mm from the tip of the mandible (fig. 32A,C). Meckel's groove was interpreted as terminating only a few millimeters anterior to the foramen (Martin and Stewart, 1977). However, the interpretation of the position of the large anterior foramen as demarcating the anterior tip of the splenial seems unjustified because in Aves (e.g.,Struthio camelus,Gavia stellata), the splenial often continues along the ventral margin anterior to the opening of Meckel's canal (where it ceases to cover the cartilage) to extend into the symphysis.

A large, anteriorly opening foramen approximately 19 mm from the tip of the jaw and a shallow groove anterior to it (fig. 31B) are visible on the left mandible of theIchthyornis dispar holotype (YPM 1450). These features appear to correspond with the groove and posterior foramen previously discussed (Gregory, 1952;Martin and Stewart, 1977). In the right jaw of YPM 1450, a foramen that may correspond with the anterior foramen lies 8.4 mm from the preserved end of the mandible, but no groove is discernable posterior to it.

The anterior portion of the groove and its termination at a second foramen are better preserved in YPM 1775, where this second foramen lies dorsal to a notch in the ventral margin of the mandible, 7.1 mm from its tip (fig. 32A,C). In YPM 1775, this anterior foramen appears to enter the medial surface of the mandible at a high angle to the axis of the groove (fig. 32A,C). It is possible that this feature is better construed as the canal for the internal mandibular artery rather than related to the passage of Meckel's cartilage, which does not curve abruptly at any point.

In sum, the posterior medial foramen that opens anteriorly may represent the anterior opening of Meckel's groove, consistent withGregory (1952) andMartin and Stewart (1977). If so, it then seems that this groove was very shallow and that Meckel's cartilage ran superficially along the medial surface of the jaw anteriorly. This superficial development was observed in a cleared and stained juvenile chicken (Gallus gallus, YPM 14517). However, is not clear that the anterior foramen represents the cartilage's exposure in the “symphysial area”, as suggested byGregory (1952); the foramen perforates the jaw at a high angle relative to the axis of the groove. Finally, even if the posterior foramen represents the anterior opening of Meckel's canal, there is no necessary relationship between this opening and the termination of the splenial; as noted in Aves, a ventral tongue of the splenial often continues anterior to the opening of Meckel's canal.

Martin and Stewart (1977) described seven mental foramina on the lateral surface of the right dentary (YPM 1450), which were not indicated byMarsh (1880). A series of six large foramina was indicated byGregory (1952) for the left jaw. On the right jaw of YPM 1450, six large and two small foramina are visible. On the anteriormost tip of the jaw (YPM 1450, YPM 1775), a large terminal foramen is developed (fig. 32B,D). The most posterior of the mental foramina was described as giving rise to a broad canal that widens posteriorly (Martin and Stewart, 1977).Gregory (1952) considered this feature artifactual, while the current analysis agrees withMartin and Stewart (1977) in considering this depression to be real. The lateral groove associated with the mental foramina seen on both the right and left mandibles of YPM 1450 is less pronounced in YPM 1775. The mental groove lies at the apex of the dorsoventrally convex lateral surface. In YPM 1775, another foramen is located ventral to the second mental foramen, just dorsal and anterior to the notch in the ventral margin.

A predentary bone, or an intersymphysial ossification, lying between the anterior tips of the dentaries (Martin, 1987), was reported forHesperornis regalis (KUVP 71012) andParahesperornis alexi (KUVP 2287), and inferred, from the shape of the mesial termination of the dentary, to be present inIchthyornis (Martin, 1987). There is no evidence in any YPMIchthyornis specimen of a predentary bone, nor of terminal facets on the dentariesMartin (1987) described as articulating with the predentary bone in Hesperornithes.

InIchthyornis, a discrete area of textured bone lies on the medial surface of the tip of the mandible in YPM 1775 (fig. 32A,C). It is interpreted as the limited area of fibrous attachment to the corresponding tip of the opposing mandibular ramous. The rami would have met at a very low angle with their symphysis developed as a restricted, primarily dorsoventral, strip of contact between the rami. The conformation of the attachment is not unreasonably compared to the small contact inSterna maxima (ossified as in all Aves) in which the area of contact at the mandibular tips is also very limited and primarily dorsoventral with almost no ventral floor. Thus, it does not appear necessary to hypothesize a predentary bone to explain the shape of the distal terminus of the preserved mandibles inIchthyornis.

Just posterior to the textured medial symphysis (YPM 1450, YPM 1775), a small foramen is visible (fig. 32A,C). A second small foramen lies in this area of textured bone in YPM 1775 that is not visible in YPM 1450 (fig. 32A,C).

Dentition

Neither the illustration of a tooth from YPM 1450, the holotype ofIchthyornis dispar, inMarsh (1880: pl. XXI) nor the reconstruction of a tooth ofIchthyornis inMartin and Stewart (1977: fig. 2) closely approximate the actual shape of the teeth inIchthyornis. The illustration inMarsh (1880) approximates only the tip rather than the complete tooth crown. Only portions of the roots and cementum are preserved in YPM 1749, and no teeth are preserved in YPM 1735, where only the shape of the empty alveoli can be observed.

Marsh commented (1880: 124) thatIchthyornis had socketed teeth with pointed, strongly recurved, compressed crowns that lacked serrations.Gregory (1952) noted that the tooth crowns were compressed, with recurved tips with sharp anterior and posterior cutting edges.Martin and Stewart (1977) reiteratedGregory's (1952) observation that the crowns were compressed and recurved posteriorly.Martin and Stewart (1977) further describe the tooth crowns as both medially and laterally convex, and that anterior edges of the teeth had a distinct shoulder that exaggerates the teeth's posteriorly recurved appearance. The teeth were supposed to rapidly expand as they entered the jaw and radiograms are cited as suggesting that the bases of the teeth may have inclined posteriorly in the tooth row to further exaggerate their recurved appearance (Martin and Stewart, 1977). Evaluation of the morphologies of YPM 1450 (fig. 31B,C) and YPM 1775 is compatible with the teeth indeed being pointed, posteriorly recurved, lacking serrations, and mediolaterally compressed with trenchant anterior and posterior edges.

In the six posteriormost teeth of the left mandible of YPM 1450, most aspects of dental morphology can be observed, although all aspects are not preserved in any single tooth portion. The first and sixth tooth portions (fig. 31C) preserve the tip of a striated and recurved tooth crown. The posterior margin of the crown is smoothly arched, rather than having the relatively straight posterior margin illustrated inMartin and Stewart (1977). The crown of the sixth exposed tooth is visibly more elongate and narrower (fig. 31C) than the stumpy morphology these authors illustrate (Martin and Stewart, 1977: fig. 2d).

In YPM 1775, a carina and a slight indentation at the anterior base of a tooth crown are visible. The teeth ofIchthyornis (crown and root) are much more compressed than illustrated inMartin and Stewart (1977: fig. 2) and lack the shoulder at the base of the crown seen in crocodilians that these authors illustrate. The root is ovoid and slightly wider that the base of the crown. The expansion interpreted byMartin and Stewart (1977) could be a result of accreted cementum. They also interpret the one exposed tooth inHesperornis regalis (YPM 1206) as having an enormously expanded root and this also appears to be cementum.

In a more recent publication (Martin and Stewart, 1999) addressing tooth implantation in avialans, complete nonavialan theropod morphologies are compared with broken teeth fromParahesperornis alexi andAlligator (Martin and Stewart, 1999: fig. 2). In fact, consideration of the YPM material suggests that the shape of the tooth crown inIchthyornis (e.g., YPM 1450) is much closer to nonavialan theropod morphologies than is suggested byMartin and Stewart (1999: fig. 2). Crocodilian teeth lack the strong mediolateral compression present inIchthyornis and other toothed theropods. There is also a distinct shoulder present, or a constriction ringing the tooth crown at its base, developed in crocodilians, that is not developed inIchthyornis (e.g., YPM 1450).

Cervical Vertebrae

Atlas: YPM 1733

The atlas is preserved only in YPM 1733 (fig. 33B) and discussed here as “1733A”. It is not currently as complete as its illustration inOdontornithes (Marsh, 1880: pl. XXXVII, fig. 1). The ventral portion (corpus atlantis;Baumel and Witmer, 1993) is preserved only on the right side. The dorsal portion (arcus atlantis;Baumel and Witmer, 1993) is currently incomplete (fig. 33B). It is unclear if the right and left laminae of the neural arch were fused on the dorsal midline as depicted. A portion of a left postzygapophysis was preserved in a separate vial and was not included in theIchthyornis victor panel mount. This process, if it belongs to the atlas (as labeled), is considerably more robust than in the Aves surveyed.

The atlas preserves a strikingly well-developed epipophysis that projects dorsal to the postzygapophysial articular surface (processus articularis caudalis;Baumel and Witmer, 1993) and extends as far as the posterior edge of this surface (Marsh, 1880: pl. XXXVII, figs. 1, 1a). A large atlantal hypapophysis is developed between two ventral tubercles (Marsh, 1880: 128,fig. 33B). Neither well-developed epipophyses or hypapophysis was observed in tinamous (e.g.,Nothura darwinii orCrypturellus noctivagus) or inGallus gallus orAnas platyrhynchos. InApteryx australis, comparatively well-developed epipophyses are present but no robust hypapophysis. By contrast, inStruthio camelus, a moderately well-developed hypapophysis is present but there is no dorsal projection of the epipophysis. Both pronounced epipophyses and a hypapophysis were observed, for example, inGavia stellata, although these are less well developed than those ofIchthyornis dispar (Marsh, 1880: 128).

Axis: YPM 1733

A nearly complete axis is represented in YPM 1733 (fig. 33B) and referred to here as “1733B”. An incomplete axis is preserved in YPM 1775 (fig. 33A) identical in morphology with that preserved in YPM 1733 (Marsh, 1880). However, in YPM 1775, the suture between the first (atlantal) centrum and the second (axial) centrum is incompletely closed (fig. 33A). This suture is closed in the slightly smaller YPM 1733 (fig. 33B). Both of these specimens are larger than the holotype ofIchthyornis dispar (see further discussion of this suture as a possible sign of the relative immaturity of YPM 1733 in the Taxonomic Revision).

The description provided inOdontornithes (Marsh, 1880) accurately represented the morphology of this element in these specimens. However, the illustration of this element (Marsh, 1880: pl. XXVII, fig. 1) is not accurate in its depiction of the bone as uncrushed. The neural arch is crushed (fig. 33B) to close the neural canal (foramen vertebrale;Baumel and Witmer, 1993). Matrix between the collapsed neural arch and the centrum indicates that this breakage is original to the specimen. Thus, the dimensions of the neural canal figured byMarsh (1880) were estimated, not observed.

The tip of the prominent hypapophysis is missing in both specimens. The anterior articular surface is exposed in YPM 1775 (fig. 33A) but covered by the atlas (fig. 33B) in YPM 1733 (Marsh, 1880). It matches its description inOdontornithes (Marsh, 1880: 128, 129). The paired pneumatic foramina noted (Marsh, 1880) are striking in their size (fig. 33B). These are not present in the axis ofHesperornis regalis; however, many pneumatic features are suppressed in diving taxa (Britt et al., 1998). They are located in the lateral surface of the neural arch laminae posterodorsal to paired protuberances on the centrum. These protuberances represent the capitula of fused cervical ribs in Aves (processus costalis;Zweers et al., 1987).

As illustrated inOdontornithes (Marsh, 1880), the epipophyses (YPM 1733) just surpass the postzygapophyses in distal extent and do not project significantly dorsally but rather are curved slightly ventrally (fig. 33B). They taper to a narrow ridge dorsally.Marsh (1880) correctly noted that they are unlike those ofSterna maxima. In that taxon, blunt processes project posterodorsally approximately the height of the posterior articular surface. Such a morphology (short, slightly ventrally projected epipophyses) is approached, for example, inNothura darwinii (Tinamidae).

The posterior articular surface the axis centrum is strongly compressed and extends slightly around the posterolateral centrum edges (YPM 1733 and YPM 1775;fig. 33A,B).Marsh (1880) considered the axis to show a kind of incipient heterocoely, and, of all preserved cervicals, it approaches this condition most closely. The lateral extension of the surface appears to approach the mediolateral convexity seen in heterocoelous vertebrae (see definition inappendix 1, character 52). However, the center of the surface is slightly depressed rather than broadly convex (fig. 33A). InHesperornis regalis and Aves, by contrast, a distinct ventral lip forms the edge of a conspicuously dorsoventrally concave surface. Also, in the heterocoelous vertebrae ofHesperornis regalis and Aves, this conformation gives the posterior articular surface the appearance of broadening laterally. This condition is not developed in the axis (fig. 33), third cervical vertebra (fig. 34), or in more posterior cervicals. Further comments on “heterocoely” are made below, as well as inappendix 1 (character 52).

Third Cervical Vertebra

Posterior Cervical Vertebrae

Thoracic Vertebrae

Posterior Thoracic Vertebrae

Three posterior thoracics lacking hypapophyses and lateral tubercles (tuberculi laterale corporis;Zweers et al., 1987) with deep, oblong lateral excavations are part of YPM 1733. A second specimen, YPM 1732, also includes three such vertebrae as well as an additional vertebra incompletely fused to the anterior end of the sacrum, whichMarsh (1880) considered a thoracic. The morphology of the vertebrae from these specimens corresponds well to each other and to their description inOdontornithes (Marsh, 1880). All but one of these vertebrae (YPM 1732, YPM 1733) were incorporated into theIchthyornis victor panel mount; they are depicted schematically in the reconstruction ofIchthyornis victor (Marsh, 1880: pl. XXXIV). This illustration gives the misleading impression that all the thoracic vertebrae lacked hypapophyses. However, as discussed, several anterior thoracic vertebrae have hypapophyses.Ichthyornis dispar andHesperornis regalis, like Aves, have hypapophyses on vertebrae of several of the anteriormost thoracics vertebrae (in addition to the posterior cervicals). To have hypapophyses on these vertebrae may be synapomorphic of at leastHesperornis regalis + Aves, and perhaps for a more inclusive avialan clade (seeappendix 1).

Two of the three posterior thoracic vertebrae from YPM 1732, as well as of those from YPM 1733, are approximately the same length as 1733G; their centra are longer than wide. However, one thoracic vertebra from each specimen is shorter than these other two posterior thoracics.Marsh (1880) identified these short vertebra as the 20th in the series, with the 21st, and last, presacral vertebra incompletely incorporated into the sacrum of YPM 1732, which was preserved with the last two thoracics in articulation with the sacrum (Marsh, 1880).

In YPM 1732, the last thoracic as well as the first (but incompletely ankylosed) sacral are shorter and broader than those lying anterior to them (fig. 40D). InSterna maxima andVanellus melanopterus, a single vertebra, shorter than those immediately anterior to it, is incorporated into the sacrum. Thus, only one short vertebra with the morphology of a thoracic is incompletely ankylosed to the sacrum in these charadriiforms. InIchthyornis, however, one incompletely ankylosed vertebra, with the morphology of a thoracic, and one free thoracic are short (YPM 1732). The lateral excavations of these shorter vertebrae inIchthyornis are ovoid (fig. 40).

Sacral Vertebrae

The sacrum ofIchthyornis is represented in three YPM specimens: the holotype ofIchthyornis dispar (YPM 1450), YPM 1733, and YPM 1732. The sacra of the holotype ofIchthyornis dispar and YPM 1732 were figured inOdontornithes (Marsh, 1880). While the illustration of theIchthyornis dispar sacrum is largely accurate (Marsh, 1880: pl. XXVII, figs. 5–7), that of YPM 1732 as representative ofIchthyornis victor (Marsh, 1880: pl. XXXII, figs. 2, 3) differs significantly from the actual material. The extremely poorly preserved partial sacrum in YPM 1733 was not figured or discussed, although many presacral vertebrae from that specimen were figured and incorporated into theIchthyornis victor panel mount.

Ten fused vertebrae comprise the sacrum of the holotype ofIchthyornis dispar (YPM 1450;fig. 41A,C). As illustrated byMarsh (1880) it is strongly crushed dorsoventrally (fig. 41C). The description of the sacrum inOdontornithes (Marsh, 1880) is also largely accurate. The anterior articular surface is depressed (though not strongly concave) and subcircular. The centra of the first three vertebrae are narrow, with the anteriormost having shallow excavations. There do not appear to be parapophyses on the first sacral vertebra. The transverse processes of the first two vertebrae are narrow and rodlike (fig. 41A). Those on the third vertebra are broader than the transverse processes of the preceding two vertebrae (fig. 41A). The transverse processes of all three of these vertebrae are directed dorsolaterally and are not as well preserved as their illustration indicates.

The following three vertebrae have very short, broad, and directly dorsally projected transverse processes (fig. 41A). By contrast, the 7th vertebra has a narrow, laterally projected transverse process and appears to also have the stump of a fused rib (processus costalis;Baumel and Witmer, 1993). The following three vertebrae have narrow transverse processes, the first of which projects directly laterally, subparallel to that of the 7th vertebra. The transverse processes of the 9th and 10th sacrals project posterolaterally, and those of the 10th sacral are located close to the posterior edge of the sacrum. Though crushed, the sacrum is visibly widest at about its midpoint. The last sacral is slightly shorter than the vertebra directly preceding it. There are well-developed prezygapophyses on the first sacral (fig. 41A) and postzygapophyses on the last sacral. The iliosynsacral grooves (sulci iliosynsacralis;Baumel and Witmer, 1993) were open dorsally (fig. 41A). The fused neural spines on the anterior vertebrae are broad at their dorsal tips, thickened by ossified tendons and forming a roughly T-shaped cross section.

The second sacrum (fig. 42) figured and discussed inOdontornithes as representative ofIchthyornis victor is from YPM 1732. Its reconstruction (Marsh, 1880: pl. XXXII, figs. 2, 3) differs from the actual material that was removed from theIchthyornis victor panel mount and reprepared (fig 42). At present, only the pelvic bones from the right side are preserved (fig. 42B,D). These include the preacetabular ilium, the anterior tip of the pubis, and an almost complete ischium. Complete pubes, ischia, and preacetabular ilia were originally depicted as present by Marsh (seefig. 42C).

The sacral series of YPM 1732 was figured (fig. 42C) as anteriorly incomplete or covered in matrix (Marsh, 1880: pl. XXXII, figs. 2, 3); however, removal from the panel mount and further preparation revealed that it is complete and well preserved (fig. 42B,D). Ten vertebrae were described as comprising the sacral series in YPM 1732 and the anteriormost attached vertebra considered byMarsh (1880) to be the last thoracic. This vertebra is here discussed as a sacral: While there is a conspicuous suture between this vertebra and the rest of the sacral series (fig. 42D), it is ankylosed to these vertebrae. Marsh considered this vertebra fused on the anterior end of the sacrum in YPM 1732 specifically to correspond to a thoracic vertebra of theIchthyornis dispar holotype (i.e., vertebra 1450D). As discussed above, the vertebra in YPM 1450 and that on the anterior end of the sacrum in YPM 1732 differ in morphology, and this assumption is considered baseless.

Twelve vertebrae actually comprise the sacrum in YPM 1732 (or 11 if the most anterior is considered to be a thoracic). The anteriormost sacral (or last thoracic ofMarsh, 1880) is a short vertebra with an incompletely closed suture between its centrum and that of the following sacral. Ossified tendons are fused to the transverse processes of this vertebra and pass posteriorly to also fuse with these processes of the second sacral (fig. 42B,D). On the left side of this vertebra, dorsal and anterior to the remains of what appears to be a small lateral excavation, a parapophysis is visible. This first sacral appears similar in morphology to the last thoracic which was found in association with it, 1732C (i.e., short compared to more anterior thoracics;fig. 40D). The apparent lack of parapophyses on the first sacral of theIchthyornis dispar holotype and their presence in YPM 1732 may suggest that this character should be also be considered polymorphic forIchthyornis dispar. The ilium overlapping one set of ribs (indicated by the presence of a parapophysis) has been considered a synapomorphy of Aves (appendix 1, character 161). However, the ilium in YPM 1732 is short, and while ribs did articulate with the first ankylosed vertebra, it does not appear to have been covered by the ilium (fig. 42B).

The centrum of the second sacral, which would appear to correspond morphologically to the first of theIchthyornis dispar holotype (YPM 1450), is more compressed than that of the first sacral of YPM 1732. The transverse processes of this vertebra, as well as those of the first and third sacrals, are narrow, rodlike, and projected directly laterally. The transverse processes of the fourth sacral appear stouter than those of the preceding vertebrae; however, their shape appears somewhat distorted by crushing.

The transverse processes of the fifth sacral are narrow, rodlike, and projected directly laterally (fig. 42D). By Marsh's count, this vertebra would correspond to the fourth sacral of the holotype ofIchthyornis dispar (YPM 1450); however, the transverse processes of the fourth sacral in YPM 1450 project dorsally and appear absent in lateral view. In YPM 1732, this morphology is developed on the following three sacral vertebrae, but not on that which would correspond to the fourth in YPM 1450.

The centra of the three vertebrae with this morphology are expanded laterally with dorsally directed transverse processes in the holotype ofIchthyornis dispar (YPM 1450;fig. 41A) and YPM 1732 (fig. 42D) while there are four inApatornis celer (fig. 41B). The presence of midsacral vertebrae that appear to lack transverse processes or have them projected directly dorsally (such that only a line of spinal nerve foramina is visible close together on the edged of compressed vertebra) is only known with confidence forIchthyornis and Aves (appendix 1, character 62). In YPM 1732, like the holotype ofIchthyornis dispar, the centra are the broadest in the middle of the series (fig. 42D).

The 9th sacral vertebra has a well-developed sacral rib that extends to contact the ilium just posterior and dorsal to the acetabulum. The 10th vertebra preserves the base of the transverse process projecting directly laterally, subparallel to that of the preceding vertebra (fig. 42D). The transverse processes of the 11th and 12th sacral vertebra are not preserved. The transverse processes of the 12th sacral appear to have arisen close to the posterior terminus of the sacrum as inIchthyornis dispar (figs. 41C,42D). Two faint tubercles are developed close to the anterior edge of this vertebra. These tubercles are also weakly developed in the holotype ofIchthyornis dispar (fig. 41C) and more conspicuous in that ofApatornis celer (fig. 14B).Marsh (1880) described and illustrated a pronounced groove on the ventral midline of the centra of the posterior portion of the sacrum (fig. 42C). There is a midline crack posteriorly and breakage in midseries that gives the centra this appearance. There may have been a slight groove on the 5th to 11th sacral vertebra, but it was clearly not as deep as figured.

The anterior and posterior articular surfaces of the 1st and 12th sacral vertebrae, respectively, are slightly concave as inIchthyornis dispar. There are well-developed prezygapophyses on the first vertebrae while no postzygapophyses are preserved on the last. Between the transverse processes of the 1st and 2nd sacral vertebrae, as well as between the 7th and 8th vertebrae, openings to the dorsal surface of the sacrum are present (fenestrae intertransversariae;Baumel and Witmer, 1993).

YPM 1733 includes two fragments of the sacral series that are severely crushed dorsoventrally. The sacrum of YPM 1733 is from a smaller individual than that of YPM 1732 but larger than theIchthyornis dispar holotype (YPM 1450). One fragment preserves what would correspond the 4th sacral of YPM 1732 and the 3rd sacral of theIchthyornis dispar holotype. In dorsal view, the coossified neural spines are relatively broad mediolaterally from attached ossified tendons. The iliosynsacral sulci are broad. The second fragment of YPM 1733 includes what would correspond in YPM 1732 to the 7th sacral through a small part of the 10th sacral, or the sixth part of the 9th in the holotype ofIchthyornis dispar (YPM 1450). The base of the rodlike transverse processes on the 9th vertebra of YPM 1732 and the 7th vertebra inIchthyornis dispar (YPM 1450) are visible. The centra in the portion preserving the 7th through 9th sacral vertebrae are broad, as in the midsacrals of both YPM 1732 and YPM 1450.

Marsh (1880) noted thatIchthyornis differed fromSterna maxima in having only 10 sacrals as opposed to 13 inSterna maxima. Total sacral number increases toward Aves (appendix 1, character 61); however, this character may vary withinIchthyornis. In theIchthyornis dispar holotype (YPM 1450), there are 10 fused sacrals while in YPM 1732 there are 12. The number of vertebrae posterior to the strut at the acetabulum, the single vertebra with a costal strut, and the number of vertebra that appear to lack transverse processes is the same in both YPM 1450 and YPM 1732. However, the number of vertebrae anterior to the first appearing to lack transverse processes is different. In the holotype ofIchthyornis dispar there are three anterior vertebrae with laterally projecting transverse processes while in YPM 1732 there are five. One appears to have been added to the anterior end and the second between the third and fourth ofIchthyornis dispar. While in YPM 1450 and YPM 1732 three midseries sacrals lack conspicuous transverse processes, four lack them in the holotype ofApatornis celer (appendix 1, character 62). Thus, it is themorphology of the “additional” vertebra (relative to theIchthyornis dispar holotype) that appears to differ betweenApatornis celer and YPM 1732, while the position of this vertebra is the same; it is at the end of the series of vertebrae with the “anterior morphology” and the beginning of those with the “midseries morphology”.

Marsh (1880) noted two further differences betweenIchthyornis andSterna maxima in the numbers of vertebrae of certain morphologies in the sacrum. He noted that the strutlike costal process (i.e., “strong transverse bar”,Marsh, 1880: 162) in the area of the acetabulum occurs on the 7th sacral inIchthyornis and on occurs on the 9th inSterna maxima and that three vertebrae are posterior to this strut inIchthyornis and four inSterna maxima. However, he was referring to the holotype ofIchthyornis dispar: By contrast, YPM 1732 (referred toIchthyornis dispar) has the costal strut on the 9th ankylosed vertebra (fig. 42D). In Aves, this strut typically occurred on the 9th or 10th sacral as mentioned byMarsh (1880) forSterna maxima.

InIchthyornis (YPM 1450, YPM 1732) andApatornis celer (YPM 1734), there are only three vertebrae (figs. 41A,42C) posterior to the strut at the acetabulum. By contrast, all Aves surveyed had more vertebrae posterior to this strut. In the tinamous and galloanserines considered, there were at least five and sometimes seven vertebrae (e.g.,Crypturellus noctivagus). As the total vertebral count is unknown forIchthyornis and cannot be compared to that in Aves, it is unclear whether additional elements were incorporated from the caudal series or whether additional new iterated elements were developed (i.e., that the total number of vertebrae increased by the addition of vertebrae with this morphology phylogenetically after the loss of caudal vertebrae).

Caudal Vertebrae

Caudal vertebrae are only represented in YPM 1732 (fig. 43) and possibly by a part of a pygostyle from YPM 1775 (fig. 44). In YPM 1732, five free caudals and the base of the pygostyle are known. These vertebrae were figured inOdontornithes (Marsh, 1880: pl. XXVIII, figs. 2–7).Marsh (1880) refers to seven caudals as preserved in YPM 1732 and figured in plate XXVIII. Five free caudals and a pygostyle are figured.Marsh (1880) reasoned that the pygostyle was comprised of two vertebrae. The tip of the pygostyle is missing for YPM 1732 (Marsh, 1880). Detailed descriptions of the individual vertebrae are provided in the text ofOdontornithes (Marsh, 1880) and were used in conjunction with plate XXVIII to identify vertebrae removed from theIchthyornis victor panel mount (fig. 2). By contrast to their illustration, the last two free caudal vertebrae, when removed from theIchthyornis victor panel mount, were joined by original matrix to each other and to the pygostyle. Apparently, the depiction of the anterior and posterior articular surfaces of the fifth, the posterior surface of the fourth, and the anterior surface of the sixth (Marsh, 1880: pl. XXVIII, fig. 6) were largely reconstructed. The transverse processes of the free caudal vertebrae are significantly less complete than their illustration (Marsh, 1880). This is especially the case of the second and third caudals (Marsh, 1880: pl. XXVIII, figs. 3, 4).

AsMarsh (1880: 166) noted, the prezygapophyses preserved on the two first caudals are extremely long and would have articulated on the dorsal surface of the preceding vertebra (fig. 43A). This is the opposite of the normal relation of pre- and postzygapophyses. Such a relationship is well developed inVanellus melanopterus (Charadriiformes), while in many other avian taxa, the pre- and postzygapophyses on the caudals are weakly developed, nonarticulating tubercles (e.g.,Nothura darwinii,Crypturellus noctivagus,Anas platyrhynchos,Columba livia). InSterna maxima,Alectura lathami,Gallus gallus, andMeleagris gallopavo, this relationship is present but slightly less developed than inIchthyornis orVanellus melanopterus. InSterna maxima, this articulation appears more conspicuous on the posterior caudals, whereas in Galliformes it is more conspicuous on the anteriormost caudals. Development of this articulation appears to be synapomorphic of at leastIchthyornis dispar + Aves. It is not developed inHesperornis regalis, although the inclination of the anterior margin of the neural spine in the first two caudals is reminiscent of the condition in Aves where this articulation is weakly developed (e.g.,Columba livia).

In YPM 1732, there is no preserved indication of articular facets for unfused chevrons. Chevrons also do not appear fused to any of the caudal vertebrae in YPM 1732. As remarked byMarsh (1880), the ventral surface of the centra bears a slight midline ridge with two shallow lateral depressions. In the fifth caudal, the last with transverse processes, this morphology is more pronounced. The complete neural arch illustrated for the fifth free caudal vertebrae is no longer present. The central articulations are all platycoelous. The central articular surfaces of the third caudal are crushed such that they appear strongly depressed (Marsh, 1880: pl. XXVII, fig. 4). The posterior articular surfaces of at least the last two free caudals are inclined to face slightly dorsally. This curvature indicated that the tail inIchthyornis (YPM 1732) would have been curved somewhat dorsally as in most of Aves (Gauthier and de Queiroz, 2001).

As the neural spines are missing in all of the free caudals, it is unknown whether they were bifid as in many Aves (e.g.,Crypturellus noctivagus,Nothura darwinii,Gallus gallus,Anas platyrhynchos). This condition is also present inHesperornis regalis, and it is possible that having at least one caudal with bifid neural spines is a synapomorphy of that taxon + Aves. The condition inApsaravis ukhaana is unknown.

The centrum articular surfaces of the caudal vertebrae, where preserved, are ovoid and flat. There is an unusual tubercle at the midpoint of the dorsal edge of the anterior and posterior articular surfaces of the first, third, and fifth caudal vertebrae. On the fourth, it is only preserved on the posterior surface, while the anterior face is broken in this area. In whatMarsh (1880) identified as the second caudal, this tubercle is not developed on the anterior surface, but it is developed on the posterior surface. Because a tubercle is also absent on the last vertebra of the sacrum, it is possible that this surface articulated with the anterior central surface of whatMarsh (1880) considered to be the second caudal; thus, the vertebraMarsh (1880) considered the first and that which he considered the second may be misidentified (reversed). However, at least the last two free caudals and the pygostyle were evidently found in articulation; it is possible that these anterior caudals were as well. Such tubercles were not observed in the surveyed Aves.

The two terminal caudal vertebrae are fused. The dorsal edge of the preserved portion of the fused terminal caudals is unbroken, while the base of this element is damaged and it lacks its distal end (fig. 43B). The element formed of fused terminal caudals was not long, and it appears to have some of the dorsally projected “plowshare” shape, as well as loss of transverse processes, consistent with the conformation of a pygostyle in Aves. The large ventral notch in this fragment is artifactual. The triangular neural canal is preserved above the slightly concave articular surface (fig. 43B). Paired foramina are developed near the dorsolateral surface of the preserved portion of the pygostyle in its lateral surface. These foramina were not illustrated inOdontornithes (Marsh, 1880). Foramina are in a similar location inVanellus melanopterus. A feature apparently topologically congruent with the foramina, a notch, is developed in the dorsal edge of the pygostyle inNothura darwinii between the neural spines of the first two partially fused vertebrae. Neither foramina nor a notch were observed inGallus gallus,Anas platyrhynchos, orColumba livia. The shape of the distal end of the pygostyle, although broken, does not suggest additional vertebrae were incorporated.

A fragment that is part of YPM 1775 could be part of a pygostyle (fig. 44). If so, it appears to differ significantly from that in YPM 1732. It also does not correspond well with the pygostyles of the surveyed Aves. This fragment includes what is interpreted as an ovoid, slightly concave, articular surface and a flat lamina that, while broken close to the ovoid articulation, has a projected lip on one edge (ventral?;fig. 44). This element deserves further scrutiny. No neural canal is preserved, and the piece narrows abruptly distal to the ovoid articular surface unlike the condition in YPM 1732.

Phalanx I:1

The element placed as phalanx I:1 in theIchthyornis victor mount appears to be, indeed, manual phalanx I:1. The element is thin and narrow, with an ovoid articular surface (fig. 56E). However, it may be from YPM 1759, which is referred toIchthyornis dispar, or it may be from YPM 1734 (the holotype ofIaceornis marshi). Phalanx I:1 is not mentioned or figured inOdontornithes (Marsh, 1880) for eitherIchthyornis or YPM 1734, which Marsh referred toApatornis celer. The element from the mount articulates better with the first metacarpal ofIchthyornis (e.g., YPM 1724) than it does with that of YPM 1734. Indeed, it appears too large to be part of YPM 1734.

As further discussed below (see Phalanx II:2), Gibb's annotated image ofIchthyornis victor fromMarsh (1880;fig. 3) placed the number “1759” on Phalanx II:2. But this element from the mount appears to be part of YPM 1734, not YPM 1759 (holotype ofIaceornis marshi; previously referred toApatornis celer) that was figured inOdontornithes asApatornis celer (Marsh, 1880: pl. XXXI, figs. 9–11). Thus, it is possible that this specimen number (i.e., YPM 1759) corresponds with the similarly shaped phalanx I:1 that had no number applied to it in Gibb's notes (fig. 3).

Phalanx II:1

The first phalanx of digit II (fig. 56A) is represented in five YPMIchthyornis dispar specimens (1463, 1726, 1755, 1759, and 1775). The morphology of these phalanges is consistent insofar as they could be compared. All are from individuals significantly larger than theIchthyornis dispar holotype and approximately the size of theIchthyornis victor holotype. One phalanx identified asIchthyornis victor is figured inOdontornithes (Marsh, 1880: pl. XXXI, figs. 16–18). The phalanx illustrated is probably from YPM 1726, which is the best preserved of the five specimens and the only one that currently matches the pictured element. This specimen was also included in theIchthyornis victor panel mount as indicated by Gibb's notes (fig. 3).

The limited comments on this element inOdontornithes (Marsh, 1880) are consistent with the morphology preserved: certain details of the articular surfaces and that the phalanx thins posteriorly. Several morphologiesMarsh (1880) did not describe or mentioned only briefly are discussed here.

A subtriangular muscle scar is developed on the proximal edge of the anterodorsal surface, with an oblong scar just distal to it on the same edge (fig. 56A). A third, larger, ovoid scar is well developed on approximately the midpoint of the dorsal surface of the anterodorsal edge of the element (fig. 56A). The scar has a raised distal edge and is moderately deep. A conspicuous scar is developed in the same position on the phalanx ofIaceornis marshi (YPM 1734), which, otherwise, little resembles that ofIchthyornis. Among the Aves surveyed (i.e.,Crypturellus noctivagus,Anhima cornuta,Anas platyrhynchos,Crax pauxi, andMeleagris gallopavo,Burhinus capensis, andSterna maxima), none had a prominent scar in the position observed inIchthyornis andIaceornis marshi (YPM 1734). A weakly developed scar did appear to be present inCrypturellus noctivagus and inSterna maxima, for example, but no such scar appears present inApsaravis ukhaana (Norell and Clarke, 2001). The distribution of this scar should be further investigated.

A strong, proximally directed process is developed on the anteroventral corner of the proximal edge of phalanx II:1 (fig. 56A). This process is developed in tinamous and Galliformes. It is not as well developed in Anseriformes. It has not been described for more basal avialans and does not appear present inApsaravis ukhaana. It is possible that this process is synapomorphic ofIchthyornis dispar + Aves, although scrutiny of more basal avialans is necessary to determine if it is synapomorphic at a more or less inclusive level.

Additionally, an internal indicus process (Stegmann, 1978) is present on the posterior edge of the distal end of the phalanx. As discussed in a comparative context byStegmann (1978) andClarke and Chiappe (2001), an internal “indicus” process is otherwise only seen in Neoaves. It is not known inApsaravis ukhaana (Norell and Clarke, 2001) or for more basal taxa. It is found to be an autapomorphy ofIchthyornis dispar (see Taxonomic Revision). An internal indicus process is not present inIaceornis marshi (fig. 56B) or inLimenavis patagonica (Clarke and Chiappe, 2001).

Phalanx II:2

Marsh (1880) did not discuss the second phalanx of the second manual digit inIchthyornis dispar, though he did comment on the morphology of this element in YPM 1734 (the holotype ofIaceornis marshi), which he referred toApatornis celer. Phalanx II:2 is represented in a partial skeleton discussed above in the context of the furcula (SMM 2503;fig. 56C). This element is significantly more robust and elongate than II:2 in YPM 1734 (named as the holotype ofIaceornis marshi; see Taxonomic Revision), for example (comparefigs. 56C and D). It bears a relatively well-developed ginglymoid articular surface for the third phalanx (ungual) of this digit (fig. 56C). Its base is excavated by strong depressions for muscular attachment. Thus, it would appear that this digit had a well-developed claw. The length of this phalanx is subequal to that of II:1.

No mention is made inOdontornithes (Marsh, 1880) of the manual phalanx II:2 ofIchthyornis. It is not represented among the YPM material. However, two manual phalanges besides II:1 were included in theIchthyornis victor panel mount placed in the positions of II:2 as well as I:1.

Gibb's notes indicate that the element placed in the mount as manual phalanx II:2 belonged to YPM 1759 (fig. 3). However, when this element was removed from the mount, it matched phalanx II:2 as figured forIaceornis marshi (YPM 1734) inOdontornithes (Marsh, 1880: pl. XXXI, figs. 9–11). YPM 1734 was referred toApatornis celer and is named as the holotype ofIaceornis marshi. The illustrated digit of YPM 1734 has a small projection on its proximal end matched by the phalanx placed as II:2 (fig. 56D). It appears that, again, Gibb's notes are misleading: The phalanx placed as II:2 ofIchthyornis and identified by Gibb as belonging to YPM 1759 appears to be II:2 of YPM 1734, which is not closely related toIchthyornis dispar (Part II, Results).

Phalanx III:1

As mentioned byMarsh (1880), phalanx III:1 inIchthyornis dispar is preserved in YPM 1775 (fig. 57D). Both the right and left phalanges are represented. The anatomy of this element does not agree with its representation in the reconstruction ofIchthyornis victor inOdontornithes (Marsh, 1880: pl. XXXIV). In that figure, it is presented as a simple, narrow element (fig. 57C), as, for example, inSterna maxima. In fact, the phalanx is a flat, robust element with a pronounced, posteriorly directed flexor process (Butendieck and Wissdorf, 1981) and is much closer in structure to the same element in Tinamidae (e.g.,Eudromia elegans andCrypturellus undulatus;fig. 57B), while inSterna maxima (and Charadriiformes generally;Stegmann, 1978) the flexor process is a diminutive (fig. 57A). InIchthyornis dispar and tinamous (and an array of other avian taxa;Stegmann, 1978) it is a broad, flangelike process. This tubercle associated with the m. flexor digiti III insertion in Aves (Chamberlain, 1943;Stegmann, 1978). Phalanx III:1 has an ovoid, slightly concave surface for articulation with metacarpal III (fig. 57D).

Characters

Of the 202 characters, 35 are ordered (i.e., 1, 8, 11, 23, 31, 52, 54, 61, 62, 66, 68, 69, 71, 76, 80, 105, 113, 117, 139, 142, 149, 153, 159, 175, 177, 180, 182, 185, 188, 192, 193, 194, 195, 196, 202). These ordered characters are also indicated inappendix 1. FollowingSlowinski (1993), multistate characters were ordered when there was evidence of a morphocline (sensuSlowinski, 1993; e.g., Metacarpal I, extensor process: [0] absent, no anteroproximally projected muscular process; [1] present, tip of process just surpassed the distal articular facet for phalanx 1 in anterior extent; [2] tip of process conspicuously surpassed the articular facet by approximately half the width of the facet itself, producing a pronounced knob; [3] tip of process conspicuously surpassed the articular facet by approximately the width of the facet itself, producing a pronounced knob).

Of the characters identified as parsimony uninformative, constant characters were included because they have previously been proposed to be synapomorphic of an avialan subclade. After being included and evaluated for the sampled taxa, these characters were either found to be ancestral for at least the included part of Theropoda (i.e., 130, fromChiappe, 1991), or all included taxa are missing data for the character or have the same state (i.e., 45, fromStidham, 1998, and 129, fromChiappe, 1991; the derived state for 129 is only possibly [seeSanz et al., 1995] present in one included taxon, which was scored “0/1”).

Of the characters identified by PAUP*4.0b8 (Swofford, 2001) as autapomorphic, some are only known in a single terminal of the ingroup, while others are present only in Dromaeosauridae and have been found in previous analysis to be synapomorphic ofArchaeopteryx lithographica + the ingroup when the resultant tree is rooted with Dromaeosauridae. In the latter case, these “0”s are assessed as autapomorphies of Dromaeosauridae for the purposes of calculating the number of parsimony-informative characters.

It has been argued that parsimony-uninformative characters (in particular autapomorphies) should be excluded, as they inflate certain tree statistics such as the consistency index (Carpenter, 1988). However, the consistency index has come to be considered more appropriate as a measure of the amount of homoplasy exhibited by included characters (e.g.,Sanderson and Donoghue, 1996) rather than as a metric of support for a particular phylogenetic hypothesis; homoplastic characters can be highly informative of phylogenetic relationship (e.g.,Sanderson and Donoghue, 1996;Kallersjo et al., 1999). Thus, inflation of the consistency index does not constitute so much an exaggeration of a support measure as an underestimation of the homoplasy exhibited by component characters. Furthermore, characters identified as parsimony uninformative are easily excluded in PAUP*4.0b8 (Swofford, 2001) to recalculate tree statistics. It is also worth noting that in phylogenetic analyses of molecular sequence data, many included characters may be parsimony uninformative, but they are, nonetheless, analyzed.

Including the greatest number of characters possible, on the other hand, brings the most information to bear on the study of morphological change across a clade, summarizes additional characters diagnosing terminal taxa, and provides a more accurate picture of branch lengths. Branch lengths may be of interest as potential indicators of the amount of morphological divergence and the number of missing taxa. Branch lengths would also be important for analyses employing maximum likelihood as an estimator with currently available methodologies (Lewis, 2001). One outcome of excluding autapomorphic characters may be partially responsible for the putative phenomenon of “character exhaustion” (Wagner, 2000) in morphological data sets, by reducing terminal branch lengths in many cases to 0.

Too often, characters proposed by other authors to be informative of avialan interrelationships, once discovered to be constant or autapomorphic for a taxon other than that which was the focus of the analysis have been excluded from analysis and undiscussed. The current analysis is also incomplete, as presumably many more uninformative characters could be included.

Taxon Sampling

A species-exemplar approach was taken to represent Aves (sensuGauthier, 1986;Gauthier and de Queiroz, 2001) in which five species were used. These five include basal parts of lineages considered to represent its two basal-most divergences: PalaeognathaePycraft, 1900, and NeognathaePycraft, 1900, based on previous hypotheses of the relationships within these clades (e.g.,Holman, 1964;Cracraft, 1974;Sibley and Ahlquist, 1990;Livezey, 1997a,1997b). Strong morphological and molecular evidence supports Palaeognathae and Neognathae as the two basal divergences of the crown clade and GalloanseraeSibley et al., 1988 as one of the two basal-most divergences within Neognathae (e.g.,Gauthier, 1986;Cracraft, 1988;Sibley and Ahlquist, 1990;Groth and Barrowclough, 1999;van Tuinen et al., 2000;Cracraft and Clarke, 2001). Four exemplars sample basal and derived parts of Galloanserae.

The sister taxon of Galloanserae, NeoavesSibley et al., 1988, was not sampled; there remains no resolution of the basal relationships within this clade (e.g.,Cracraft and Clarke, 2001). Without a hypothesis of ingroup relationships, a supraspecific Neoaves terminal representing most of 9000+ species of birds (del Hoyo et al., 1992–1999) would be scored as polymorphic for many, if not most, characters variable for Aves and any exemplars elected would only sample the smallest part of this highly diverse clade. It was decided that whatever exemplars were elected could not with any confidence be deemed to appropriately bracket the ancestral condition. Unfortunately, because of the absence of a Neoavian terminal(s), the states ancestral to Aves may be inaccurately represented in the current analysis for some characters. This issue is further discussed in the Conclusion.

The five species-exemplars for crown clade Aves includeCrypturellus undulatus (YPM 11564), one of the forest-dwelling tinamous that have been placed as basal within Tinamidae (S. Bertelli, personal commun.), which was used as the exemplar for PalaeognathaePycraft, 1900. Of NeognathaePycraft, 1900,Chauna torquata (YPM 6046, AMNH-3616) andAnas platyrhynchos (YPM 2230, YPM 14369, YPM 14344, AMNH-5847) were used to represent AnseriformesFürbringer, 1888;Crax pauxi (YPM 2104) andGallus gallus (YPM 2106, YPM 6705) were used to represent GalliformesGarrod, 1873. The included anseriforms and galliforms were chosen to sample both basal divergences (i.e.,Crypturellus,Chauna, andCrax) and deeply nested taxa (i.e.,Anas andGallus) from within these clades based on previous phylogenetic hypotheses (e.g.,Holman, 1964;Cracraft, 1974;Sibley and Ahlquist, 1990;Livezey, 1997a,1997b).

Supraspecific OTUs were used for “Lithornis” and the outgroup terminal “Dromaeosauridae”. These two terminals were scored from species considered to comprise them (see Materials and Methods). Both DromaeosauridaeMatthew and Brown, 1922, andLithornisOwen, 1840, have been supported as monophyletic (e.g.,Houde, 1988;Gauthier, 1986; although seeNorell et al., 2001). In both of these cases, these supraspecific terminals were used to compensate for missing data and because of the availability of multiple specimens. If there was variation among the material studied (see Material and Methods), this was represented as polymorphism (e.g., as “0 and 1”). While not an ideal approach (e.g.,Prendini, 2001;Simmons, 2001), the alternative course of evaluating and including as distinct terminals an array of dromaeosaurids and lithornithids was outside the scope of the present analysis. The specimens that formed the basis for the scorings of the two terminals just discussed are indicated below.

The terminal “Lithornis” (Owen, 1840) is scored from study ofLithornis plebius (USNM-336534, AMNH-21902),Lithornis promiscuus (USNM-336535, USNM-424072, AMNH-21903),Lithornis celetius (USNM 290601, USNM-290554, USNM 336200, YPM-PU-23485, YPM-PU-23484, YPM-PU-23483, YPM-PU-16961), and supplemented by the description of this material provided byHoude (1988). The paraphyly of the “Lithornithidae” has been proposed (Houde, 1988). However, the “Lithornithidae” includedParacathartes andPseudocrypturus, as well asLithornis; the monophyly ofLithornis itself has not been disputed.Paracathartes andPseudocrypturus material was not scored for the current analysis.

TheIchthyornis dispar terminal was scored from all YPM material assessed to be part of that species in the current analysis (see Part I; listed intable 1), as well as from SMM 2305 and BMNH A905 (see Taxonomic Revision). Two other terminals were employed in a single analysis (seeIchthyornis dispar in Results below) to investigate the placement ofIchthyornis dispar including and excluding YPM 1732: (1)Ichthyornis dispar as known from the holotype and all referred specimens, exclusive of YPM 1732, and (2)Ichthyornis dispar as known from YPM 1732 alone. YPM 1732 is considered part ofIchthyornis dispar (see Part I) as many morphologies of the femur, tibiotarsus, and sacrum can be compared directly to theIchthyornis dispar holotype. The comparison indicates they are morphologically identical except with regard to the number of sacral vertebrae, which differs. YPM 1732 was used in the scoring of theIchthyornis dispar composite terminal in all but this one analysis.

In addition to theIchthyornis dispar terminal, the other morphologically distinct Late Cretaceous Niobrara Chalk specimens previously considered part ofIchthyornis andApatornis (Marsh, 1880) were evaluated:Ichthyornis (Guildavis) tener (Marsh, 1880; new clade, Part I) was scored from the holotype, YPM 1760;Ichthyornis (Austinornis) lentusMarsh, 1877b (Marsh, 1880; new clade, Part I) was scored from the holotype, YPM 1796;Apatornis celer Marsh, 1873a, 1873b, was scored from the holotype, YPM 1451;Iaceornis marshi (new clade, new species; Part I) was scored from the holotype specimen, YPM 1734 (which was previously referred toApatornis celer;Marsh, 1880), and from the illustration of the pelvis inOdontornithes (Marsh, 1880), as this element is currently missing (Clarke, 2000a; Part I). To explore support for the placement of theIaceornis marshii holotype specimen (YPM 1734) excluding the unverifiable characters scored from the depiction of the missing pelvis inOdontornithes (Marsh, 1880), the terminal “Iaceornis-sacrum” was swapped in (see Results, below).

Other taxa were included to sample basal Avialae.Hesperornis regalis was scored primarily from study of the holotype (YPM 1200) and referred YPM specimens (1206, 1207, 1476), as well as from the description of that taxon inMarsh (1880),Witmer and Martin (1987),Bühler et al. (1988), andWitmer (1990).Baptornis advenus was scored primarily fromMartin and Tate (1976), although the holotype of that taxon (YPM 1465) was also consulted.Apsaravis ukhaana (Norell and Clarke, 2001) was scored from the holotype specimen, IGM 100/1017.Patagopteryx deferrariisi was scored from MACN-N-03 (holotype), MACN-N-10, MACN-N-11, and MACN-N-14, as well as fromChiappe (1996).Vorona berivotrensis was scored fromForster et al. (1996).Confuciusornis sanctus was scored based on the study of numerous IVPP and GMV specimens referenced inHou (1997) andChiappe et al. (1999).

Enantiornithes (Walker, 1981; converted clade name,Sereno, 1998) was represented by taxa referred to it by previous authors (e.g.,Zhou et al., 1992;Chiappe and Calvo, 1994;Sanz et al., 1995;Chiappe, 1995a,1996;Norell and Clarke, 2001). The four taxa included as exemplars were chosen because (1) they are known from relatively complete and/or multiple specimens, (2) they sample Early and Late Cretaceous identified parts of the clade, and (3) they are geographically and morphologically diverse.

The four exemplar taxa chosen are the following:Cathayornis yandica (Zhou et al., 1992) was scored from the holotype specimen (IVPP V-9769A, B) and two referred specimens (IVPP 10890, IVPP 10916).Concornis lacustris (Sanz and Buscalioni, 1992;Sanz et al., 1995) was scored from the holotype specimen, LH 2814, and fromSanz et al. (1995).Neuquenornis volans was scored fromChiappe and Calvo (1994).Gobipteryx minuta was scored fromElzanowski (1974,1977,1995),Chiappe et al. (2001), and from the description of the holotype of “Nanantius valifanovi” (Kurochkin, 1996), considered to be a junior synonym ofGobipteryx minuta (Chiappe et al., 2001). Unfortunately, because relationships among Enantiornithes remain largely unresolved (e.g.,Padian and Chiappe, 1998;Chiappe, 2001,2002), sampling taxa closest to the base of the clade, as recommended in exemplar choice (e.g.,Prendini, 2001 and references therein) was problematic.

Outgroup terminalArchaeopteryx lithographica (Meyer, 1861) was scored based on study of the London and Berlin specimens and from descriptions provided inWellnhofer (1974,1993),Ostrom (1976),Witmer (1990), andElzanowski and Wellnhofer (1996). Dromaeosauridae (Matthew and Brown, 1922; sensuGauthier, 1986) was represented primarily by studied specimens ofDeinonychus antirrhopus,Dromaeosaurus albertensis, andVelociraptor mongoliensis cited inOstrom (1969),Norell et al. (1992),Colbert and Russell (1969),Currie (1995), andNorell and Makovicky (1997,1999), as well as from descriptions provided in those publications.

The data set was analyzed in PAUP*4.0b8 (PPC;Swofford, 2001). All searches were branch and bound. Several settings were altered from the PAUP defaults in all searches: “amb-” in the “Parsimony Settings” menu was selected so that internal branches with a minimum length of 0 were collapsed to form a soft polytomy; by contrast, the PAUP* default is to collapse only internal branches with a maximum length of 0. Additionally, when interpreting entries with more than one state, ambiguity (e.g., “state 1 or 2”) was distinguished from polymorphism (e.g., “states 1 and 2”). Bootstrap support values from 1000 replicates (10 random sequence additions per replicate) were computed with all the same settings as in the other branch and bound searches (see above). The bootstrap scores greater than 50% for the all analyses are reported at the internodes infigures 62–63.

Ichthyornis disparMarsh, 1872b

Marsh (1873b,1880) consideredIchthyornis to be most closely related to Hesperornithes, forming part of his lineage of “toothed birds”, the “Odontornithes”. Subsequently, most authors placedIchthyornis as more closely related to Aves than to Hesperornithes (e.g.,Martin, 1983;Cracraft, 1986,1988;Chiappe, 1995a,2001). Others, however, have suggested thatHesperornis is more closely related to Aves thanIchthyornis (e.g.,Elzanowski, 1995) or have followedMarsh (1873b,1880) in considering Hesperornithes +Ichthyornis to be a clade (e.g.,Elzanowski et al., 2000).

Five unambiguous synapomorphies (59, 62, 73, 92, 98) placeIchthyornis dispar, from the Late Cretaceous Niobrara Chalk of Kansas (Marsh, 1872b,1880), as more closely related to Aves than Hesperornithes (fig. 62). These five synapomorphies supportingIchthyornis dispar + Aves relative to Hesperornithes include: (59:1) thoracic vertebrae with ossified connective tissue bridging between transverse processes, (62:1) midseries sacral vertebrae with dorsally projected transverse processes that give the appearance of being absent, (73:1) a pneumatic foramen on midline on dorsal surface of the sternum, (92:1) the presence of a lateral process of the coracoid (discussed inMartin, 1987), and (98: 1) lack of a depression at the medial opening of n. supracoracoideus foramen (evaluated to have a different distribution by previous authors; e.g.,Chiappe and Calvo, 1994). Of these five synapomorphies, three (i.e., 59, 62, 73) are newly brought to bear on basal avialan relationships.

Five unambiguous autapomorphies ofIchthyornis dispar are identified in the phylogenetic analysis (i.e., 52, 66, 103, 132, 152; each is discussed in the Taxonomic Revision of Part I): (52:0) amphicoelous cervical vertebrae, (66:1) proximal free caudal vertebrae with well-developed prezygapophyses clasping the dorsal surface of preceding vertebra, (103:1) unprojected acromion on scapula, (132:1) subequal posterior and distal dimensions of the distal articular surface of ulna, and (152:1) internal index process present. Of these characters, two (i.e., 66, 152) are also seen in Charadriiformes and some other neoavians but not in the included Aves. This distribution, and the absence of these morphologies in other basal avialans, is consistent with the hypothesis (e.g.,Marsh, 1880) that several derived aspects of the anatomy ofIchthyornis dispar may be convergent on charadriiform morphologies. It does not, however, support these structures, or “shorebird” morphologies generally, as ancestral to Aves, as has often been argued (Feduccia, 1995,1999).

As noted in Part I, YPM 1732 is morphologically consistent with the holotype and all other referred specimens ofIchthyornis dispar except with regard to the number of ankylosed sacral vertebrae. YPM 1732 is considered on this basis to be a part ofIchthyornis dispar, as discussed in Part I. Based on this referral, theIchthyornis dispar terminal is scored as polymorphic for total sacral number. Although total sacral number is often considered intraspecifically conservative, such variation has been described. For example, similar to the case of YPM 1732, a large individual ofVelociraptor mongoliensis has been described (Norell and Makovicky, 1999) with an additional partially “sacralized” caudal. Additionally, intraspecific variation in sacral number was also early noted in breeds of the domestic pigeon,Columba livia (Darwin, 1859).

To investigate the effect of considering YPM 1732 (comprised of thoracic and caudal vertebrae, sacrum, and partial hind limb) notIchthyornis dispar, the “Ichthyornis dispar” terminal was removed and replaced with two terminals. One of the swapped-in terminals representedIchthyornis dispar as scored from holotype and referred material with the exception of YPM 1732; the second swapped-in terminal was scored only from YPM 1732.

Analysis resulted in 2MPTs (L: 395; CI: 0.67; RI: 0.81; RC 0.55) with the “Ichthyornis-YPM 1732” terminal placed in the same position as theIchthyornis dispar terminal in the primary analysis. YPM 1732 was placed closer to Aves than toIchthyornis dispar but outside of the polytomy includingIaceornis marshi,Apatornis celer andGuildavis tener. The strict consensus of these two trees, which differ only in enantiornithine interrelationships, is shown infigure 63. The rest of the resultant topology is the same as in the primary analysis.

YPM 1732 is placed closer to Aves than the rest of the YPM material by a derived character state for character 61 (number of ankylosed sacral vertebrae). It is not surprising that it is not supported as most closely related to the otherIchthyornis dispar material because, as discussed in Part I, eight of nine identified autapomorphies ofIchthyornis dispar are not preserved in YPM 1732 and one character optimized as an autapomorphy ofIchthyornis dispar is only preserved in YPM 1732. This character (66:1; proximal free caudal vertebrae with well-developed prezygapophyses clasping the dorsal surface of preceding vertebra) cannot be scored for anyIchthyornis dispar specimens other than YPM 1732. Thus, when YPM 1732 is included as a separate terminal, it is optimized as an autapomorphy of YPM 1732.

The unambiguous synapomorphies uniting “Ichthyornis-YPM 1732” + Aves are the same as those for the “Ichthyornis dispar” terminal (which includes YPM 1732) + Aves, with one exception; character 59:1 (thoracic vertebrae with ossified connective tissue bridging between transverse processes) is ambiguously optimized. Character 59 is not preserved in the holotype or any referred specimens ofIchthyornis dispar besides YPM 1732. And it was this material that was scored for the “Ichthyornis dispar-YPM 1732” terminal.

AustinornisMarsh, 1877band Pangalliformes (New Provisional Clade Name)

Austinornis lentus (Ichthyornis lentus ofMarsh, 1877b) is a distal tarsometatarsus from the Late Cretaceous (?) of Texas (Marsh, 1877b,1880).Marsh (1880) consideredIchthyornis (Austinornis) lentus to be part ofIchthyornis, a hypothesis that is not supported by the present analysis.Martin and Stewart (1982) suggested thatIchthyornis (Austinornis) lentus was not part ofIchthyornis but offered no hypothesis of its phylogenetic position.Shufeldt (1915) commented thatIchthyornis (Austinornis) lentus had galliform affinities.

Because no stratigraphic and little locality information for theAustinornis lentus holotype is available, its putative Cretaceous age should be further investigated (see Part I). The present analysis is the first phylogenetic analysis to include this taxon.

Austinornis lentus is placed in a trichotomy with the included galliforms (fig. 62). It is supported as part of at least the galliform stem clade by one unambiguous synapomorphy, (201:1) asymmetrical development of the edges of the trochlea of metatarsal III (Mayr, 2000). The phylogenetic position ofAustinornis lentus is, otherwise, supported by the presence of a hierarchical set of unambiguous synapomorphies as part ofVorona berivotrensis + Aves,Patagopteryx deferrariisi + Aves, Ornithurae,Hesperornis regalis + Aves. It is missing data, however, for unambiguous synapomorphies of Aves itself and for Neognathae. Thus, its placement as part of the avian crown clade (indeed, as a neognath and a galliform;fig. 62) is supported by a single character.

The galliform stem clade, thatAustinornis is supported as a part of, is named in Part I. If new data on the distribution of Cretaceous outcrops in Collin County, Texas and archival data support the provenance ofAustinornis lentus from these outcrops, then there is evidence of the Pangalliformes in the Cretaceous, but not necessarily for the presence of any part of the crown, Galliformes.

Iaceornis marshi, new clade, new species

The holotype specimen ofIaceornis marshi, a partial postcranial skeleton (YPM 1734) from the Late Cretaceous Niobrara Chalk of Kansas, was previously referred toApatornis celer (Marsh, 1880; see Part I). The previous proposed relationships of YPM 1734, as discussed asApatornis celer, are discussed under the “Apatornis celer” heading below. It is also possible that YPM 1734 elements have been mistakenly scored asIchthyornis in other analyses; there is evidence of parts of YPM 1734 figured inOdontornithes (Marsh, 1880) being mistaken for parts ofIchthyornis (e.g.,Sereno and Rao, 1992; see Part I).

Iaceornis marshi is placed outside of Aves (fig. 62) by four unambiguous synapomorphies (fig. 63): (74:1) pneumatic foramina between sternal rib articulations on sternum; (90:1) pneumatization of the coracoid (lost inAnas platyrhynchos); (148:1) metacarpal III surpasses II in distal extent (intermediate inLithornis); (179:1) distal condyles of the tibiotarsus equal in anterior extent (optimized as ancestral to Avialae, but as lost in basal avialae). It is possible that increased pneumatization of the skeleton is a single character (as discussed inGauthier, 1986) and that characters 74 and 90 are linked. However, each of several other characters concerning pneumatization has distinct distributions. For example, the presence of pneumatic foramen on the dorsal midline of the sternum is a synapomorphy ofIchthyornis + Aves; additionally, the kinds of postcranial and cranial pneumatization vary significantly across Avialae (Witmer, 1990;Britt et al., 1998). Because distinct air sacs pneumatize the skeleton, different pneumatic features may be expected to have different points of origin related to the evolution of these distinct air sacs (Britt et al., 1998).

Iaceornis marshi shared four derived characters with Aves relative toIchthyornis dispar (fig. 63). These characters are the following: (77:1) paired intermuscular ridges on sternum, (142:2/3) anterior extent of extensor process surpasses articulation of phalanx I:1 by more than half the width of this articulation, (147:1) intermetacarpal space terminates distal to the end of metacarpal I, (180: 2) ossified supratendinal bridge on tibiotarsus. An ossified supratendinal bridge has traditionally been considered to diagnose Aves relative to its nearest fossil outgroup (e.g., historicallyIchthyornis;Cracraft, 1986). Its presence inIaceornis marshi indicates that it is synapomorphic of a more inclusive clade that must include outgroups of Aves. Thus, its presence in other fragmentary tibiotarsi from the Cretaceous should no longer be the basis for necessarily referring these elements to Aves. The absence of a supratendinal bridge inLithornis (placed within Panpalaeognathae: Aves) is unambiguously optimized as a derived loss of this feature, which could be discovered to unite this taxon with ratites within which this feature is also lost (e.g., Cracraft, 1979).

Iaceornis marshi is diagnosed by two local autapomorphies: (82:1) a strongly tapering, or pointed, omal tip of the furcula, and (104: 1) a hooked scapular acromion process. Character 82 is also seen within Aves (e.g., in Anseriformes) and character 104 is also seen in more basal taxa such asApsaravis ukhaana (Norell and Clarke, 2001) and inLithornis (in which the acromium is otherwise markedly different from that ofIaceornis marshi andApsaravis ukhaana).

The characters optimized as autapomorphies ofIaceornis marshi and supporting its placement as the sister taxon of Aves, are the same whether or not the illustration of the sacrum inOdontornithes (Marsh, 1880) is included as part of theIaceornis marshi holotype.

Guildavis tener (new clade name; converted species name [Marsh, 1880])

Ichthyornis (Guildavis) tener (Marsh, 1880) is an extremely fragmentary sacral series from the Late Cretaceous Niobrara Chalk Formation of Kansas.Guildavis tener was placed in an unresolved polytomy withApatornis celer andIaceornis marshi as the sister taxon of Aves.Marsh (1880) consideredIchthyornis (Guildavis) tener more closely related toIchthyornis dispar than to Aves. However, it is here placed as more closely related to Aves than toIchthyornis dispar (figs. 62,64). This position is supported by one character: (161:1) the ilium overlaps at least one set of ribs, as indicated by a parapophyses visible on the first sacral. It is placed outside Aves by one character. A flat, round anterior articular surface of the centrum of the first sacral indicates that the thoracic series was not entirely heterocoelous (55:0) inGuildavis tener. Thus, a completely heterocoelous thoracic series (55:1) is a synapomorphy of Aves relative toGuildavis tener.

Apatornis celerMarsh, 1873b

The holotype and only known specimen ofApatornis celer (Marsh, 1873b,1880; see Part I: Taxonomic Revision) consists of an incomplete sacrum (lacking pelvic bones) from the Late Cretaceous Niobrara Chalk Formation of Kansas.Apatornis celer has no scored characters in common withIaceornis marshi (pelvic elements but not the sacral series) orGuildavis tener. AlthoughGuildavis tener andApatornis celer are known exclusively from fragmentary sacra, they cannot be evaluated for any of the same included characters and effectively share no “parts” in common. (As discussed in Part I, new specimens are needed to determine whether one or more of these names may refer to the same taxon).

Apatornis celer was considered byMarsh (1873b,1880) to be most closely related toIchthyornis dispar as part of his “Ichthyornithes”, and byBrodkorb (1967) as part of his “Ichthyornithiformes”.Howard (1955) suggested thatApatornis celer was most closely related to a fossil taxon that was considered to have anseriform “affinities” and that is now placed as part of Aves (Presbyornithidae; e.g.,Livezey, 1997b). This conclusion was based in large part, if not entirely, on study of the single referred specimen, here designated the holotype ofIaceornis marshi (YPM 1734). Similarly, whenMartin (1987) remarked that the elongate acromion seen in “Apatornis celer” andAmbiortus dementjevi suggested that they may be closely related, he must have been referring to YPM 1734 (holotype ofIaceornis marshi) as theApatornis celer holotype does not include a scapula.

The systematic position ofApatornis celer, as more closely related to Aves than toIchthyornis dispar (fig. 65) is supported by one character (62:2) the presence of four or more midseries sacral vertebrae with diminutive and dorsally projected transverse processes.Ichthyornis dispar has but three such vertebrae. Its placement outside of Aves is also supported by one character, a synapomorphy of Aves relative toApatornis celer (fig. 65). This synapomorphy of Aves absent inApatornis celer is the presence of 15 or more fused sacral vertebrae (61:3). The morphology of the incomplete anterior sacral vertebrae ofApatornis celer evidences the presence of only from 11 to possibly 13 fused sacral vertebrae (see Part I).