Caudipteryx as a non−avialan theropod rather than a flightless bird

Caudipteryx zoui is a small enigmatic theropod known from the Early Cretaceous Yixian Formation of the People’s Re− public of China. From the time of its initial description, this taxon has stimulated a great deal of ongoing debate regarding the phylogenetic relationship between non−avialan theropods and birds (Avialae) because it preserves structures that have been uncontroversially accepted as feathers (albeit aerodynamically unsuitable for flight). However, it has also been pro− posed that both the relative proportions of the hind limb bones (when compared with overall leg length), and the position of the center of mass in Caudipteryx are more similar to those seen in extant cusorial birds than they are to other non−avialan theropod dinosaurs. This conclusion has been used to imply that Caudipteryx may not have been correctly in− terpreted as a feathered non−avialan theropod, but instead that this taxon represents some kind of flightless bird. We re− view the evidence for this claim at the level of both the included fossil specimen data, and in terms of the validity of the re− sults presented. There is no reason—phylogenetic, morphometric or otherwise—to conclude that Caudipteryx is anything other than a small non−avialan theropod dinosaur.


Introduction
The non−avialan theropod dinosaur Caudipteryx zoui ( Fig. 1) was described by Ji et al. (1998) from the Early Cretaceous Yixian Formation of Liaoning Province, People's Republic of China (Ji et al. 1998;. Along with another taxon from the same deposits, Protar− chaeopteryx robusta Ji and Ji, 1997, both fossils added signifi− cantly to our understanding of the relationship between birds (Avialae) and non−avialan theropods, because both preserve integumentary structures uncontroversially interpreted as feathers (Ji et al. 1998;Padian 2001;Padian et al. 2001;Prum and Williamson 2001;Xu et al. 2001). Although the feathers preserved in these taxa were certainly not aerodynamically suitable for active flight (Rayner 2001), they have been inter− preted as providing clear evidence that the origination of these complex integumentary structures evolved prior to the phylo− genetic divergence of Avialae (Archaeopteryx and later forms). Subsequent cladistic analyses have also supported the contention that Caudipteryx is a member of Maniraptora, close to (but not within) the phylogenetic divergence of birds (Avialae) (Currie et al. 1998;Ji et al. 1998;Holtz 1998;Sereno 1999;Norell et al. 2001), closely related to Oviraptor and its kin (Oviraptoridae; Fig. 2).
Despite some dissention regarding a relationship between birds and theropods (reviewed by Chatterjee 1997, andFe− duccia 1999; see also Prum , 2003, no quantitative analyses (phylogenetic or otherwise) have been published to date in support of the hypothesis that the evident similarites between the two groups can be explained as a result of con− vergence. Hence, the only currently available alternative hy− pothesis states that birds (Avialae) did not diverge from within non−avialan theropods, but from another, as yet un− specified taxon (Feduccia 1999).
As as result of this ongoing debate regarding the origina− tion of Avialae, Jones et al. (2000) presented the results of a quantitative analysis of hind limb and body proportions, con− cluding that both the hind limbs and position of the centre of mass of Caudipteryx are more similar to extant "cursorial" (or "ground living"; including flightless) Neornithes (i.e., modern birds sensu Cracraft 1988) than they are to non− avialan theropods. On the basis of their analysis, Jones et al. (2000; see also Ruben and Jones 2001) suggested that previ− ous interpretations of Caudipteryx as a feathered non−avialan theropod could be incorrect.
Because of the evident discrepancy between reported morphological trends and the conclusions of phylogenetic analyses, we revisit in this paper the analysis of Jones et al. (2000). In addition to highlighting a number of significant problems with their measurement data (Appendix 1), we demonstrate by use of a separate, and more complete, set of limb measurements (Appendix 2) that the hind limbs of Caudipteryx are not significantly different from those of other known non−avialan theropods.

Limb proportions revisited
Assumptions of function and phylogeny.- Jones et al. (2000) presented the results of a morphometric analysis of non−avialan theropod and avialan hind limb proportions on the basis of a data set comprising 24 "cursorial" (their use of the term) extant birds (Neornithes) and 40 non−avialan thero− pod and ornithopod dinosaurs. They presented statistical re− gressions between limb and trunk lengths (Fig. 3) and con− cluded that the hind limb structure of Caudipteryx provides evidence that this taxon had a locomotor strategy similar to secondarily flightless Neornithes. The implication of this study being that because non−avialan theropods and Neo− rnithes had different locomotor strategies (reflected in their body shapes and limb proportions), the two groups are likely not related, and hence Caudipteryx cannot be considered simply as a non−avialan theropod with feathers. This conclu− sion was subsequently cited in both technical (Ruben and Jones 2001) and popular literature (Gould 2000) because it appears to provide a direct empirical challenge to the hypoth− esis of a "bird-dinosaur" relationship. From the outset, we would argue that simply because two groups have different locomotor strategies, they are not necessarily unrelated. Many groups of modern birds hop when on the ground, for example, while some others prefer to run; all passerine birds, however, are still considered closely related to one another (Barker et al. 2004). In addition to problematic assumptions of function and its relevance to phylogeny, we also highlight four further signifi− cant problems with the analysis (and hence conclusions) of Jones et al. (2000). These are: (1) assumptions of non−compa− rable hind limb function between non−avialan theropods and Neornithes; (2) accuracy of included specimen data used as a basis of conclusions; (3) calculation and use of trunk lengths as approximations for overall body size; and (4) calculation of regression statistics and the subsequent significance of results. Furthermore, Jones et al.'s (2000) calculations of centre of mass in Neornithes and non−avialan theropods are biased by assumptions concerning the position and extent of soft part anatomy in taxa that are closely related to avialans-the au− thors admit these were based on the skeletal reconstructions presented in G. Paul's (1988) Predatory Dinosaurs of the World.
Hind limb and tail: centre of mass and total leg length.- Jones et al. (2000) presented two linear regression analyses (that we discuss below) on the basis of their original morpho− metric data (supplementary information to their publication that can either be downloaded from www.nature.com or pro− vided electronically by GJD [gareth.dyke@ucd.ie]). In the second of two graphs (Jones et al. 2000: fig. b), effective hind limb lengths of terrestrial birds (Neornithes), non−avialan theropods and ornithopod dinosaurs are plotted against total trunk length (i.e., in their analysis this was defined as the length from the first dorsal vertebra to the midpoint of the ischium). Effective hind limb length was used by Jones et al (2000) because of a supposed difference in the contribution of the segments of the hind limb to terrestrial locomotion be− tween non−avialan theropods and Neornithes. As pointed out, for example by Gatesy (1990Gatesy ( , 1991Gatesy ( , 1995, reduction of the tail and the development of the caudofemoral musculature along the transition between non−avialan theropods and avialans led to a forward shifting of the relative centre of mass in the latter group (Christiansen 1999;Farlow et al. 2000;Christiansen and Bonde 2002). As a consequence, the more acutely angled femur seen in Neornithes contributes less to the total effective length of the hind limb (Gatesy 1990) than is the case in non−avialan dinosaurs. In correspondence, the femora of Neornithes are shorter and stouter to preserve bending and torsional strength (Gatesy 1991;Carrano 1998). Recognizing this difference, Jones et al. (2000) did divide their measure− ment data set accordingly but for the regression calculations made the a priori assumption, contradictory to the available skeletal evidence uniting Caudipteryx with Oviraptoridae (see above), that this taxon is a member of Avialae (i.e., a bird). As a result, in their second graph for Caudipteryx, they included only measurements for the distal segments of the hind limb ( Fig. 3). Caudipteryx plots out with the birds because only data from these taxa were included in the analysis. Jones et al. (2000) between hind limb proportions and estimated trunk length are extremely prob− lematic. Total trunk length has not been considered seriously as a proxy for overall body size since the work of Böker (1935). The measurement, and significance, of this quantity is difficult to assess because not only do the numbers of dor− sal vertebrae vary within both non−avialan theropods and ex− tant Neornithes (e.g., Mayr and Clarke 2003;Dyke et al. 2003), but there are serious problems with measurement of this quantity in many of the museum specimens cited by Jones et al. (2000). Differential preservation of fossils makes estimation of exact parameters such as trunk length problem− atic; separation of vertebral discs during fossilization, for ex− ample, will add significant error to a measurement of dorsal vertebrae. Jones et al. (2000) are unclear as to whether such factors were taken into account in their measurements of total trunk lengths.

Measuring trunk length in dinosaurs.-The additional comparisons made by
Specimen data.- Jones et al. (2000) presented measure− ments for segments of the hind limb and trunk length in a va− riety of dinosaur taxa. However, these measurements are ex− tremely hard to reconcile with the actual specimens from which they were taken (Appendix 1). There are a number of aspects to this problem. First, as discussed above, to accu− rately measure trunk length a number of assumptions would have to have been made with regard to the length of the verte− bral discs. Second, there is a clear problem in identifying the number of dorsal vertebrae (relative to thoracics) that are preserved in many fossil specimens, especially when they are smashed. This point is illustrated by Currie andZhao (1993: 2057) who stated that "the 10 th presacral vertebra of IVPP 10600 [Sinraptor] is identified as a cervical, although it is morphologically identical to the 10 th presacral of Allosaurus which is a dorsal. The identification is based on the anatomy associated with the rib". Hence, the simple evaluation of iso− lated vertebral elements in dinosaurs without identifying cor− responding ribs will not give an realistic impression of trunk length. Third, and most problematically, we have identified a number of cases where Jones et al. (2000) provide measure− ments (to a resolution, in some cases, of 1 mm) for bones that do not exist-they are not preserved with the specimen numbers indicated. A second issue is sampling. The measurement sample presented by Jones et al. (2000) cannot be considered to be an unbiased tabulation of non−avialan theropod taxa. Notable by their absence, for example, are Archaeopteryx and Sino− rnithoides. The type, and only known specimen, of Sino− rnithoides was deleted (Jones et al. 2000) because it is pur− ported to be a juvenile; however, Russell andDong (1993: 2164) indicate that "the animal was immature but approach− ing maturity upon death". Interestingly, Sinornithoides, has been placed phylogenetically within Troodontidae and is therefore puportedly more closely related to avialans (e.g., Holtz 1994a; Gauthier 1986; Sereno 1997Sereno , 1999Norell et al. 2001) than the majority of the taxa sampled by Jones et al. (2000). In addition, this taxon was reported to fall out on the "bird line" before removal by Jones et al. (2000), yet no other troodontids appeared in their study. Conversely, another taxon, Eustreptospondylus, which is known to be a subadult (Molnar et al. 1990) was included in the final analysis. Rea− sons for the exclusion of Archaeopteryx from the analysis re− main unclear; presumably because this taxon has been shown to have been volant (Rayner 1991(Rayner , 2001. However, since at the time, before the discovery of Jeholornis (Zhou and Zhang 2002), it was the only well−preserved avialan with a long tail, clear definition of its locomotor capabilities seem crucial to the Jones et al. (2000) analysis. 104 ACTA PALAEONTOLOGICA POLONICA 50 (1), 2005 Total hindlimb length (mm) "Effective hindlimb" length (mm) Trunk length (mm) Trunk length (mm) Fig. 3. The graphs presented by Jones et al. (2000). A. Total hind limb length against trunk length. B. Effective hind limb length against trunk length. Regression statistics based on our re−analysis are given in Table 1 (see Jones et al. 2000 for original statistics). This figure redrawn with per− mission from Nature (Jones et al. 2000), copyright (2000), Macmillan Mag− azines Ltd.
Finally, we note severe difficulties with measurements of trunk lengths reported by Jones et al. (2000) for three speci− mens of Caudipteryx (Appendix 1) as one of us (MAN) has spent significant time studying these specimens. We feel that the measurement of trunk length are at best imprecise, and at worst (e.g., in the case of NGMC 97−9a which preserves only a few fragments of the dorsal vertebrae and no ilia) hypo− thetical.
Phylogenetic control.-Despite the number of phylogenetic studies that have supported the placement of Caudipteryx within Oviraptoridae (e.g., Currie et al. 1998;Holtz 1998;Clark et al. 1999;Sereno 1999;Norell et al. 2001), only one other example (Ingenia) of these taxa was considered by Jones et al. (2000); no analyses were presented comparing ei− ther the hind limbs or trunk length of Caudipteryx to existing (and largely well−preserved) specimens such as Oviraptor (e.g., IGM 100/42). Further, and as discussed above, speci− men IGM 100/30 of Ingenia lacks almost all of its dorsal ver− tebrae. Using specimens on loan to the AMNH from the IGM, we took measurements of hind limbs and estimated trunk lengths for two exceptionally well−preserved ovirap− torids, IGM 100/1002 and IGM 100/973 (Khaan; Clark et al. 2001). Although both of these specimens are preserved in al− most complete articulation, we noted differences in up to 20 percent when trunk length was measured based on the total extent of the dorsal vertebral series compared to taking indi− vidual measurements from each vertebral centrum. Given this percentage uncertainty when working even with well− preserved fossil material, the accuracy of the measurements presented by Jones et al. (2000) remains unclear.
Reanalysis of Jones et al. (2000) Ignoring all the assumptions we have highlighted above, we reproduced the results presented by Jones et al. (2000) by use of their data. Following their methods, we calculated linear regressions for each data subsample to the exclusion of Caudipteryx. This taxon was then overlain onto the resultant regression lines.
Having replotted both total and effective hind limb lengths against trunk length, we then used a standard f−test (as done by Jones et al. 2000) to test for significant differ− ences between the slopes and intercepts of the regression lines. Results show that for both "total" and "effective" hind limb length, there is significant difference between the slopes of regression lines, although their intercepts are different (Table 1). Jones et al. (2000) recombined their "theropod" and "bird" subsets for further analysis. To test the significance of this fur− ther assumption, we conducted another standard t−test in order to make pairwise comparisons between the intercepts of the three regression lines and did find significant differences be− tween the lines for ornithopod dinosaurs and those for thero− pods and birds (Table 1). Separation of the measurement data for ornithopods may be supportable, this is likely not the case for theropods or birds.
Although the question of the relative limb proportions of Caudipteryx is interesting (Christiansen and Bonde 2002), any consideration of this problem must incorporate rigorous phylogenetic control, especially with regard to included data for Neornithes. Although terrestriality has evolved at least six times within extant bird clades (Gatesy 1991) these events are not directly comparable because they are disparate phylogenetically.

Hind limb proportions of Caudipteryx
In order to further test the hypothesis of Jones et al. (2000) -the hind limbs of Caudipteryx are significantly different from those of non−avialan theropod dinosaurs, more similar to those of terrestrial birds-we assembled a data−set of osteological measurements (Appendix 2). Because of the nu− merous measurement problems discussed above, we did not consider further the parameter of total trunk length. Our data set of measurements for both birds and theropods consists of the component segments of the hind limb, obtained either by direct measurement of specimens or from the relevant litera− ture (e.g., Magnan 1922;Böker 1935;Gatesy 1991;Hazle− hurst 1992;Holtz 1994b;Gatesy and Middleton, 1997;Dyke 2000;Dyke and Rayner 2002;Nudds et al. 2004). The measurement data were subdivided according to the phylogenetic rationale outlined above and plotted the three component segments of the hind limb against total leg length. On this basis, and considering the length of the fe− mur against total leg length, a very well−defined linear cor− relation is recovered (Fig. 4). Significant differences in this plot can be ascertained between the principal divisions of the data as defined, theropods (r 2 = 0.90), birds (including Archaeopteryx; r 2 = 0.81) and oviraptorosaurs (including Caudipteryx; r 2 = 0.98). Both non−avialan theropods and avialans exhibit a wide range of femur lengths (Gatesy and Middleton 1997), but in general the length of this element is well−correlated with the total length of the leg. Non−avialan theropods are distributed across the trend line in a manner which does approximate recent phylogenetic hypotheses for the group (Fig. 2). The ornithomimids (e.g., Archaeo− rnithomimus, Gallimimus) with long overall leg length and femur length cluster on the right−hand side of the trend line; oviraptorosaurs (with the exception of the much larger specimen IGM 100/973), including the three specimens of Caudipteryx, cluster at the base of the trend in the left hand side of the diagram (Fig. 4). In these taxa, the femur contrib− utes about one−third of the total leg length, as is seen in many Neornithes as well as the basal avialan Archaeopteryx and the maniraptoran Protarchaeopteryx robusta (Ji et al. 1998).
Our plots of tibia length against total leg length also re− veal two distinct trends within the bird and theropod data (Table 2; Fig. 4). Much of this variation, however, is con− tained within the ratites and non−avialan theropods other than oviraptors (including Caudipteryx). The non−avialan thero− pod included in our data with the shortest tibia to total leg length ratio is Protarchaeopteryx; Caudipteryx clusters with other small oviraptorosaurs (again with the exception of IGM 100/973) at the base of the trend lines along with some of the smaller ornithomimid specimens (e.g., Gallimimus), Saurornithoides and Archaeopteryx (Fig. 4). Regression co− efficients for the two principal subsets of the data are signifi− cant (Table 2), but given the position of Caudipteryx within the basal convergence of the two trend lines (dividing our measurement data into non−avialan theropods and Avialae), this taxon cannot be definitively grouped within either sub− sample.
Our data for metatarsal lengths (i.e., either the tarsometa− tarsus in Avialae or metatarsal III in non−avialan theropods) vary widely both within, and between, taxa (Fig. 4). Non− avialan theropods are distributed all across this graph; the three Caudipteryx specimens group with one specimen of Gallimimus and Saurornithoides (Fig. 4). Again, the non− avialan theropod with the shortest metatarsal III compared to total leg length is Protarchaeopteryx.

Conclusions
By use of proportional comparisons between hind limb and trunk lengths, Jones et al. (2000) purported to demonstrate that Caudipteryx zoui had both a locomotor strategy and limb proportions similar to extant "cursorial" birds. Jones et al. (2000) claimed that interpretations of specimens of Caudi− pteryx (based on phylogenetic analyses) as a small feathered non−avialan theropod should be reevaluated in light of these results-in other words, the overwhelming number of osteo− logical similarities evident between Caudipteryx and non− avialan theropods are not the result of evolutionary relation− ship.
We have shown that the majority of the conclusions pre− sented by Jones et al. (2000) are based on the a priori assump− tion that Caudipteryx is an avialan and that Avialae is unre− lated to non−avialan theropod dinosaurs. The most important conclusion made by Jones et al. (2000), that Caudipteryx had a locomotor strategy similar to that of extant "cursorial" birds, is dependant on the fact that the limbs of this taxon are treated as if it were a bird prior to inclusion in the analysis. Although Jones et al. (2000) did not directly claim that Caudipteryx is actually related to one of the diverse extant clades of Neo− rnithes that are "cursorial", they did imply that this taxon dem− onstrates some sort of "trend" or parallelism with extant birds in its "bauplan" (Ruben and Jones 2001).
We have cast significant doubt upon both the primary specimen data and conclusions presented by Jones et al. (2000). The majority of the non−avialan theropod specimens measured by these workers are shown to be either too incom− plete to allow for replication of their measurements, or sim− ply do not exist. Furthermore, our own analysis, using much additional measurement data and incorporating phylogenetic control, supports the view that the locomotor capabilities of neornithines are similar to their closest non−avialan theropod relatives, including taxa such as Caudipteryx.

Appendix 1
List of taxa As discussed in the text, because recent phylogenetic studies have demonstrated that non−avialan theropods are relevant to the issue of avialan origins and body plan evolution, we re− view here the specimens of non−avian theropods cited by Jones et al. (2000). Measurements taken from these speci− mens were used in our attempts to reproduce the graphs and conclusions of Jones et al. (2000). Please note that through− out this section specimens numbers are listed as cited by Jones et al. (2000).
Afrovenator UC OBA 1.-The original figure published by Sereno et al. (1994) indicates that the thoracic column of Afro− venator is extremely fragmentary. Indeed, as few as four ver− tebral elements may be preserved (Sereno et al. 1994) making measurement of trunk length impossible for this taxon.
Albertosaurus AMNH 5458.-This is an excellent specimen preserving all of the relevant bones for the study of Jones et al. (2000). However, the femur length reported (1025 mm) is identical to that given by Russell (1970: table 1) where only an estimate is provided. It is further unclear how a measure− ment for trunk length for this taxon was derived, as this quan− tity was not reported by Matthew and Brown (1923). AMNH 5458 has been on display and behind glass at the AMNH for more than 40 years thus rendering any measurement of this specimen impossible.
Ceratosaurus USNM 4735.-According to Gilmore (1920), the actual number of vertebrae in the dorsal column is un− known. Gilmore (1920) notes that in the mounted reconstruc− tion of this specimen at least one additional vertebra is included.
Carnotaurus.-No museum number was reported by Jones et al. (2000) for this taxon. Presumably, reference is made to MACN CH 894 since this is the only described specimen of Carnotaurus. Although the vertebral column is complete in this taxon, Bonaparte et al. (1990: 31) state that the tibiae are "represented only by their proximal parts", and that no meta− tarsal bones were found with the specimen. Yet Jones et al. (2000) provide lengths for both metatarsal III and tibia for Carnotaurus.
Coelophysis AMNH 7224.-The metatarsals of this speci− men are reconstructed (MAN, personal observations), and as a result of flattening it is hard to estimate the total number of dorsal vertebrae. Hence measurements of these quantities are problematic.
Daspletosaurus AMNH 5438.-Inclusion of this taxon in this morphometric analysis (as well as that of Jones et al. 2000) is impossible because AMNH 5438 consists of only a sacrum, a right femur and a single metatarsal.
Deinonychus MCZ 4371.-This specimen includes a very well−preserved hind limb and pelvis. However, as noted by Ostrom (1976: 2) and Peter Makovicky (personal communi− cation 2002), the dorsal vertebrae are not well enough pre− served to allow accurate measurement. Even by use of the Deinonychus reconstruction given in Ostrom (1976: 3) we were unable to reconcile the measurement of 601 mm given by Jones et al. (2000) for trunk length. We estimate that this length was approximately 503 mm.
Dilophosaurus UCMP 37302.-According to Welles (1984), many of the vertebrae in this specimen are extremely crushed thus making any measurement of trunk length problematic. For instance, Welles (1984: 113) states in the description of dorsal 2 that: "in lateral view, the centrum is 78 mm long above and 70 below". Similar distortions as a result of preser− vation are also reported for dorsals 5 and 6 (Welles 1984: 116), and in dorsal six: "this and the next three were rotated 180 de− grees to the right so that their spines pointed ventrally. The centrum is crushed just below the center, the arch is pushed forward. The centrum is similar to the preceeding but its length has been increased from an estimated 88 mm to 113 mm by the crushing" (Welles 1984: 116). Most of the pre− served vertebrae of Dilophosaurus show clear variance be− tween dorsal and ventral centrum lengths (Welles 1984).
Elaphrosaurus HMN Gr S 38−44.-Janensch (1920: 225) described only 10 dorsal and 7 cervical vertebrae for this specimen. In addition, a number of the vertebrae in this spec− imen have been substantially reconstructed (Peter Mako− vicky, personal communication 2002). Hence measurement of trunk length in this taxon is impossible.
Eoraptor PVSJ 512.-This specimen is reasonably complete and includes both hind limbs and a presacral series (MAN, personal observations). However, since a detailed osteological treatment of this taxon has not yet been published, it is difficult to verify the measurements reported by Jones et al. (2000).
Eustreptospondylus OUM J13558.-This is a reasonably well−preserved specimen that includes the hind limbs (in− cluding the feet) and pelvis. However, a number of the verte− brae are reconstructed, and hence the entire vertebral series may not be complete. In addition, this specimen is a juvenile (Molnar et al. 1990).
Gorgosaurus NMC 2120.-This is an excellent and nearly complete specimen (Lambe 1917;Russell 1970). However, the poor preservation of the femur and metatarsal III led Lambe (1917: 76) and later Russell (1970: table 1) to report only approximate measurements for these elements. The trunk region of this specimen is also incompletely preserved. Rus− sell (1970: table 1) did not provide a measurement for this part of the skeleton (yet he does for other tyrannosaur specimens in the same table). Lambe (1917: 24) indicated that the centra of dorsals 6-9 are not well enough preserved to measure.
Herrerasaurus PVL 2566.- Reig (1963) noted that the ver− tebral column of this specimen is incomplete. A measure− ment of total trunk length for this specimen is impossible.
Ingenia IGM 100/30.-The type specimen of this taxon is a well−preserved skeleton but including only a few fragments of the dorsal vertebrae. As is visible in the figure provided by Psihoyas (1994: 211; MAN, personal observations), it is im− possible to reconstruct an accurate trunk length for this speci− men (as measured by Jones et al. 2001).
Lilliensternus HMN R1291.-This taxon consists of two partial skeletons. Reconstructions of Lilliensternus (derived from Huene 1934) are based on a composite of the two dif− ferently sized specimens (Glut 1997). Although the limbs are complete, both lack a complete sacrum. The only dorsals that are preserved are numbered 1-3 and 12-14 (Huene 1934), al− though some other isolated bone fragments may also form part of this series. Rowe and Gauthier (1990) suggested that Lilliensternus may be a subadult (on the basis of the lack of fusion between the tarsus and the tibia).
Sinosauropteryx NIGP 127587.-This specimen is well enough preserved for all of the measurements reported by Jones et al. (2000) to be replicated.
Sinraptor IVPP10600.-This specimen is complete in all ar− eas measured by Jones et al. (2000) (Currie and Zhao 1993). Yet, see our comments above about distinguishing dorsal from cervical vertebrae.
Staurikosaurus MCZ 1669.-Although the skeleton of this taxon is relatively complete, Colbert (1970) and personal ob− servation (MAN) indicate that there are no metatarsals pre− served. This measurement was nevertheless included by Jones et al. (2000).
Struthiomimus AMNH 5339.-This is a nearly complete specimen (Osborn 1917) from which all of the relevant mea− surements can be taken.
Syntarsus QVM QG/1. -Raath (1969) notes that at least the anterior three dorsal vertebrae of this specimen are not pre− served: "the first vertebra preserved in the present specimen is dorsal 4" (Raath 1969: 2). The remainder of the hind limb and pelvic elements are complete and can be measured.
Tyrannosaurus CM 9380.-According to Osborn (1906: 282), this specimen lacks a number of dorsal vertebrae. Al− though Osborn (1906) does indicate that metatarsal III is complete, examination of his figure (and accompanying ta− ble ;Osborn 1906: 282) shows that only the distal end of this element is preserved.
Velociraptor GI 100/25.-This is a nearly complete speci− men that has never been adequately described. All of the ele− ments that are measured by Jones et al. (2000), are preserved on this specimen. However, their measurements do not cor− respond with the actual specimen. For instance they list the femur length as 200 mm, when in fact it is 185 mm in length. The tibia is listed by Jones et al. (2000) as 210 mm in length when in fact it is 225 mm (231 with the astragalus), metatar− sal III is 108, not the 95 mm reported. Similarly the twisted nature of the specimen makes accurate (to 1 mm resolution) measurement of the dorsal series impossible.
Yangchuanosaurus CV 00215.-This specimen is reasonably complete, having a good vertebral series, but incomplete hind limbs (Molnar et al. 1990). Both the metatarsals and feet are unknown for CV 00215 (Molnar et al. 1990;Philip Currie, personal communication 2002), hence hind limb length can− not be calculated for this taxa. Glut (1997) further indicated that this specimen may be a subadult.
Caudipteryx IVPP (uncatalogued).-The correct museum number for this specimen is BPM 001 . While this specimen is very well−preserved, the acual num− ber of dorsal vertebrae is uncertain: "there appear to be only 9 thoracic vertebrae" (Zhou et al. 2000: 246). As is the case in other specimens of Caudipteryx (see below), the ilia are disarticulated from the sacral vertebrae which are crushed beneath them. It is therefore impossible to have an accurate impression of the relationship between the acetabulum and the dorsal vertebral column. Although the limb bones are well−preserved, a number of the measurements given by Jones et al. (2000) differ from those provided in the original specimen description . Caudipteryx V 12344.-The correct museum number for this specimen is IVPP 1240. The same problem in identifying the number of thoracic vertebrae in BPM 001 also applies to this specimen. In addition, the sacrum is crushed and ob− scured by the displaced blade of the ilia thus making accurate measurement of sacral length impossible. As a consequence, it is not possible to measure the trunk length in this specimen (the relationship between the acetabulum and the dorsal se− ries cannot be ascertained). Although the hind limbs are well−preserved, again the measurements for this elements given by Jones et al. (2000) differ from those presented in the original specimen description .