Pedal Proportions of Poposaurus gracilis: Convergence and Divergence in the Feet of Archosaurs

Transcription

1 THE ANATOMICAL RECORD 297: (2014) Pedal Proportions of Poposaurus gracilis: Convergence and Divergence in the Feet of Archosaurs JAMES O. FARLOW, 1 * EMMA R. SCHACHNER, 2 JOHN CODY SARRAZIN, 3 HENDRIK KLEIN, 4 AND PHILIP J. CURRIE 5 1 Department of Geosciences, Indiana-Purdue University, Fort Wayne, Indiana 2 Department of Biology, University of Utah, Salt Lake City, Utah 3 Department of Mechanical Engineering, University of Utah, Salt Lake City, Utah 4 Saurierwelt Pal aontologisches Museum, D-92318, Neumarkt, Germany 5 Department of Biological Sciences, University of Alberta, Edmonton, Alberta, Canada ABSTRACT The crocodile-line basal suchian Poposaurus gracilis had body proportions suggesting that it was an erect, bipedal form like many dinosaurs, prompting questions of whether its pedal proportions, and the shape of its footprint, would likewise mimic those of bipedal dinosaurs. We addressed these questions through a comparison of phalangeal, digital, and metatarsal proportions of Poposaurus with those of extinct and extant crocodileline archosaurs, obligate or facultatively bipedal non-avian dinosaurs, and ground birds of several clades, as well as a comparison of the footprint reconstructed from the foot skeleton of Poposaurus with known early Mesozoic archosaurian ichnotaxa. Bivariate and multivariate analyses of phalangeal and digital dimensions showed numerous instances of convergence in pedal morphology among disparate archosaurian clades. Overall, the foot of Poposaurus is indeed more like that of bipedal dinosaurs than other archosaur groups, but is not exactly like the foot of any particular bipedal dinosaur clade. Poposaurus likely had a digitigrade stance, and its footprint shape could have resembled grallatorid ichnotaxa, unless digit I of the foot of Poposaurus commonly left an impression. Anat Rec, 297: , VC 2014 Wiley Periodicals, Inc. Key words: Poposaurus; archosaur pedal morphology; vertebrate ichnology The early Mesozoic adaptive radiation of archosaurs produced a diversity of reptilian life forms; these animals were variously large or small; aquatic, terrestrial, or aerial; herbivorous or carnivorous; ponderous or cursorial (Ewer, 1965; Sereno, 1991; Heckert and Lucas, 2002; Dzik, 2003; Ziegler et al., 2003; Nesbitt, 2003, 2011; Desojo and Baez, 2005; Parker et al., 2005; Sen, 2005; Fraser and Henderson, 2006; Li et al., 2006; Jalil and Peyer, 2007; Kubo and Benton, 2007, 2009; Schoch, 2007; Peyer et al., 2008; Ezcurra, 2010; Heckert et al., 2010; Sues and Fraser, 2010; Brusatte et al., 2011a; Nesbitt et al., 2010, 2013a,b; Padian et al., 2010; Butler et al., 2011; França et al., 2011; Desojo et al., 2011; Lecuona and Desojo, 2011; Galton, 2012; Grant sponsor: National Science Foundation; Grant number: EAR; Grant sponsor: Dinosaur Society; Grant sponsor: NSERC; Grant number: ; Grant sponsors: American Association of Anatomists Scholar, American Association of Anatomists. *Correspondence to: James O. Farlow, Department of Geosciences, Indiana-Purdue University, 2101 East Coliseum Boulevard, Fort Wayne, IN Fax: farlow@ipfw.edu Received 26 March 2013; Accepted 22 November DOI /ar Published online 13 January 2014 in Wiley Online Library (wileyonlinelibrary.com). VC 2014 WILEY PERIODICALS, INC.

2 POPOSAURUS GRACILIS PEDAL PROPORTIONS 1023 Kammerer et al. 2012; Li et al., 2012; Niedzwiedzki et al., 2012; Parrish, 2012; Trotteyn et al., 2012, 2013; Belen Von Baczko and Ezcurra, 2013; Ezcurra et al., 2013; Desojo et al., 2013; Irmis et al., 2013; Kubo and Kubo, 2013; Langer et al., 2013; Niedziedzki et al., 2013; Peecock et al., 2013; Sereno et al., 2013; Stocker and Butler, 2013; Turner and Nesbitt, 2013). Some early Mesozoic archosaurs were bipeds, variously facultative or obligate (Kubo and Kubo, 2012). Such animals were at or near the base of the adaptive radiation of dinosaurs and their close relatives (avemetatarsalians/ornithodirans/bird-line archosaurs: Sereno and Arcucci, 1993, 1994; Benton, 1999, 2012; Langer, 2003; Butler et al., 2007; Marsicano et al., 2007; Nesbitt et al., 2007, 2009a,b; Padian, 2008; Martinez and Alcober, 2009; Rinehart et al., 2009; Butler, 2010; Irmis, 2011; Brusatte et al., 2011b; Nesbitt, 2011; Brusatte, 2012; Langer et al., 2013) but also evolved among non-ornithodiran/crurotarsan/crocodile-line archosaurs (Nesbitt and Norell, 2006; Nesbitt, 2007, 2011; Piechowski and Dzik, 2010; Sennikov, 2012; Sookias and Butler, 2013; Weinbaum, 2013). A particularly well-known crocodile-line form with body proportions suggesting bipedalism is Poposaurus gracilis from the Late Triassic of the western United States (Weinbaum and Hungerb uhler, 2007; Gauthier et al., 2011; Parker and Nesbitt, 2013). Phylogenetic and functional morphological considerations suggest that this archosaur (Fig. 1) was an erect, striding biped (Gauthier et al., 2011; Schachner et al., 2011; Bates and Schachner, 2012; Kubo and Kubo, 2013). At least superficially, the foot of Poposaurus has a very dinosaur-like gestalt (Fig. 2E). This observation prompts a more detailed, quantitative comparison of the pedal proportions of Poposaurus with those of other archosaurs. Was this crocodile-line archosaur indeed a dinosaur-mimic in details of its foot structure? If so, to what dinosaurs is its foot most similar? And what is the likelihood that footprints made by such a dinocroc might be mistaken for dinosaur footprints, and vice versa? MATERIALS AND METHODS This study builds upon previous comparisons of pedal proportions within and across groups of dinosaurs and ground birds (Farlow and Lockley, 1993; Farlow, 2001; Fig. 1. Skeleton of Poposaurus gracilis (YPM VP ) in field jackets, left lateral view. Courtesy of the Peabody Museum of Natural History, Yale University, New Haven, Connecticut, USA.

3 1024 FARLOW ET AL. Fig. 2. Gallery of diagrammatic images of pedal skeletons of selected archosauriform reptiles. Scale bars 5 50 mm. Left/right symmetry of specimens indicated by labeling digit II. (A I) Crocodile-line archosaurs. (A) Proterosuchus vanhoepeni (redrawn from Cruickshank [1972]). (B) Revueltosaurus callenderi (PEFO 34561). (C) Stagonolepis robertsoni (redrawn from Walker [1961]). (D) Ticinosuchus ferox (redrawn from Krebs [1965]). (E) Poposaurus gracilis (YPM VP ). (F) Incomplete foot of Shuvosaurus inexpectatus (TTU-P09001). (G) Postosuchus alisonae (UNC 15575). (H) Terrestrisuchus gracilis (redrawn from Crush [1984]). (I) Protosuchus richardsoni (redrawn from Colbert and Mook [1951]). (J S) Bird-line archosaurs. (J) Silesaurus opolensis (redrawn from Dzik [2003]). (K) Coelophysis bauri (MNA V3320). (L) Tyrannosaurus rex (BHI-6230). (M) Dinornis robustus (NMNZ S.28225). (N) Plateosaurus longiceps (Skelett 45, MB.R ). (O) Leptoceratops gracilis (CMN 8889). (P) Thescelosaurus garbanii (Bugenasaura infernalis [cf. Boyd et al., 2009]; LACM 33542). (Q) Parksosaurus warreni (ROM 804). (R) Tenontosaurus tilletti (OU 10132). (S) Brachylophosaurus canadensis (MOR 794). Smith and Farlow, 2003; Farlow et al. 2006, 2012, 2013) by adding new data (especially for crocodile-line archosaurs), correcting errors in earlier versions of the data base, and by analyzing the data in different ways than previously. We analyzed metatarsal (or tarsometatarsal), digital, and pedal phalangeal lengths, and in some instances phalangeal distal widths, for a diverse set of living and extinct species (Figs. 2, 3; institutional abbreviations in Table 1) of basal archosauriforms, basal suchians (including Poposaurus), basal crocodyliforms, crocodylians, basal dinosauromorphs, non-avian theropods, basal birds, living and extinct ground birds (palaeognaths as well as neognaths), and bipedal or potentially bipedal basal sauropodomorphs (Eoraptor, prosauropods [cf. Mallison, 2011]) and ornithischians (Lesothosaurus, ornithopods, basal ceratopsians [cf. Senter, 2007; Lee et al., 2011]). Several issues of specimen preparation or preservation, or of data reporting, complicated our analyses. For fossil forms the number of specimens with complete phalangeal counts is often limited, and due to the vagaries of preservation it is not always possible to make all the desired measurements on all pedal phalanges. The smaller and/or more distal non-ungual phalanges are frequently missing, and the unguals themselves often have significant portions of their distal tips missing. For osteological specimens of modern taxa in museum collections, dried tissues frequently envelop metatarsals and phalanges, making it difficult or impossible to measure phalangeal dimensions. In osteological specimens of extant crocodylians, the horny claw sheaths often are so strongly attached to their underlying unguals that they cannot easily be removed for measuring the unguals. Finally, published descriptions of

4 POPOSAURUS GRACILIS PEDAL PROPORTIONS 1025 Fig. 3. CT scan of the foot of a large (total animal length from snout to tail tip 3 m), male Alligator mississippiensis, Rockefeller Wildlife Refuge, Louisiana. Note that soft tissues surrounding the digits result in greater splaying of the digits than would be expected from the skeleton alone. Scale bar 5 20 mm. fossil feet that report measurements of pedal phalanges generally are not explicit about how the measurements were made; in our experience, differences in phalangeal proportions due to different measurement protocols are most striking for ungual rather than non-ungual phalanges. We therefore did parallel analyses using two different data sets. Data set 1 consisted of lengths of the nonungual phalanges of digits I, II, and III, and the first three phalanges of digit IV, and the length of metatarsal III. The measurements in data set 1 were made by ourselves, or taken from the literature, the latter without worrying about the details of how measurements were made by other authors. In some cases, phalanx lengths were estimated from scaled drawings of foot skeletons. Data set 1 therefore provided an extensive coverage of taxa, and a large number of specimens, but at the expense of measuring fewer bones, and risking operator error in measurements due to different procedures possibly being followed by different authors. (The cost of measuring fewer bones may not be excessive, however; Farlow et al. [2013] found that a canonical variate analysis [CVA] based on the dimensions of just the first two phalanges of digits II IV performed almost as well as a CVA based on the dimensions of all the phalanges of those digits in discriminating among the foot skeletons of tridactyl ratites.) Data set 2 consisted of measurements made by ourselves, or by others following our explicit measurement protocols (Fig. 4), of the lengths of metatarsal III and all pedal phalanges, and of the distal widths of phalanges II2, III2, and IV2. As in previous studies (Farlow and TABLE 1. Institutional abbreviations Abbreviation Institution Location AM Auckland War Memorial Museum Auckland, New Zealand AMNH American Museum of Natural History New York BHI Black Hills Institute of Geological Research Hill City, South Dakota CEUM Utah State University Eastern Prehistoric Museum Price, Utah CM Canterbury Museum Christchurch, New Zealand CMN and GSC Canadian Museum of Nature Ottawa, Ontario, Canada DMNH Denver Museum of Nature and Science Denver, Colorado FMNH Field Museum of Natural History Chicago, Illinois IRSNB Institut Royal des Sciences Naturelles de Belgique Brussels, Belgium LACM Natural History Museum of Los Angeles County Los Angeles, California MB Museum f ur Naturkunde Berlin, Germany MCF Museo Municipal Carmen Funes Neuquen, Argentina MNA Museum of Northern Arizona Flagstaff, Arizona MOR Museum of the Rockies, Montana State University Bozeman, Montana NHM Natural History Museum London, England NMNZ Museum of New Zealand Te Papa Tongarewa Wellington, New Zealand OU & OMNH Sam Noble Oklahoma Museum of Natural History Norman, Oklahoma PEFO Petrified Forest National Park Arizona PU Princeton University (Yale Peabody Museum) New Haven, Connecticut PVSJ Museo de Ciencias Naturales, Universidad de San Juan San Juan, Argentina QM Queensland Museum Brisbane, Australia ROM Royal Ontario Museum Toronto, Ontario, Canada RTMP Royal Tyrrell Museum of Palaeontology Drumheller, Alberta, Canada SMA Sauriermuseum Aathal Aathal, Switzerland TTU Texas Tech University Lubbock, Texas UCMP University of California Museum of Paleontology Berkeley, California U Illinois University of Illinois Museum of Natural History Urbana, Illinois UNC University of North Carolina Chapel Hill, North Carolina USNM National Museum of Natural History, Smithsonian Institution Washington, D.C. YPM Peabody Museum of Natural History, Yale University New Haven, Connecticut

5 1026 FARLOW ET AL. Fig. 4. Single phalanges of Poposaurus illustrating protocols for measuring pedal phalanges. (A) Lengths of non-ungual phalanges. (B) Distal widths of non-ungual phalanges. (C) Lengths of ungual phalanges. Lockley, 1993; Farlow, 2001; Smith and Farlow, 2003; Farlow et al., 2006, 2012, 2013), lengths of non-ungual phalanges were measured from the dorso-ventral midpoint (or close to it) of the concave proximal edge of the bone to the dorsoventral midpoint of the convex distal end of the bone; where possible, this measurement was made on both the medial and lateral sides of the bone, and the measurements averaged. The maximum transverse width of non-ungual phalanges was measured across the distal articular end of the bone. Ungual lengths were measured in a straight-line fashion from the dorsoventral midpoint of the concave proximal edge of the bone to its tip, again averaging measurements of the medial and lateral sides, when possible. Frequently composite measurements of the left and right feet of a single individual were made, when a complete set of measurements could not be made on the left or right foot alone. Data were analyzed in several ways. Multivariate analyses were employed to analyze phalangeal lengths (and sometimes widths). Principal components analyses (PCAs) were done using log-transformed data and a covariate matrix. Two versions of PCA were performed. The first used data set 1 and the log-transformed lengths of phalanges I1, II1-2, III1-3, and IV1-3. To enable inclusion of taxa in which digit I has been lost, 5 mm was added to lengths of phalanx I1 of all specimens prior to log transformation. The second PCA employed data set 2 and the lengths of all phalanges of digits I-IV, and also the distal widths of phalanges II2, III2, and IV2 (which are positioned about halfway out along the lengths of their respective toes). We added 5 mm to the lengths of phalanges I1 and I2 to allow inclusion of taxa in which digit I has been lost. A problem arises with some birds (particularly some moa species) in which there are only four phalanges (including the ungual) in digit IV. Typically in ground birds and dinosaurs the smallest bone in digit IV is the final non-ungual phalanx, IV4, which can be very short (e.g. in Rhea and Pterocnemia). As in a previous study (Farlow et al., 2013), in forms with only four phalanges on digit IV we treated the ungual as phalanx IV5, and assigned phalanx IV4 a length of zero. In our PCA with data set 2, instead of using separate log-transformed lengths of both IV3 and IV4 (which would have necessitated discarding specimens in which IV4 had zero length), we used the log-transformed summed lengths of IV3 and IV4. Cluster analyses of phalangeal lengths and widths using data set 2 were performed after first logtransforming all variables used in the analysis, computing the mean of all the log-transformed variables, and finally scaling each log-transformed variable by subtracting the mean of all the log-transformed variables in the analysis from each log-transformed variable. Cluster analyses used between-groups linkage employing the squared Euclidean distance. Finally, simple bivariate comparisons were made whenever possible, to increase the sample size involved in the comparison. Because the number of specimens representing any given taxon in any particular bivariate comparison was highly variable (some species, such as Alligator mississippiensis, were represented by many specimens, but most species were represented by one or a few specimens), no attempt was made to test statistically significant differences among groups in bivariate comparisons. Instead we looked for consistent visual patterns in which data cases for particular groups plotted in different regions of graphs. For some bivariate graphs, we log-transformed data, to better show points for both very large and very small forms. In other bivariate graphs, we did not log-transform data before plotting, but rather restricted the plotted data to forms comparable to or smaller than Poposaurus in size. This was a strictly ad hoc decision, based on which form of graph was judged better to show how Poposaurus compared with other taxa. All data analyses were performed in IBM SPSS version 20. The data upon which analyses were based are available from Farlow. In the graphs that follow (Figs. 5-8, 10, 11), limitations in the number of different symbols the plotting routines permitted us to use forced some uncomfortable choices in labeling points. We tried to use consistent symbols and color schemes (Table 2). Where possible to do so without creating excessive clutter, we also labeled points (or groups of points) for individual genera (Table 2). Some of the groups used in the graphs are paraphyletic or otherwise idiosyncratic. Thus Archaeopteryx was given its own symbol, and not included with other basal birds, partly because of the intrinsic interest of this form, and partly because of controversy over exactly how it relates to non-avian theropods and avialians (Xu et al., 2011b; Lee and Worthy, 2012).

6 Fig. 5. Principal components analyses of phalangeal proportions. (A) Log-transformed lengths of phalanges I1, II1 2, III1 3, and IV1 3 of non-avian dinosaurs, basal birds, ground birds, and crocodile-line archosaurs (data set 1; label abbreviations in Table 2). PC 1 is positively associated with all phalanx lengths, making it a proxy for overall size. Positive values of PC 2 are associated with a relatively long phalanx I1, and less strongly with relatively long non-ungual distal phalanges of digits II-IV (IV3, II2, IV2, and III3). Negative values of PC 2 are associated with relatively long proximal phalanges of digits II-IV (IV1, III1, II1, and III2). The data are separated into an upper band of forms in which digit I is present and a lower band of forms in which digit I is absent. In this comparison, Poposaurus is most like basal sauropodomorphs ( prosauropods ) and basal ornithopods ( hypsilophodontids ), but also continues the trend defined for crocodylians. (B) Log-transformed lengths of all phalanges, and the distal widths of phalanges II2, III2, and IV2, of non-avian dinosaurs, ground birds, Revueltosaurus, and Poposaurus (data set 2). PC 1 is positively associated with all phalanx lengths but now also with distal widths, once again making it a proxy for overall size. Positive values of PC 2 are associated with relatively long phalanges I1 and I2 and less strongly with relatively long phalanges from the middle portions of digits II IV. Negative values of PC 2 are associated with relatively broad phalanges, and with relatively long first phalanges and unguals of digits II IV. The data are separated into an upper band of forms in which digit I is present, and a lower band of forms in which digit I is absent. In this comparison, Poposaurus is most like non-avian theropods, basal sauropodomorphs, the moa Megalapteryx, and small to midsized ornithischians. Taxon abbreviations in Table 2.

7 1028 FARLOW ET AL. TABLE 2. Abbreviations of genus or group names (alphabetical order) used in Figures 5-8, 10, and 11, and the group names associated with each taxon in the figure symbol keys Abbreviation Genus or group Al Allosaurus (Allosauroids) Alb Albertosaurus (Tyrannosauroids) Ali Alligator (Crocodylians) An Anomalopteryx (Dinornithiforms) Apte Apteryx (Struthioniforms) Apto Aptornis (Neoaves) Ar Ardeotis (Neoaves) Arc Archaeoceratops (Ornithischians) Ari Araripesuchus (Basal Mesoeucrocodylians) Au Aucasaurus (Ceratosaurs) Bm Bambiraptor (Dromaeosaurids) Ca Cariama (Neoaves) Cai Caiman (Crocodylians) Cas Casuarius (Struthioniforms) Cer Cerasinops (Ornithischians) Ch Chunga (Neoaves) Chi Chirostenotes (Oviraptorosaurs) Chr Chanaresuchus (Basal Archosauriforms) Chry Chrysolophus (Galloanserae) Cm Camptosaurus (Ornithischians) Co Confuciusornis (Basal Avialians) Cro Crocodylus (Crocodylians) Cyr Cyrtonyx (Galloanserae) Das Daspletosaurus (Tyrannosauroids) De Deinonychus (Dromaeosaurids) Di Dilophosaurus (Coelophysoids) Dn Dinornis (Dinornithiforms) Dr Dromaius (Struthioniforms) Dry Dryosaurus (Ornithischians) Eud Eudromia (Tinamiforms) Eup Eupodotis (Neoaves) Eupa Euparkeria (Basal Archosauriforms) Eur Euryapteryx (Dinornithiforms) Gav Gavialis (Crocodylians) Ge Genyornis (Galloanserae) Gg Gorgosaurus (Tyrannosauroids) Glm Gallimimus (Ornithomimosaurs) Gs Gastornis (Galloanserae) Ha Hadrosaurids (several genera; Ornithischians) Het Heterodontosaurus (Ornithischians) Hng Hongshanornis (Basal Ornithuromorphans) Hy Haya (Ornithischians) Ibe Iberomesornis (Basal Avialians) Ig Iguanodon and Mantellisaurus (Ornithischians) Lg Lagerpeton (Basal Dinosauromorphs) Lp Leptoceratops (Ornithischians) Lph Lophura (Galloanserae) Lu Lufengosaurus (Basal Sauropodomorphs) Me Mesitornis (Neoaves) Mg Megalapteryx (Dinornithiforms) Mi Microceratops (Ornithischians) Mo Montanoceratops (Ornithischians) Ms Massospondylus (Basal Sauropodomorphs) Or Orodromeus (Ornithischians) Ost Osteolaemus (Crocodylians) Ot Otis (Neoaves) Oth Othnielosaurus (Ornithischians) Pa Pachyornis (Dinornithiforms) Pal Palaeociconia (Neoaves) Ph Phasianus (Galloanserae) Abbreviation Pk Pl PL Plsh Post Pr Prt Ps Pso Pst Ptg Ptx Re Rhn Rhs Sa Sat Scn Sgt Sgx Sh Sn St Ste Str Te Th To Tu Tx Ty TABLE 2. (continued) Genus or group Parksosaurus (Ornithischians) Plateosaurus (Basal Sauropodomorphs) Proctor Lake ornithopod (Ornithischians) Paleosuchus (Crocodylians) Postosuchus (Basal Suchians) Protosuchus (Basal Crocodyliforms) Proterosuchus (Basal Archosauriforms) Psilopterus (Neoaves) Psophia (Neoaves) Psittacosaurus (Ornithischians) Patagopteryx (Basal Ornithuromorphans) Protopteryx (Basal Avialians) Revueltosaurus (Basal Suchians) Rhynchotus (Tinamiforms) Rhea and Pterocnemia (Struthioniforms) Saurosuchus (Basal Suchians) Saturnalia (Basal Sauropodomorphs) Scansoriopteryx (Dromaeosaurids) Sagittarius (Neoaves) Saurophaganax (Allosauroids) Shuvosaurus (Basal Suchians) Sinornis (Basal Avialians) Struthio (Struthioniforms) Steneosaurus (Basal Mesoeucrocodylians) Struthiomimus (Ornithomimosaurs) Tenontosaurus (Ornithischians) Thescelosaurus (including Bugenasaura; Ornithischians) Tomistoma (Crocodylians) Turnix (Neoaves) Typothorax (Basal Suchians) Tyrannosaurus (Tyrannosauroids) The genera listed here are not the only ones used in the analyses, but merely the ones that could be labeled on the graphs. Color conventions for points belonging to groups in graphs employing data set 1: green 5 crocodyliforms; dark blue 5 crocodile-line archosaurs other than crocodyliforms; black 5 basal dinosauromorphs; red 5 non-avian theropods and Archaeopteryx; light blue 5 basal sauropodomorphs; light gray 5 paleognaths; yellow 5 ornithischians. There are a few departures from these color conventions in graphs employing data set 2. The relationships of neornithine paleognaths are particularly controversial (cf. Livezey and Zusi, 2007; Harshman et al., 2008; Bourdon et al., 2009; Phillips et al., 2010; Smith et al., 2013). Different analyses agree in finding the two extant rhea species (Rhea americana plus Pterocnemia pennata), the emu (Dromaius novaehollandiae) plus the cassowaries (Casuarius spp.), and the various moa species (dinornithiforms), respectively, to constitute clades. In contrast, the relationships among tinamous, moa, and the extant ratites vary markedly among studies. In our graphs we will use separate symbols for lithornithids, tinamiforms, and dinornithiforms. We will label emu, cassowaries, rheas, ostrich, and kiwi collectively as struthioniforms (Clements, 2007) but for convenience more than anything else. Indeed, a recurrent theme in our analyses (as in Farlow et al., 2013) will be how kiwi (Apteryx) consistently differ from other extant ratites in pedal proportions. To create a three-dimensional digital model of the distal hindlimb and pes of Poposaurus, the left femur, tibia, fibula, metatarsals, and pes were scanned with a Polhemus FastSCAN cobra laser scanner (

8 POPOSAURUS GRACILIS PEDAL PROPORTIONS 1029 com), which generated a high-resolution, digital, threedimensional point cloud. All data post processing and image creation was performed using Geomagic Studio 12 ( As each element was scanned in multiple parts, it was necessary to re-align individual scans to produce a single 3D mesh of each bone. Individual scans were first manually aligned using the pick points function, after which their alignment was refined using the surface-matching algorithm on overlapping areas. Once all scans of an element were maximally registered, mesh repair and wrapping were applied to seal the two halves, to fix any non-uniformities in the point cloud data, and to ultimately produce congruent water tight meshes for further utilization (for more information on 3D models and mesh production see the Supplemental Information of Bates and Schachner [2012]). Special attention was paid to the calcaneal/astrageal and phalangeal scans, where multiple bones were fossilized as one unit and could not be separated during preparation without risking extensive damage. After the appropriate scans were mated, the available exposed surface of each constitutive bone was systematically removed from the overall scan computationally, and the holes left by the unexposed surface area in the fossil clump were repaired in Geomagic. Some liberty was taken in reconstructing surfaces where the scans were incomplete, but anatomical insight and comparison with the opposite limb allowed for increased accuracy. Once a complete surface mesh was created for each fossilized bone, the entire lower limb was rearticulated into the various configurations for further analysis, based upon wear facets and the original pedal articulation of the fossil specimen. Anatomical abbreviations for features of our digital model of the Poposaurus foot are in Table 3. Abbreviation a c f mt p t TABLE 3. Anatomical abbreviations Meaning of Abbreviation Astragalus Calcaneum Fibula Metatarsal Phalanges Tibia We used the proportions of the foot skeleton of Poposaurus to reconstruct what a pedal footprint of this reptile would look like, using standard procedures (Baird, 1957). The reconstructed footprint was then compared with several Triassic reptilian ichnotaxa to see if any known forms are similar. RESULTS Phalangeal and Digital Proportions in Poposaurus and Other Archosaurs Phalangeal proportions. PCA (data set 1) of logtransformed lengths of phalanges I1, II1-II2, III1-III3, and IV1-IV3 (Table 4; Fig. 5A) identifies a first principal component strongly associated with overall foot size that accounts for the bulk of data variance, as found in previous analyses of this kind (Farlow, 2001; Farlow et al., 2012, 2013). A much smaller, but still important, second principal component distinguishes forms that have a digit I, and also relatively long more distal non-ungual phalanges (positive values of PC 2), from forms that lack digit I and also have relatively long more proximal nonungual phalanges (negative values of PC 2). Plotting PC 2 from PC 1 separates specimens into two bands that diverge with increasing overall foot size. The lower band (negative values of PC 2) includes many extant struthioniforms (Casuarius, Dromaius, Pterocnemia, Rhea; Struthio was excluded from this analysis (and most other analyses in this study) because it lacks digit II as well as digit I), some tinamous (Eudromia), the dromornithid Genyornis, bustards (Ardeotis, Eupodotis, Neotis, Otis) and buttonquail (Turnix), most ornithomimosaurs (Dromiceiomimus, Gallimimus, Ornithomimus, Struthiomimus), and very large ornithopods (Iguanodon, Mantellisaurus and several genera of hadrosaurs). The divergence of this lower band from the upper band with increasing animal size is largely a result of the way phalanx 1 length was log-transformed in order to include taxa lacking digit I; with very large ornithopods in which digit I has been lost (cf. Moreno et al., 2007), the artificial 5-mm length of phalanx I1 becomes trivial compared with the lengths of phalanges in digits II IV. The upper band (positive values of PC 2) includes many genera of archosaurs and their close relatives from TABLE 4. Principal components analysis (using a covariance matrix) of log-transformed lengths of phalanges I1 (5 mm were added to all lengths of this bone prior to log-transformation to enable inclusion of specimens in which digit I has been lost), II1-II2, III1-III3, and IV1-IV3 of basal archosaurs and their close relatives, basal suchians, basal mesoeucrocodylians, crocodylians, basal dinosauromorphs, non-avian dinosaurs, and ground birds (number of specimens 5 593) Parameter Component 1 Loading (Raw [Rescaled]) Component 2 Loading (Raw [Rescaled]) I1 Length (0.308) (0.940) II1 Length (0.971) (20.186) II2 Length (0.956) (0.158) III1 Length (0.974) (20.192) III2 Length (0.972) (20.070) III3 Length (0.968) (0.077) IV1 Length (0.966) (20.195) IV2 Length (0.978) (0.092) IV3 Length (0.961) (0.180) Eigen values (% of variance) (84.291) (12.209) Cumulative variance explained (%) Kaiser-Meyer-Olkin measure of sampling adequacy ; Bartlett s test of sphericity: chi-square , p <

9 1030 FARLOW ET AL. several major groups: basal archosauriforms (Euparkeria), basal suchians (Poposaurus, Revueltosaurus, Shuvosaurus, Stagonolepis, Ticinosuchus), basal crocodyliforms (Protosuchus), basal mesoeucrocodylians (Araripesuchus, Steneosaurus), crocodylians (Alligator, Caiman, Crocodylus, Osteolaemus, Paleosuchus, Tomistoma), basal dinosauromorphs (Lagerpeton), coelophysoids (Coelophysis, Dilophosaurus, Megapnosaurus, Segisaurus), ceratosaurs (Aucasaurus), allosauroids (Allosaurus), basal coelurosaurs (Compsognathus, Juravenator, Sinosauropteryx, Tanycolagreus), tyrannosauroids (Albertosaurus, Daspletosaurus, Gorgosaurus, Tarbosaurus, Tyrannosaurus), the ornithomimosaur Harpymimus, alvarezsaurids (Albinykus, Mononykus), oviraptorosaurs (Caudipteryx, Chirostenotes, Conchoraptor, Ingenia, Khaan), troodontids (Sinornithoides, Troodon), dromaeosaurids (Bambiraptor, Deinonychus, Neuquenraptor, Scansoriopteryx, Sinornithosaurus, Velociraptor), Archaeopteryx and other basal avialians (Bohaiornis, Confuciusornis, Iberomesornis, Jeholornis, Protopteryx, Sinornis, Zhouronis), basal ornithuromorphans (Changmaornis, Gansus, Hongshanornis, Yanornis, Yixianornis), kiwi (Apteryx), moa (Anomalopteryx, Dinornis, Euryapteryx, Megalapteryx, Pachyornis), several members of Galloanserae (Gastornis and many species of currasows, pheasants, grouse, and quail), several taxa of Neoaves (Aptornis, Cariama, Chunga, Foro, Gallirallus, Idiornis, Juncitarsus, Mesitornis, Rhynochetos, Psilopterus, Psophia, Sagittarius, Salmila), basal sauropodomorphs (Gyposaurus, Lufengosaurus, Massospondylus, Plateosaurus), and several taxa of small to medium-sized ornithischians (Lesothosaurus, Heterodontosaurus, Agilisaurus, Haya, Orodromeus, Othnielosaurus, Parksosaurus, Thescelosaurus, Yandusaurus, Tenontosaurus, Camptosaurus, Psittacosaurus, Archaeoceratops, Cerasinops, Leptoceratops, Microceratops, Montanoceratops). Within the upper cluster, data for some groups represented by multiple specimens fall in a series of roughly parallel bands, all of which show more positive values of PC 2 with increasing animal size. The lowest band is for moa, with non-avian theropods, crocodylians, basal sauropodomorphs, and small to medium-sized ornithischians ( hypsilophodontids, Tenontosaurus, basal ceratopsians) plotting progressively higher. Poposaurus plots among the basal sauropodomorphs and ornithischians but also is arguably along the trend defined by crocodylians. PCA using data set 2 yields analogous results (Table 5). PC 1 again is associated with animal size, and again accounts for more than half the data variance, but not as much as in PCA using data set 1. The cumulative percentage of variance explained by principal components 1 and 2 is about the same for data sets 1 and 2, and so PC 2 accounts for substantially more of the data variance in data set 2 than in data set 1. As in PC 2 of data set 1, positive values are associated with having a digit I, and also relatively long phalanges from the middle portions of toes (that is, phalanges between the most proximal phalanx and the ungual). Negative values of PC 2 are associated with relatively long proximal phalanges of digits II-IV, as in PC 2 of data set 1, but also with relatively long unguals, and with relatively broad phalanges (particularly in large ornithopods [Moreno et al., 2007]). Plotting PC 2 against PC 1 of data set 2 (Fig. 5B) again separates the data into two bands that diverge with increasing values of PC 1 (animal size), again largely on the basis of whether digit I is absent (bustards, most tridactyl struthioniforms, most ornithomimosaurs, Iguanodon, Mantellisaurus, and hadrosaurids) or present (most non-avian theropods, kiwi [Apteryx], seriemas [Cariama, Chunga], adzebill [Aptornis], moa, basal sauropodomorphs, smaller ornithopods, basal ceratopsians, Revueltosaurus, and Poposaurus). Pedal proportions of Poposaurus are most like those of non-avian theropods, the moa Megalapteryx, the ornithischians Tenontosaurus, Thescelosaurus, and Leptoceratops, and the crurotarsan Revueltosaurus. TABLE 5. Principal components analysis (using a covariance matrix) of log-transformed phalanx lengths of all phalanges of digits I IV (5 mm were added to all lengths of phalanges I1 and of I2 prior to logtransformation to enable inclusion of specimens in which digit I has been lost), and distal widths of phalanges II2, III2, and IV2, in Revueltosaurus, Poposaurus, non-avian dinosaurs, and ground birds (number of specimens 5 108) Parameter Component 1 Loading (Raw [Rescaled]) Component 2 Loading (Raw [Rescaled]) I1 Length (0.431) (0.894) I2 Length (0.404) (0.906) II1 Length (0.924) (20.257) II2 Length (0.952) (0.097) II2 Distal Width (0.947) (20.249) II3 Length (0.933) (20.143) III1 Length (0.909) (20.250) III2 Length (0.818) (0.025) III2 Distal Width (0.909) (20.373) III3 Length (0.904) (0.101) III4 Length (0.958) (20.178) IV1 Length (0.910) (20.299) IV2 Length (0.942) (0.125) IV2 Distal Width (0.943) (20.283) Combined IV3 Length 1 IV4 Length (0.956) (0.152) IV5 length (0.966) (20.170) Eigen values (% of variance) (68.577) (25.415) Cumulative variance explained (%) Kaiser-Meyer-Olkin measure of sampling adequacy ; Bartlett s test of sphericity: chi-square , p <

10 POPOSAURUS GRACILIS PEDAL PROPORTIONS 1031 Aspects of these relationships can be further elaborated, with larger sample sizes, by selected bivariate comparisons. Principal component 2 in PCA of both data sets is closely associated with the relative length (or absence) of digit I. For those forms that retain digit I, plotting phalanx I1 length against the length of digit III excluding the ungual (Fig. 6A; data set 1) separates the various taxa into a series of distinct bands. Moa, kiwi, the lithornithid Pseudocrypturus, some tinamous, many galloanseriforms and Neoaves, Patagopteryx, Rahonavis, Archaeopteryx, Protarchaeopteryx, and most non-avian theropods plot together in a band of points with a phalanx I1 that is present, but relatively short compared with the length of digit III. Basal archosauriforms (Chanaresuchus, Euparkeria, Proterosuchus), basal suchians (Postosuchus, Revueltosaurus, Shuvosaurus, Stagonolepis, Ticinosuchus, Typothorax), basal mesoeucrocodylians (Araripesuchus, Steneosaurus), crocodylians (Alligator, Caiman, Crocodylus, Gavialis, Osteolaemus, Paleosuchus, Tomistoma, Wannaganosuchus), basal sauropodomorphs, and small to medium-sized ornithischians (Lesothosaurus, Heterodontosaurus, basal ornithopods, basal ceratopsians) comprise a heterogeneous group with a particularly long phalanx I1 compared to the length of digit III. Poposaurus (itself a basal suchian) falls within this last group, plotting among basal sauropodomorphs and hypsilophodontids, but is aligned with the trend defined by crocodylians and other non-ornithodirans. After the presence or absence of digit I, two of the variables with the greatest loadings on PC 2 (but with opposite signs) in data set 1 are the lengths of phalanges II2 (positive association) and IV1 (negative association) (Fig. 6B). Large ornithopods (hadrosaurs, Iguanodon, and Mantellisaurus) and several kinds of ground birds (bustards, Gastornis, Genyornis, many moa, and many struthioniforms) have a relatively long phalanx IV1 compared with the length of phalanx II2 (exaggeratedly so in the ostrich Struthio, in which digit II is absent, and so II2 length is assigned a value of zero). A variety of crocodile-line archosaurs, non-avian theropods, basal sauropodomorphs, ground birds, and ornithischians form a dense main sequence of points in which the length of IV1 relative to that of II2 is more modest. Poposaurus plots among this latter group of points but plots noticeably lower than the trend defined by crocodylians. In data set 2, one of the most important contrasts in PC 2 after the presence or absence of digit I involves the relative widths of phalanges. Plotting the distal width of phalanx III2 against the length of the same phalanx (Fig. 7A) separates the very stout-toed large ornithopods from other non-avian dinosaurs, ground birds, Poposaurus, and Revueltosaurus. Non-avian theropods and many ground birds are also slimmer-toed than most small to mid-sized ornithischians, and Poposaurus is more like non-avian theropods than nearly all ornithischians in this comparison. Another aspect of PC 2 in data set 2 is the length of unguals relative to the lengths of phalanges from the middle portions of digits (Fig. 7B). Most ornithischians have relatively longer unguals than do theropods (including ground birds) and Revueltosaurus, although small to mid-sized ornithopods overlap the theropod points. Poposaurus plots among the larger birds, close to points for both non-avian theropods and hypsilophodontids. One feature in which Poposaurus is distinct from many non-avian dinosaurs (but like galliform birds) is in having a rather short ungual on digit I compared with the length of phalanx I1 (Fig. 8A). In contrast, Revueltosaurus, Plateosaurus, basal avialians, basal ornithuromorphans, kiwi, moa, and some other ground birds have a relatively long ungual I2 compared with the length of phalanx I1. A cluster analysis of the scaled lengths of the phalanges of digits I-IV, along with the scaled distal widths of phalanges II2, III2, and IV2, in birds, dinosaurs, Revueltosaurus, and Poposaurus yields a very large dendrogram (Fig. 9). At low levels in the dendrogram, there are several phylogenetically reasonable groupings: large, narrow-toed moa (Dinornis plus Pachyornis); cassowaries (Casuarius); rheas (Pterocnemia and Rhea); emus (Dromaius); bustards (Ardeotis); dromaeosaurids (Bambiraptor, Deinonychus); ornithomimids (Gallimimus, Struthiomimus); specimens of Iguanodon bernissartensis; many hadrosaur specimens. There are also some phylogenetically satisfying clusters at higher levels in the dendrogram. Most specimens of extant large struthioniforms (cassowaries, emu, rheas) cluster together (cf. Farlow et al., 2013), as do Iguanodon, Mantellisaurus, and hadrosaurids. Other clusters are phylogenetically more problematic. Some allosaurs and tyrannosaurs (Allosaurus, Gorgosaurus, Tyrannosaurus) cluster together (cf. Farlow et al., 2013), and also link with moa, some ornithopods, and Revueltosaurus. Kiwi (Apteryx) cluster with the small moa Megalapteryx, the adzebill Aptornis, and Aucasaurus. Tenontosaurus plausibly clusters with the Camptosaurus and Leptoceratops, but also with the basal sauropodomorph Plateosaurus. Bustards (a group within Neoaves) cluster with extant large ratites, as do ornithomimosaurs (thereby living up to their nickname of ostrich mimics ). Large ornithopods form a cluster that links with Genyornis and with all other groups combined. Although phalangeal proportions work reasonably well at grouping congeneric (and often even conspecific) forms together, they seem to be phylogenetically unreliable at higher taxonomic levels. Of greatest interest for the present study, Poposaurus groups with the nonavian theropods Dilophosaurus, Chirostenotes, and one specimen of Gorgosaurus. Digital proportions. Comparisons of the overall lengths of digits II-IV (Fig. 10) show ornithischians generally to have relatively long digits II and IV compared with the length of digit III, while most non-avian theropods and ground birds have proportionally shorter digits II and IV compared with digit III. Dromaeosaurids, troodontids (cf. Zanno et al., 2011), basal sauropodomorphs Shuvosaurus, Revueltosaurus, and to a lesser extent crocodylians, have a relatively high digit IV/digit III length ratio, but without a correspondingly high digit II/digit III length ratio (Fig. 10A,B). Poposaurus plots among pileups of points for large theropods, oviraptorosaurs, the basal bird Confuciusornis, moa (Dinornis, Megalapteryx), the quail Cyrtonyx, and hypsilophodontids. In those ornithischians that have a digit I, this digit is generally fairly long, relative to the length of digit III (Fig. 8B). The same is true of Revueltosaurus and some Meosozic birds. In contrast, non-avian theropods and

11 1032 FARLOW ET AL. Fig. 6. Bivariate phalangeal proportions across all groups (data set 1). (A) Phalanx I1 length vs. digit III length (excluding the ungual) in archosaurs that retain digit I. Data are log-transformed to spread out cases of smaller forms; no particular structural relationship between the two variables is assumed. Poposaurus plots among or along the trends defined by basal sauropodomorphs ( prosauropods ), small to medium-sized ornithischians, and crocodylians. (B) Phalanx II2 length vs. phalanx IV1 length in archosaurs of about the same size as Poposaurus or smaller; Poposaurus has a relatively short phalanx IV1 in this comparison.

12 POPOSAURUS GRACILIS PEDAL PROPORTIONS 1033 Fig. 7. Bivariate phalangeal proportions in Poposaurus, Revueltosaurus, and dinosaurs similar to or smaller than Poposaurus in size (data set 2). (A) Phalanx III2 distal width vs. length of the same phalanx. Poposaurus is most like non-avian theropods, ground birds, and hypsilophodontids. (B) Ungual III4 length vs. phalanx III2 length. Ornithischians tend to have relatively longer unguals than do theropods (including ground birds). Poposaurus is theropod-like in this feature, but plots close to points for hypsilophodontids.

13 1034 FARLOW ET AL. Fig. 8. Size and proportions of digit I (data set 2) in those forms in which this toe is present. In both panels, data are log-transformed to spread out points for smaller taxa; no particular structural relationship between variables is assumed. (A) Ungual I2 length vs. phalanx I1 length in dinosaurs (including birds), Revueltosaurus, and Poposaurus. Poposaurus has a relatively short I2, while moa, kiwi, Revueltosaurus, and Plateosaurus have a long I2. (B) Digit I length vs. digit III length in small to medium-sized dinosaurs, Revueltosaurus, and Poposaurus. Poposaurus is similar to Revueltosaurus and small to mid-sized ornithischians in having a relatively long digit I.

14 POPOSAURUS GRACILIS PEDAL PROPORTIONS 1035 moa have a relatively short digit I. Poposaurus plots near the lower edge of the ornithischian points. Metatarsal/digital proportions. Crocodylians and most other crocodile-line archosaurs in our sample, along with Dilophosaurus, allosauroids, dromaeosaurids, Chirostenotes, moa other than Dinornis, basal sauropodomorphs most small ornithischians, Tenontosaurus, Camptosaurus, Iguanodon, and Mantellisaurus have a relatively long digit III compared with the length of metatarsal III (or the tarsometatarsus) (Fig. 11A,B). In contrast, most tyrannosaurs and ornithomimosaurs, as well as Dinornis and most other ground birds, have a proportionally shorter digit III (or proportionally longer metatarsus). (There is, however, a complication in such comparisons, created by the fact that in birds the length of metatarsal III will be inflated, compared to that of non-avian archosaurs, due to inclusion of ankle bones in the length of the tarsometatarsus). Poposaurus continues a trend defined by crocodylians (Fig. 11A) and is otherwise most similar to Deinonychus, prosauropods, and small ornithischians. Other pedal features. Metatarsal I length varies greatly relative to the lengths of metatarsals II IV in archosaurs and their close relatives (Figs. 2, 3). It is substantially shorter than the other metatarsals in basal archosauriforms like Chanaresuchus (Romer, 1972), Proterosuchus, and Euparkeria (Ewer, 1965) but is also relatively quite short or even absent in theropods (including ground birds) and derived ornithopods. In contrast, metatarsal I is only a little shorter than the other metatarsals in crocodyliforms (e.g. Protosuchus, Terrestrisuchus, and Alligator). Poposaurus is one of several phylogenetically heterogeneous taxa (along with forms like Stagonolepis, Ticinosuchus, Shuvosaurus, Plateosaurus, Tenontosaurus, basal ornithopods [ hypsilophodontids ], and Leptoceratops) in which the relative length of metatarsal I is neither particularly long nor short. However, the aggregate length of the phalanges of digit I (especially the shortness of ungual I2 [Fig. 8A]), added to the modest length of metatarsal I, causes the first toe of Poposaurus to terminate well proximally to the distal termination of digits II IV (Fig. 2E). This configuration differs from that of crocodylians and their close relatives, basal sauropodomorphs, basal ceratopsians, Tenontosaurus, and some basal ornithopods, in which the very large ungual on digit Fig. 9. Cluster analysis of scaled lengths of all phalanges of digits I- IV, and the distal widths of phalanges II2, III2, and IV2 in Revueltosaurus, Poposaurus, non-avian dinosaurs, and ground birds (data set 2: N see Table 1 for institutional abbreviations). As in the PCA (Fig. 5B), 5 mm were added to the lengths of phalanges I1 and I2 before scaling so that taxa in which digit I is lost could be included. Phalangeal measurements were scaled by subtracting the mean of all logtransformed phalanx lengths and widths from each log-transformed phalanx length or width. Feet of conspecific or congeneric specimens tend to cluster together. At higher levels are some clusters which are phylogenetically problematic, including a cluster composed of moa, most non-avian theropods, Poposaurus, Revueltosaurus, kiwi, seriema, adzebill (Aptornis), and smaller ornithischians, and a cluster composed of ornithomimosaurs, rheas, emu, cassowaries, and bustards. More reasonable is a cluster composed of Iguanodon, Mantellisaurus, and hadrosaurids.

15 1036 FARLOW ET AL. Fig. 10. Relative digit lengths. (A) Length of digit II (excluding the ungual), and the aggregate lengths of phalanges 1 3 of digit IV, as percentages of the length of digit III (excluding the ungual), for all groups (data set 1). Compared with the length of digit III, most ornithischian dinosaurs tend to have relatively longer digits II and IV than do most other forms. Troodontids, dromaeosaurids, and basal sauropodomorphs, and to a lesser extent some struthioniforms (particularly Casuarius), have a relatively long digit IV length compared with the length of digit II. Poposaurus plots near the lower edge of points for crocodylians, surrounded by points for non-avian theropods, kiwi, tinamous, moa, and hypsilophodontids. (B) Digit II and digit IV lengths, both as percentages of the length of digit III (all including ungual lengths), in non-avian dinosaurs, birds, Revueltosaurus, and Poposaurus (data set 2). Most ornithischians are separated from other groups by having relatively long digits IV and especially II. Poposaurus plots among points for large non-avian theropods, oviraptorosaurs, moa, the basal bird Confuciusornis, the quail Cyrtonyx, and basal ornithopods.

16 POPOSAURUS GRACILIS PEDAL PROPORTIONS 1037 Fig. 11. Length of digit III as a function of metatarsal III (or tarsometatarsus) length. (A) Digit III length excluding the ungual; all groups (data set 1), forms comparable to, or smaller than, Poposaurus in size. Poposaurus continues a trend defined by crocodylians, in which digit III is relatively long compared with the length of metatarsal III; basal sauropodomorphs ( prosauropods ), dromaeosaurids, and hypsilophodontids also fall along this trend. (B) Digit III length including the ungual; forms across the entire size range of dinosaurs, birds, Revueltosaurus, and Poposaurus (data set 2). Data are logtransformed to spread out cases of smaller forms; no particular structural relationship between the variable is assumed. Poposaurus is similar to the more basal theropods (Herrerasaurus, Dilophosaurus, Allosaurus, Saurophaganx), some non-avian coelurosaurs (Deinonychus, Chirostenotes), hypsilophodontids, Tenontosaurus, Iguanodon (and Mantellisaurus), basal ceratopsians, kiwi, and moa other than Dinornis in having a relatively long digit III compared to the length of metatarsal III.