A taxonomic and phylogenetic re-evaluation of Therizinosauria (Dinosauria: Maniraptora)

Transcription

1 Journal of Systematic Palaeontology, Vol. 8, Issue 4, December 2010, A taxonomic and phylogenetic re-evaluation of Therizinosauria (Dinosauria: Maniraptora) Lindsay E. Zanno The Field Museum, 1400 S Lake Shore Drive, Chicago, Illinois 60605, USA (Received 11 May 2009; accepted 29 October 2009) Therizinosaurians are among the most poorly understood dinosaurs. Their unusual morphology and fragmentary fossil record has precluded a synthetic understanding of the group since their remains were first discovered over 60 years ago. Although the clade was recently substantiated as a monophyletic group of maniraptoran theropods, little foundational work has been conducted at the species level. A recent plethora of therizinosaurian discoveries including the most complete primitive and derived members recovered to date permits an alpha taxonomic and phylogenetic re-evaluation of the clade. The phylogenetic analysis presented is the most comprehensive yet conducted for Therizinosauria, and provides a foundation for scrutinizing previous definitions of Therizinosauria, Therizinosauroidea and Therizinosauridae. Here, support is provided for the maintenance of all three taxa; however, Therizinosauria is redefined and Falcarius is excluded from Therizinosauroidea. In addition, the previously described therizinosauroids, Beipiaosaurus, Enigmosaurus, Suzhousaurus, Segnosaurus and Therizinosaurus, are rediagnosed and photodocumented. In contrast to other analyses, the ingroup topology recovered in this study suggests intermediate (therizinosauroid) status for Neimongosaurus, Erliansaurus and Enigmosaurus (based on relatively primitive pelvic morphology), despite the derived forelimb anatomy evident in the former two taxa. Here, the largebodied taxa Nothronychus and Nanshiungosaurus brevispinus are recovered as therizinosaurids. This discrepancy indicates a relatively complex pattern of mosaic evolution, which may ultimately be found to correlate with body-size trends in the clade. This work also reviews the chronostratigraphic and biogeographic distribution of therizinosaurian taxa and putatively referred elements and finds no compelling evidence of the clade outside of Asia and North America, nor for the referral of therizinosaurian materials from Kazakhstan to cf. Neimongosaurus. Time calibration of ingroup relationships indicates a pre-turonian dispersal event is needed to account for the presence of therizinosaurids in the Late Cretaceous of North America and Asia; this conclusion supports previous hypotheses of a Laurasian faunal interchange event during the Albian. Keywords: Segnosauria; Theropoda; Dinosauria; systematics; biogeography; evolution Introduction Therizinosaurians comprise a poorly understood clade of coelurosaurian dinosaurs that are largely marginalized as evolutionary misfits in palaeoecological and systematic studies of Theropoda. In contrast to the predatory adaptations exhibited by most theropod clades, therizinosaurians are characterized by traits indicative of a plant-based diet, such as: small, tightly packed, leaf-shaped cheek teeth; medial deflection of the tooth row and development of a lateral shelf on the dentary; incisiform rostral dentition (basal forms); development of a rostral rhampotheca (derived taxa); a broad abdomen and pelvis, indicating massive gut capacity; and secondary loss of cursorial adaptations (e.g. reversal to a functionally tetradactyl pes). In keeping with their unusual and paradoxical anatomy, therizinosaurians have endured one of the most convoluted taxonomic histories within Dinosauria. Pre-cladistic hypotheses regarding their classification vary widely and include placement within nearly all of the major dinosaurian subclades. Early systematic studies were equally problematic, producing widely differing topological placement (Fig. 1) and suffering from a historically inadequate fossil record and an unexpected degree of convergence with other herbivorous dinosaur clades. Several authors have provided excellent reviews of the history of therizinosaurian classification. The most comprehensive review to date is that of Zanno (2004); however, Barsbold & Maryańska (1990), Clark et al. (1994, 2004), Maryańska (1997, Xu et al. (2001a) and Li et al. (2007) also provide excellent summations. A long history of taxonomic controversy notwithstanding, Therizinosauria is now unequivocally regarded as a monophyletic subclade of coelurosaurian theropods, and all recent phylogenetic analyses that include therizinosaurians corroborate this topology. Although the hypothesis that therizinosaurians are coelurosaurian theropods is widely accepted, few phylogenetic studies have been constructed with the explicit intent lzanno@fieldmuseum.org ISSN print / online Copyright C 2010 The Natural History Museum DOI: /

2 504 L. E. Zanno Figure 1. Previously hypothesized outgroup relationships of Therizinosauria. A, Barsbold & Maryanska (1990); B, Russell & Dong (1993); C, Sereno (1999a), only the most inclusive clade including Ornitholestes shown; D, Makovicky & Sues (1998); E, Sues (1997); F, Holtz (1998); G, Rauhut (2003) (see Fig. 7 for outgroup topologies from TWiG analyses). Some monophyletic clades of OTUs combined into single branch in reproduced trees. Therizinosauria shown in grey. of testing their outgroup relationships, and fewer still have attempted to resolve species-level ingroup relationships. Several otherwise taxonomically comprehensive studies of coelurosaurian phylogeny lack therizinosaurians altogether (e.g. Gauthier 1986; Holtz 1994; Forster et al. 1998; Xu et al. 1999a, 2000; Lu et al. 2004; Senter et al. 2004; Kim et al. 2005). The consequence of this exclusion is that nearly all modern hypotheses of coelurosaurian evolution lack the data provided by the fossil record of this unique maniraptoran group. Recently, a plethora of therizinosaurian discoveries has propelled a new understanding of therizinosaurian anatomy and diversity. From their once poor fossil representation, therizinosaurians are rapidly becoming one of the most diverse theropod clades, with at least 14 species described from Cretaceous deposits of the USA, China and Mongolia. Additionally, several putative therizinosaurian elements have been described from other countries in Asia, Europe and North America. Within the past eight years alone, more new therizinosaurian species have been documented than were known during the preceding half-century of palaeontological research. Two recent, highly significant discoveries are the most primitive and most complete therizinosaurian yet known Falcarius utahensis from the lower Cedar Mountain Formation (Barremian), Utah (Kirkland et al. 2005; Zanno 2006, 2010; Smith et al. in press) as well as the most complete derived therizinosaurian recovered to date, Nothronychus graffami (Gillette et al. 2001, 2005; Zanno et al. 2009). Additional highlights include the first feathered therizinosaurian (Xu et al. 1999b, 2003, 2009), the first non-asian therizinosaurians (Kirkland & Wolfe 2001; Kirkland et al. 2004, 2005; Zanno 2004, 2006, 2010; Gillette et al. 2001, 2005), the first ontogenetic sample of a therizinosaurian species (Zanno & Erickson 2006) and the first therizinosaurian bone bed (Kirkland et al. 2004, 2005; Zanno 2004, 2006, 2010). The result of these critical discoveries has been a dramatic increase in the amount of information applicable to morphological, functional, palaeoecological and phylogenetic investigations of therizinosaurians. Yet, to date, few advances have been made in our understanding of therizinosaurian anatomy and evolution. The intent of this study is to address these problems by investigating the impact of recent therizinosaurian discoveries on the evolutionary relationships of this enigmatic group of theropods. Results pertaining to the outgroup relationships of therizinosaurians and their palaeobiological and evolutionary implications are published in a separate manuscript (Zanno et al. 2009), which describes the North American therizinosaurid No. graffami and presents a phylogeny founded on the characters developed here. Primary goals of this manuscript focus upon the ingroup taxonomy and systematics of Therizinosauria and include: (1) documentation of current accessibility and condition of therizinosaurian holotypes and referred materials; (2) photodocumentation of therizinosaurian species thus far poorly represented in the literature; (3) commentary on the utility of features previously used in therizinosaurian diagnoses and rediagnosis of several species in light of more recent discoveries; (4) evaluation of previously proposed phylogenetic definitions for Therizinosauria and Therizinosauroidea on the basis of a new comprehensive phylogenetic analysis (Therizinosauridae is redefined in Zanno et al. 2009); (5) commentary on the validity of characters previously used to support hypotheses of ingroup therizinosaurian relationships; and (6) derivation of palaeobiogeographical and evolutionary implications resulting from these findings.

3 Re-evaluation of Therizinosauria 505 Institutional abbreviations AMNH: American Museum of Natural History, New York, USA; BMNH: Natural History Museum, London, UK; CCMGE: Chernyshev s Central Museum of Geological Exploration, St Petersburg, Russia; CEUM: College of Eastern Utah Prehistoric Museum, Price, Utah, USA; FRDC: Fossil Research and Development Centre, Third Geology and Mineral Resources Exploration Academy of Gansu Province, Lanzhou, China; IGM: Geological Institute of the Mongolian Academy of Sciences, Ulaan Bataar, Mongolia (= PST, GIN and PST GIN AN MNR); IVPP: Institute of Vertebrate Palaeontology and Palaeoanthropology, Chinese Academy of Sciences, Beijing, China; IZK: Laboratory of Paleobiology, Institute of Zoology, National Academy of Science, Alma-Ata, Kazakhstan; LH: Long Hao Geologic and Paleontological Research Centre, Department of Land Resources of Nei Mongol, Hohhot, China; MCF PVPH: Museo Carmen Funes, Paleontología de Vertebrados, Plaza Huincul, Argentina; MSM: Mesa Southwest Museum, Mesa, Arizona, USA; NGMC (= GMV): National Geological Museum of China, Beijing, China; CMN (= MNC): Canadian Museum of Nature, Ottawa, Ontario; PIN: Paleontological Institute, Russian Academy of Sciences, Moscow, Russia; TMP: Royal Tyrrell Museum of Palaeontology, Drumheller, Alberta; UMNH: Utah Museum of Natural History, Salt Lake City, Utah, USA; YPM: Peabody Museum of Natural History, New Haven, Connecticut, USA; ZIN PH: Palaeoherpetological Collection, Zoological Institute, Russian Academy of Sciences, St Petersburg, Russia. Material and methods Specimen availability The following re-evaluation considers only the 15 named therizinosaurian species (Table 1). Unfortunately, a large percentage of representative specimens, including many holotype specimens, were inaccessible for the duration of this study. Materials pertaining to Falcarius, Beipiaosaurus, Alxasaurus, No. mckinleyi, No. graffami, Enigmosaurus, Erliansaurus, Neimongosaurus, Segnosaurus, Therizinosaurus, Nanshiungosaurus brevispinus, Suzhousaurus and several Therizinosauria incertae sedis were examined first hand and/or from casts of specimens. Although some materials were available for study, a large portion of published specimens of Alxasaurus, Therizinosaurus and Nanshiungosaurus were not accessible. In addition, all specimens of the therizinosaurids Erlikosaurus and Nanshiungosaurus bohlini were inaccessible for the duration of this project; therefore comparisons for these taxa were made using the literature. The taxonomy of the putative therizinosaurian Eshanosaurus is not revisited here (see Xu et al. 2001; Clark et al. 2004; Kirkland et al. 2005; Barrett 2009). Although Eshanosaurus may represent a therizinosaurian, as suggested by several authors (Zhao & Xu 1998; Xu et al. 2001; Barrett 2009), comparative mandibles from Segnosaurus, Erlikosaurus and Alxasaurus were unavailable during this study; thus the taxonomic affinity of Eshanosaurus is not formally addressed here. Specific information on the availability of therizinosaurian specimens is detailed below. Falcarius utahensis Kirkland et al., All UMNH materials referred to Falcarius were included in this study. Specimens collected by CEUM from a second locality, although examined, are not considered here, as it has not yet been determined if these materials are referable to Falcarius utahensis. For additional information on available materials, see the Systematic Palaeontology section below. Beipiaosaurus inexpectus Xu et al., 1999b. The majority of the original Beipiaosaurus inexpectus holotype materials (IVPP V11559) described by Xu et al. (1999b) are available for study at the IVPP and remain in good condition. A caudal vertebra, ungual, a putative ischial fragment, fragments of the right forelimb, and several unidentified fragments possibly representing skull material are not currently associated with the rest of the holotype. Additional, subsequently discovered holotype materials (IVPP V11559) described by Xu et al. (2003) are referenced from the literature. Alxasaurus elesitaiensis Russell & Dong, For the duration of this study, the majority of Alxasaurus elesitaiensis holotype materials (IVPP 88402, two individuals) could not be located (X. Xu pers. comm., June 2006); thus only a fraction of the holotype individual was examined first hand. Available holotype specimens included: two sacral? vertebrae, five dorsal vertebrae, 14 caudal vertebrae, 10 chevrons, proximal dorsal ribs, two appendicular fragments, distal carpal 1, an unidentified carpal (radiale?), and two manual unguals (likely I-II and II-III). Many of these elements were not marked with the holotype accession number IVPP 88402; however, they are consistent with elements listed by Russell & Dong (1993). Currently accessible referred elements from IVPP include three dorsal centra, one-and-a-half sacral vertebrae, proximal portions of metacarpal I and II, distal portion of manual PII-I, proximal PI-II and PIII-IV, additional manual phalanges fragments, distal fragments of metatarsals, and proximal and distal fragments of pedal phalanges. Several unlabelled elements could not be confidently associated with a particular specimen, including proximal metacarpal I, proximal metacarpal II and proximal manual PIII-III. Neimongosaurus yangi Zhang et al., 2001 and Erliansaurus bellamanus Xu et al., 2002a. During this project, the holotype of Neimongosaurus yangi (LH V0001)

4 506 L. E. Zanno Table 1. Stratigraphic, biogeographic and taxonomic data for therizinosaur species. Chronostratigraphic data from Barsbold (1983), Jerzykiewicz & Russell (1991), Wolfe & Kirkland (1998), Swisher III et al. (1999, 2002), Shuvalov (2000), Tang et al. (2001), von Itterbeeck et al. (2005), Greenhalgh et al. (2006), Albright et al. (2007a, b) and Zhu et al. (2007). Species Occurrence Age Country Nominal reference Other references Falcarius utahensis Yellow Cat Mb. Cedar Mountain Fm. >125 Ma Barremian USA, Utah Kirkland et al Zanno 2006, 2010 Beipiaosaurus inexpectus Lower Yixian Fm. 125 Ma Barremian People s Republic of China, Liaoning Alxasaurus elesitaiensis Bayin Gobi Fm. Albian People s Republic of China, Nei Mongol Zizhiqu Enigmosaurus mongoliensis Bayan Shire Fm. Cenomanian-Turonian Mongolian People s Republic Suzhousaurus megatheroides Xinminpu (= Xinminbao) Group?Aptian-Albian People s Republic of China, Gansu Xu et al. 1999b Xu et al Russell & Dong 1993 Barsbold 1983 Barsbold & Perle 1980 Li et al Nothronychus mckinleyi Moreno Hill Fm. Middle Turonian USA, New Mexico Kirkland & Wolfe 2001 Nothronychus graffami Tropic Shale Early Turonian USA, Utah unpubl. data Albright et al. 2002; Gillette et al. 2001, 2005 Erliansaurus bellamanus Iren Dabasu Fm. Campanian-Maastrichtian People s Republic of China, Nei Mongol Zizhiqu Neimongosaurus yangi Iren Dabasu Fm. Campanian-Maastrichtian People s Republic of China, Nei Mongol Zizhiqu Nanshiungosaurus bohlini Nanshiungosaurus brevipsinus Xinminpu (= Xinminbao) Group Xu et al. 2002a Zhang et al. 2001?Aptian-Albian People s Republic of China Dong & Yu 1997 Yuanpu Fm. (= Nanxiong) Campanian-Maastrichtian People s Republic of China, Guandong Erlikosaurus andrewsi Bayan Shire Fm. Cenonmanian-Turonian Mongolian People s Republic Segnosaurus galbinensis Bayan Shire Fm. Cenonmanian-Turonian Mongolian People s Republic Therizinosaurus cheloniformis Nemegt Fm. Early Maastrichtian Mongolian People s Republic Dong 1979 Perle 1981 Barsbold & Perle 1980; Barsbold 1983; Clark et al Perle 1979 Barsbold & Perle 1980; Barsbold 1983 Maleev 1954 Barsbold 1976, 1983; Perle 1982

5 Re-evaluation of Therizinosauria 507 and a portion of the holotype of Erliansaurus bellamanus (LH V0002) were under study at the University of Chicago. Currently unaccessioned research casts of Neimongosaurus and Erliansaurus provided by the University of Chicago were also used as comparative material during this project. Several elements not noted by Zhang et al. (2001) and Xu et al. (2002a) were included in these cast materials, yet were not ascribed to one taxon or the other. This complication, combined with the similar morphology of these taxa and the presence of duplicate elements (such as three different right femora), suggests more material from these taxa exists than has been noted previously in the literature. This hampered the revision of these species and is further discussed below. Enigmosaurus mongoliensis Barsbold, The holotype specimen of Enigmosaurus mongoliensis (IGM 100/84) is in good condition and includes the sacrum, ventral portions of both ilia, complete right and left pubes and complete left ischium. A number of other fragmentary elements not mentioned in the original descriptions of the holotype are labelled with the same specimen number, including a crushed proximal femur, a large femoral shaft (tibia?), dorsal ribs, distal humerus, and a putative radius and proximal ulna, all in poor condition. Although the large femur is clearly not associated with the holotype individual, it can be assigned to Therizinosauria incertae sedis. The remaining postcranial elements might reasonably pertain to the holotype. Nanshiungosaurus bohlini Dong & Yu, The current whereabouts of the holotype of Nashiungosaurus bohlini (IVPP V11116) are unknown (X. Xu pers. comm., June 2006). Therefore comparisons with N. bohlini were made using the descriptions and figures in Dong & Yu (1997), which consist of the first nine cervical vertebrae in ventral view (Dong & Yu 1997, p. 91, fig. 1) and the 10th cervical in ventral and dorsal views (Dong & Yu 1997, p. 94, fig. 2). Nothronychus mckinleyi Kirkland & Wolfe, 2001 and Nothronychus graffami Zanno et al., Cast materials of the holotype of No. mckinleyi (MSM P-2117) were examined for this study, including several elements not mentioned in Kirkland & Wolfe (2001). These elements are detailed in the Systematic Palaeontology section below. A second specimen of Nothronychus, No. graffami (UMNH VP 16420), was discovered penecontemporaneously with No. mckinleyi and represents a new, more robust species from a lower stratigraphic horizon. No. graffami (Zanno et al. 2009) is the most complete therizinosaurid known, missing only the skull, a majority of the cervical series and a few elements of the distal extremities. All elements of No. graffami were examined at first hand. Suzhousaurus megatherioides Li et al., All holotype materials of Suzhousaurus megatherioides (FRDC- GSJB-99) were examined first hand. The referred specimen, FRDC-GSJB , is referenced from Li et al. (2008). Nanshiungosaurus brevispinus Dong, Only the pelvis of Nashiungosaurus brevispinus (IVPP V4731) can be located at the IVPP (X. Xu pers. comm., June 2006). The holotype presacral vertebrae are not accessible. The pelvis of N. brevispinus has suffered postcollection damage and is reconstructed with painted plaster in several areas. The distal left preacetabular blade is missing, as are the distal portions of the left pubis and ischium illustrated in Dong (1979). Two cervical and a single dorsal vertebra are illustrated in Dong (1979, figs 1 3). Dong (1979, pl. 2) also figures a photograph of the holotype in left lateral view and three photographs of different vertebrae in ventral, cranial, or caudal view. These figures, as well as written descriptions, were used to supplement first-hand observation of inaccessible elements. Erlikosaurus andrewsi Barsbold & Perle, The holotype of Erlikosaurus andrewsi (IGM 100/1111) includes the only well-preserved therizinosaurian skull. Unfortunately, the skull of Erlikosaurus (as is also the case with the mandible of Segnosaurus and Alxasaurus) is currently on loan to A. Perle (C. Tsogtbaatar pers. comm., November 2006) and was not accessible during this study, nor was any of the postcranial material of Erlikosaurus. Thus comparisons with Erlikosaurus were made using the detailed work of Clark et al. (1994) on the skull anatomy, and descriptions and figures provided by Barsbold & Perle (1980), Perle (1981) and Barsbold (1983). Segnosaurus galbinensis Perle, There are numerous problems associated with specimens referred to Segnosaurus galbinensis, including postcollection damage, missing elements from the holotype, incorrect identification of referred elements, and more than one individual and/or taxon bearing the same specimen number. Currently it is easier to review those specimens available for study than those that are presently inaccessible. Accessible elements of the holotype (IGM 100/80) include a severely damaged right ilium with sacrum missing left sacral ribs (with damage to the preacetabular portion such that it cannot properly be conjoined with the remainder of the ilium), pubis missing proximal portion, and ischium with damage to proximal portion. The holotype mandible was not accessible during this study, nor were any forelimb, hind limb or axial elements. Accessible paratype elements consist of the following: IGM 100/81, nearly complete left tibia with astragalus and calcaneum, and left fibula (noted as right elements by Barsbold & Perle 1980); IGM 100/82, nearly complete left ilium, with damage to dorsal iliac blade, shaft fragment of right femur, distal tibia with astragalus, possible distal tarsal (unnumbered), and calcaneum; and IGM 100/83, complete

6 508 L. E. Zanno left humerus, radius, and ulna, fragmentary proximal right humerus, damaged right radius, fragmentary cervical neural arch. Additional elements bearing the specimen number 100/82, yet not noted in Perle (1979), were found in the IGM collection. These include a proximal left femur and a large postacetabular portion of left ilium of a different individual with ischiadic peduncle. Whereabouts of the paratype pes, fibula, ribs, ischium and pubis (IGM 100/82), and scapulocoracoid and manual phalanges (IGM 100/83) are unknown (C. Tsogtbaatar, pers. comm. November 2006). Therizinosaurus cheloniformis Maleev, Therizinosaurus cheloniformis is known from several specimens, including the holotype (PIN ; Maleev 1954) and four referred specimens housed at the IGM representing the forelimb (Barsbold 1976) and hind limb (Perle 1982). None of these specimens were available for examination for the duration of this study. Several Therizinosaurus specimens are part of a travelling exhibit that was not on display during this project. The whereabouts of other IGM materials are unknown (C. Tsogtbaatar pers. comm., November 2006). A privately owned cast of the right manus of Therizinosaurus with unguals was used as comparative material, although the cast appears to be a composite of IGM 100/15, 100/16, and/or 100/17. Descriptions and figures provided by Maleev (1954), Barsbold (1976) and Perle (1982) were also used in this study. Taxonomic considerations The discovery of increasingly primitive and complete therizinosaurians has rendered most original diagnoses ineffective; thus nearly all therizinosaurian diagnoses are in need of re-evaluation. A species-level taxonomic re-evaluation of Therizinosauria is critical at this juncture because present species diagnoses include many synapomorphic and/or symplesiomorphic features which have hindered recent attempts to diagnose taxa effectively (e.g. Kirkland & Wolfe 2001; Li et al. 2007). The diagnoses given here provide updated differentiae for known therizinosaurian species. Unfortunately, the poor quality, fragmentary nature, and inaccessibility of many therizinosaurian specimens proved an impediment to the rediagnosis of several taxa. For those taxa that could not be properly evaluated, yet are clearly in need of a viable rediagnosis, see Remarks. When a revised diagnosis for a previously described taxon is published for the first time herein, the subheading is listed as Revised Diagnosis. If a previously proposed diagnosis is considered effective and is maintained, a citation for the first publication of that diagnosis is provided (e.g. Diagnosis [Xu et al. 2002a] ). Every attempt was made to maintain viable differentiae provided by earlier authors; however, the retention of previously proposed differentiae for therizinosaurian species generally did not prove useful. Cited characteristics within a rediagnosis are indicative of a maintained differentia. Wherever possible the rediagnoses presented below highlight autapomorphic morphology; however, the fragmentary nature of most therizinosaurian species and the inaccessibility of holotype and referred materials often precluded the identification of enough autapomorphic features to create a stable diagnosis. In these instances, the diagnoses were supplemented with features differential at the comparative ingroup level in addition to autapomorphies. All diagnostic features listed in the diagnoses provided here are figured as reference points for future work, except those differentiating Falcarius, which has already been extensively photo-documented in the literature (Kirkland et al. 2005; Zanno 2006, 2010) and both species of the genus Nothronychus, for which an illustrated diagnosis has already been published (Zanno et al. 2009, electronic supplementary material). Several authors have proposed taxon-based or phylogenetic definitions for the supertaxa Therizinosauria, Therizinosauroidea and Therizinosauridae, and their junior subjective synonyms, Segnosaurischia, Segnosauria, Segnosauridae, Enigmosauridae, Erlikosauridae and Nanshiungosauridae (Perle 1979; Barsbold & Perle 1980; Russell & Dong 1993; Dong & Yu 1997; Russell 1997; Sereno 1998, 1999a, Xu et al. 1999b, 2001, 2002a; Ryan & Russell 2001; Zhang et al. 2001; Clark et al. 2004; Kirkland et al. 2005; Zanno 2006). Yet these definitions have never been evaluated in light of a comprehensive, speciesrich phylogeny, which should form the foundation for the construction of a phylogenetic taxonomy de facto. Here previously proposed definitions are scrutinized in light of node stability and the degree of morphological differentiation illustrated by character change at nodes. Phylogenetic definitions are constructed following the rules proposed by the International Code of Phylogenetic Nomenclature (Phylocode), Version 4b (Cantino & de Queiroz 2007). Neither the Phylocode (2007, Article 4.3) nor the International Rules of Zoological Nomenclature (1999, Article 9.8) recognizes the validity of new taxonomic names and/or definitions proposed solely via the internet (e.g. Taxon- Search); thus definitions proposed via this format are not considered valid. Following the Phylocode (2007, Recommendation 6.1A, B), rank-based taxa are denoted by [R] in the Systematic Palaeontology section to help avoid confusion between these and phylogenetic-based taxa. Operational taxonomic units Base operational taxonomic units (OTUs) are taken from the Turner et al. (2007a) version of the Theropod Working Group (TWiG), which includes three therizinosaurians: Segnosaurus, Erlikosaurus and Alxasaurus. In order to represent all currently known therizinosaurian species, 11 OTUs were added: Therizinosaurus cheloniformis, Nanshiungosaurus brevispinus,

7 Re-evaluation of Therizinosauria 509 Nanshiungosaurus bohlini, Suzhousaurus megatherioides, No. mckinleyi, No. graffami, Enigmosaurus mongoliensis, Erliansaurus bellamanus, Neimongosaurus yangi, Beipiaosaurus inexpectus and Falcarius utahensis. The analysis was further expanded by the addition of Mahakala ömnogövae, Dilong paradoxus, Coelurus fragilis, Yanornis martini, Jeholornis prima, Protarchaeopteryx robustus, Caenagnathus collinsi, Caenagnathus sternbergi (but see below) and Hagryphus giganteus. The default matrix (Appendix 1) represents a 34% increase in OTUs over the base TWiG; however, ad hoc pruning of the matrix during analysis resulted in the use of smaller datasets using various subsets of available OTUs (see Results). A large percentage of therizinosaurian materials were inaccessible for the duration of this project. Many specimens are damaged and the whereabouts of others are unknown. A detailed review of the therizinosaurian materials examined for this analysis is provided (see Specimen Availability) and specifics regarding potentially problematic OTUs are provided below. Therizinosaurians. With the exception of the poorly known taxon Eshanosaurus (Xu et al. 2001), all named therizinosaurian species are included in this analysis as OTUs. The contention regarding the classification of Eshanosaurus is not likely to be resolved until additional materials are recovered (see Geographic Distribution of Therizinosauria). Furthermore, including Eshanosaurus would fail to test currently competing hypotheses regarding its evolutionary relationships because this analysis is restricted to coelurosaurian theropods, which would constrain Eshanosaurus to this clade (and probably Therizinosauria) by default. A syntype of the tyrannosaur Alectrosaurus olseni (AMNH 6368) from the Iren Dabasu Formation is now recognized as belonging to Therizinosauria (Mader & Bradley 1989). In a phylogenetic analysis restricted to forelimb material, Zanno (2006) found the specimen to be closely related to Erliansaurus, also from the Iren Dabasu Formation, yet the fragmentary nature of the specimen and poor preservation do not permit referral to this taxon. Moreover, compared to Erliansaurus, AMNH 6368 represents a significantly larger and more gracile individual; thus all that can be confidently inferred about this specimen is that it represents an intermediate-grade therizinosaurian. AMNH 6368 was not included within any OTU; nor was it coded as an individual taxon in this analysis, as it does not possess unique character combinations (Wilkinson 1995). The fragmentary therizinosaurian Nanshiungosaurus bohlini (IVPP V11116) was coded based on descriptions and photos provided by Dong & Yu (1997, pp. 91, 94). As noted by Li et al. (2007), it is unlikely that IVPP V11116, from the Early Cretaceous of the Xinminpu Group (Mazongshan, Gansu Province, northwestern China), represents the genus Nanshiungosaurus from the Late Cretaceous Yuanpu Formation (south-eastern China); however, this specimen also cannot currently be referred to Suzhousaurus, known from potentially correlative sediments. Although IVPP V11116 is not presently accessible for study and is not well documented in the literature, the specimen was included here as a separate OTU in an attempt to test its phylogenetic relationships. Perle (1982) referred the therizinosaurian hind limb IGM 100/45 to Therizinosaurus cheloniformis based on its stratigraphic and geographic proximity to a forelimb referred to this taxon (IGM 100/15). Although IGM 100/45 cannot unequivocally be referred to Therizinosaurus due to an absence of overlapping elements, this specimen was included in the Therizinosaurus OTU. Oviraptorosaurians. Although North American oviraptorosaurians are known from only fragmentary specimens, their unique morphology relative to Asian oviraptorosaurians renders them a critical component of the clade. North American oviraptorosaurian taxonomy is contentious, but is represented by unique taxa, including several potentially synonymous species (e.g. Caenagnathus collinsi, Caenagnathus sternbergi, Chirostenotes pergracilis and Elmisaurus elegans). No record could be located of the specimens representing the Chirostenotes OTU in various iterations of the TWiG; however, mandibular characters are coded for this taxon in TWiG matrices, leading to the likelihood that Chirostenotes includes specimens described as the genus Caenagnathus. Preliminarily, a conservative approach was taken by removing the mandibular characters from the OTU Chirostenotes, and coding the two species of Caenagnathus (C. collinsi and C. sternbergi) as independent OTUs. However, C. sternbergi and C. collinsi differ in only a single character state (#70) and add a large percentage of missing data to the analysis (C. collinsi 7% complete, C. sternbergi 6%). Thus both species were combined as Caenagnathus sp. and character 70 was noted as polymorphic for this OTU. Several authors argue for the synonymization of Chirostenotes and Caenagnathus (Currie & Russell 1988; Sues 1997); however, there is currently no more compelling evidence for their synonymization than for their distinction. When coded as a distinct OTU, Caenagnathus is comprised of a unique set of character combinations; as such, its inclusion is warranted either as a distinct taxon or as part of the Chirostenotes hypodigm. Iterations of the analysis under both conditions were conducted here to examine the difference in topology (see Results). Senter et al. (2004) considered Incisivosaurus (IVPP V13326, lower Yixian Formation; Xu et al. 2002b) a subjective junior synonym of Protarchaeopteryx (NGMC 2125; Chaomidianzi Formation; Ji & Ji 1997; Ji et al. 1998) based on the possession of incisiform first premaxillary teeth, peg-like second premaxillary teeth, lanceolate

8 510 L. E. Zanno maxillary teeth, and an edentulous rostral dentary in both taxa, and suggested maintaining taxonomic distinction only at the species level. The potential congeneric status of these species is irrelevant for the purposes of phylogenetic analysis, and these taxa are maintained as distinct OTUs in this study. Other coelurosaurians. Turner et al. (2007a) follow Makovicky et al. (2005) in considering Neuquenraptor (MCF PVPH 77) a junior synonym of Unenlagia (MCF PVPH 78); this coding is maintained here. Following Turner et al. (2007a), Microraptor gui is not considered part of the Microraptor zhaoianus hypodigm (these two taxa plus Crypotovolans pauli were considered junior subjective synonyms in Senter [2007]). The generic distinction of Gorgosaurus and Albertosaurus (as per Holtz 2004) is maintained; thus the OTU formerly listed as Albertosaurus libratus in the TWiG is listed as Gorgosaurus libratus, as in Senter (2007). A partial dentary has been referred to the type specimen of Coelurus agilis (YPM 2010), which has been synonymized with Coelurus fragilis (Ostrom 1980; Carpenter et al. 2005). Carpenter et al. (2005) questioned the association of the dentary because of its slender morphology. Due to its ambiguous association, a conservative approach was taken and the dentary was not coded within the Coelurus OTU. Outgroups. Following the general practice of the TWiG, Allosaurus fragilis is maintained as the outgroup taxon in this analysis. Despite its consistent use as an outgroup taxon for phylogenetic studies of coelurosaurian theropods, Allosaurus fragilis is poorly figured in the current literature. First-hand observation of specimens accessioned at the UMNH necessitated the revision of certain character codes and led to repolarization of several characters (see Appendix 2: Annotated List of Added and Modified Characters). Characters. The base character set of 251 characters was adopted from a recent TWiG analysis (Turner et al. 2007a). Base TWiG character 110 (number of sacral vertebrae) was revised according to Turner et al. (2007b). Base TWiG characters 39, 40, 84, 138 and 142 were revised based on Senter (2007). Different base versions of the TWiG data matrix are used in both studies, hence character numbers vary between Senter (2007) and this analysis. Likewise, Turner et al. (2007b) and this paper use different revised versions of the base characters 136, 148 and 151; therefore derived states for these characters may represent different character conditions. Explanations for changes made to the base TWiG in this analysis are given in Appendix 2 (Annotated List of Added and Modified Characters). In total 97 characters were added to the base TWiG, resulting in a 348 character database (complete character list and data matrix provided in Appendix 1). These additions (50% novel, 50% adapted from cited sources) represent a 39% increase in character information. The database used in this study is nearly identical to the base TWiG in data per skeletal region (see Table 2), although the new character database contains a higher percentage of postcranial (75% versus 69%) characters overall, as is necessary given the relative over-abundance of therizinosaurian postcranial materials. Commentary on newly added characters is also provided in Appendix 2. Phylogenetic methods The dataset was constructed in MacClade v. 4.0 (Maddison & Maddison 2000) and analysed using TNT (Tree analysis using New Technology) v. 1.1 (Goloboff et al. 2008). Most parsimonious trees (MPTs) were obtained via heuristic search methods on 1000 replicates of Wagner trees with random addition sequences and then subject to TBR (tree bisection-reconnection) swapping methods holding 10 trees per replicate. Subsequent rounds of TBR swapping were performed to capture additional MPTs. All characters were equally weighted. Base TWiG characters 165 and 215 were excluded from this analysis because they are redundant, with revisions to characters 230 and 66 respectively. A conservative approach was taken in the ordering of characters. In an attempt to accommodate missing data in the study without inflating tree length and reduce apriori assumptions regarding character homology, multistate characters were designated additive (ordered) only if they represent nested states of change within a transformation series for which the relationship and polarity of the characters are understood. Characters 27, 37, 40, 68, 76, 78, 97, 106, 113, 157, 163, 168, 253, 303, 308, 309, 310, 334, 342 and 345 were designated additive in this study. Ambiguous nodes were collapsed following Rule 1 of Coddington & Sharff (1994). Maximum agreement subtrees (Finden & Gordon 1986) calculated using TNT were used to identify labile taxa and common topology among all MPTs. Resampling methods were also conducted in TNT and included standard bootstrapping (1000 replicates; Felsenstein 1985) and symmetrical resampling (1000 replicates; Goloboff et al. 2003). Mesquite (Maddison & Maddison 2008) and MacClade v. 4.0 (Maddison & Maddison 2000) were used for ad hoc data collection including examining character optimization and distribution, and generating tree statistics. Systematic palaeontology Dinosauria Owen, 1842 Saurischia Seeley, 1887 Theropoda Marsh, 1881 Coelurosauria von Huene, 1914

9 Re-evaluation of Therizinosauria 511 Table 2. Character support by skeletal region. Data from this study is compared to the base TWiG (Turner et al. 2007a). The percent increase in characters is calculated by dividing the number of new characters per skeletal region by the total number of characters per skeletal region in this dataset, not as compared to those listed in Turner et al. 2007a. New characters refer to both novel characters and characters added to the base TWiG from cited sources. Cranial Mandibular Dentition Axial Pectoral Forelimb Pelvic Hind limb Dataset %total %incr. %total %incr. %total %incr. %total %incr. %total %incr. %total %incr. %total %incr. %total %incr. Rank This analysis Rank Turner et al. 2007a

10 512 L. E. Zanno Therizinosauria Russell, 1997 nomen cladi conversum 1980 Segnosauria[R] Barsbold & Perle 1992 Segnosaurischia[R] Dong 1998 Therizinosauridae Sereno 1999a Therizinosauridae Sereno 2001 Therizinosauroidea Zhang et al. 2002a Therizinosauroidea Xu et al. Definition. The most inclusive clade containing Therizinosaurus cheloniformis but not Tyrannosaurus rex, Ornithomimus edmontonicus, Mononykus olecranus, Oviraptor philoceratops or Troodon formosus. Heterodefinitional junior synonyms. Therizinosauroidea (Zhang et al. 2001); Therizinosauroidea (Xu et al. 2002a); Therizinosauridae (Sereno, 1998); Therizinosauridae (Sereno, 1999a). Subjective junior synonyms. Segnosaurischia[R] (Dong, 1992); Segnosauria[R] (Barsbold & Perle 1980). Remarks. Russell (1997) coined the name Therizinosauria and provided a branch-based definition and diagnosis for the clade. Russell s (1997) definition is given as Alxasaurus, Enigmosaurus, Erlikosaurus, Nanshiungosaurus, Segnosaurus, Therizinosaurus, and all taxa more closely related to them than to oviraptorosaurians, ornithomimids and troodontids. The inclusion of six therizinosaurian species as internal specifiers is now superfluous and renders the definition unstable if one of these genera should be reclassified or considered a junior synonym (Padian & May 1993; Padian et al. 1999; Sereno 1999b). In an attempt to promote stability, the redefinition provided here uses a single internal qualifier: the type species, Therizinosaurus cheloniformis. The utility of multiple external specifiers is also contentious (Padian et al. 1999) and carries the same potential problems as multiple internal specifiers; however, as Sereno (1998, 1999b) noted, this problem can be minimized by using nested rather than basal taxa (or clades, e.g. Russell 1997) to guard against potential discrepancies in taxonomic content even under circumstances of varying topology. Furthermore, multiple external specifiers are currently necessary as outgroup relationships of Therizinosauria remain somewhat unstable. Recent phylogenetic studies have suggested close evolutionary relationships between therizinosaurians and oviraptorosaurians, alvarezsaurids, ornithomimosaurians and/or troodontids (see Table 3); thus the use of multiple external Table 3. Previous hypotheses of therizinosaur outgroup relationships. Shown are selected recent phylogenetic analyses of Coelurosauria. Abbreviations: AL, Alvarezsauridae; AV, Avialae; D, Dromaeosauridae; OR, Ornithomimosauria; OV, Oviraptorosauria; TH, Therizinosauria; TR, Troodontidae; TY, Tyrannosauridae. (Therizinosauria/Oviraptorosauria) Monophyly TWiG base Other Partial skeletal analyses Novas et al Xu et al Frankfurt & Chiappe 1999 Xu et al Rauhut & Xu 2005 Zanno, 2006 Turner et al. 2007a,b Xu & Wang 2004 Göhlich & Chiappe 2006 Holtz et al Norell et al. 2001, 2006 Holtz & Osmólska 2004 Xu & Zhang 2005 Rauhut, 2003 Novas & Pol 2005 Novas & Pol 2002 Kirkland et al Holtz, 2001 Mayr et al Makovicky & Sues 1998 Calvo et al Holtz, 1998 Xu & Norell 2004 Sues, 1997 Makovicky et al. 2003, 2005 Hwang et al. 2002, 2004 Clark et al Xu et al. 1999b, 2002b, c (Therizinosauria(Alvarezsauridae(Oviraptorosauria(Paraves)))) This paper (and Zanno et al. 2009) Senter, 2007 Other topologies Russell & Dong 1993 Sereno 1997, 1998 Sereno, 1999a Paul, 1984 Barsbold & Maryańska 1990 Maryańska et al ((O/TR)TH)OR)) ((OR/TH)(TY(OV(AV(TR/DR))))) (((AL/OR)TH)(TY(OV(AV(TR/D))))) ((Ornithischia)(TH)(Saurischia)) ((Sauropodamorpha)(TH)(Theropoda)) ((AV/OV)TH)

11 Re-evaluation of Therizinosauria 513 qualifiers for the clade is apropos (Phylocode 2007, Recommendation 11E). Here Mononykus is chosen because of its derived status within Alvarezsauridae. Emendment of Therizinosauria is conducted under the guidelines of the PhyloCode (2007), which permit unrestricted emendations when the definitional type is maintained and the intent is to preserve conceptualization of the clade as intended by the original author (Art. 15.1, 15.8, 15.11). Both conditions are met here. Both Therizinosauridae and Therizinosauroidea were given phylogenetic definitions postdating Russell (1997) that rendered them junior heterodefinitional synonyms to Therizinosauria. Therizinosauroidea[R] was coined by Russell & Dong (1993) as a rank elevation for Therizinosauridae[R] (Maleev, 1954). Russell & Dong (1993, p. 2121) provide a taxon-based definition for Therizinosauroidea[R] as the clade encompassing Alxasaurus and therizinosaurids, and added an extensive diagnosis. In their description of the primitive therizinosauroid Beipiaosaurus, Xu et al. (1999b) assigned this taxon to Therizinosauroidea[R] although a phylogenetic definition expanding Therizinosauroidea beyond that given by Russell & Dong (1993) was not provided. Xu et al. (2001) further assigned the enigmatic taxon Eshanosaurus to the clade Therizinosauroidea[R]. Thus it is clear that Xu et al. (1999b) intended the taxon Therizinosauroidea to be more inclusive than might be interpreted from Russell & Dong (1993). Accordingly, Zhang et al. (2001) provided the first formal phylogenetic definition for Therizinosauroidea: all coelurosaurians closer to Therizinosaurus than to Ornithomimus, Oviraptor, Velociraptor or Neornithes. However, in doing so they rendered Therizinosauroidea a junior synonym of Therizinosauria (sensu Russell, 1997), leaving Therizinosauroidea open for redefinition (Art. 15.5). Clark et al. (2004) corrected this mistake by emending the definition of Therizinosauroidea as the least inclusive clade containing Therizinosaurus and Beipiaosaurus. Having composed their manuscript prior to the publication of Clark et al. (2004), Kirkland et al. (2005) assigned the basal-most therizinosaurian Falcarius to Therizinosauroidea sensu Zhang et al. (2001) and this assignment is maintained in Zanno (2006, 2010). Here the nodebased definition provided by Clark et al. (2004) is followed, which serves to exclude Falcarius from Therizinosauroidea on the basis of its relatively primitive morphology. Further support for this reassignment is provided in Zanno (2010) and in the synapomorphy list provided in Appendix 3. Falcarius utahensis Kirkland et al., 2005 Holotype. UMNH VP 15000, partial juvenile braincase. Referred specimens. Over 3000 complete or partial elements have been collected from the holotype locality and referred to Falcarius utahensis. Currently only a small fraction of these materials have been prepared and are available for study. Although the majority of the skeleton is represented in the collection, portions of the axial column and a large portion of the skull are currently unknown. Accession numbers and identifications for catalogued referred specimens can be found in Kirkland et al. (2005), Zanno (2006, in 2010) and Smith et al. (in press). More recent materials are on file at the UMNH. Occurrence. Crystal Geyser Quarry (CGQ) approximately 12 miles SE of Green River, UT; lower Yellow Cat Member of the Early Cretaceous Cedar Mountain Formation; minimally dated at ma (Greenhalgh et al. 2006). Diagnosis. A therizinosaurian bearing the following autapomorphies: antorbital fenestra extends onto lateral margin of nasal; facet for postorbital on frontal rostral to postorbital process; expansive, deeply depressed, and highly pneumatic subcondylar and subotic recesses each possessing multiple pneumatic fossae; inflated basisphenoid with hypertrophied basisphenoidal recess; rostral dentary teeth conical, extending to, at minimum, the fifth dentary tooth position; cupped, dorsoventrally elongate rostral teeth lacking denticles; median ridge dividing ventral sulcus of cervical vertebrae; infraprezygapophyseal fossa on cranial dorsal vertebrae divided into three accessory fossae; accessory caudal centrodiapophyseal lamina on cranial dorsal vertebra; hypertrophied, obliquely oriented humeral entepicondyle, with concave caudal margin, and well-defined groove extending proximally up humeral shaft; flexor tubercle on manual PI-II with deep, distally positioned collateral ligament pits; proximal pubic tubercle (sensu Hutchinson 2001) well developed and caudolaterally oriented, occurring at cranialmost margin of acetabulum; midseries chevrons possessing markedly distinct cranial tubercles (Zanno 2010). Remarks. Falcarius utahensis is the most basal and most complete therizinosaurian yet recovered and is known from a mass death assemblage estimated to hold hundreds of individuals of various growth stages. The anatomy of this taxon is the best documented among therizinosaurians, having been extensively reviewed in three separate manuscripts (Zanno 2006, 2010; Smith et al. in press) following its initial description (Kirkland et al. 2005). Due to the comprehensive literary treatment of this taxon, a re-evaluation of the genus Falcarius is unnecessary. The original diagnosis presented in Kirkland et al. (2005) was emended in Zanno (2010). Therizinosauroidea Russell & Dong, 1993 (sensu Clark et al., 2004)

12 514 L. E. Zanno

13 Re-evaluation of Therizinosauria 515 Definition. The least inclusive clade containing Beipiaosaurus inexpectus and Therizinosaurus cheloniformis (Clark et al. 2004). Remarks. Under the ICZN, Maleev (1954), author of Therizinosauridae, is credited with establishing Therizinosauroidea (Russell & Dong 1993; Clark et al. 1994; Zhang et al. 2001; Xu et al. 2001, 2002a; Kirkland et al. 2005; Zanno 2006, 2010; Li et al. 2007). However, under the Phylocode (2007, Note 9.8A.2), Therizinosauroidea is attributable to those who coined it (i.e. Russell & Dong 1993), as referenced in Xu et al. (1999b). Since its establishment as a higher rank taxon for Therizinosauridae[R] (Russell & Dong 1993), Therizinosauroidea[R] has become the taxon of choice for the most inclusive grouping of therizinosaurians. As discussed above, the lack of a formal phylogenetic definition upon establishment (Russell & Dong 1993) and the obscure publication of both the branch-based, phylogenetically defined Therizinosauria by Russell (1997) and the heterodefinitional taxon Therizinosauridae by Sereno (in a table, Sereno 1998, p. 65; and as a footnote, Sereno 1999c, p. 2147) resulted in referral of several recent basal therizinosaurian discoveries to Therizinosauroidea (sensu Zhang et al. 2001) rather than Therizinosauria (sensu Russell, 1997) or Therizinosauridae (sensu Sereno 1999a) (see Xu et al. 1999b; 2002a; Zhang et al. 2001; Kirkland et al. 2005). Whereas it might seem appropriate to conserve one of these more widely used taxa for the group, maintaining Therizinosauria on the basis of precedence follows the general intent of the Phylocode (2007, Article 14.2). Furthermore, the high degree of morphological diversity encompassed by known therizinosaurian taxa, together with the morphological disparity between Falcarius and more derived therizinosaurians, renders the maintenance of all three taxa of taxonomic and practical utility. This can only be accomplished if Therizinosauria is used for the most inclusive branch-based clade. Finally, the nodebased phylogenetic definition provided here encompasses the historical taxonomic content of Therizinosauroidea as given by most authors, including those conducting explicit phylogenetic studies (Xu et al. 1999b; Clark et al. 2004), those recently adding to the diversity of the clade (Russell & Dong 1993; Xu et al. 1999b; 2002a; Kirkland & Wolfe 2001; Zhang et al. 2001), and those using a historical conceptualization of the clade (e.g. Perle 1979, 1981; Barsbold & Perle 1980; Barsbold 1983; Barsbold & Maryańska 1990; Clark et al. 1994; 2004; Maryańska, 1997; Dong & Yu 1997). The only exceptions are the most recent conceptualizations of the clade (Kirkland et al. 2005; Zanno 2006, 2010; Li et al. 2007; Senter 2007). Beipiaosaurus inexpectus Xu et al., 1999b Holotype. IVPP V11559, immature individual (Xu et al. 1999b) represented by cranial material, rostral portion of mandible, and nearly complete postcranial skeleton. Additional materials of the holotype were discovered subsequently and published in a separate manuscript (Xu et al. 2003). These include the right ilium missing the postacetabular portion, both ischia, two sacral vertebrae, 30 caudal vertebrae including pygostyle. Occurrence. Sihetun locality near Beipiao, Liaoning, China. Lower Yixian Formation, Early Cretaceous (Barremian, 125 Ma; Swisher III et al. 1999, 2002; Zhu et al. 2007). Revised diagnosis. (Fig. 2A E). A therizinosaurian bearing the following autapomorphies: elongate lateral articular surface on manual phalanx I-I (Xu et al. 1999b); obturator process of ischium sinusoidal, with ventrally deflected distal portion; ischial boot approximately twice craniocaudal depth of distal shaft; low ridge on cranial femoral shaft extending proximally from medial condyle. Possession of four fused caudal dorsals and a pygostyle incorporating up to seven caudal vertebrae may also be autapomorphic for this species, as well as the presence of a prominent triangular flange extending from ventrolateral surface of metacarpal I, which although found on other theropod taxa is likely of autapomorphic morphology in Beipiaosaurus and may be homologous with rectangular buttress on advanced taxa. Remarks. Beipiaosaurus is consistently recovered as one of the most basal therizinosaurians (Xu et al. 1999b; Clark et al. 2004; Kirkland et al. 2005; Zanno 2006, 2010; Li et al. 2007; Senter 2007). In the initial and still most extensive description of the taxon, Xu et al. (1999b) provided a Figure 2. Diagnostic features of the holotype of Beipiaosaurus (IVPP V11559) (A-F) and Alxasaurus (IVPP 88402) (G).A, ventral view of dorsal vertebrae, showing fusion; B, left lateral view of caudal vertebrae with pygostyle; C, ventral view of right metacarpal I, showing triangular flange; D, ventral view of right manual phalanx I-I, showing enlarged lateral condyle; E, lateral view of right ilium and ischium, showing ratio of ischial boot to distal shaft and sinusoidal obturator process; F, cranial view of right femur, showing craniomedial ridge on shaft; G, cranial and left lateral views of chevron. Abbreviations: af, articular facet on chevron; cv, caudal vertebrae; dvf, fused dorsal vertebrae; ib, ischiadic boot; il, ilium; lc, lateral condyle; lfb, lateral flare of cranial chevron blade; mc, medial condyle, MCI, metacarpal one; mfr, ridge on cranial femoral shaft; op, obturator process; PI-I, first phalanx of digit one; py, pygostyle; rf, right femur; ris, right ischium; and tf, ventrolateral triangular flange. Photographs not to scale.

14 516 L. E. Zanno differential diagnosis focusing on its primitive morphology relative to other therizinosaurians (e.g. relatively large skull, functionally tridactyl pes, bearing proximally compressed first metatarsal, unexpanded preacetabular process of ilium, elongate manus; tibia elongate relative to femur; and proximally compressed metatarsus). Although apropos at the time of its composition, the original diagnosis has been rendered insufficiently differential by the discovery of the more basal taxon Falcarius, necessitating re-evaluation of Beipiaosaurus. Although none of the character states provided in this rediagnosis are recognized on other therizinosaurians, some may represent an early stage in a continuum rather than discrete morphologies (e.g. ischial boot twice width of distal shaft; triangular flange on metacarpal I), whereas the absence of others may be preservational (e.g. presence of a pygostyle). Additional features are not autapomorphic among coelurosaurians, yet are included because they are presently apomorphic (and thus differential) among therizinosaurians (e.g. cranial ridge on distal portion of femur, also present on several oviraptorosaurians, but not on more basal therizinosaurians). Alxasaurus elesitaiensis Russell & Dong, 1993 Holotype. IVPP (larger individual), right dentary and most of postcranial skeleton including five disarticulated cervical vertebrae, two cervical ribs, seven caudal disarticulated vertebrae, three right and three left dorsal ribs, sacrum, four sacral ribs, 21 caudal vertebrae, 15 chevrons, scapula, left and right coracoids, left and right humeri, radius, left and right ulnae, left and right distal carpal 1, distal carpal 2, ulnare?, radiale, metacarpals I-III, manual PI-I-II, II-I-III, III-I, III-II, left and right ilia, left and right ischia, left and right femora (Russell & Dong 1993). Referred specimens. IVPP (small individual), three articulated caudal cervicals and/or cranial dorsals, four articulated dorsal vertebrae; IVPP (immature individual), five dorsal vertebrae, first sacral, second sacral, right metacarpal I, left metacarpal III, manual PI-I-I-II, II-I- II-III, III-II, and III-IV, left postacetabular portion of ilium, left and right femora, left and right tibiae, left and right fibulae, MTI-IV, pedal PII-I-III, III-I-IV, IV-I-V, IVPP (scattered vertebrae, ribs, and appendicular material), IVPP 88510, appendicular material. Occurrence. Bayin Gobi Formation (lacustrine facies), Inner Mongolia, Albian (Jerzykiewicz & Russell 1991). Remarks. Russell & Dong (1993, p. 2108) provided the original diagnosis for Alxasaurus: dentary teeth number approximately 40; symphyseal region of dentary bears teeth; ribs not fused to cervical centra; ligament pits often well developed in manual phalanges; ilium not greatly shortened craniocaudally; preacetabular ala moderately expanded, ungual shorter than or subequal to first phalanx in pedal digits II-IV. The original diagnosis suffers from the same problems as that given for Beipiaosaurus in that it is composed of plesiomorphies which, in light of recent discoveries, are no longer of use as differentiae (see also Clark et al. 2004). Although clearly necessary, the rediagnosis of Alxasaurus must await the availability of the type materials (see Materials and Methods). A single autapomorphic feature can be recognized on the type materials available for this study: cranial blade of proximal chevrons with profound lateral flare at region of bifurcation into articular facet (Fig. 2F G). Neimongosaurus yangi Zhang et al., 2001 Holotype. LH V0008: partial braincase; cranial end of right dentary; majority of axial column except atlas; mid and caudal dorsal vertebrae; distal-most caudal vertebrae; pectoral girdle and forelimb elements, including the left and partial right scapulocoracoids, furcula, both humeri, and left radius; pelvic girdle and hind limb elements including partial left and right ilia, both femora and tibiae, left distal tarsals three and four, and partial left pes. The axial column and pes were discovered in articulation. Referred specimens. LH V 0008, sacrum with six coossified vertebrae articulated with both ilia. Occurrence. Sanhangobi, Sunitezuoqi, Nei Mongol Autonomous Region (20 km south-west of Erlian city); Iren Dabasu Formation, Late Cretaceous (Campanian- Maastrichtian; von Itterbeeck et al. 2005). Remarks. Zhang et al. (2001, p. 39) provided the following diagnosis for Neimongosaurus: cranial caudal vertebrae with circular fossa under transverse process; radius with prominent biceps tuberosity; proximal pedal phalanges with well developed heels; tibia with extremely long fibular crest exceeding half total tibia length; and lateral surface of preacetabular process twisted to face dorsally; and caudal vertebrae are characterized by widely divergent prezygapophyses (distribution poorly known among other therizinosauroids). Of these characters, cranial caudal vertebrae with a circular fossa under the transverse process are also found in other therizinosaurians (e.g. No. mckinleyi and No. graffami) as well as in oviraptorosaurians (e.g. Nomingia, IGM 100/119), and proximal pedal phalanges with welldeveloped heels occur in No. mckinleyi and No. graffami. Possession of an elongate fibular crest exceeding half the total length of the tibia (approximately 57% in Neimongosaurus) also does not significantly differ from the condition in Nothronychus (approximately 55% in No. graffami). The presence of a preacetabular process in

15 Re-evaluation of Therizinosauria 517 which the lateral surface is twisted so as to face dorsally is a derived therizinosaurian feature (e.g. Nothronychus, Segnosaurus, Nanshiungosaurus brevispinus), probably the result of lateral deflection of the preacetabular portion of the ilium. Zhang et al. (2001) note that widely divergent caudal prezygapophyses are potentially diagnostic, yet the prezygapophyses of Neimongosaurus are not unusual with respect to the condition in other therizinosaurians (Zanno in press). The radius of Neimongosaurus does possess a more proximally located and pronounced biceps tubercle than is known for other therizinosaurians (e.g. Segnosaurus, Falcarius), so at present this taxonomic distinction appears to be valid. Neimongosaurus is not rediagnosed here (see Erliansaurus below). Erliansaurus bellamanus Xu et al., 2002a Holotype. LH V 0002: five vertebrae, left forelimb lacking carpus, partial right ilium and fragments of the ischium and pubis, right femur, both tibiae, right fibula, several partial metatarsals. Occurrence. Sanhangobi, Sunitenzuoqi, Nei Mongol Autonomous Region (20 km south-west of Erlian city); Iren Dabasu Formation, Late Cretaceous (Campanian- Maastrichtian; von Itterbeeck et al. 2005). Remarks. Xu et al. (2002a, p. 230) provided the following diagnosis for Erliansaurus: enlarged nutritional foramen on cranial caudals; crest-like caudal trochanter on humerus bordered medially by oval depression; rugose swelling located dorsal to ischiadic peduncle on lateral surface of ilium; fibular proximal end with caudal margin significantly higher than cranial margin; and distally located hypertrophied cranial trochanter on fibula. The humerus of Erliansaurus was not examined firsthand; however, an oval depression in the region of the caudal trochanter is present on other therizinosaurians (e.g. Falcarius, Neimongosaurus) and other coelurosaurians (e.g. Coelurus, Ornitholestes), although not coupled with a caudal trochanter as in Erliansaurus. The crestlike morphology of the caudal trochanter in Erliansaurus is unknown in other taxa. The ilium of Erliansaurus is severely crushed and badly preserved. Thus the rugose swelling located dorsal to the ischiadic peduncle is most likely to be the craniolateral edge of the cubic tuberosity present on other derived therizinosaurians (Segnosaurus, Enigmosaurus and Nothronychus), rather than an autapomorphy as suggested by Xu et al. (2002a). Enlarged nutrient foramen on cranial caudal vertebrae and proximal fibula with the caudal margin significantly higher than the cranial margin are autapomorphic features, whereas a distally located hypertrophied cranial trochanter on fibula is also known in Nothronychus. The Erliansaurus bellamanus holotype comes from the same quarry as the holotype and referred material of Neimongosaurus yangi (Xu et al. 2002a). Research cast material obtained for both Erliansaurus and Neimongosaurus (see Specimen Availability above) indicate the presence of more than two individuals, although this cannot be confirmed in the literature. Xu et al. (2002a) note the presence of a cranial caudal vertebra among the holotype materials for Erliansaurus bellamanus possessing an incompletely fused neural arch. This feature, coupled with the larger size of several Erliansaurus elements, indicates that this taxon achieved significantly larger adult body sizes than Neimongosaurus. The argument of Xu et al. (2002a) for the presence of two species is compelling if the materials belong to only two individuals and the elements associated with each taxon have been properly assigned. Quarry maps showing the distribution of elements referred to these two taxa could not be obtained. Thus these taxa could not be re-evaluated with any confidence. Enigmosaurus mongoliensis Barsbold, 1983 Holotype. IGM 100/84; pelvic girdle lacking dorsal aspect of ilia and distal ischia. Occurrence. Bayan Shire Formation (Baynshirenskaya), Khara Khutul, south-east Mongolia, Cenomanian-Turonian (Shuvalov 2000). Revised diagnosis. A therizinosaurian bearing the following autapomorphies: dorsoventrally shallow pubic boot with pronounced, subequal cranial and caudal processes; pubic boot fused, enlarged medially, with medial expansion forming V-shaped trough (adopted from Barsbold 1983); medial fusion of obturator process and pubic apron forming tetraradiate process (Fig. 3). Remarks. The original diagnosis for Enigmosaurus (Barsbold, 1983) contains two characters: pubis and ischium narrow; and cranial presymphyseal region of distal pubis with elongated indentation. Possession of a [craniocaudally] narrow pubis and ischium is plesiomorphic for the group. The second feature (elongated indentation of the pubic boot) is incorporated in the revised diagnosis presented here. A key point not previously noted in the literature is that the holotype pelvis of Enigmosaurus mongoliensis is clearly abnormal, possessing numerous resorption pits and areas of bone remodelling. As these features appear to be restricted to the dorsal iliac blade, the ilium has not been referenced in this diagnosis. However, if the abnormal condition of the ilium is the result of advanced ontogenetic stage in this individual, this finding could have implications for the unusually extreme medial fusion of the

16 518 L. E. Zanno Figure 3. Diagnostic features of Enigmosaurus mongoliensis holotype pelvis (IGM 100/84). A, cranial view; B, dorsal view; C, oblique caudal/left lateral view; and D, right lateral view, showing shallow pubic boot with subequal cranial and caudal processes and abnormalities in the dorsal ilium; E, inset of tetraradiate obturator process in dorsal view; and F, pubic boot in oblique cranial/right lateral view, showing fused v-shaped trough. Abbreviations: a1 4, abnormal holes 1 4; at, antitrochanter; cbt, cubic tuberosity of postacetabular process; cp, caudal process of ischium; cpb, caudal process of pubic boot; crpb, cranial process of pubic boot; dmi, dorsal margin of ilium; ilp, iliac peduncle of pubis; lis, left ischium; lp, left pubis; nsr, neural spine ridge, sacrum; op, obturator process; ris, right ischium; rp, right pubis; sv, sacral vertebra; tr, medial trough on pubic boot. Scale bar 10 mm (D only). obturator process and pubic boot, both of which are considered autapomorphic in this diagnosis. Further study of the holotype is necessary to determine the nature and extent of these influences. A large left femur (estimated length 105 cm) is labelled IGM 100/84, yet is clearly not associated with the pelvis due to its massive size, which more closely approximates that of Segnosaurus.

17 Re-evaluation of Therizinosauria 519 Suzhousaurus megatherioides Li et al., 2007 Holotype. FRDC-GSJB-99; an associated partial postcranial skeleton, including 10 partial dorsal vertebrae, dorsal rib fragments, right scapulocoracoid, right humerus, isolated pubic peduncle of left ilium, nearly complete left and fragmentary right pubes. Paratypes. FRDC-GSJB ; partial postcranial skeleton, including three dorsal vertebrae, sacrum, six caudal vertebrae, dorsal ribs, chevrons, nearly complete left ilium (missing ventral-most preacetabular process), nearly complete left pubis (missing boot), complete left ischium, left femur, distal right femur. Occurrence. Yujingzi Basin, north-western Gansu Province, China. Xinminpu (= Xinminbao) Group, Early Cretaceous (Aptian-Albian, Tang et al. 2001). Revised diagnosis. Suzhousaurus can be differentiated from all other therizinosaurians by the following autapomorphies: gently rounded craniodorsal margin of preacetabular process of ilium (Li et al. 2008); obturator foramen, craniocaudal width greater than dorsoventral height; obturator notch on pubis deep (approximately half the maximum dorsoventral length of ischiadic peduncle of pubis) and wide (approximating craniocaudal dimension of proximal pubic shaft); pubis short, maximum length nearly subequal with maximum dorsoventral height of ilium; ischiadic shaft distal to ventral caudal process deflected caudally. Remarks. The original diagnosis for Suzhousaurus given by Li et al. (2007) consists of two features: shallow, poorly demarcated glenoid fossa with prominent rounded and striated tumescence on dorsomedial surface of its scapular portion; and pubis with strongly concave cranial margin (Li et al. 2007, p. 541). Based upon newly described paratype materials, Li et al. (2008) added the following features to their diagnosis: transverse expansion of the distal aspect of dorsal, sacral and cranial caudal neural spines; laterally deflected, thin, and flat preacetabular process of ilium; and smoothly curved craniodorsal margin of preacetabular process of ilium (Li et al. 2008, pp ). Transverse expansion of the caudal aspect of the distal neural spine is present on the dorsal vertebrae of Falcarius (also concomitant with bifurcation in therizinosaurians). Both cranial and caudal expansion is present in the smallbodied taxon Neimongosaurus, although the feature is more subtle in this species. At first inspection this feature appears to be related to ossified reinforcement along the dorsal axial series, which may be correlated with increased body size in some clades. This characteristic is found in largebodied theropod taxa such as Allosaurus fragilis and Tyrannosaurus rex. However, as the dorsal neural spines are also caudally bifurcate in Falcarius and Mononychus, this feature also appears to carry phylogenetic significance. The circumglenoid region in the holotype of Suzhousaurus megatherioides (FRDC-GSJB-99) is damaged; thus although it can be ascertained that the glenoid faces ventrally in this taxon, poor demarcation of the articular surface of the glenoid cannot be confirmed. Even if this character can be demonstrated on additional specimens of Suzhousaurus, No. mckinleyi also exhibits poor differentiation of the glenoid, therefore this feature is not autapomorphic. Additionally, transverse thickening of the scapulocoracoid at the contact between these elements is common in derived therizinosaurians and noted to be extreme in Therizinosaurus (Barsbold 1976, 1983), indicating it may be related to large adult body-size. Although Barsbold s (1976) description of this feature in Therizinosaurus suggests that it mirrors the condition of Suzhousaurus, the scapulocoracoid of Therizinosaurus could not be examined first hand and has never been figured in medial view. Thus the contention that the degree of thickening in Suzhousaurus is sufficiently distinct to be considered autapomorphic cannot presently be verified or invalidated. A laterally deflected, [transversely] thin preacetabular process is probably a characteristic of derived therizinosaurians and is documented in several taxa, including Segnosaurus, Nanshiungosaurus brevispinus and No. graffami. The body of this process is flat (i.e. not medially or laterally curved) in at least N. brevispinus, and may also be in No. graffami and Beipiaosaurus although crushing in these specimens renders this feature ambiguous. The smoothly curved craniodorsal margin of the preacetabular process noted by Li et al. (2008) remains a potential autapomorphy of this taxon. The craniodorsal aspect of the preacetabulum forms a more acute angle in No. graffami and N. brevispinus, the only taxa in which it is completely preserved. However, the condition is unknown in many taxa including Falcarius, Neimongosaurus, Enigmosaurus and Segnosaurus, making comparison difficult. Li et al. (2007, 2008) noted that the pubis of Suzhousaurus exhibits a greater degree of concavity than is present in all other therizinosaurians for which this element is known. While the pubis of Suzhousaurus is indeed more concave than Falcarius, Beipiaosaurus and Enigmosaurus, examination of the pubes of Segnosaurus (IGM 100/80) and No. graffami (UMNH VP 16420) reveal an almost identical degree of curvature; thus this condition cannot be considered autapomorphic in Suzhousaurus. In sum, prior diagnoses provided for Suzhousaurus based on holo- and paratype materials contains only a single valid autapomorphy, therefore a rediagnosis is presented here. Therizinosauridae Maleev, Segnosauridae Perle 1983 Enigmosauridae Barsbold

18 520 L. E. Zanno 1997 Nanshiungosauridae Dong & Yu 2001 Erlikosauridae Ryan & Russell 2001 Therizinosauridae Zhang et al. 2002a Therizinosauridae Xu et al Therizinosauridae Clark et al. Definition. The least inclusive clade containing Nothronychus (UMNH VP 16420), Segnosaurus galbinensis (IGM 100/80), Erlikosaurus andrewsi (IGM 100/111) and Therizinosaurus cheloniformis (PIN ) (Zanno et al. 2009). Subjective junior synonyms. Enigmosauridae[R] (Barsbold, 1983), Nanshiungosauridae[R] (Dong & Yu 1997), Erlikosauridae[R] (Ryan & Russell 2001); Therizinosauridae (Zhang et al. 2001); Therizinosauridae (Xu et al. 2002a); Therizinosauridae (Clark et al. 2004). Remarks. The taxon Therizinosauridae has undergone several proposed modifications in recent years. Zanno et al. (2009) provide an extensive review of the definitional history of the taxon and provide a slightly modified definition for the clade. Their definition, based on a comprehensive phylogenetic analysis derived from this study, includes Nothronychus (UMNH VP 16420) the most complete derived therizinosaurian known to date as an internal specifier and removes Nanshiungosaurus on the basis of the current inaccessibility of specimens. Here, Suzhousaurus and Enigmosaurus are considered outside Therizinosauridae based on the recovery of these species as sister-taxon to the clade in the agreement subtree of all MPTs, or when the labile taxon Therizinosaurus is pruned from the analysis (see Results). As in Zanno et al. (2009), the clades Enigmosauridae[R] (Barsbold, 1983), Nanshiungosauridae[R] (Dong & Yu 1997) and Erlikosauridae[R] (Ryan & Russell 2001) are considered junior synonyms of Therizinosauridae. Alxasauridae[R] (Russell & Dong 1993), although currently consisting of only a single species and of dubious future utility, cannot be rendered a junior synonym of Therizinosauridae and remains valid. Nothronychus Kirkland & Wolfe, 2001 Type species. No. mckinleyi Kirkland & Wolfe, Occurrence. Moreno Hill and Tropic Shale formations, Late Cretaceous, early-to-middle Turonian (Wolfe & Kirkland 1998). Diagnosis. A therizinosaurian bearing the following autapomorphies: scapula with dorsoventrally tapering blade; distinctly subcircular obturator process; ovoid, dorsoventrally elongate obturator foramen (maximum length greater than 200% of maximum width); contact between pubis and ischium restricted to proximal half of obturator process; and presence of a deep notch between ventral aspect of obturator process and cranial ischial shaft (Zanno et al. 2009). Remarks. The genus Nothronychus is currently represented by two species, No. mckinleyi and No. graffami,from the upper Cretaceous of New Mexico and Utah respectively. The type species, No. mckinleyi was the first definitive therizinosaurian to be discovered outside of Asia. Discovery of a second species referable to the genus No. graffami from slightly older sediments necessitated revising the diagnosis of No. mckinleyi into genus- and species-level characters. Zanno et al. (2009) provided a thorough discussion of the methodology used in creating the diagnoses for the genus Nothronychus, the species No. mckinleyi and No. graffami. Their diagnoses are followed here. Nothronychus mckinleyi Kirkland & Wolfe, 2001 Holotype. MSM P-2117, partially disarticulated skeleton including isolated teeth, partial braincase, two potential cranial fragments, fragments of eight cervical vertebrae, one cranial dorsal vertebra, one dorsal centrum, caudal vertebrae 6 or 7, cervical and dorsal ribs, fused gastralia, chevron fragment, left scapula, right humerus, complete right ulna, unidentified antebrachial shaft, distal metacarpal II, manual PI-I and PII-I, both ischia, both tibiae, right fibula, partial left metatarsal, pedal PI-I, PIII-III, PIV-II, pedal unguals I-IV. Occurrence. Moreno Hill Formation, southern Zuni Basin, Catron Co., New Mexico, Late Cretaceous, Middle Turonian (Wolfe & Kirkland 1998). Diagnosis. No. mckinleyi can be differentiated from all other therizinosaurians by the following autapomorphies: ventral notch between obturator process and iliac shaft craniocaudally narrow; and from No. graffami by the following combination of features: platycoelous cranial caudal centra; heel on caudoventral surface of caudal centra hypertrophied (approximately one third the dorsoventral length of the caudal centrum); glenoid lacking caudal buttress; caudal process of ischium poorly defined and elongate, extending to shaft opposite obturator process (Zanno et al. 2009). Remarks. Based on comparisons with the new species No. graffami, Zanno et al. (2009) revised the holotype of No. mckinleyi to include previously unidentified elements (left pedal PI-I, dorsal centrum, and chevron fragment), refinement of previously identified elements (metacarpals, manual phalanges, and a caudal vertebra), and misidentified elements (pedal PIV-III and PIV-IV, reidentified as PIII-III

19 Re-evaluation of Therizinosauria 521 and PIV-II and two manual unguals now recognized to be pedal). Nothronychus graffami Zanno et al., 2009 Holotype. UMNH VP 16420, partial skeleton missing skull, most cervical vertebrae, several dorsals, several caudals. Occurrence. Tropic Shale, Late Cretaceous (Early Turonian) approximately 4.5 metres below the Lowest Stratigraphic Datum of the Mammites nodosoides ammonite zone (Albright et al. 2007a, b), southern margin of Kaiparowits Basin, Kane County, Utah. Diagnosis. No. graffami can be differentiated from all other therizinosaurians by the following autapomorphies: neural spines of sacrum ankylosed into dorsoventrally low continuous ridge with individual spines unidentifiable; pubic boot with small yet pronounced knob-like caudal process; ventral margin of pubic boot dorsally convex between cranial and caudal process; caudal process of ischium dorsoventrally restricted (less than 5% of total length of ischium), subtriangular, and dorsally positioned; caudal process of ischium located proximal to obturator foramen. No. graffami can be differentiated from No. mckinleyi by the following four features: strongly amphicoelus cranial caudal centra; heel on caudoventral surface of caudal centra poorly developed; glenoid with pronounced caudal buttress; ventral notch between obturator process and iliac shaft craniocaudally wide (Zanno et al. 2009). Remarks. No. graffami is the most complete derived therizinosaurian known to date and serves as an important reference point for comparing the anatomy of more poorly preserved species. Thus its inclusion as an internal specifier for the taxon Therizinosauridae bolsters character support and stability of this node. Nanshiungosaurus brevispinus Dong, 1979 Holotype. IVPP V 4731, ilium missing ventral preacetabular process, proximal portions of the pubis and ischium, mostly complete vertebral series from axis to first caudal vertebrae. Occurrence. Yuanpu (Nanxiong) Formation, Dapingcun, Shuikou Commune, (Dong 1979, English translation by Will Downs), Guandong, China, Upper Cretaceous (Dong 1992). Remarks. Lacking access to the holotype vertebrae and without preservation of the distal pubis and ischium, Nanshiungosaurus brevispinus is difficult to diagnose. The diagnosis provided by Dong (1979) was based on the assignment of N. brevispinus to Titanosaurinae. This misidentification combined with the discovery of other derived therizinosaurian taxa renders Dong s (1979, p. 343) differentiae undiagnostic: short neck with platycoelous cranial cervicals, pleurocoels undeveloped, neural spines low, and caudal series not distinctly bifid; 12 cervical vertebrae with centrum length 2.5 times that of dorsal centra; ten dorsal vertebrae with platycoelous centra of equivalent height and length and shallow pleurocoels; neural spines low and transversely broadened with a broad apex; five fused sacral centra with short unified neural spines, inflated apices, and saddle-shaped depressions; ilium low with extremely welldeveloped narrow and elongated preacetabular process; pubic peduncle of ilium straight and robust; pubis linear with thick lateral margin and closed obturator foramen; ischium thinly plate-shaped with expansive and fused distal ends; acetabulum large and circular. A single feature from the original diagnosis opisthocoelous caudal cervical centra is retained here as a potential autapomorphy, as it is undocumented in other therizinosaurians. The reconstruction of the pubis and ischium provided by Dong (1979, p. 345) is likely to be incorrect in that it portrays the ischium terminating at the obturator process rather than extending beyond, as in all other therizinosaurians. Erlikosaurus andrewsi Barsbold & Perle, 1980 Holotype. IGM 100/111, nearly complete skull with mandible, several disarticulated fragmentary cervical vertebrae, left humerus, articulated right pes missing proximal ends of two metatarsals (Clark et al. 1994). (Barsbold & Perle [1980] note a left and right pes with the holotype; Perle [1981] notes only a pes). Occurrence. Baysheen Tsav locality (= Baynshin, Bayshin, or Bainshin, Tsav; Clark et al. 1994), Bayan Shireh Svita (Barsbold & Perle 1980; Perle 1981) (= Baynshirenskaya Svita; Tsybin & Kurzanov 1979; Shuvalov & Chkikvadze 1979), south Gobi (Ömnogöv) Aimak, Mongolia (Clark et al. 1994). Upper Cretaceous, Cenomanian to Turonian (Shuvalov 2000). Remarks. The original diagnosis for Erlikosaurus by Barsbold & Perle (1980, p. 190): medium-sized [therizinosaurian] with laterally compressed pedal unguals does not adequately diagnose this taxon. Barsbold (1983, p. 105) added two characters to the diagnosis: mandibular teeth small and closely positioned; and rostral portion of lower jaw edentulous. However, these characters are more widespread within the clade. Clark et al. (1994) redescribed the skull of Erlikosaurus and provided a comprehensive diagnosis on the basis of this study. However, as noted by Clark et al. (1994), the near complete absence

20 522 L. E. Zanno of comparative cranial materials prevented assessment of these features as autapomorphic, synapomorphic or plesiomorphic. The recent discovery of therizinosaurian cranial materials pertaining to Falcarius (Zanno 2010) and No. mckinleyi (Kirkland & Wolfe 2001) is shedding light on the utility of these features in differentiating Erlikosaurus. The following features noted as possible autapomorphies by Clark et al. (1994, pp ) are now known to have a broader distribution among therizinosaurians: edentulous premaxilla with sharp, vertical ventrolateral edge; vomer extremely elongate and extending caudally to meet cultriform process; parabasisphenoid with extremely large pneumatic spaces; external auditory meatus restricted ventrally by lateral expansion of braincase; medial wall of antorbital fossa extensive; and maxilla with medially inset dentition and few nutrient foramina on caudal part of facial process. The taxonomic distribution of the following features (Clark et al. 1994, pp ) remains unknown: extremely elongate nares due to regression of the maxilla; antorbital fossa with well-developed over-hanging lip; medial wall of antorbital fossa imperforate; caudal process of jugal covering cranial surface of quadratojugal; and a homodont maxillary dentition of numerous (23) small, lanceolate, coarsely serrated, unrecurved, transversely flattened teeth constricted at the base. Unfortunately, the skull of Erlikosaurus could not be accessed during this study and the holotype postcranial materials cannot be located (see Specimen Availability), preventing a re-evaluation of Erlikosaurus. Segnosaurus galbinensis Perle, 1979 Holotype. IGM 100/80 (incorrectly noted as 100/90 in Barsbold [1983, English translation by C. Siskron & S. P. Welles]); mandible, incomplete humerus, complete radius and ulna, manual phalanges including one ungual, nearly complete pelvis, incomplete right femur, 10 cranial caudals, 15 caudal caudals, dorsal and ventral rib fragments (Perle 1979). Barsbold (1983, English translation by C. Siskron & S. P. Welles) also lists carpal elements with the holotype, although their presence could not be confirmed. Paratypes. IGM 100/81, left tibia and fibula (noted as right in Barsbold & Perle 1980); IGM 100/82, right femur, tibia, fibula, tarsals III and IV, metatarsals I V, five pedal phalanges including one pedal ungual, rib fragments, ilium, proximal ischium and distal pubis; IGM 100/83 left scapulocoracoid, humerus, radius, ulna, manual unguals, cervical fragment (Perle 1979; Barsbold & Perle 1980). Barsbold (1983) noted the presence of two additional referred specimens, IGM 100/87 and IGM 100/88, consisting of fragments of a postcranial skeleton including a pelvis. Given the presence of several apparent typographical errors regarding specimen numbers in the manuscript, these may be the known paratypes IGM 100/82 and IGM 100/83. Holotype occurrence. Bayan Shireh Svita (= Baynshirenskaya [Barsbold 1983], Bayan Shiren Svita [Maryańska 1997], and Bayaanshiree [Shuvalov 2000]) Upper Cretaceous (Cenomanian-Turonian; Shuvalov 2000) locality Amtgay (Amtgai), Mongolia, and Khara-Khutul, Bayshin- Tsav and Urilbe-Khuduk, Mongolia (Barsbold & Perle 1980; Barsbold 1983 [referred]). Referred occurrence. Khara-Khutul, Bayshin-Tsav and Urilbe-Khuduk, SE Mongolia (Barsbold & Perle 1980; Barsbold 1983). Revised diagnosis. A therizinosaurian bearing the following autapomorphies: well-developed coronoid process on mandible (not figured); straight-shafted humerus lacking cranially deflected distal aspect; dorsal margin of ilium with pronounced ventral overhang; caudal process of ischium caudally extensive, nearly 50% of craniocaudal depth of obturator process; hypertrophied epiphyses on distal metatarsals (noted by Perle, 1979, yet could not be confirmed first hand; not figured). Segnosaurus can be differentiated from Nothronychus by the following two characters (ischium with subrectangular obturator process; subcircular obturator foramen); from Enigmosaurus by the following three characters (dorsoventrally deep obturator process that does not fuse medially with its counterpart; unfused pubic boot; craniocaudally wide distal pubic shaft); and from both Nothronychus and Enigmosaurus in possessing a deep brevis fossa and lacking a well developed caudal projection on pubic boot (Fig. 4). Remarks. Perle (1979), Barsbold & Perle (1980) and Barsbold (1983) all provided diagnoses for Segnosaurus galbinensis. The following diagnosis was given by Barsbold & Perle (1980, p. 188): large segnosaurid with noncompressed pedal unguals. Barsbold (1983, p. 104) added several characters: mandibular teeth small, compactly positioned, slightly bent cranially and straight caudally; cranial jaw with small diastema; and pedal unguals slightly laterally compressed. Both diagnoses are insufficient to differentiate this taxon from other therizinosaurians as they focus primarily on perceived variations in the degree of pedal ungual compression between Segnosaurus and Erlikosaurus. Whereas the degree of compression of the pedal unguals may currently distinguish Erlikosaurus, the lack of severe pedal ungual compression in Segnosaurus closely resembles the condition in other therizinosaurians. Perle (1979, pp ) provided the following diagnosis for Segnosaurus: mandible elongate and dorsoventrally narrow; teeth compact; cranial dentary teeth reduced in size, straight; scapulocoracoid massive, fused; deltopectoral crest hypertrophied, manual phalanges dorsoventrally compressed, lacking collateral ligament pits; PI-II sharply recurved and laterally compressed; preacetabular portion of ilium broad; postacetabular portion of ilium reduced, with

21 Re-evaluation of Therizinosauria 523 Figure 4. Diagnostic features of Segnosaurus galbinensis. A, cranial view of referred humerus (IGM 100/80); and B, medial view of humerus showing straight profile; C, right lateral view of holotype ilium with sacrum (IGM 100/80) with inset, showing ventral overhang of dorsal margin and D, dorsal view of ilium and sacrum; E, lateral view of composite left pelvis, referred ilium (IGM 100/82), holotype ischium and pubis (IGM 100/80), showing caudally extensive caudal process of ischium, subrectangular, dorsoventrally deep obturator process, subcircular obturator foramen, craniocaudally wide distal pubic shaft; F, cranial view of referred left proximal femur (IGM 100/82); and G, cranial view of referred right distal tibia with astragalus (IGM 100/82). Abbreviations: apa, ascending process of astragalus; as, astragalus; at, antitrochanter; cbt, cubic tuberosity on postacetabular process; crpb, cranial process of pubic boot; ct, caudal tuberosity of humerus; dmi, dorsal margin of ilium; dpc, deltopectoral crest; ec, ectepicondyle; en, entepicondyle; fh, femoral head; fifi, M. iliofemoralis insertion on femur; flt, lesser trochanter; gt, greater trochanter; hh, humeral head; ib, ischiadic boot; it, internal tuberosity; mtt, medial tuberosity on cranial distal tibia; ns, neural spine, of, obturator foramen; op, obturator process; pil, pubic peduncle of ilium; pra, preacetabular process; rc, radial condyle, and uc, ulnar condyle. Scale bar 7 mm (A-B), 10 mm (E-G), C-D not to scale.

22 524 L. E. Zanno cubic tuberosity; iliac peduncle of pubis elongate; ischiadic shaft flattened, with distally positioned obturator process contacting pubis; ascending process of astragalus extends lateral to tibia, overlapping fibula; and functionally tetrapodal pes. Although most of these features are too widespread among therizinosaurians to be useful differentiae, Perle s (1979) diagnosis represents a summary of the distinguishing features characterizing derived species and may be useful in the creation of an apomorphy-based diagnosis for Therizinosauridae. Dorsoventral compression of the manual phalanges and the morphology of the teeth may represent autapomorphies of Segnosaurus; however they are not included in the revised diagnosis until they can be confirmed first hand. Therizinosaurus cheloniformis Maleev, 1954 Holotype. PIN , three manual unguals (two incomplete), metacarpal fragment, several rib fragments (see Remarks) (Maleev 1954). Referred specimens. Three additional specimens were referred to Therizinosaurus by Barsbold (1976): IGM 100/15, right and left scapulocoracoid (left poorly preserved), right and left humeri, right antebrachium, left ulna, two right distal carpals, right metacarpus with phalanges of the second digit including ungual, fragmentary dorsal ribs, gastralia; IGM 100/17, manual ungual missing distal portion; and IGM 100/16, manual ungual missing proximal portion. Perle (1982) also referred a hind limb (IGM 100/45) from the same beds in Khermin Tzav to Therizinosaurus. The specimen consists of femoral fragments, tibia, astragalus, calcaneum, a distal tarsal, four metatarsals, nearly complete digit one, complete digits two and four missing unguals, and presumed PIII-II from third digit of pes. Holotype occurrence. Nemegtskaya Svita; Nemegtu region, Nemegt Formation, south-west Mongolia (Maastrichtian; Barsbold 1983). Referred occurrence. Khermin Tzav, Nemegt Formation Omnogov Aimak, SW Mongolia. Revised diagnosis. A therizinosaurian bearing the following autapomorphies: deltopectoral crest on humerus hypertrophied, exceeding 60% length of humerus; metacarpal I greater than two-thirds length of metacarpal III; metacarpal I with enlarged medial crest connecting medial distal condyle and proximal medial lobe; manual unguals transversely compressed (transverse dimension less than 20% dorsoventral height in cross-section), poorly recurved, and hypertrophied, proximodistal length to dorsoventral height greater than 4:1; and manual unguals lack ventrally projecting flexor tubercle in lateral view (Fig. 5). Remarks. Since its original description in 1954, Therizinosaurus has been rediagnosed twice by Barsbold (1976, 1983). The diagnosis presented here encompasses more quantitative versions of two characters retained from iterations of earlier diagnoses: ungual of giant size, compressed laterally, narrowing distally, with a slight bend (Barsbold 1976, p. 78) (laterally compressed unguals sensu Barsbold [1983, p. 101]); and humerus with strongly enlarged deltopectoral crest (Barsbold 1976, p. 78) (deltopectoral crest highly developed sensu Barsbold [1983, p. 101]). The remaining features noted by Barsbold (1976, p. 78, 1983, p. 101) (listed below) are not diagnostic for Therizinosaurus: scapulocoracoid with shortened scapular blade and enlarged acromion; humeral epiphyses transversely wide; humerus distally expanded and oriented in same plane as proximal humerus; ulna with semilunar proximal articular surface; sigmoidal radius; semilunate carpus; metacarpal bones of unequal transverse width (i.e. metacarpal I short and dorsoventrally flattened, metacarpal II elongate and metacarpal III reduced, thin); and phalanges of the second manual digit shortened. Perle (1982) referred a fragmentary hind limb (IGM 100/45) to Therizinosaurus cheloniformis. IGM 100/45 exhibits multiple therizinsaurian synapomorphies including: the apparent reduction of the astragalar body, development of a craniomedial tuberosity on the distal tibia, proximodistally short and laterally deflected ascending process of the astragalus, and functionally tetradacyl pes. Based on its morphology, geographic proximity to and stratigraphic continuity with the holotype specimen of Therizinosaurus cheloniformis (Perle 1982), IGM 100/45 is reasonably referred to this taxon. However, this material is excluded from the diagnosis presented here in order to prevent future complications if the hind limb is ultimately found to represent another taxon. The rib material originally considered part of the holotype (Maleev 1954) was identified by Rozhdestvensky (1970) as pertaining to Sauropodomorpha and is not considered here. Therizinosauria incertae sedis Nanshiungosaurus bohlini Dong & Yu, 1997 Holotype. IVPP V11116, articulated series of 11 cervicals (including atlas), five dorsals, fragmentary ribs. Occurrence. Gongpoquan Basin, Mazongshan Area, Gansu Province, uppermost Xinminbao Group (late Early or early Late Cretaceous, Tang et al. 2001). Remarks. The features originally used to diagnose N. bohlini (Dong & Yu 1997) are now known to represent other therizinosaurians (e.g. Falcarius, Neimongosaurus,

23 Re-evaluation of Therizinosauria 525 Figure 5. Diagnostic features of Therizinosaurus cheloniformis. Casts are a composite of referred materials IGM 100/15, 100/16, and/or 100/17. A, right lateral view of right manual unguals I-III, showing poor recurvature and hyperelongation; B, dorsal view of right manus, showing ratio of first to third metacarpals; C, dorsal view of right PI-II, showing extreme transverse compression; and D, lateral view of right MCI, showing enlarged medial crest. Abbreviations: dit, dorsal intercondylar tuberosity; ft, flexor tubercle; MCI-MCIII, metacarpals I-III; mc, medial condyle; mcr, medial crest; PI-II, first manual ungual; PII-III, second manual ungual; PIII-IV, third manual ungual. Scale bar 10 mm (A-B), C-D not to scale. Nothronychus) including possession of platycoelous vertebral centra; low and craniocaudally long cervical neural spines; large zygapophyses situated dorsal to the neural canal; and neural spines of dorsals reduced. Dong & Yu (1997) recognize the presence of ventral keels on the first four dorsals in N. bohlini. To date the only therizinosaurian for which these elements are preserved is Neimongosaurus, in which the cranial dorsal centra are reported as lacking a ventral keel. However, the primitive taxon Falcarius has a minimum of two ventrally keeled cranial dorsals and may possess up to four (Zanno 2010), so the apomorphic status of this feature appears unlikely. As noted by Li et al. (2007), the taxonomic status of N. bohlini is uncertain. Whereas the authors offer an argument for excluding N. bohlini from the genus Nanshiungosaurus, they also note that a lack of available overlapping elements currently precludes assignment to its potential subjective synonym Suzhousaurus. Additional specimens of Suzhousaurus may resolve these issues. As noted by Li et al. (2007), two options are available for IVPP V11116: (1) new specimens of Suzhousaurus will reveal that this taxon is indistinguishable from IVPP V11116 and the resultant name will be Suzhousaurus bohlini; or (2) IVPP V11116 is unique, in which case a nomen novum will need to be erected to form a new combination for bohlini. Phylogenetic analysis Results Analysis of the new phylogenetic dataset generated by this study resulted in 756 MPTs (TL = 1225), reached after a single subsequent round of TBR swapping. Strict consensus of the recovered MPTs (Fig. 6) illustrates the relatively high degree of resolution recovered here, including lower-level relationships. In contrast to nearly all recent studies, none of the MPTs recovered a monophyletic Therizinosauria/Oviraptorosauria. Here Oviraptorosauria is posited as the sister taxon to Paraves under strict consensus. Forcing a monophyletic Therizinosauria/Oviraptorosauria requires an additional eight steps in this analysis. Forcing a monophyletic relationship between Therizinosauria and Alvarezsauridae requires an additional six steps, and forcing various topologies producing an exclusive common ancestor for all three clades (independent of Paraves) requires a minimum of 11 and up to 18 additional steps. Resolution of derived therizinosaurians and ornithomimosaurs is poor under strict consensus, with all therizinosaurians more derived than Neimongosaurus unresolved (except for a monophyletic Nothronychus), and a large basal polytomy within Ornithomimosauria. A maximum agreement subtree (MAST) was used to identify causes of ambiguity, which include the labile therizinosaurians Nanshiungosaurus bohlini, Segnosaurus galbinensis, Erlikosaurus andrewsi, Therizinosaurus cheloniformis (Fig. 7) and the basal ornithomimosaur Pelecanimimus polydon, and identify common topology among all MPTs. Upon subsequent investigation it was determined that N. bohlini could be excluded from subsequent analyses a priori (i.e. safe taxonomic reduction, Wilkinson 1995). Lability of N. bohlini results from a high percentage of missing data (97%) rather than character conflict as indicated by

24 526 L. E. Zanno Figure 6. Strict consensus of 756 most parsimonious trees (tree length 1225). Numbers above lines represent Bremer values; numbers below line represent symmetrical resampling GC values (left) and standard bootstrap frequencies (right, in bold). Node numbers represent taxa as follows: (1) Coelurosauria; (2) Maniraptora; (3) Paraves; (4) Tyrannosauroidea; (5) Compsognathidae; (6) Ornithomimosauria; (7) Therizinosauria; (8) Alvarezsauridae; (9) Oviraptorosauria; (10) Aves; (11) Troodontidae; (12) Dromaeosauridae; (13) Unenlagiinae. Unambiguous synapomorphies shown in Appendix 3. tree length stability after its exclusion from the dataset. However, Segnosaurus, Erlikosaurus, Therizinosaurus and Pelecanimimus contain unique character combinations and cannot be pruned in the total evidence approach used here. Ingroup lability of the taxa Erlikosaurus and Therizinosaurus is not surprising. Due to the unknown whereabouts of associated postcranial materials, few of the newly added postcranial characters could be coded for Erlikosaurus. Thus, in this study, the taxon is represented predominantly from cranial material, which is unknown for the majority of other therizinosaurians. Highly transformed forelimb materials and a morphologically conservative hind limb represent Therizinosaurus, and forelimb elements display a suite of autapomorphic features representing extreme stages of observable phylogenetic trends within the clade. Thus although the derived position of Therizinosaurus is evident and may be anchored in future studies through the use of these features, the discrete and additive quantification of these trends to produce such a result can only be accomplished by arbitrary state division, a method not employed here. The hind limb materials

25 Re-evaluation of Therizinosauria 527 Figure 7. Time-calibrated phylogeny of Therizinosauria based on relationships recovered in the maximum agreement subtree of all MPTs. Shown temporal ranges represent uncertainties in chronostratigraphy, not actual species ranges. Segnosaurus, Erlikosaurus and Therizinosaurus are shown as unresolved therizinosaurids due to their lability in this study. Minimum temporal hypothesized faunal interchange events between North America and Asia in the Late Cretaceous shown. referred to Therizinosaurus lack a unique suite of character states and thus are of little use in discerning its taxonomic affinities. In this study Therizinosaurus is the only labile taxon to be posited outside Therizinosauridae (between Neimongosaurus and Enigmosaurus) among MPTs. Thus the taxon is the sole culprit for the irresolution recovered for derived therizinosauroids under strict consensus of MPTs. Pruning Therizinosaurus from the matrix a priori anchors Suzhousaurus and Enigmosaurus as successive outgroups to Therizinosauridae (this topology is also recovered in the MAST, see below), but does not resolve ambiguity within the derived subclade. However, pruning Therizinosaurus does dramatically reduce the number of MPTs recovered from 756 to 108. Segnosaurus possesses a highly transformed humerus that compares poorly to other therizinosaurians in that it possesses a straight rather than sigmoid shaft, little expansion of the distal end relative to the shaft, no expansion of the medial aspect of the humerus, and a poorly developed entepicondyle. In the absence of these features the humerus of Segnosaurus is more akin to ornithomimosaurs (and in some respects troodontids) than to other therizinosaurians. Although Segnosaurus can clearly be identified as a therizinosaurid, character conflict with other members of the subclade renders its position unstable. In several respects, the pelvis of Segnosaurus is similar to Nothronychus (particularly the ischia); however, it is not yet clear if these similarities reflect synapomorphies or a morphological grade early in the evolution of therizinosaurids, as these features are unknown in other derived taxa. The MAST (from which Therizinosaurus, Erlikosaurus and Segnosaurus are removed a posteriori after MPT generation; Fig. 7) also supports a more basal position of Suzhousaurus and Enigmosaurus (successive sister taxa to Therizinosauridae) and a more derived position for Nothronychus and Nanshiungosaurus brevispinus. The poor condition, fragmentary nature and unknown whereabouts of Asian therizinosaurid materials are major impediments to phylogenetic resolution among the subclade Therizinosauridae, and are unlikely to be overcome without the addition of new discoveries, especially for the taxa Therizinosaurus, Segnosaurus and Erlikosaurus.

26 528 L. E. Zanno Despite the increase in therizinosaurian resolution gained from pruning N. bohlini, ornithomimids remain poorly resolved in this study. Maximum agreement subtrees identified Pelecanimimus as a primary source of ambiguity. A priori pruning of Pelecanimimus results in nearly complete resolution of Ornithomimosauria under strict consensus (also recovered by the MAST), another nine-step drop in tree length, and reduction of MPTs (from 756 to 378). Iterations of the control analysis were run coding Chirostenotes and Caenagnathus collinsi as independent OTUs in order to investigate the impact of these alternatives on tree topology. Coding Caenagnathus as a separate OTU resulted in a shorter tree score yet caused reduction in resolution among oviraptorosaurians and significantly reduced frequency support among internal oviraptorosaurian subclades. Under majority rule consensus, Chirostenotes remains within a derived subclade of Oviraptorosauria whereas Caenagnathus is recovered at the base of the clade. Senter (2007) also reported this discrepancy when the taxa were coded as separate OTUs. The contention put forth that wide topological distance casts doubt on the synonymy of these taxa (Senter 2007, p. 1) is unwarranted. Many of the features lacking on the dentary of Caenagnathus that, in turn, support a more primitive position among oviraptorosaurians (e.g. downturned rostral dentary, shortened dentary ramus, rostral displacement of external mandibular fenestra; see Senter 2007) can just as easily be reversals attributed to secondary elongation of the mandible. Furthermore, the dentary of Caenagnathus does not resemble that of Incisivosaurus, widely regarded as the most basal member of the clade. The topological discrepancy in these taxa when coded separately is interesting and probably points to mosaic evolution in the skeletons of North American oviraptorosaurians relative to their Asian counterparts, specifically with regard to skull morphology. The dromaeosaurid topology recovered here differs significantly from that of most recent version of the base TWiG (Turner et al. 2007b). Turner et al. (2007b) included four additional paravian taxa as well as several character states not incorporated in this analysis; thus, with regard to paravian relationships, their analysis is based on more comprehensive evidence. The matrix was tested using standard bootstrapping (Felsenstein 1985) and symmetrical resampling (Goloboff et al. 2003) methods to investigate clade support and robusticity (Fig. 6). The clades Therizinosauria (bootstrap frequency [BF] = 53), Therizinosauroidea (BF = 97), therizinosauroids more derived than Beipiaosaurus (BF = 96) and therizinosauroids more derived than Alxasaurus (BF = 77) are among a handful recovered among all of Coelurosauria with greater than 50% frequency in this study. Symmetrical resampling using GC values illustrates a similar pattern. Here, Therizinosauroidea (GC = 97) and all therizinosaurians more derived than Beipiaosaurus (GC = 99) have greater support values than any other coelurosaurian subclade except the monophyletic Mononykinae (GC = 100) and Tyrannosauridae (GC = 100). Thus the current matrix establishes strong character support for the monophyly of Therizinosauria, Therizinosauroidea and basal therizinosaurian tree topology. Comparisons with previous analyses Ingroup relationships of Therizinosauria. To date, only four phylogenetic studies have attempted to resolve ingroup relationships among more than three therizinosaurian species, and all were conducted within the past five years (Table 4). Although the analysis of Zanno (2006) contains 10 therizinosaurian OTUs, that study was intended to describe the morphological evolution of the therizinosaurian forelimb and is therefore considered only in a marginal context here. Among the four comprehensive analyses, Kirkland et al. (2005) is too poorly resolved to be considered further. Clark et al. (2004), Li et al. (2007) and Senter (2007) include the most resolved hypotheses of therizinosaurian ingroup relationships and form the basis of comparison with this work. The therizinosaurian topology recovered in this analysis differs from previous hypotheses in several regards (Fig. 8) most notably in the positions of Neimongosaurus, Erliansaurus, Nothronychus, Suzhousaurus and Nanshiungosaurus brevispinus, and the taxonomic composition of Therizinosauridae. In this study, Neimongosaurus, Erliansaurus and Enigmosaurus represent intermediategrade therizinosaurians restricted from Therizinosauridae following both the phylogenetic definition proposed by Zhang et al. (2001) and the modified definition proposed in Zanno et al. (2009) whereas Nothronychus is posited as a therizinosaurid and Suzhousaurus is posited as the sistertaxon to Therizinosauridae (Fig. 7). Thus, as noted by Zanno et al. (2009), this study provides the first systematic support for the proposal of Xu et al. (2002a) regarding the intermediate status of Neimongosaurus and Erliansaurus and the proposal by Kirkland & Wolfe (2001) regarding the derived status of Nothronychus. By comparison, Therizinosauridae is maximally inclusive in Clark et al. (2004) and Senter (2007), including all therizinosaurians other than Falcarius (Senter 2007 only), Beipiaosaurus, Alxasaurus and Enigmosaurus (Clark et al only), and containing equivalent taxonomic content under both definitions. In Clark et al. (2004), Li et al. (2007) and Senter (2007), Neimongosaurus is considered a therizinosaurid; the latter two studies also include Erliansaurus within Therizinosauridae. As Zanno (2006) demonstrated, the humerus of Neimongosaurus is among the most derived of any therizinosaurian, exhibiting strongly expanded proximal and distal aspects, a hypertrophied internal tuberosity delineated by a pronounced notch, and a tubercle on the cranial face of the distal humerus, yet lacking a well developed caudal tubercle on the humeral shaft. However,

27 Re-evaluation of Therizinosauria 529 Table 4. Statistical summary of previously hypothesized therizinosaur ingroup relationships. Phylogenetic analysis including at least two therizinosaur OTUs (i.e. ingroup or partial ingroup analyses) shown. This list does not include TWiG based analyses other than Kirkland et al. (2005) and Senter (2007). Reference Character list Therizinosaur OTUs (bold taxa pruned from final tree) Taxa/characters #MPTs/tree score Tree stats This paper 28% new (= 50% novel; 50% adapted from cited sources) Falcarius Neimongosaurus 76/ /1225 CI = % TWiG (= Turner et al. 2007a; 10% modified) Beipiaosaurus Enigmosaurus Ratio: 4.6 RI = 0.70 Alxasaurus N. brevispinus RC = 0.25 No. mckinleyi N. bohlini No. graffami Erlikosaurus Suzhousaurus Segnosaurus Erliansaurus Therizinosaurus Li et al % (Clark et al. 2004) Falcarius Neimongosaurus 17/ /70 CI = 0.67 Beipiaosaurus Enigmosaurus Ratio: 2.4 HI = 0.33 Alxasaurus N. brevispinus RI = 0.70 No. mckinleyi Erlikosaurus RC = 0.47 Suzhousaurus Segnosaurus Erliansaurus Therizinosaurus Senter % new characters (= 15% novel; 21% from cited sources) Falcarius Neimongosaurus 85/ /1224 CI = % TWiG (Kirkland et al. 2005; 18% modified) Beipiaosaurus N. brevispinus Ratio: 4.2 HI = 0.64 Alxasaurus Erlikosaurus RI = 0.76 No. mckinleyi Segnosaurus RC = 0.27 Erliansaurus Therizinosaurus Zanno % novel Falcarius Neimongosaurus 13/32 35/46 CI = % adapted from cited sources Beipiaosaurus Erlikosaurus Ratio: 2.5 RI = 0.78 Alxasaurus Segnosaurus RC = 0.56 No. mckinleyi Therizinosaurus Erliansaurus Alectrosaurus Kirkland et al % novel Falcarius No. mckinleyi 58/ /679 CI = % TWiG (Hwang et al. 2004) Beipiaosaurus Erlikosaurus RI = 0.74 Alxasaurus Segnosaurus RC = 0.31 Clark et al Xu et al. 1999b Beipiaosaurus Neimongosaurus 15/ /63 CI = 0.75 Alxasaurus Erlikosaurus Ratio: 2.7 RI = 0.75 No. mckinleyi Segnosaurus RC = 0.56 Enigmosaurus Therizinosaurus N. brevispinus Eshanosaurus Xu et al. 1999b 14% novel Beipiaosaurus Therizinosauridae 8/84 CI = % from cited sources Alxasaurus Ratio: /133 RI = 0.65 RC = 0.46

28 530 L. E. Zanno Figure 8. Previously hypothesized ingroup relationships of Therizinosauria. A, Xuet al. (1999b); B, Xu & Wang (2004); C, Clarket al. (2004), strict consensus; D, Clark et al. (2004) Adams consensus; E, Zanno (2006); F, Li et al. (2007); G, Senter (2007); H, Kirkland et al. (2005). Therizinosauria (as defined here) shown in grey. the pelvic and hind limb morphology is intermediate in Neimongosaurus, suggesting a more basal position as was reported by Xu et al. (2002a). Neimongosaurus lacks the following features present in more derived therizinosaurians: dorsal surface of postacetabular process of ilium hyper-rugose, with hypertrophied caudal tuberosity; pubic peduncle of ilium severely compressed craniocaudally, with transverse dimension more than twice craniocaudal depth; ischiadic peduncle of ilium forms hypertrophied, spherical boss; cranial tuberosity on exposed distal tibia; proximal metatarsal IV lacking caudal flange buttressing metatarsal III, and metatarsus less than 39% length of tibia.

29 Re-evaluation of Therizinosauria 531 Relative over-representation of pectoral and forelimb characters can result in derived placement for this taxon. This appears to be the case in Senter s (2007) analysis, where the therizinosaurid placement of Neimongosaurus is supported only by unambiguous synapomorphies of the pectoral girdle and forelimb fused scapulocoracoid; and width of distal humerus relative to length. In Clark et al. (2004) and Li et al. (2007) the derived position of Neimongosaurus is supported by a rostrally edentulous dentary and laterally facing glenoid fossa. Rostral edentulism remains a synapomorphy of Neimongosaurus and therizinosaurids under the topology presented here. However, the glenoid is primitive in Neimongosaurus, facing ventrally rather than laterally. A laterally facing glenoid characterizes only the therizinosaurids Nothronychus, Suzhousaurus and Therizinosaurus. Neimongosaurus and Erliansaurus are relatively smallbodied taxa from the Iren Dabasu Formation, for which dates have not yet been firmly established (micropalaeontological studies [Ma 1994; von Itterbeeck et al. 2005] support a Campanian-Maastrichtian age, whereas macrovertebrates support a Cenomanian-early Campanian age [Jerzykiewicz & Russell 1991; Currie & Eberth 1993; Godefroit et al. 1998]). Although numerous hypotheses can be put forth to explain the discrepant morphology of Neimongosaurus, two are regarded as most plausible here: either (1) the pectoral girdle and forelimb were a centre of rapid adaptation occurring early in the evolution of therizinosaurians (preceding the development of the characteristic therizinosaurian pelvis and hind limb), indicating that Neimongosaurus and/or Erliansaurus represent true intermediate-grade taxa; or (2) the more primitive pelvis and hind limb of these species is attributable to secondary reduction of body size, which would require adaptive coupling of body size and several of the pelvic and hind limb characters used in this analysis. In the latter case Neimongosaurus and/or Erliansaurus would be expected to nest within Therizinosauridae in future studies, pending discovery of transitional species that demonstrate secondary reversal of pelvic and hind limb characters. The phylogenetic hypothesis put forth by Li et al. (2007; Fig. 8F) is somewhat anomalous in that Nothronychus and Suzhousaurus are recovered as a monophyletic clade more basal than Alxasaurus. The monophyly of Nothronychus and Suzhousaurus in Li et al. (2007) is supported by a single purported synapomorphy (length of deltopectoral crest less than one-third total length of humerus). The deltopectoral crest is 42 46% the total length of the humerus in the two species of Nothronychus; therefore, this character is incorrectly coded. No evidence of a monophyletic relationship between Nothronychus and Suzhousaurus was recovered in this study. The evolutionary relationships of Nothronychus have been fairly contentious, yet only recently subject to empirical study. Kirkland & Wolfe (2001) suggest an intermediate morphology for N. mckinleyi yet included this species within Therizinosauridae. Xu et al. (2002a, p. 3) excluded No. mckinleyi from Therizinosauridae based on the presumed absence of an expanded and laterally deflected preacetabular process of ilium, straight dorsal margin of ilium, and blade-like pedal unguals. However, No. graffami demonstrates that the North American genus possesses numerous therizinosaurid features, including the altiliac condition. Li et al. (2007) consider the Early Cretaceous taxon Suzhousaurus to represent a basal therizinosaurian based on the absence of a caudal tubercle on the humerus (also lacking in Neimongosaurus and Nothronychus). Here, Suzhousaurus is supported as sister-taxon to Therizinosauridae (MAST and/or Therizinosaurus pruned; see Results). The derived placement of Nothronychus and Suzhousaurus in this study is based on possession of the following derived synapomorphies: dorsal flange on dorsal scapular blade (present only for Suzhousaurus); glenoid fossa extends onto external surface of scapulocoracoid; ulnar shaft straight (unknown in Suzhousaurus); pubic boot cranially oriented with little to no caudal process; ischiadic peduncle and antitrochanter form hypertrophied, spherical boss (not in Suzhousaurus); pubic shaft mediolaterally flattened; distal pubic shaft enlarged, more than twice craniocaudal depth of proximal shaft; mediocranial distal tibia exposed, expressed as cranial tuberosity (unknown for Suzhousaurus); and metatarsus less than 36% length of tibia (unknown for Suzhousaurus). In contrast to this study, Senter (2007) recovered Nanshiungosaurus as the basalmost therizinosaurid (Fig. 8G); however, no synapomorphies are presented for the most inclusive clade of therizinosaurians lacking Nanshiungosaurus, so it is unclear what characters exclude this taxon from occupying a more nested position. Here Nanshiungosaurus is recovered as a derived therizinosaurian based on pelvic morphology (see above). Although Enigmosaurus was found to be an unstable taxon in Clark et al. (2004), it is posited outside of Therizinosauridae in this study, primarily due to the absence of a flattened pubic shaft (also absent in Erliansaurus) and lack of a cranially oriented pubic boot. Fusion of the pubis and ischium via the obturator process indicates a close evolutionary relationship to Therizinosauridae; however, this feature is unknown in other therizinosauroids (e.g. Neimongosaurus, Erliansaurus and Alxasaurus) and may have a broader distribution than currently recognized. Outgroup relationships of Therizinosauria. Perhaps the most discrepant finding of this study is that it challenges the widely regarded hypothesis of a monophyletic Therizinosauria/Oviraptorosauria clade (see Table 3). The anatomy of the basalmost therizinosaurian Falcarius directly contradicts many of the characters previously purported to be synapomorphic for this grouping and demonstrates that basal therizinosaurians possessed many

30 532 L. E. Zanno primitive coelurosaurian features not currently known in basal oviraptorosaurians. Pruning Falcarius from the analysis results in the re-establishment of a monophyletic relationship between therizinosaurians and oviraptorosaurians; thus this taxon is critical to the recovery of the topology presented here. The absence of a monophyletic Therizinosauria/Oviraptorosauria clade was also recovered by Senter (2007) and, although apparently not tested, was probably due to the inclusion of Falcarius in the study. Unambiguous derived synapomorphies reported in previous analyses as supporting a monophyletic Therizinosauria/Oviraptorosauria clade include: edentulous premaxilla (Makovicky & Sues 1998); caudal margin of naris nearly reaching or overlapping rostral border of antorbital fossa (Xuet al. 2002b [i.e. TWiG]); rim around antorbital fossa well developed (Sues 1997); reduced basipterygoid processes (Sues 1997; Makovicky & Sues 1998; Rauhut 2003; Xu et al. 2002b [i.e. TWiG]); suborbital fenestra reduced or absent (Xu et al. 2002b [i.e. TWiG]); ectopterygoid lateral to palatine (Sues 1997); palatine not tetraradiate (Rauhut 2003); symphyseal region of dentary inflected medially toward midline [U-shaped rostral symphysis] (Makovicky & Sues 1998; Xu et al. 2002b [i.e. TWiG]); cranial part of dentary flexed medially [in dorsal view] (Rauhut 2003); moderate number of small teeth (25 30) in dentary (Xu et al. 2002b [i.e. TWiG]); two pairs of pleurocoels in cervical vertebrae (Rauhut 2003); prezygapophyses of distal caudal vertebrae not elongate (Rauhut 2003); neural spines on distal caudals form low ridge (reversal) (Xu et al. 2002b [i.e. TWiG]); distal caudal centra short (Sues 1997; Makovicky & Sues 1998; Rauhut 2003); proximal end of chevrons of proximal caudals craniocaudally short with cylindrical shaft (reversal) (Xu et al. 2002b [i.e. TWiG]); pronounced lip on manual unguals (Sues 1997; Makovicky & Sues 1998; Xu et al. 2002b [i.e. TWiG]); preacetabular portion of ilium deep relative to postacetabular portion (Makovicky & Sues 1998; Rauhut 2003). Several of these features are absent in basal therizinosaurians and/or oviraptorosaurians thus cannot represent synapomorphies. Zanno et al. (2009) reviewed these features and presented a dozen additional convergences among derived members of the clades that are absent in basal members. Although appearing morphologically primitive in several respects relative to their Asian counterparts, North American (NA) oviraptorosaurians are recovered in this study as more derived than Incisivosaurus, Protarchaeopteryx and Caudipteryx (Fig. 6). Therefore, although NA oviraptorosaurians share a number of features exclusively with therizinosaurians, these characters are refuted as synapomorphies by the nested topology of the former clade in this analysis. Such features include cervical centra with two pairs of pneumatic foramina (also in Avimimus); absence of accessory antorbital fenestra; coracoid subcircular, caudoventral process poorly developed (also in Avimimus); subcondylar recess absent; and ligament groove on caudal femoral head reduced (also in Avimimus). In addition, two oviraptorosaurian genera, Avimimus and Ingenia, share features exclusively with therizinosaurians. Features shared between therizinosaurians and Avimimus alone include hypertrophied diapophyseal facets on cranial dorsal vertebrae (not in Falcarius); and transverse processes of cranial dorsal vertebrae pendent. Features shared exclusively with Ingenia include distal scapula blade tapered; deltopectoral crest more than 60% total length of humerus (also in Mononykus); and ratio of metacarpal II/(phalanx II-I + phalanx II-II) greater than 100%. The unusually high degree of convergence between derived therizinosaurians and oviraptorosaurians suggests that similar adaptive drivers were propelling morphological evolution within these clades. The convergence exhibited between all therizinosaurians and several species of oviraptorosaurians (e.g. Avimimus) is particularly intriguing. Evolutionary studies from a functional and/or palaeobiological perspective are needed to add broader significance to these findings and establish causal factors that may be responsible for the patterns identified here. This study also recovered several homoplastic characters present on therizinosaurians and alvarezsaurids that are refuted as homologies by the anatomy of the basalmost members of these clades, including: opisthocoelous cervical vertebrae; femoral head confluent with neck, lacking raised ventral rim; transverse width of distal humerus significantly expanded relative to transverse width of shaft; medial aspect of distal humerus expanded, subtriangular; entepicondyle displaced medially relative to ulnar condyle; and lateral fossa of proximal tibia shallow. The discovery of primitive, Neocomian members of the clades Therizinosauria and Oviraptorosauria is adding critical information to studies of character transformation within these clades and helping to resolve their unusual degree of convergence. However, the current rarity of alvarezsaurid remains may be hindering our understanding of the evolutionary history of this group. Although this study did not recover exclusive common ancestry among therizinosaurians, oviraptorosaurians and/or alvarezsaurids, as was recovered in Xu & Wang (2004), a high degree of morphological similarity between several members of these clades may present evidence for a close evolutionary relationship. Several characters that cannot currently be refuted by ancestral state reconstruction remain possible synapomorphies linking at least two of these clades; these include: basipterygoid processes reduced; premaxillary symphysis rounded; suborbital fenestra reduced; increased number of cervical vertebrae; caudolateral margins of ventral cervical centra with prominent crests; neural spines on dorsal vertebrae caudally bifurcated; cranial portion of proximal fibula dorsally elevated; and proximal fibula narrows caudally in dorsal view. However, as was noted in this study (see Results), forcing a monophyletic relationship between any combination

31 Re-evaluation of Therizinosauria 533 of Therizinosauria, Alvarezsauridae and Oviraptorosauria requires a significant number of additional steps, suggesting that the basal topology recovered here is well supported by the present data. Alternatively these features may represent basal maniraptoran symplesiomorphies subsequently lost in paravians. Persistent problems Although the phylogenetic analysis presented here is one of the most comprehensive yet conducted for Therizinosauria, species-level relationships within the derived subclade Therizinosauridae are not adequately resolved and a high degree of homoplasy remains a compounding factor in the support of basal tree topology. Several factors are likely to account for much of our inability to resolve evolutionary relationships among derived therizinosaurians. The most significant of these include: (1) inaccessibility, damage and potential loss of holotype materials; (2) paucity of therizinosaurian cranial remains; and (3) fragmentary specimens with non-overlapping elements. The discovery of additional, well-preserved specimens and/or the rediscovery of currently missing elements of the derived therizinosaurians Segnosaurus, Nanshiungosaurus, Therizinosaurus and Erlikosaurus will be necessary to increase resolution and support at internal nodes within Therizinosauridae. Primary tree topology among non-paravian coelurosaurian clades remains tenuously supported. This ambiguity is due to the lack of transitional taxa at basal nodes and may be resolved with the addition of characters and taxa, whereas character conflict at the base of Ornithomimosauria as documented by this study (due primarily to the mosaic morphology of Pelecanimimus) will probably remain until the discovery of additional basal ornithomimosaurs. The high degree of adaptive convergence (possibly related to palaeobiological factors) between nonparavian coelurosaurians obscures our ability to identify phylogenetic signals amongst members of these clades. This is especially true for derived therizinosaurians, oviraptorosaurians and alvarezsaurids, which, as demonstrated by this study, possess an unusually high degree of morphological convergence. The exceedingly poor fossil record of North American oviraptorosaurians, lack of a postcranial skeleton for the basalmost oviraptorosaurian Incisivosaurus, and poor condition of the holotype of Protarchaeopteryx robustus are all factors precluding morphological comparisons with the anatomy of therizinosaurians in particular. Similarly the sparse fossil record of alvarezsaurids (including a lack of pre-turonian species) is currently hindering our understanding of the phylogenetic relationships of this clade. Future study of nonparavian coelurosaurian clades with an eye to identifying homoplastic morphology as related to palaeobiological factors is needed to advance our understanding of the evolution and morphological transformation of these clades. In addition, character correlation techniques will probably prove useful in helping to tease out the functional and/or dietary significance of convergent characteristics amongst coelurosaurian lineages and identifying homoplastic transformations (see Zanno et al. 2009). Finally, the re-evaluation and addition of several known Late Jurassic and Early Cretaceous coelurosaurians may yield additional insight into morphological trends within coelurosaurian subclades and their early diversification as well as provide increased understanding of character evolution within basal coelurosaurian nodes. Discussion Morphological trends Zanno (2010) reviewed several morphological trends evident within Therizinosauria. However, these interpretations are general and based on prior definitions of the major subclades Therizinosauroidea and Therizinosauridae rather than species as the phylogenetic study presented here was not yet completed. Here the discussion of the trends identified in Zanno (in press) is augmented with specific reference to character development based on the phylogenetic relationships recovered in this study. Zanno (2010) noted the presence of unserrated, incisiform and elongate rostral teeth in basal therizinosaurians and oviraptorosaurians and hypothesized that this condition may have been present in a shared common ancestor based on prior phylogenetic studies suggesting a monophyletic relationship between these clades. The broad phylogenetic relationships recovered here (i.e. therizinosaurians and oviraptorosaurians representing successive basal maniraptoran clades) suggest that unserrated and enlarged incisiform rostral teeth (and marked heterodonty) appeared early in the evolution of maniraptorans and do not represent an apomorphy restricted to these clades. If correctly optimized as a basal maniraptoran condition, this feature is predicted to be found in basal members of other maniraptoran subclades (e.g. alvarezsaurids, paravians). Regarding basal therizinosaurians, a high degree of morphological change is evident between Beipiaosaurus and Falcarius, although these taxa are approximately coeval. The basalmost therizinosauroid Beipiaosaurus possesses several derived features not present in Falcarius, including: coarsely serrated cheek teeth; expanded preacetabular portion of ilium; reduced postacetabular portion of ilium; subequal pubic and ischiadic peduncles of ilium; elongate ischium with distal boot; lateral ridge on ischiadic shaft; tibia/femur ratio approximately 1; and shortening of the distal caudal vertebrae. These features indicate a high degree of adaptive change occurring in the pelvis between basal therizinosaurians and basal therizinosauroids. Although the fragmentary condition of most derived therizinosaurian specimens makes identifying the timing

32 534 L. E. Zanno of character development within the clade to the level of species difficult, several patterns regarding the sequence of character acquisition can be hypothesized based on the ingroup topology recovered in this study. Following a high degree of development in the pelvis of basal therizinosauroids (e.g. Beipiaosaurus), adaptive change appears to be concentrated in the forelimbs, ankle and pes, which are highly transformed in the intermediate therizinosauroids Neimongosaurus and Erliansaurus; whereas pelvic morphology in these taxa is relatively conservative. Derived therizinosauroids and basal therizinosaurids (e.g. Nothronychus, Suzhousaurus and Enigmosaurus) exhibit a second concentration of character development in the pelvis, with continued transformation of the forelimb and hind limb. In addition to general trends identified here, specific morphological features can be identified as characterizing increasingly derived therizinosaurian subclades. These are listed along with the basalmost therizinosaurian for which the trait is known; however, the fragmentary condition of most derived therizinosaurian specimens creates minimal estimates on the most inclusive clade for which the character can be attributed. Cranial characteristics are especially problematic given the near lack of cranial materials known for therizinosaurians. Features present in Alxasaurus and more derived therizinosauroids include: increase in crown symmetry; elongation of neural spines on dorsal vertebrae; markedly shortened distal caudal vertebrae; loss of fused semilunate carpal (DCI and DCII); and metatarsals reduced in length. Features present in Neimongosaurus and/or Erliansaurus and more derived therizinosauroids include: development of broad, U-shaped rostral dentary; internal tuberosity of humerus hypertrophied and delineated from humeral head by notch, cranial tubercle on distal humerus; caudal tuberosity on humeral shaft; medial aspect of distal humerus expanded, entepicondyle located well medial to ulnar condyle; distal metacarpal III ginglymoid in dorsal view, with straight shaft; manual phalanges reduced in length; caudodorsal rugosity on postacetabular portion of ilium; femoral head dorsally inclined, region bridging head and greater trochanter depressed and craniocaudally constricted; lesser trochanter of femur cylindrical, terminates ventral to greater trochanter; fibular crest on tibia elongate; lateral condyle of astragalus reduced; cranial trochanter on fibula distally located at midshaft; and first metatarsal contacts tarsus, pes functionally tetradactyl. Features present in Enigmosaurus and therizinosaurids include: six sacral vertebrae; hyperrugose cubic tuberosity on postacetabular process of ilium; pubic peduncle of ilium severely flattened craniocaudally, caudally recurved, and more than triple the length of the ischiadic peduncle; obturator process of ischium contacts pubis; contact between ischium and pubis sinusoidal; loss of lateral tubercle on pubis proximal to acetabulum; pubic shaft cranially concave; and development of a ventrocaudal tuberosity on ischium. Features present in Nothronychus and more derived therizinosaurids include: foreshortening of basicranial region; increasing basicranial pneumaticity; enclosure of basisphenoidal recess by peripheral intramembraneous ossification (Smith et al. in press); lateral reorientation of glenoid; manual unguals enlarged (more than twice the length of penultimate phalanx) and transversely flattened; hyperelongate preacetabular portion of ilium; antitrochanter and ischiadic peduncle of ilium indistinguishable, forming a spherical boss; loss of caudal component of pubic boot; and pubic shaft flattened. Features present in derived therizinosaurids other than Nothronychus include: caudal extension of the vomers; development of a coronoid process on the dentary; downturned and edentulous rostral dentary; secondary reduction in number of dentary teeth; hyperelongate manual unguals; and ascending process of astragalus braces cranial fibular shaft. Geographic distribution of Therizinosauria North America. Three species and two genera of therizinosaurians are known from the western USA: Falcarius utahensis from the Lower Cretaceous Cedar Mountain Formation of Utah, No. mckinleyi from Upper Cretaceous Moreno Hill Formation of New Mexico and No. graffami from the Upper Cretaceous Tropic Shale of Utah. Several isolated elements from the Late Cetaceous of Alberta, Canada have been referred to the clade (Russell 1984; Currie 1992; Ryan & Russell 2001). These include two isolated frontals. One of these specimens, CMN 12355, was regarded as deinonychosaurian gen. et. sp. indet. by Sues (1978), and as cf. Erlikosaurus by both Currie (1987; see also Currie 1992 in which the specimen is apparently listed mistakenly as CMN 12349), and Ryan & Russell (2001). The second specimen is TMP , regarded as cf. Erlikosaurus by Currie (1992). Other possible therizinosaurian elements from Alberta include an ungual (TMP , cf. therizinosaurid, Currie 1992) from the Dinosaur Park Formation, an astragalus (therizinosaurid indet., Russell 1984) from the Hell Creek Formation and a cervical vertebra (TMP , therizinosaurid indet., Ryan & Russell 2001) from the Scollard Formation. Referral of these finds has been questioned by several authors (Barsbold & Maryańska 1990; Clark et al. 1994; Maryańska 1997), although no anatomical evidence has been put forth to refute these assignments. Moreover it is unclear whether it is referral of these elements to Therizinosauria indet. or to cf. Erlikosaurus that is in doubt. Asia. China and Mongolia are the only Asian countries from which therizinosaurians have been named. However, several putative therizinosaurian elements have

33 Re-evaluation of Therizinosauria 535 been reported from Russia, Uzbekistan, Kazakhstan and Japan. Since 1979 eight species of therizinosaurian have been described from China, six of these within the past decade alone. China currently has the highest known diversity of therizinosaurians and the longest documented temporal span for the clade (with or without the alleged Early Jurassic therizinosaurian, Eshanosaurus). Chinese therizinosaurians include: Beipiaosaurus inexpectus from the Yixian Formation (Neocomian); Alxasaurus elesitaiensis from the Bayin Gobi Formation (Aptian/Albian); Erliansaurus bellamanus and Neimongosaurus yangi from the Iren Dabasu Formation (Campanian/Maastrichtian); Nanshiungosaurus brevispinus from the Yuanpu (Nanxiong) Formation (Campanian/Maastrichtian); the problematic taxa Nanshiungosaurus bohlini from the uppermost Xinminbao Group (late Early Cretaceous) and Eshanosaurus deguchiianus from the lower Lufeng Formation (Early Jurassic, Hettangian; Luo & Wu 1994). Eshanosaurus deguchiianus was described by Xu et al. (2001) from an isolated, incomplete left mandible with teeth (IVPP V11579). If referral of Eshanosaurus to Therizinosauria is correct and the specimen is correctly dated as Early Jurassic, its existence extends the phylogenetically estimated divergence time of the nine major coelurosaurian clades by at least 50 million years, producing lengthy ghost lineages. As Kirkland et al. (2005) noted, Eshanosaurus possesses derived therizinosaurian features (e.g. lateral shelf on the dentary; downturned mandibular symphysis) absent in the most primitive North American therizinosaurian Falcarius, known from approximately 80 million years later. Although the combination of morphological and stratigraphical incongruence casts doubt on the assignment of Eshanosaurus to Therizinosauria, this evidence is not definitive. The existence of the Late Jurassic paravian Archaeopteryx produces million year ghost lineages across major coelurosaurian clades, yet its taxonomic affinities are unquestioned. Furthermore, it is conceivable that Falcarius is a late-surviving member of a basal clade of therizinosaurians that persisted in North America until the Early Cretaceous, a conclusion consistent with the observation that Beipiaosaurus, an Asian therizinosaurian of comparable age to Falcarius, is morphologically advanced relative to the Utah taxon (Zanno 2010). Finally, a recent study (Barrett 2009) disputed the referral of the dentary to Sauropodomorpha, and Eshanosaurus possesses many features unknown in ornithischians (Xu et al. 2001) and thus cannot confidently be referred to these taxa. Overall, the identity of Eshanosaurus remains enigmatic and unlikely to be satisfactorily resolved until additional materials are recovered. Putative therizinosaurian materials from China include portions of the Alectrosaurus olseni hypodigm from the Iren Dabasu Formation (Mader & Bradley 1989), several specimens from the Iren Dabasu Formation housed at the Erenhot Dinosaur Museum (Currie & Eberth 1993) and Chilantaisaurus zheziangensis from the Tangshang Formation (Dong 1979; Barsbold & Maryańska 1990). Mader & Bradley (1989) assigned one of the syntypes of Alectrosaurus olseni (AMNH 6368) to Therizinosauridae (Fig. 9). AMNH 6368 includes a proximally and distally damaged right humerus, a manual phalanx and a manual ungual. First-hand examination of the material suggests it is referable to Therizinosauria and appears to be intermediate in morphology. Currie & Eberth (1993) also referred several additional isolated elements from the Iren Dabasu to Therizinosauria (including a dentary with teeth). They referred these elements specifically to Erlikosaurus, Segnosaurus and possibly Enigmosaurus based on a hypothesis of biostratigraphical correlation between the Iren Dabasu and the Bayan Shire Formation of Mongolia (see also Jerzykiewicz & Russell 1991). However, von Itterbeeck et al. (2005) suggested a Campanian-Maastrichtian age for the Iren Dabasu Formation based on charophyte and ostracod fossils, which would indicate closer temporal correlation with the Nemegt Formation. AMNH 6368 is not referable to the Iren Dabasu therizinosauroid Neimongosaurus, based on the relatively slender proportions, prominent caudal trochanter and poorly defined internal tuberosity of the humerus. Presence of the latter character also precludes its referral to the Bayan Shire taxon Erlikosaurus and the Nemegt therizinosauroid, Therizinosaurus, as does its more moderately developed deltopectoral crest. Segnosaurus possesses a remarkably straight humeral shaft, more so than AMNH 6368, with the latter exhibiting moderate deflection of the proximal and distal aspects. Furthermore, although the distal aspect of the humerus of AMNH 6368 is crushed, this element appears notably narrow craniocaudally relative to that of Segnosaurus. The humerus is more gracile than that of Erliansaurus and possesses a different caudal trochanter morphology, yet the holotype of Erliansaurus bellamanus is a juvenile (Xu et al. 2002a) and AMNH 6368 is approximately 50% larger; therefore ontogenetic variation cannot be excluded as an explanation for these differences. Referral to the Bayan Shire taxon Enigmosaurus (for which the humerus is unknown) is also possible. The material identified by Currie & Eberth (1993) as therizinosaurian has not yet been examined first hand by the author, so no observations regarding the referral of these elements are made here. Chilantaisaurus zheziangensis is known from a proximal portion of a right tibia, incomplete metatarsals II, III and IV and pedal unguals (Dong 1979, translated by Will Downs) (although Barsbold & Maryańska [1990] and Maryańska [1997] only noted the presence of pedal digits II and III) and was originally referred to Megalosauridae (Dong 1979). Chilantaisaurus zheziangensis is questionably referred to this genus, which also contains the species C. tashuikouensis and C. maortuensis (Barsbold

34 536 L. E. Zanno Figure 9. Therizinosaurian humerus (AMNH 6368) (A-C) and manual unguals (AMNH 6554) (D) formerly part of Alectrosaurus olseni hypodigm. A, cranial view showing poorly defined internal tuberosity; B, medial view showing moderately developed deltopectoral crest and cranial deflection and gracile morphology of distal humerus and morphology of caudal trochanter; and C, caudal view; D, lateral view. Abbreviations: ct, caudal tuberosity of humerus; dpc, deltopectoral crest; hh, humeral head; it, internal tuberosity; and uc, ulnar condyle. Scale bars: 5 mm (A-C); 3 mm (D). & Maryańska 1990). Extreme lateral compression and large size of the pedal unguals are indicative of Therizinosauria; however, examination of the proximal portion of the tibia is needed for a confident referral to the clade. To date the Cretaceous deposits of Mongolia have produced four therizinosaurian species: Erlikosaurus andrewsi (Perle 1981), Enigmosaurus mongoliensis (Barsbold 1983) and Segnosaurus galbinensis (Perle 1979) from the Late Cretaceous Bayan Shire Formation, and Therizinosaurus cheloniformis (Maleev 1954), the first therizinosaurian recovered, from the Upper Cretaceous Nemegt Formation (Maastrichtian). Several isolated therizinosaurian elements from Mongolia are known, including an exceptionally preserved left pubis from Baishin Tsav locality I-IV lacking a specimen number (listed as Field # 93) discovered in the IGM collections and referable to cf. Segnosaurus (pers. obs.), as well as a fibula from the same locality (Field # 08) which may also be referable to this taxon. Finally, a considerable amount of as yet undescribed therizinosaurian material was collected during the Japan Mongolian Joint Palaeontological Expeditions in , including cranial, axial and appendicular elements (Watabe & Suzuki 2000a, b) and a well preserved pelvis (M. Watabe pers. comm., December 2006). A partial braincase, several teeth and a humeral shaft of a therizinosaurian were reported from the Mifune Group (Upper Cretaceous) on Japan s Kyushu Island (Ikegami & Tomida 2005). Reported basisphenoidal pneumaticity, as well as the presence of a posterior trochanter on the humeral shaft, is strong evidence supporting assignment of this material to Therizinosauria, as suggested by Ikegami & Tomida (2005). Putative therizinosaurian materials are also known from Uzbekistan, Karakalpakstan, Tajikistan, Kazakhstan and Russia. Although much of this material is represented by large, isolated unguals of questionable affinity, additional postcranial and cranial elements potentially referable to Therizinosauria have been described. Averianov (2007) referred several elements from the Santonian Bostobe Formation of Kazakhstan to Therizinosauria; these include a right femur (CCMGE 601/12457), tooth (ZIN PH 35/49), two proximal portions of left femora (ZIN PH 24/49 and 37/49), and pedal phalanx (ZIN PH 25/49). Averianov (2007) also cited therizinosaurian material from the Turonian Khirkindek Formation, including a juvenile sacral centrum (ZIN PH 32/49, partial left metacarpal III (ZIN PH 1/49), and two proximal portions of left femora (ZIN PH 38/49, and 39/49). Nessov (1995) referred the right femur noted above from the Bostobe Formation (CCMGE 601/12457) to Tarbosaurus. Averianov (2007) assigned it to cf. Neimongosaurus sp. based on presumed similarity to this taxon.