A brief history of Massospondylus: its discovery, historical taxonomy and redescription of the original syntype series Paul M. Barrett1,2* & Kimberley E.J. Chapelle2,3 1Fossil Reptiles, Amphibians and Birds Section, Natural History Museum, Cromwell Road, London SW7 5BD, U.K. 2Evolutionary Studies Institute, University of the Witwatersrand, Private Bag 3, WITS 2050, Johannesburg, South Africa 3Department of Anatomical Sciences, Stony Brook University, Nicolls Rd, Stony Brook, NY, 11794, U.S.A. Received 24 June 2024. Accepted 27 November 2024 INTRODUCTION Massospondylus carinatus Owen, 1854 is one of the most fully documented Early Jurassic sauropodomorph dino- saurs. The anatomy of several key specimens has been described in detail (Gow 1990; Sues et al. 2004; Chapelle & Choiniere 2018; Barrett et al. 2019), and it has been the focus of numerous studies on growth rate (Chinsamy 1993; Erickson et al. 2001), ontogeny (Gow 1990; Gow et al. 1990; Reisz et al. 2005, 2010; Chapelle et al. 2020a, 2021), repro- ductive behaviour (Kitching 1979; Grine & Kitching 1987; Zelenitsky & Modesto 2002; Reisz et al. 2012; Stein et al. 2019), feeding and tooth replacement (Raath 1974; Barrett 2000; Barrett & Upchurch 2007; D’Emic et al. 2013) and locomotion (Bonnan & Senter 2007; Neenan et al. 2019; Chapelle et al. 2020b), providing an exemplar for work on other early sauropodomorph taxa. Referred specimens have been considered so abundant that Massospondylus forms the basis of an eponymous biostratigraphical assemblage, the Massospondylus Assemblage Zone, which has been used for intra- and interbasinal correlations across southern Africa (e.g. Bond 1965, 1973; Cooper 1981; Kitching & Raath 1984; Bordy et al. 2020; Viglietti et al. 2020) and elsewhere (Attridge et al. 1985; Apaldetti et al. 2011). However, our current familiarity with M. carinatus masks a lengthy taxonomic history, and the taxon was poorly characterized and rarely discussed until relatively recently (see reviews in Cooper 1981; Sues et al. 2004; Yates & Barrett 2010; Barrett et al. 2019). This neglect was due, at least in part, to the poor quality of the original syntype series, which consisted of isolated, and often broken, elements of uncertain association that were destroyed during World War II. Here, we re-evaluate the collection of material that yielded the syntypes of M. carinatus Owen, 1854, Leptospondylus capensis Owen, 1854 and Pachyspondylus orpenii Owen, 1854, as well as the material that Richard Owen and others referred to each of these taxa historically, based on the plaster casts of these speci- mens that survive in museum collections in London and Cape Town. We also review the history of this formerly obscure, but now well-known, early dinosaur taxon. A HISTORY OF MASSOSPONDYLUS CARINATUS AND ASSOCIATED TAXA For most of the nineteenth century, South Africa was a British colonial possession (then known as the ‘Cape Colony’): consequently, many of the fossils found in the region at this time were transported back to London, where they were deposited in the British Museum (now the Natural History Museum), the Hunterian Museum of the Royal College of Surgeons and other U.K. collections. In 1846, four brothers relocated from Ireland to settle in South Africa. They were the sons of Charles Edward Herbert Orpen (1791–1856), an Edinburgh-trained physi- cian and eye surgeon (https://www.geni.com/people/ ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 97 Palaeontologia africana 58: 97–131 — ISSN 2410-4418 [Palaeontol. afr.] Online only Permanently archived on the 6th of December 2024 at the University of the Witwatersrand, Johannesburg, South Africa The article is permanently archived at: https://hdl.handle.net/10539/43016 *,‡Author for correspondence. E-mail: p.barrett@nhm.ac.uk Massospondylus carinatus Owen, 1854 is one of the first dinosaurs to have been described from outside Europe and was based on material collected from what is now the upper Elliot Formation of the Free State province, South Africa. The species was included in various taxonomic reviews during the late nineteenth and early twentieth centuries but no additional material was referred beyond the original syntype series and it remained poorly known. This was exacerbated by the destruction of the syntypes during World War II. From the 1970s onward, fieldwork in the upper Stormberg Group of South Africa, Lesotho and Zimbabwe led to the discovery of many new sauropodomorph dinosaur specimens that have been referred to the taxon (often uncritically) that have been used to shed further light on the anatomy, palaeobiology and biostratigraphical utility of Massospondylus carinatus. Here, we review the taxonomic history of this species, provide updated descriptions of the syntypes (based on surviving casts) and use apomorphies to identify these specimens more accurately. Keywords: Massospondylus carinatus, Dinosauria, Sauropodomorpha, Orpen Collection, upper Elliot Formation, historical taxonomy. Palaeontologia africana 2024. ©2024 Paul M. Barrett & Kimberly E.J. Chapelle. This is an open-access article published under the Creative Commons Attribution 4.0 Unported License (CC BY4.0). To view a copy of the license, please visit http://creativecommons.org/licenses/by/4.0/. This license permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. The article is permanently archived at: https://hdl.handle.net/10539/43016 https://orcid.org/0000-0003-0412-3000 https://orcid.org/0000-0002-9991-0439 http://creativecommons.org/licenses/by/4.0/ https://hdl.handle.net/10539/43016 mailto:p.barrett@nhm.ac.uk https://hdl.handle.net/10539/43016 https://www.geni.com/people/Dr-SV-PROG-Charles-Edward-Orpen/6000000015671553805 Dr-SV-PROG-Charles-Edward-Orpen/6000000015671553 805): Francis (Frank) Henry Samuel Orpen (1824–1893), Charles Sirr Orpen (1826–1887), Joseph Millerd Orpen (1828–1923), and Richard John Newenham Orpen (1830–1913). Joseph Orpen (Fig. 1) became a British government official, working first as a land surveyor and later holding various administrative posts across south- ern Africa (South Africa, Lesotho and Zimbabwe) (Orpen 1964): https://www.geni.com/people/Joseph-Millerd- Orpen/6000000018953033632). During his surveying work, Joseph and his brothers acquired a series of 56 postcranial bones ‘… in the Drakenberg [sic] range of mountains near Harrismith …’ from ‘… a secondary formation, probably of the age of the New Red Sandstone in Europe’ (Owen 1854, p. 97) in either 1853 (J.M. Orpen in Seeley 1895a) or 1854 (Owen 1854). These specimens were sent to their father, then based in East London (Eastern Cape, South Africa), who forwarded them to Richard Owen, then curator of the Hunterian Museum (Charles Orpen, a fellow of the Royal College of Surgeons, is noted as the donor of the material with his sons as the collectors: Owen 1854). Given Joseph Orpen’s long life, he was able to provide a further account of these discoveries to Harry Govier Seeley in a letter written in 1889, noting that ‘The spot where I obtained some large bones of a saurian about 1853, which my father sent home, was on a hill capped by sandstone on the east boundary of the farm Beaucherf, in the district of Harrismith, on the watershed of the Drakensberg. Below the sandstone is a chocolate-colour- ed shale. I think more of the bones would be found on the spot by excavating. The fossils were on the east face of the beacon-hill which is northwest of Beaucherf House’ (Orpen in Seeley 1895a, p. 102). Although the locality name ‘Beaucherf ’ was used in some historical accounts (e.g. Haughton 1924), the farm is currently known as Beauchef Abbey 215 (Kitching & Raath 1984; Sues et al. 2004; Yates & Barrett 2010) and is located within the administrative district of Harrismith, Free State province, South Africa at 28°30’9.91”S and 29°9.1’29”E (Kitching & Raath 1984; P. Viglietti, pers. comm., 2021). Following additional stratigraphic work, the ‘New Redstone’ units mentioned by Owen (1854) were subsequently allocated to the ‘Stormberg Series’, being mentioned first as the ‘Red Beds’ (e.g. Haughton 1924) but later formalized as the Elliot Formation of the Stormberg Group (see review in Viglietti et al. 2020). Although these deposits were formerly regarded as Late Triassic in age (e.g. Haughton 1924; Kermack 1974), it was later recognized that they straddle the Triassic–Jurassic boundary (Olsen & Galton 1984; Olsen & Sues 1986). Kitching & Raath (1984) reported that the only exposures visible at Beauchef Abbey were from the upper Elliot Formation, which is currently regarded as latest Triassic (Rhaetian) to Early Jurassic (Hettangian–Sinemurian) in age, with most of its known fossil localities situated in the Early Jurassic portions of the unit (Sciscio et al. 2017; Bordy et al. 2020; Viglietti et al. 2020). Owen (1854) registered the Orpen Collection at the Hunterian Museum (Figs 2, 3A), assigning the specimens to a continuous series of catalogue numbers (No[s].): Nos. 331–386. One of these (No. 386) was compared closely with the jaw of a gavial, but no formal referral was made to any taxon, although Owen’s comparative observations (1854, p. 106) make it clear that he regarded it as croco- dile-like; however, Seeley (1895a) later suggested that this specimen was a saurischian chevron. Huene (1906) proposed that No. 386 represented conjoined ischial shafts, which seems plausible on the basis of his figure, but as this specimen was not referred to any particular taxon, has no type status, and is no longer available (neither the original nor a cast exists), is it not considered further herein. By contrast, Nos. 331–385 formed the basis for three new taxa – Massospondylus carinatus Owen, 1854, Pachyspon- dylus orpenii Owen, 1854 and Leptospondylus capensis Owen, 1854 (given in order of page priority) – providing the syntype series of each and a list of specimens that were referred to the three species with varying degrees of confidence (Table 1). Owen (1854) refrained from refer- ring his new taxa to any of the reptile groups recognized at the time, noting instead that they combined features seen in lizards, crocodiles and dinosaurs. Massospondylus carinatus was named based on five cervi- cal vertebrae (Nos. 331–335) that Owen (1854: pp. 97–98) identified unambiguously as belonging to the taxon and that should be regarded as its syntype series (Yates & Barrett 2010). Many other specimens from the Orpen Collection (Nos. 336, 337, 349, 350, 352, 354–356, 358–364), 98 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 Figure 1. Portrait of Joseph Millerd Orpen (from Orpen 1964). https://www.geni.com/people/Dr-SV-PROG-Charles-Edward-Orpen/6000000015671553805 https://www.geni.com/people/Joseph-Millerd-Orpen/6000000018953033632 https://www.geni.com/people/Joseph-Millerd-Orpen/6000000018953033632 including dorsal vertebrae, scapulae, humeral fragments, an ilium, a pubis (misidentified as a coracoid, see below), femora and tibiae, were also mentioned by Owen (1854: 98, pp. 101–103) as either Massospondylus carinatus?, Massospondylus or Massospondylus? (no species designa- tion in either of the latter cases) and these should be regarded as referrals (Yates & Barrett 2010), rather than part of the type series (Table 1). Owen (1854) mentioned that his new genus name referred to the elongation of the cervical vertebrae (‘masson’ Gr. = longer [comparative of makros]; ‘spondylos’ Gr. = vertebra) and it seems likely that the species name refers to the well-developed keel (carina) present on their ventral surfaces (although Owen did not mention its etymology). The second new taxon, Pachyspondylus orpenii, was also based on a set of vertebrae, in this case one dorsal, one sacral and five caudals (Nos. 338–342, 345, 346), which form the syntype series (Owen 1854: pp. 99–100). Other specimens (Nos. 343, 344, 351, 353, 357, 366–384), includ- ing caudal vertebrae, a humerus, a pubis (misidentified as a scapula, see below), an ischium (misidentified as a coracoid, see below) and an assortment of metacarpals, metatarsals and phalanges, were referred to either Pachyspondylus or Pachyspondylus? (with no species desig- nation in either case) and should be regarded as referrals (Owen 1854, pp. 100–105; Table 1). Owen’s (1854) genus name refers to the stoutness of the vertebrae (‘pachys’ Gr. = thick; ‘spondylos’ Gr. = vertebra) and the species name ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 99 Figure 2. Colour lithograph of the palaeontological gallery at the Hunterian Museum in the Royal College of Surgeons of England, London. Artists, T.H. Shepherd and E. Radclyffe. This file comes from Wellcome Images, a website operated by Wellcome Trust, a global charitable foundation based in the United Kingdom (© Wellcome Trust: https://wellcomeimages.org/indexplus/image/V0013494.html). https://wellcomeimages.org/indexplus/image/V0013494.html presumably honours one or more members of the Orpen family. Finally, Leptospondylus capensis was based on a syntype series consisting of only two caudal vertebrae (Nos. 347 and 348; Owen 1854, p. 100) and a single ungual phalanx was referred to Leptospondylus? (no species designation: No. 385; ibid., p. 105; Table 1). In this instance, Owen’s (1854) genus name refers to the slenderness of the syntype vertebrate (‘leptos’ Gr. = slender; ‘spondylos’ Gr. = verte- bra) and, although he omitted the etymology, his species name clearly reflects their provenance. Owen (1854) provided no illustrations of these speci- mens nor any detailed justifications for dividing them among the three taxa that he had erected, noting only that 100 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 Figure 3. Historical image of the Hunterian Museum in the building of the Royal College of Surgeons of England. A, lithograph of the exterior of the Royal College of Surgeons, London. Engraving after T. H. Shepherd. This file comes from Wellcome Images, a website operated by Wellcome Trust, a global charitable foundation based in the United Kingdom (© Wellcome Trust: https://wellcomeimages.org/indexplus/im- age/V0013491.html). B, photograph of the Royal College of Surgeons after a WWII bombing raid on the night of May 10th/11th 1941. The museum was directly struck by a high-explosive bomb which hit Room V and adjacent areas, destroying the osteological collections, including the Orpen Collec- tion. Courtesy of Westminster City Archives (with permission). https://wellcomeimages.org/indexplus/image/V0013491.html https://wellcomeimages.org/indexplus/image/V0013491.html ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 101 Table 1. Catalogue of the Orpen Collection formerly held in the Hunterian Museum of the Royal College of Surgeons, London. This shows the original catalogue number, Owen's (1854) taxonomic statements and his original anatomical identifications. The taxonomic status of each element is shown, along with the NHMUK and SAM catalogue numbers for the surviving casts corresponding to the original specimens. Catalogue Owen taxonomy Status Owen identification NHMUK No. SAM No. 331 Massospondylus carinatus Syntype Caudal vertebra PV R3027 SAM-PKC-958 332 Massospondylus carinatus Syntype Caudal neural arch PV R3028 (partim) SAM-PKC-959 333 Massospondylus carinatus Syntype Caudal vertebra PV R3028 (partim) SAM-PKC-960 334 Massospondylus carinatus Syntype Partial centrum PV R3028 (partim) SAM-PKC-961 335 Massospondylus carinatus Syntype Caudal centrum PV R3028 (partim) SAM-PKC-962 336 ?Massospondylus carinatus Referred Dorsal centrum PV R1312 SAM-PKC-963 337 ?Massospondylus carinatus Referred Dorsal centrum NA NA 338 Pachyspondylus orpeni Syntype Caudal vertebra PV R3035 (partim) NA 339 Pachyspondylus orpeni Syntype Caudal centrum NA NA 340 Pachyspondylus orpeni Syntype Caudal centrum NA NA 341 Pachyspondylus orpeni Syntype Caudal neural arch NA NA 342 Pachyspondylus orpeni Syntype Caudal vertebra PV R3035 (partim) NA 343 ?Pachyspondylus Referred Caudal vertebra PV R3035 (partim) NA 344 ?Pachyspondylus Referred Caudal vertebra NA NA 345 Pachyspondylus orpeni Syntype Dorsal vertebra PV R3035 (partim) NA 346 Pachyspondylus orpeni Syntype Sacral PV R3035 (partim) NA 347 Leptospondylys capensis Syntype Caudal centrum PV R3025 NA 348 Leptospondylys capensis Syntype Caudal vertebra NA NA 349 Massospondylus Referred Left scapula PV R3029 (partim) SAM-PKC-964 350 Massospondylus Referred Left scapula PV R3029 (partim) SAM-PKC-965 351 Pachyspondylus Referred Left scapula PV R3036 (partim) NA 352 Massospondylus Referred Right coracoid PV R3030 SAM-PKC-966 353 ?Pachyspondylus Referred Coracoid PV R3036 (partim) NA 354 ?Massospondylus Referred Humerus PV R3031 (partim) SAM-PKC-967 355 No name given NA ?Humerus NA NA 356 ?Massospondylus Referred Humerus PV R3031 (partim) SAM-PKC-968 357 Pachyspondylus Referred ?Humerus NA NA 358 ?Massospondylus Referred Ilium PV R3032 (partim) SAM-PKC-969 359 Massospondylus Referred Ischium PV R3032 (partim) SAM-PKC-970 360 ?Massospondylus Referred Right femur PV R3033 (partim) SAM-PKC-971 361 Massospondylus Referred Left femur PV R3033 (partim) SAM-PKC-972 362 Massospondylus Referred Right femur PV R3033 (partim) SAM-PKC-973 363 Massospondylus Referred Right tibia PV R3034 (partim) SAM-PKC-974 364 Massospondylus Referred Tibia PV R3034 (partim) NA 365 No name given NA Left tibia NA SAM-PKC-975 366 ?Pachyspondylus Referred Tarsal PV R3037 (partim) NA 367 ?Pachyspondylus Referred Metatarsal PV R3037 (partim) NA 368 ?Pachyspondylus Referred Metatarsal PV R3037 (partim) NA 369 ?Pachyspondylus Referred Metatarsal PV R3037 (partim) NA 370 ?Pachyspondylus Referred Metatarsal NA NA 371 ?Pachyspondylus Referred Metatarsal NA NA 372 ?Pachyspondylus Referred Metatarsal NA NA 373 ?Pachyspondylus Referred Metatarsal PV R3037 (partim) NA 374 ?Pachyspondylus Referred Metatarsal PV R3037 (partim) NA 375 ?Pachyspondylus Referred Phalanx PV R3037 (partim) NA 376 ?Pachyspondylus Referred Phalanx NA NA 377 ?Pachyspondylus Referred Phalanx NA NA 378 ?Pachyspondylus Referred Phalanx PV R3037 (partim) NA 379 ?Pachyspondylus Referred Phalanx PV R3037 (partim) NA 380 ?Pachyspondylus Referred Phalanx PV R1312a SAM-PKC-976 381 ?Pachyspondylus Referred Phalanx NA NA 382 ?Pachyspondylus Referred Ungual phalanx PV R3037 (partim) NA 383 ?Pachyspondylus Referred Ungual phalanx PV R3037 (partim) NA 384 ?Pachyspondylus Referred Ungual phalanx NA NA 385 ?Leptospondylus Referred Ungual phalanx PV R3026 NA 386 No name given NA ?Dentaries NA NA these decisions were based on size, proportions and preservational features. Indeed, as noted by Sues et al. (2004, p. 240), many of his referrals seem to have been made ‘somewhat arbitrarily’, especially as most of these additional specimens have no anatomical overlap with those in the aforementioned syntype series. Moreover, it is unknown if any of the specimens were originally associ- ated or articulated, whether they were all found in close proximity, or if they were recovered from the same stratum. Owen (1854) suggested that some of the individ- ually catalogued fragments might originally have been parts of the same bone (for example, he proposed that Nos. 354–356 were parts of the same humerus) but he offered no further comments on the number of individu- als represented. Following Owen’s establishment of these taxa, they were mentioned only rarely during the remainder of the nineteenth century, with brief, scattered references appearing in the literature (see below). Given the intense interest in other extinct reptiles at this time, it is tempting to speculate that this lack of engagement reflects Owen’s (1854) rather perfunctory descriptions and his decision to gloss over the affinities of these taxa (he later referred Massospondylus to Crocodylia but provided no reasons for doing so; e.g. Owen 1860a, p. 271; Owen, 1860b, p. 164). The fragmentary nature of the material might have con- tributed to this ambivalence but, given the large number of similarly poorly preserved specimens of other taxa that were mentioned frequently in the early palaeontological literature, this seems unlikely. Huxley (1867) mentioned Massospondylus, Pachyspondylus and Leptospondylus in his first paper on South African dinosaurs, noting in passing that they differed from his new taxon ‘Euskelesaurus’ [sic] (now considered a nomen dubium; Yates 2003a; Yates & Kitching 2003), without offering further comment on their relationships, while Jones (in Tate 1867) included them in a faunal list alongside other Stormberg taxa. Owen (1880) referred to all three taxa in comparisons with dicynodont material, and this seems to have been his last publication to mention them. Lydekker (1888) re-examined the Orpen Collection in order to compare these specimens with new fossil reptile material from the ‘Maleri Beds’ of central India. He appears to have been the first to recognize the dinosaur- ian nature of Massospondylus and Pachyspondylus and suggested that these taxa were probably theropods and members of Anchisauridae (N.B., many early discoveries of non-sauropod sauropodomorph taxa were originally regarded as members of various theropod sub-groups until the mid-twentieth century). This classification was accepted by later authors (e.g. Zittel 1887–90; Marsh 1889a, b). Lydekker (1888) figured two of the specimens (No. 336, a cervical vertebra; No. 380, a pedal phalanx) and opined that Massospondylus and Pachyspondylus might be synonymous, giving priority to the former, although his discussion omitted Leptospondylus. Finally, Lydekker suggested that Owen’s (1854) description was taxonomi- cally invalid and proposed instead that his own publica- tion should: 1) be regarded as the official source of the binomen; and 2) that the two specimens he figured (Lydekker 1888, fig. 3) be designated as the types of Massospondylus carinatus. However, Owen’s (1854) publi- cation and descriptions are clearly valid under the rules of the International Code of Zoological Nomenclature (International Commission on Zoological Nomencla- ture 1999) so Lydekker’s proposals are invalid (and were ignored by the majority of later workers). Subsequently, the same figures of the vertebra (No. 336) were repub- lished (Lydekker in Nicholson & Lydekker 1889, fig. 1067; Lydekker 1890a, fig. 58) alongside affirmations of the anchisaurid affinities of Massospondylus and the type status of the vertebra (No. 336: with the phalanx, No. 380, being relegated to referred specimen status). The saurischian affinities of Massospondylus were elaborated upon by Seeley (1892) who compared it further to ‘Euskelesaurus’ [sic] (Seeley 1894a,b). An isolated vertebra (IM G.281/1-u) and tooth (NHMUK PV R4190) from the ‘Maleri Beds’ and ‘Tikli Beds’, respectively, were assigned to Massospondylus around this time, leading to the erection of two new species – M. hislopi Lydekker, 1890b and M. rawesi Lydekker, 1890b – which extended the distribu- tion of the genus to India (see also Lydekker in Nicholson & Lydekker 1889, fig. 957). Seeley (1895a) was the first to provide a detailed, illus- trated account of the Orpen Collection, as well as a critical evaluation of Owen’s earlier conclusions. Following his examination of the material, Seeley (1895a) concluded that: 1) at least three individuals were represented; 2) that most of the individual bones belonged to a single taxon – Massospondylus carinatus; and 3) that Massospondylus was a ‘megalosaurian saurischian’ (ibid., p. 103). Seeley (1895a) corrected several of Owen’s anatomical misidentifications and suggested that the names Pachyspondylus and Lepto- spondylus should be ‘… held for the present in abeyance …’ (ibid.: pp. 103–104), as he presumably regarded them as indistinguishable from Massospondylus. Indeed, Seeley referred most of the Orpen Collection to the latter taxon, simplifying Owen’s (1854) complex scheme of syntypes and referred specimens. Seeley (1895a) did not designate a lectotype specimen or comment on Owen’s proposed taxonomy but regarded the elements that he described as parts of an extended syntype series for M. carinatus. Seeley’s revision of M. carinatus was published alongside an account of several other bones collected from the ‘Stormberg beds’ of ‘… the Telle River, north of the Witte Bergen in the Mattisi country …’ by Alfred Brown, which included several vertebrae, a pair of femora and phalan- ges. These specimens (NHMUK PV R3302) were used as the basis for a new species, Massospondylus browni Seeley, 1895b. (NB, although this binomen was first used in Seeley [1892] it was published without an accompanying descrip- tion, illustration or diagnosis, so was a nomen nudum prior to 1895). Nopcsa (1901) listed M. carinatus and M. browni in his synopsis of known dinosaur species and followed Lydekker and Seeley in regarding them as ‘anchisaurid megalosaurs’. Huene (1906) was the next to tackle the relationships of Massospondylus and provided many new figures of the Orpen specimens (based on casts held in the University of Tübingen), gave additional descriptions and corrected 102 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 several of Seeley’s anatomical mistakes. Although he did not discuss the taxonomy of Owen’s Stormberg species, Huene clearly considered them as synonyms, as he listed all of the specimens as M. carinatus. Following compari- sons with other early dinosaurs, Huene (1906) grouped Massospondylus, Thecodontosaurus and Anchisaurus into the Thecodontosauridae, which he considered to be a theropod subgroup. He also commented on the status of M. browni (referring the material to two species of Thecodontosaurus) and suggested that neither M. hislopi nor M. rawesi were dinosaurs. Consequently, Huene (1906) restricted the genus to the original Orpen Collec- tion material. Broom (1911) suggested that that Orpen Collection represented two individuals, one of which was M. carinatus and the other potentially referrable to his new species M. harriesi (see below). He also agreed with previous authors that Pachyspondylus and Leptospondylus were likely invalid and in referring M. browni to Theco- dontosaurus (Broom 1911). The next two decades witnessed the discovery of many other dinosaur specimens from the Stormberg Group of South Africa, leading to the naming of two new species of Massospondylus. The first of these, M. harriesi Broom, 1911, was based on material collected from the upper Elliot Formation of Foutanie Farm, near Fouriesburg in the Free State province (cadastral unit 331), whose type material included incomplete fore- and hind limb elements (SAM-PK-3394) (Broom 1911; Kitching & Raath 1984). An isolated pes (Huene 1911) and several partial skeletons, one with skull material (SAM-PK-5135), were referred to this species (Haughton 1924). A partial postcranial skele- ton originally referred to M. browni by Van Hoepen (1920) was also included in M. harriesi (Haughton 1924). The second new species, M. schwarzi Haughton, 1924, was based on a partial hind limb (SAM-PK-5134) from the Elliot Formation of Makomoreng, Eastern Cape province (Haughton 1924; Kitching & Raath 1984). Haughton (1924) provided a brief review of M. carinatus, based on Huene’s (1906) account and the plaster casts held in the Iziko South African Museum, but despite having exten- sive new collections of dinosaur material at his disposal he did not refer any new specimens to the taxon. During this period, Huene (1914) erected the new family Massospon- dylidae for M. carinatus, M. harriesi and another South African taxon Aetonyx palustris Broom, 1911. Haughton (1924) added a further Stormberg taxon, Dromicosaurus gracilis Van Hoepen, 1920, to this group. In his benchmark review of saurischian dinosaur taxonomy and systemat- ics, Huene (1932) retained M. carinatus, M. schwarzi and M. harriesi as valid species, but abandoned Massospon- dylidae and placed them within Thecodontosauridae, along with Thecodontosaurus (= Massospondylus) browni and several other taxa. Otherwise, Massospondylus was not much discussed during the mid-twentieth century, although it was consistently included in taxonomic compilations (e.g. Lull 1910; Williston 1925; Romer 1956, 1966; Steel 1970; White 1973). Perhaps the most famous event in the history of Massospondylus carinatus is the destruction of the original syntype series, although only a few details have been provided in the palaeontological literature (Attridge 1963; Steel 1970; Sues et al. 2004; Yates & Barrett 2010). As noted previously, the Orpen Collection, and many other palaeontological specimens, were housed in the Hunte- rian Museum of the Royal College of Surgeons (Owen 1854), where six conjoined galleries and other associated facilities functioned as multipurpose teaching, research and collections spaces (Keith 1910; Allen 1974; Figs 2, 3A). The fossil reptile specimens formed part of the compara- tive osteological collection, most of which was stored in Room V of the museum, with the other rooms being dedi- cated to various aspects of human anatomy (Keith 1910; Cave 1941). Most of the fossil reptile specimens were held in a series of wall-mounted cases around the north and west sides of Room V, close to examples of modern reptile skeletons; other significant specimens of recent and fossil mammals, birds and fish were stored nearby (Keith 1910; Cave 1941; Fig. 2). Following the onset of World War II, precautions were taken to protect the collections in the event of bombing raids, including the removal of some collections to remote locations and strengthening of the museum basements for on-site storage underground (Cave 1941; Wakeley 1965). However, many specimens could not be moved due to either a lack of sufficient storage space or their large size (Cave 1941). Between October 1940 and April 1940 shell fragments from German bombing raids damaged the roof of the museum but did not lead to any losses from the collection (Cave 1941). On 11 October 1940 a bomb exploded on Portugal Street, immediately adjacent to the museum and directly outside Room V, shattering many of the windows and roofs around the building but causing only minor damage to the osteological specimens, although details of the speci- mens affected were not recorded (Cave 1941). Just a few days later, on 16 October 1940, a land mine exploded in nearby Lincolns Inn Fields, which caused considerable structural damage to the building and affected specimens in the human anatomical collections (Cave 1941). Other bombing incidents in October 1940 and February 1941 caused only minor building damage or temporary closures of the building (Cave 1941). However, on the night of May 10th/11th 1941, the museum was struck directly by a high-explosive bomb that hit Room V and adjacent areas, destroying this part of the building and most of its contents, while incendiary bombs gutted other areas (Cave 1941; Wakeley 1965; Fig. 3B). Falling debris broke through into the basement storage rooms, leading to the destruction of the osteological collections, which were stored in this area, a problem exacerbated by heavy rain- fall, which flooded the site (Cave 1941; Wakeley 1965). The material affected included all the fossil reptile material catalogued by Owen (1854) as well as many other scientifi- cally important specimens that had been integral to the development of medicine, anatomy and zoology. As noted by Cave (1941, p. 8): ‘The magnitude of this disaster to the College, to the nation and to the medical and scien- tific worlds needs no emphasis’. Among these casualties were Massospondylus, Pachyspondylus and Leptospondylus. Of the 550 reptile specimens included in the compara- tive collection, only 23 survived (Cave 1941). Fortunately, ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 103 casts of selected specimens from the Orpen Collection remained in the Natural History Museum (London, U.K.), Iziko South African Museum (Cape Town, South Africa) and the Institut und Museum für Geologie und Paläon- tologie, Universität Tübingen (Tübingen, Germany). These casts, and the line-drawings provided by Seeley (1895a), Huene (1906) and others, are the only surviving records of this material (see below). Subsequent to the work of Broom, Haughton and Van Hoepen, no significant discoveries of Massospondylus material were reported from South Africa for many years, although it is likely that material continued to accumulate in Cape Town, Pretoria and Bloemfontein during the mid-twentieth century. It is tempting to speculate that work on South African dinosaurs was overshadowed by work on the abundant and important synapsid material from the Karoo Basin, which formed the primary research focus for many palaeontologists in the region at this time. However, material continued to be referred to Massospon- dylus from other southern African countries with reports from the Karoo-aged sediments of the mid-Zambezi, Mana Pools and Tuli basins of Zimbabwe (e.g. Attridge 1963; Bond 1965, 1973; Raath et al. 1970; Raath 1974; Cooper 1981; Munyikwa 1997) and from the Stormberg Group of Lesotho (Ellenberger et al. 1964; Attridge & Charig 1967; Anonymous 1969; Kermack 1974). Many of the Zimbabwean specimens, which consist almost exclu- sively of postcranial remains, were described by Cooper (1981) in a detailed, well-illustrated monograph. By contrast, the material collected from Lesotho, which is housed in collections in Cape Town, Paris and London, has not yet been described in detail, with the exception of a complete but crushed skull (SAM-PK-K1314), which has been figured on many occasions (e.g. Attridge & Charig 1967; Attridge et al. 1985; Norman 1985) but only described in part (Barrett & Yates 2006). Galton & Cluver (1976) provided a taxonomic review of many southern African sauropodomorph species in which they proposed a wide-ranging series of synony- mies and shuffled genera between two of the family-level taxa that were in use at this time: Anchisauridae (‘narrow- footed prosauropods’) and Plateosauridae (‘broad-footed prosauropods’). They referred Massospondylus to Plateo- sauridae, in contrast to previous authors, but followed Huene, Haughton and others in restricting Massospon- dylus carinatus to its syntype series and in regarding Pachyspondylus orpeni and Leptospondylus capensis as its junior subjective synonyms. Similarly, Galton & Cluver (1976) regarded M. harriesi as a valid taxon and referred many other taxa and specimens to its hypodigm, includ- ing Aetonyx palustris, Gryponyx africanus, Dromicosaurus gracili, Thecodontosaurus dubius and a referred specimen of M. browni. However, no rationale was given for these decisions other than noting superficial resemblances in foot anatomy. By contrast, the type specimen of M. browni was regarded as a nomen dubium (Anchisauridae incertae sedis) and was excluded from Massospondylus. Based on these taxonomic changes, Galton & Cluver (1976) noted that Massospondylus was the most widely distributed genus of sauropodomorph in southern Africa. Alongside his description of the Zimbabwean material, Cooper (1981) performed a comprehensive review of upper Stormberg ‘prosauropod’ species and agreed with Galton & Cluver (1976) that the taxonomy was oversplit. Cooper (1981) compared the Zimbabwean sample with the type specimens of South African taxa and noted that none of them differed substantially. He concluded that the minor differences noted by earlier authors were not taxo- nomically informative and could be regarded as either taphonomic, geographic, individual or ontogenetic varia- tion. Moreover, following the principle of date priority he regarded Massospondylus carinatus as the valid binomen for all this material, leading him to refer a vast number of specimens to this taxon for the first time in its long history. Many species became junior subjective synonyms of Massospondylus, including not only those taxa listed by Galton & Cluver (1976), but also M. browni, M. harriesi, Gryponyx taylori, Gr. transvalensis, Aristosaurus erectus, Thecodontosaurus minor and Gyposaurus capensis. In addi- tion, Cooper (1981) suggested that Lufengosaurus and Yunnanosaurus should also be referred to Massospondylus. This resulted in a trans-continental distribution covering South Africa, Lesotho, Zimbabwe and China. Cooper (1981) also reinstated Massospondylus to Anchisauridae and provided the first in-depth palaeobiological analysis of the taxon, envisaging it as an active, bipedal, endother- mic carnivore, although most recent authors regard it as either herbivorous or facultatively omnivorous (Galton 1985; Barrett 2000). Cooper’s (1981) detailed descriptions superceded the earlier descriptions of Seeley (1895a) and Huene (1906), and his monograph became a standard reference on Massospondylus postcranial anatomy for anatomical and phylogenetic studies. However, this work was limited by several factors: 1) the lack of cranial material that could be referred confidently to the genus; 2) the fact that few associated skeletons were available to him; and 3) the assumption that the Zimbabwean material was conspeci- fic with the South African taxon named by Owen (1854). Cooper (1981) did not provide a new diagnosis of the taxon nor use unique features to refer specimens to the hypodigm, relying instead on overall similarity to build his taxonomic framework. This led to a situation in which it was generally assumed that all lightly-built sauropodo- morph remains from the upper Stormberg Group were referrable to Massospondylus (e.g. Kitching & Raath 1984; Gow 1990; Gow et al. 1990) and there were no further assessments of upper Elliot sauropodomorph taxonomy until the 2000s. As a result, hundreds of sauropodomorph specimens collected from the upper Elliot and Clarens formations between the 1970s–2000s were simply assumed to be Massospondylus, labelled as such in collections in Cape Town, Johannesburg, Pretoria, Grahamstown and Bloemfontein, and largely ignored: many of these speci- mens require re-examination. However, a strong case was never made for referring any of these specimens to M. carinatus based on comparisons with the syntypes: rather, a set of nested inferences was built by Cooper (1981) and others that linked the priority of this name (over others) with stratigraphic and geographic overlap 104 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 and general anatomical similarity. It is entirely possible that the set of specimens now considered referrable to this taxon are in no way related to the specimens of the syntype series, but that habitual usage of the name shifted understanding of the taxon’s hypodigm. As a result, the application of the name M. carinatus to all of this more recently recovered material might have been more seren- dipitous than taxonomically justified. Systematic collecting efforts in the 1970s–1980s, led by Alfred Crompton (SAM), James Kitching (BPI), Christo- pher Gow (SAM then BPI) and Roger Smith (SAM), resulted in the recovery of many new sauropodomorph specimens from the upper Elliot Formation of the Free State and Eastern Cape provinces, with crucial discoveries of skeletons including complete skulls. Preliminary descriptions of these specimens referred all of them to M. carinatus (Cooper 1981; Kitching & Raath 1984; Gow 1990; Gow et al. 1990), a conclusion followed by all subse- quent authors (e.g. Sues et al. 2004). A clutch of eggs was also referred to the genus (Kitching 1979) although its significance was not fully appreciated until later (Grine & Kitching 1987; Zelenitsky & Modesto 2002; Reisz et al. 2005, 2010, 2012; Stein et al. 2019; Chapelle et al. 2020a). Although these new specimens offered the opportunity for a thorough re-assessment of the genus, Gow (1990; Gow et al. 1990) did not comment on the diagnosis of the taxon but simply accepted the prior conclusions of Cooper (1981) and others. Sues et al. (2004) were the first authors to provide a character-based diagnosis of M. carinatus, based on cranial anatomy, and these authors also highlighted the need to stabilize its taxonomy further, as well as providing the first thorough description of four well- preserved skulls. Building on this work, Barrett & Yates (2006) provided more details of another referred skull and proposed BP/1/4934 as the neotype of M. carinatus (Yates & Barrett 2010). The latter specimen includes a skull and almost complete postcranial skeleton, enabling it to serve as a more useful exemplar than the unavailable, unassociat- ed and fragmentary syntypes described by Owen (1854) thereby improving its taxonomic stability. During this period, a referred partial skull was shown to differ from other examples known and was established as the type specimen of a second species, M. kaalae (Barrett 2009). Subsequently, more detailed studies of the neotype post- cranium (Barrett et al. 2019) and a referred skull (Chapelle & Choiniere 2018) permitted the development of stricter, character-based diagnoses. In parallel with these South African discoveries, new material from other countries was also referred to Masso- spondylus. These included specimens from the Lower Jurassic Kayenta Formation of the U.S.A. (Attridge et al. 1985) and Cañón del Colorado Formation of Argentina (Martínez 1999), and from the Late Triassic Upper Maleri Formation of India (Kutty et al. 1987), all of which suggested a trans-continental distribution and a possible use for the genus in global stratigraphic correlations. However, re-assessment of the Indian specimen identi- fied it as guibasaurid sauropodomorph (Novas et al. 2011), and the Argentinean material was later recognized as a distinct (but closely related) taxon, Adeopapposaurus mognai (Martínez 2009). For several years the Kayenta Formation specimens were used as exemplars for the genus in biomechanical and taxonomic studies (e.g. Crompton & Attridge 1986; Galton 1990), but additional work has shown that they also represent a different taxon, Sarahsaurus aurifrontalis (Rowe et al. 2011; Marsh & Rowe 2018). Cooper’s (1981) suggestion that Lufengosaurus and Yunnanosaurus were junior subjective synonyms of Massospondylus has not been supported by any other study (e.g. Galton 1990; Galton & Upchurch 2004) and the historical referrals of the Indian specimens ‘M. rawesi’ and ‘M. hislopi’ (see above) have also been rejected (Huene 1906; Galton & Upchurch 2004; Carrano et al. 2010). Conse- quently, the distribution of Massospondylus sensu stricto is considered restricted to southern Africa. More rigorous character-based taxonomic approaches, and a clearer understanding of the composition of the Massospondylus hypodigm, have also allowed the descrip- tion of two new sauropodomorph taxa from the upper Elliot Formation of South Africa and Lesotho based on material formerly referred to Massospondylus – Ignavu- saurus rachelis (Knoll 2010) and Ngwevu intloko (Chapelle et al. 2019). In addition to basic anatomical and taxonomic work, the discovery of this new, more complete material stimulated phylogenetic work and Massospondylus has been included in numerous cladistic studies on sauropodomorph sys- tematics (e.g. Galton 1990; Sereno 1999; Benton et al. 2000; Yates 2003b, 2004, 2007; Yates & Kitching 2003; Galton & Upchurch 2004; Kutty et al. 2007; Smith & Pol 2007; Upchurch et al. 2007; Martínez 2009; Knoll 2010; Sertich & Loewen 2010; Apaldetti et al. 2011, 2013; Novas et al. 2011; Pol et al. 2011, 2021; Yates et al. 2011; Otero & Pol 2013; McPhee et al. 2015, 2018; Peyre de Fabrègues & Allain 2016, 2020; Wang et al. 2017; Chapelle & Choiniere 2018; Müller et al. 2018; Chapelle et al. 2019; Rauhut et al. 2020; Regalado Fernández & Werneburg 2022). Its inferred relationships have varied through time, dependent on the size of the dataset, the taxa included in the analysis and the source material for scoring anatomical characters. For example, many earlier studies relied on Cooper’s (1981) descrip- tions for postcranial characters, but the lack of associated material and doubts on its conspecificity render these scores problematic. Similarly, the holotype skull of Ngwevu intloko would have been conflated with those of Massospondylus in many analyses. Early non-numerical analyses identified Massospondylus as an early-diverging member of a monophyletic Prosauropoda and led to the resurrection of Massospondylidae as a distinct monotypic family (e.g. Galton 1990). Subsequently, numerical analy- ses have generally recovered Massopondylus in a masso- spondylid clade whose composition varies but that most frequently includes Adeopapposaurus, Coloradisaurus, Leye- saurus and Lufengosaurus as close relatives (e.g. Smith & Pol 2007; Upchurch et al. 2007; Yates 2007; Apaldetti et al. 2011, 2013; Otero & Pol 2013; Wang et al. 2017; Peyre de Fabrègues & Allain 2020; Rauhut et al. 2020; Pol et al. 2021). In spite of this work, much remains to be done. It remains far from clear if all the specimens currently referred to Masso- spondylus genuinely pertain to the taxon (as currently ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 105 defined) or if other taxa remain hidden within its hypo- digm. Moreover, the exact dating of these specimens is often uncertain and needs to be re-assessed in light of new constraints on the chrono- and lithostratigraphy of the upper Stormberg Group (e.g. Bordy et al. 2020; Viglietti et al. 2020). This is important as it has consequences for the use of this material in biostratigraphical correlations (both within and between basins) as well as its impact on palaeo- biological studies (as the large sample size currently avail- able might be a chimera). For convenience, key events in the history of Massospondylus are summarized in Fig. 4. THE ORPEN COLLECTION: DESCRIPTIONS AND COMPARISONS Here, we present descriptions of the Orpen Collection based on the surviving casts in the collections of NHMUK and SAM (we did not have the opportunity to examine those at GPIT firsthand). Sadly, these represent only part of the original collection, and, in the absence of published figures, the other specimens originally listed by Owen (1854) are now completely lost to science. None of these institutions appears to hold much archival information on these casts, and it is not clear how or why they were procured. In the case of the NHMUK set, the registers record the fact that two of the casts (NHMUK PV R1312, R1312a) were made in the museum sometime in 1888 whereas the others were purchased in April 1902 from C.B. Firth. The casts housed in SAM were obtained around 1906 from R.F. Damon (Z. Skosan, pers. comm.), the owner of a Dorset-based fossil dealership and natural history cast-making business. Those in GPIT were acquired by Friedrich von Huene (I. Werneburg, pers. comm.). Although the casts represent the original syntype series of these taxa, these ‘plastotypes’ cannot act as name-bearing type specimens as artificial physical replicas are ineligible for nomenclatural acts (see International Code on Zoolog- ical Nomenclature 1999, Article 72.5; see also Yates & Barrett 2010). A summary of the current identifications of each of these specimens is provided in Table 2. Axial skeleton Cervical vertebrae Owen (1854, pp. 97–98) regarded Nos. 331–335 as caudal vertebrae, but Seeley (1895) correctly noted that they are from the cervical series. Seeley (1895, pp. 104–106, fig. 1) provided an accurate description of No. 331 (updated herein), which he retained within Massospondylus carinatus. By contrast, Seeley (1885) excluded Nos. 332–335 from M. carinatus and regarded these vertebrae as referable to a different taxon, although no evidence was provided to support this proposal (nor were they referred to any other named taxon). Huene (1906, p. 133) dismissed Seeley’s suggestion and reinstated the vertebrae in the type series, provided further descriptions and figured Nos. 331–335 (ibid., figs 44, 45, pl. XIII [XX], figs 6–8). Measurements of the cervical vertebrae are provided in Table 3. The centrum of No. 331 (NHMUK PV R 3027; SAM-PKC- 958; Fig. 5A–F) is complete but is slightly deformed and its articular surfaces have been damaged. The anterior articular surface is sub-circular in outline and is gently concave (Fig. 5A); the posterior articular surface is also shallowly concave but has a more shield-shaped outline that is taller than wide, with a straight dorsal margin and lateral margins that descend in a straight line for a short distance before merging ventrally (Fig. 5C). In lateral view, the ventral margin of the centrum is strongly concave and the lateral surfaces are anteroposteriorly concave and convex dorsoventrally (Fig. 5B, D). A distinct break-in-slope, marked by a thin ridge divides the lateral surfaces into larger dorsal and smaller ventral portions. There are no pneumatic openings. A low, sub-triangular parapophysis is present on the anterior-most part of the lateral centrum surface, at approximately mid-centrum height. The lateral surfaces merge ventrally to form a sharp midline keel that extends for the full length of the centrum (Fig. 5F). The centrum has a total length/anterior articular surface height ratio of 2.43. A robust, posterior centrodiapophyseal lamina (PCDL) is present, which arises from the posterodorsal corner of the centrum and merges with the posterior margin of the transverse process to create a rounded ledge that over- hangs the lateral surface of the centrum (Fig. 5B, D). This ledge is almost horizontal, rising only slightly dorsally as it extends anteriorly. There is some indication of a low, ridge-like paradiapophyseal lamina (PPDL). The PPDL and PCDL form the boundaries of an extensive, shallow, triangular centrodiapophyseal fossa (CDF). The neural arch extends for the full length of the centrum (Fig. 5B, D, E). It is almost complete but lacks the right prezygapophysis, left transverse process and left postzygapophysis. In anterior view, the neural canal opening is sub-circular (Fig. 5A), but posteriorly it is sub-elliptical, with the long axis of this ellipse oriented transversely (Fig. 5C). The prezygapophysis extends anterior to the anterior articular surface for a short distance. Its articular surface is ovate, flat to very gently convex and is bevelled to face dorsomedially at an angle of approximately 45° to the horizontal (Fig. 5B, D). The base of the prezygapophysis has a narrow triangular cross- section. An interprezygapophyseal lamina (TPRL) links the two prezygapophyses and forms the floor of a shallow concavity that is confluent with an incipient prespinal fossa posteriorly (Fig. 5E). Low, ridge-like spinoprezyg- apophyseal laminae (SPRL) extend from the dorsal margins of the prezygapophyses to the anterior margin of the neural spine (Fig. 5E). A low lamina (epipophyseal- prezygapophyseal lamina: EPRL) extends posterodor- sally from the prezygapophysis to meet the anterior margin of the postzygapophysis, dividing the lateral surface of the neural arch into distinct dorsal and ventral portions (Fig. 5B, D). A short, wing-like transverse process is present, which is situated just anterior to the midpoint of the vertebra. It has a flattened, elliptical cross-section and a short, rounded triangular outline in dorsal view (Fig. 5E). The postzygapophyses extend for a short distance beyond the centrum posteriorly and have flat, elliptical 106 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 107 Figure 4. Chronological timeline summarising the taxonomic history of Massospondylus spp. 108 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 Table 2. Current identifications of the specimens within the Orpen Collection (see text for further details). As some of the specimens are not represented by casts, Huene's (1906) identifications of these specimens are retained for now (indicated with '*'). articular surfaces that face ventrolaterally, forming an angle of approximately 45° with the horizontal (Fig. 5B, D). A low, ridge-like epipophysis is present on the dorsal surface of the postzygapophysis and extends for approxi- mately half to two-thirds of its length but does not reach the tip of the process (Fig. 5D). The base of the postzyg- apophysis has a flattened cross-section. Very short spino- postzygapophyseal laminae (SPOL) are present. A dorso- ventrally elongate, slit-like postspinal fossa is situated at the junction of the postzygapophyses (Fig. 5C). The neural spine has a sub-rectangular outline in lateral view, with a straight, dorsally inclined anterior margin, a gently convex and horizontally extending dorsal margin and a short, posteroventrally sloping posterior margin (Fig. 5B, D). It is laterally compressed and lacks a spine table. Seeley (1895) did not describe or figure Nos. 332–335 as he regarded them as uninformative and potentially referrable to a different species. No. 332 (NHMUK PV R3028 partim; SAM-PKC-959; Fig. 5H) consists of a broken neural arch lacking the tips of the prezygapophyses, the transverse processes and the summit of the neural spine. It is dorsoventrally lower than the neural arch of No. 331 and more elongate, but its poor preservation prevents description of many features. It differs from No. 331 in lacking epipophyses, having a proportionately longer neural spine and in possessing a more rounded, promi- nent EPRL. No. 333 (NHMUK PV R3028 partim; SAM-PKC-960; Fig. 5G) consists of the posterior part of a vertebra, including a partial neural arch. The preserved part of the neural arch is essentially identical to that of No. 332, with the exception that a ridge-like epipophysis is present. Its centrum is low and elongate, with a rounded, horizontally inclined PCDL extending anteriorly from its posterodorsolateral corner. In ventral view, the centrum is strongly constricted transversely and its ventral surface is smoothly rounded transversely, lacking a ventral midline keel. In posterior view, the articular surface is slightly taller than wide and is deeply concave. No. 334 (NHMUK PV R3028 partim; SAM-PKC-961; Fig. 5I–K) is a fragment of a spool-like centrum, consisting of the ?posterior part only. Its preserved articular surface is shield-shaped in outline and shallowly concave (Fig. 5I). A prominent midline keel is present on its ventral surface (Fig. 5K; ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 109 Table 3. Vertebral measurements in millimetres. Asterisks (*) indicate incomplete bones; dashes ('–') indicate inapplicable or substantially incomplete measurements. Cervical vertebrae NHMUK catalogue number: PV R3027 PV R3028 PV R3028 PV R3028 PV R3028 (partim) (partim) (partim) (partim) Owen (1854) catalogue number: 331 332 333 334 335 Centrum, length 102 – – – 88 Centrum, anterior height 42 – – – 46 Centrum, anterior width 44 – – – 44 Centrum, posterior height 46 – 39 ?54 42 Centrum, posterior width 46 – 37 ?44 38 Neural arch, total height (including spine) 58 – – – – Neural spine, anteroposterior length 47 124* – – – Dorsal vertebra NHMUK catalogue number: PV R1312 Owen (1854) catalogue number: 336 Centrum, length 70 Centrum, anterior height 50 Centrum, anterior width 41 Centrum, posterior height 53 Centrum, posterior width 44 Sacral vertebra NHMUK catalogue number: PV R3035 (partim) Owen (1854) catalogue number: 346 Centrum, length 50* Centrum, posterior height 29 Centrum, posterior width 44 Caudal vertebrae NHMUK catalogue number: PV R3035 (partim) PV R3035 (partim) PV R3035 (partim) PV R3035 (partim) PV R3025 Owen (1854) catalogue number: 338 342 343 345 347 Centrum, length 68 43 50* 56 48 Centrum, anterior height 59 50 – 54 59 Centrum, anterior width 49* 38 – 42 52 Centrum, posterior height (including chevron facet) 63 50 30 53 57 Centrum, posterior width 61 38 31 38 53 110 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 Figure 5. Casts of cervical vertebrae from the Orpen Collection. These represent the original syntype series of Massospondylus carinatus. A–F, NHMUK PV R3027, catalogue number 331, mid-cervical vertebra in anterior (A), left lateral (B), posterior (C), right lateral (D), dorsal (E) and ventral (F) views. G, NHMUK PV R3028, catalogue number 333, mid-cervical vertebra in left lateral view. H, NHMUK PV R3028, catalogue number 332, partial mid-cervical neural arch in dorsal view. I–K, NHMUK PV R3028, catalogue number 334, mid-cervical centrum in anterior (I), right lateral (J) and ventral (K) views. incorrectly identified as a dorsal vertebra by Huene [1906]). Finally, No. 335 (NHMUK PV R3028 partim; SAM-PKC-962; not figured) consists of a small, badly preserved centrum. Both the anterior and posterior artic- ular surfaces are slightly taller than they are wide, are shallowly concave and have shield-shaped outlines. The lateral surfaces of the centra are longitudinally concave and dorsoventrally convex and merge ventrally, meeting to form a sharp midline keel that extends for the full length of the vertebra. Mound-like swellings on the lateral surfaces that are situated a short distance posterior to the anterior margin might represent parapophyses. The centrum has a length/anterior height ratio of 1.91. None of the cervical vertebrae display any of the features currently regarded as autapomorphic for M. carinatus, namely centrum length/height ratios >7.0 (for anterior cervical vertebrae) and neural spines of anterior cervi- cal vertebrae bearing hook-like expansions anteriorly (Barrett et al. 2019) (although the latter feature might be ontogenetically variable: K.E.J.C., pers. obs.). Neverthe- less, the neural arch of No. 332 does bear a strong resem- blance to those of M. carinatus (Barrett et al. 2019) and Ngwevu intloko (Chapelle et al. 2020), suggesting that it might pertain to an indeterminate massospondylid. However, none of the other cervicals bear features synapomorphic of Massospondylidae, such as possessing centrum length/height ratios >3.0 (e.g. Yates 2007; Apaldetti et al. 2013). No. 331 bears several features diag- nostic of Sauropodomorpha and several slightly more exclusive clades therein (following Yates, 2007; Apaldetti et al., 2013), some of which are also present in Nos. 333– 335, notably: epipophyses forming tall ridges (Sauro- podomorpha); anterior cervical vertebrae with centrum length/height ratios >2.5 (Efraasia+more derived sauropodomorphs); ventral midline keel present on anterior cervicals (Sarahsaurus+more derived sauro- podomorphs); and a concave anterior articular facet on centrum (excluding it from Sauropoda). This combination of features suggests that Nos. 331 and 333–335 repre- sent indeterminate massopodan sauropodomorphs, but further identification is not possible with the available material. Dorsal vertebra A single dorsal vertebral centrum is preserved among the casts (No. 336; NHMUK PV R1312; SAM-PKC-963; Fig. 6A), which was retained in Massospondylus carinatus by Seeley (1895, p. 106) and was figured by Lydekker (1888, fig. 3; 1890a, fig. 58). Lydekker (1890a) deemed this specimen the type as he suggested that Owen’s original description was inadequate; however, Owen (1854) clearly established the name on the basis a syntype series without designating a holotype (see above). Huene (1906, pl. XIII [XX], fig. 9) figured this element and noted four other dorsal vertebrae in the collections of the Hunterian Museum (Nos. 337, 348 [a syntype of Leptospondylus capensis], and two unnumbered specimens; ibid. fig. 46), but casts of these do not survive (see Table 1). Huene’s (1906) descriptions and figures are not sufficient to comment further on the identification of these four specimens. Measurements of the dorsal vertebrae are provided in Table 3. In lateral view, the centrum has lateral surfaces that are anteroposteriorly concave and dorsoventrally convex (Fig. 6A). It has a concave ventral margin, and no foramina or fossae are present. The lateral surfaces are separated from the ventral surface by distinct breaks of slope, but these do not form strong bounding ridges. The ventral surface is strongly concave anteroposteriorly but almost flat mediolaterally. It is smooth and bears neither a ventral keel nor a ventral groove. The anterior and posterior articular surfaces are both flat to very shallowly concave. Each is dorsoventrally taller than mediolaterally wide and has a sub-elliptical outline with its long axis trending dorsoventrally. The basal part of the neural arch is also preserved and its anterior part flares slightly laterally (in the region of the parapophyses) but few other details are preserved. The transverse cross-section through the base of the neural arch produces a neural canal with an elon- gate, hourglass-shaped outline in dorsal view. The neural canal excavates the dorsal surface of the centrum slightly. The centrum anterior height/length ratio of 1.4 is within the range seen in M. carinatus but excludes it from Ignavu- saurus (Barrett et al. 2019). The vertebra possesses no useful diagnostic features, although its size does suggest that it is from a dinosaur, so No. 336 is regarded as Dino- sauria indet. herein. Sacral vertebra No. 346 (NHMUK PV R3035 partim: Fig. 6B, C) is a sacral centrum missing its ?anterior articular surface. Although it is one of the syntypes of Pachyspondylus orpenii (see Owen 1854), it was referred to Massospondylus carinatus by Seeley (1895, p. 106, fig. 2) and Huene (1906, p. 133, pl. XIV [XXI], fig. 1). Measurements of the sacral vertebrae are provided in Table 3. In lateral view, the ventral margin of the centrum is only slightly concave so that it has a sub-rectangular outline (Fig. 6C). The lateral surfaces are dominated by large sub-triangular processes, situated centrally, which repre- sent the broken bases of the sacral ribs; the rest of the lateral surface is shallowly concave longitudinally. The lateral and ventral surfaces are separated by a gentle break-of-slope, rather than distinct ridges and the ventral surface is mediolaterally constricted with respect to the articular ends. There is neither a groove nor a keel on the ventral surface, which is gently convex mediolaterally and shallowly concave anteroposteriorly. The posterior articular surface has a kidney-shaped outline, with a dorsal margin that is gently invaginated and with straight to slightly convex lateral and ventral margins (Fig. 6B). The articular surface is shallowly concave and does not appear to be broken, so it either represents the posterior- most sacral or a sacral from an individual in which sacral fusion was incomplete. The specimen possesses no diagnostic features and is referred to Dinosauria indet., though based solely on its size and its general resemblance to other dinosaur sacral vertebrae. ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 111 112 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 Figure 6. Casts of dorsal, sacral and caudal vertebrae from the Orpen Collection. A, NHMUK PV R1312, catalogue number 336, dorsal vertebra centrum in left lateral view. B–C, NHMUK PV R3035 (partim), catalogue number 346, sacral vertebra in posterior (B) and left lateral (C) views. D, NHMUK PV R3025, catalogue number 347, proximal caudal vertebra in anterior view. E–F, NHMUK PV R3035 (partim), catalogue number 338, proximal caudal vertebra in anterior (E) and left lateral (F) views. G–H, NHMUK PV R3035 (partim), catalogue number 342, proximal caudal vertebra in anterior (G) and left lateral (H) views. I–J, NHMUK PV R3035 (partim), catalogue number 343, middle caudal vertebra in anterior (I) and left lateral (J) views. K–L, NHMUK PV R3035 (partim), catalogue number 345, middle caudal vertebra in anterior (K) and left lateral (L) views. Caudal vertebrae Seeley (1895, p. 107) transferred the caudal vertebrae that formed the syntype series of both Pachyspondylus orpenii and Leptospondylus capensis (see Owen 1854) to Massospondylus carinatus but regarded them as represent- ing the remains of more than one individual. Many of the caudal vertebrae listed by Owen (1854) do not appear to have been cast (Nos. 339–341, 344, 348) and as they were not figured are known only from Owen’s and Huene’s (1906) brief descriptions. Owen (1854) identified No. 345 as a dorsal centrum, but its large haemal arch facets iden- tify it as a caudal, whereas Huene (1906) regarded No. 348 as a dorsal (see above). Of the five available casts, three are proximal caudals (Nos. 338, 342, 347) while the other two are from the middle or distal part of the tail (Nos. 343, 345). Seeley (1895, fig. 3) figured only one caudal (No. 338); Huene (1906) figured Nos. 338, 343, 345, 347 (ibid., pl. XIV [XXI], figs 2–5) as well as four unnumbered vertebrae that are otherwise unknown and are presumably lost (ibid., fig. 47). Measurements of the caudal vertebrae are provided in Table 3. The proportions of the proximal caudals vary slightly, reflecting their different sizes (as noted by Seeley [1895]) and/or slightly different positions within the tail, although they are otherwise similar (Fig. 6D–L). No. 347 (NHMUK PV R3025; Fig. 6D) is the most complete, consisting of a centrum and partial neural arch; No. 338 (NHMUK PV R3035 partim; Fig. 6E, F) includes only a small section of the neural arch base; and No. 342 (NHMUK PV R3035 partim; Fig. 6G, H) is similarly preserved to the latter. No. 347 lacks large haemal arch facets and might represent caudal 1 or 2, though the positions in the caudal series of all the other preserved vertebrae are unknown. Nos. 342 and 347 have centra that are approximately square-shaped in lateral view (Fig. 6H), whereas that of No. 338 is more elongate and rectangular (Fig. 6F). All three proximal caudals have lateral surfaces that are anteroposteriorly concave and dorsoventrally convex and that generally lack foramina, ridges or other features. However, two short, stout vertebral laminae are present in No. 338: a sub-vertical anterior centrodiapophyseal lamina (ACDL) and obliquely-inclined PDCL. The lateral surfaces blend into the ventral surfaces around gentle breaks-in-slope, without forming distinct ridges. In ven- tral view, the centra have a slightly constricted hourglass- shaped outline and the ventral surface is longitudinally concave and mediolaterally convex. Nos. 342 and 347 lack any ventral keel or groove; No. 338 possesses a short midline buttress that is confluent with the anterior articu- lar surface, but that only extends posteriorly for a short distance before merging into the ventral surface. The articular surfaces of all three vertebrae are shallowly concave and have sub-elliptical outlines with their long axes oriented vertically (Fig. 6D, E, G). None of the verte- brae bear prominent anterior haemal arch facets, but Nos. 338 and 342 exhibit well-developed posterior facets, so the posterior margin of the centrum is extended much farther ventrally than its anterior margin in lateral view (Fig. 6F, H). In posterior view, the haemal arch facets have a crescentic outline. Neural canal openings are visible in No. 347 and are sub-circular anteriorly and elliptical posteriorly (with the long axis of this ellipse oriented transversely; Fig. 6D). The bases of the transverse processes are preserved in Nos. 342 and 347: these are dorsoventrally compressed, sheet-like and situated centrally on the neural arch. They appear to have extended horizontally, rather than dorsolaterally, although that of No. 347 seems to have been rotated so that its dorsal surface faced anterodorsally. All other neural arch processes are broken and offer no informa- tion. No. 343 (NHMUK PV R3035 partim; Fig. 6I, J) is a partially complete middle to distal caudal lacking its anterior end and most of the neural arch; No. 345 (NHMUK PV R3035 partim; Fig. 6K, L) is a middle caudal centrum. In most respects the centra are similar to those of the proximal caudals, differing only in their proportions, as the middle caudals are more elongate and are longer than they are tall. The lateral surfaces of the centra are saddle-shaped and blend gradually into the ventral surfaces (Fig. 6J, L). In ventral view, the centra are constricted and there is some indication of a shallow ventral midline groove in No. 345. The articular surfaces are shallowly concave (Fig. 6I, K): those of No. 345 are taller than wide and have a sub- elliptical outline, whereas the posterior articular surface of No. 343 is sub-circular (the anterior facet is not preserved). No. 345 possesses a small anterior haemal arch facet and much larger posterior facet, both of which have a crescen- tic outline. Haemal arch facets are absent in No. 343 suggesting that it is from a more distal part of the tail. However, although broken, transverse processes were still present in No. 343: these were dorsoventrally flattened and extended laterally. With one exception (No. 338), none of the caudal verte- brae possess features that are synapomorphic for any clade although their overall morphology and size is consistent with identification as a dinosaur, and they are regarded as Dinosauria indet. herein. The presence of prominent vertebral laminae in No. 338 is consistent with identification as either a theropod or sauropodomorph dinosaur (Wilson 1999; Galton & Upchurch 2004). Pectoral girdle and forelimb Scapula At least three of Owen’s (1854, pp. 101–102) specimens represent portions of scapulae (Nos. 349, 350 and 359) although one was incorrectly identified as a left ischium (ibid., p. 102). Seeley (1895, pp. 103 and 109–110, fig. 5) incorrectly identified all three specimens as ischial fragments, but this was corrected by Huene (1906). Owen (1854, p. 101) correctly noted that that No. 349 is the proximal part of a left scapula and he, Seeley (1985, p. 103) and Huene (1906, p. 135, pl. XIV [XXI], figs 8, 9) opined that Nos. 349 and 350 represented different parts of the same individual element; however, although this is plausible on the basis of size, the mixed nature of the assemblage prevents confirmation of this suggestion. No. 359 was figured by Huene (1906, fig. 49). A further speci- men, No. 357, which was originally listed as a ?humerus ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 113 of Pachyspondylus (Owen 1854), was identified as a left scapula by Huene (1906, p. 135) but was not figured and no cast is available. The chimeric nature of the assemblage is exemplified by the presence of several proximal scapu- lae (Nos. 349, 357, 359), which not only differ in size (see Table 4 for measurements) but are all left scapulae, thus representing a minimum of three individuals. The scapulae are described with their long axes oriented horizontally. In lateral view, the proximal end of the scapula forms a dorsoventrally expanded plate with a sub-pentagonal outline (No. 349 [NHMUK PV R3029 partim; SAM-PKC-964; Fig. 7A]; No. 359 [NHMUK PV R3032 partim; SAM-PKC-970; Fig. 7B]). Both scapulae are lacking the anterodorsal corner of the proximal plate, which could account of their earlier identification as ischia. The lateral surface of the proximal plate is smoothly concave both anteroposteriorly and dorsoventrally and the posterior margin of this concavity is formed by a distinct, rounded rim that is confluent with the acromial ridge dorsally and that expands to buttress the glenoid ventrally. This rim also separates the lateral surfaces of the proximal plate and scapula shaft. The concavity is deeper in the dorsal part of the proximal plate than ventrally. As preserved, the proximal plate expands further ventrally than dorsally with respect to the margins of the scapula shaft but given the damage to the anterodorsal region it is possible that the dorsal part of the plate was originally more extensive. The base of the acromion process forms an obtuse angle of approximately 120° with the dorsal margin of the shaft. The anterior margin of the proximal plate forms the articular surface for the coracoid and is straight and vertically inclined. It is separated from the ventral margin of the proximal plate, which is very slightly concave and represents the glenoid, by an angle of approximately 120° (No. 349; Fig. 7A) or 150° (No. 359; Fig. 7B). 114 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 Table 4. Pectoral girdle, forelimb and manus measurements in millimetres. Asterisks (*) indicate incomplete bones; dashes ('–') indicate inapplicable or substantially incomplete measurements. Scapula NHMUK catalogue number: PV R3029 (partim) PV R3029 (partim) PV R3032 (partim) Owen (1854) catalogue number: 349 350 359 Proximal plate, maximum dorsoventral height 105* – 88* Proximal plate, maximum anteroposterior length 74 – 58 Proximal plate, maximum mediolateral width (at dorsal margin of glenoid fossa) 39 – 29 Shaft, dorsoventral height 53 – 34 Distal expansion, maximum dorsoventral height – 71* – Humerus NHMUK catalogue number: PV R3031 (partim) PV R3031 (partim) Owen (1854) catalogue number: 354 356 Length (as preserved) 140* 78* Proximal expansion, maximum mediolateral width 100 – Distal expansion, maximum mediolateral width – 89 Radial (lateral) condyle, maximum anteroposterior – 35 Metacarpal I NHMUK catalogue number: PV R3037 (partim) Owen (1854) catalogue number: 374 Length, dorsal margin 36 Length, ventral margin 53 Proximal end, dorsoventral height 45 Proximal end, maximum mediolateral width (ventral margin) 33 Distal end, dorsoventral height 36 Distal end, maximum mediolateral width 18 Metacarpal V NHMUK catalogue number: PV R3037 (partim) Owen (1854) catalogue number: 366 Length 34 Proximal end, maximum dorsoventral height 25 Proximal end, maximum mediolateral width 25 Distal end, maximum dorsoventral height 20 Distal end, maximum mediolateral width 21 Manual unguals NHMUK catalogue number: PV R3037 (partim) PV R3037 (partim) PV R3026 Owen (1854) catalogue number: 382 383 385 Length 47* 55* 32* Proximal articular surface, height 36 40 20 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 115 Figure 7. Casts of pectoral girdle, forelimb and manus bones from the Orpen Collection. A, NHMUK PV R3029 (partim), catalogue number 349, left proximal scapula in lateral view. B, NHMUK PV R3032 (partim), catalogue number 359, left proximal scapula in lateral view. C, NHMUK PV R3029 (partim), catalogue number 350, left distal scapula in lateral view. D–F, NHMUK PV R3031 (partim), catalogue number 354, right proximal humerus in anterior (D), posterior (E) and proximal (F) views. G–I, NHMUK PV R3031 (partim), catalogue number 356, right distal humerus in anterior (G), poste- rior (H) and distal (I) views. J–L, NHMUK PV R3037 (partim), catalogue number 374, right metacarpal V in dorsal (J), left lateral (K) and distal (L) views. M, NHMUK PV R3037 (partim), catalogue number 366, ?left metacarpal V in dorsal view. N, NHMUK PV R3026, catalogue number 385, ungual phalanx in lateral view. O, NHMUK PV R3037 (partim), catalogue number 382/383, ungual phalanx in lateral view. P, NHMUK PV R3037 (partim), catalogue number 382/383, manual ungual phalanx in lateral view. In medial view, the proximal plate is shallowly concave dorsoventrally in its dorsal part but becomes flat to slightly convex ventrally. There is some indication of a short, horizontally directed groove that would have been confluent with the articular surface for the coracoid, situated immediately dorsal to the expanded glenoid region. The surface of the proximal plate merges smoothly with the shaft surface. In anterior view, the proximal plate is strongly expanded mediolaterally in its ventral part to form the articular surface for the coracoid and the glenoid fossa. The glenoid surface has a sub-triangular outline, with the apex of this triangle extending ventrolaterally and the surface is gently concave. it is separated from the articular surface for the coracoid by a marked ridge of bone and distinct break-in-slope. The coracoid articulation has a rounded, sub-elliptical outline with its long axis trending dorsoventrally, but is broken dorsally. The preserved surface is shallowly concave. Dorsally, the proximal plate thins considerably in mediolateral thickness and its medial margin describes a shallow sinusoidal curve. In both Nos. 349 and 359 the base of the scapular shaft is dorsoventrally narrow and its dorsal and ventral margins converge to extend parallel to each other over their preserved lengths (Fig. 7A, B). The lateral surface of the shaft is strongly convex dorsoventrally, whereas the medial surface is either weakly convex (No. 349) or flat (No. 359), imparting either a narrow elliptical or ‘D’- shaped transverse cross-section, respectively. No. 350 (NHMUK PV R3029 partim; SAM-PKC-964; Fig. 7C) is a badly preserved fragment of a scapula blade missing parts of its dorsal and ventral margins. Its medial and lateral surfaces are both gently convex so that anteri- orly it has a tall, narrow elliptical cross-section. This distal part of the blade was clearly expanded with respect to the anterior part of the scapular shaft, but its incompleteness prevents determination of whether this expansion was symmetrical or not, or the relative degree of expansion with respect to either the anterior shaft or proximal plate. Nos. 349, 350 and 359 are superficially similar to the neotype of M. carinatus (Barrett et al. 2019) but similar scapulae are also present in a variety of other sauropodo- morphs (e.g. Adeopapposaurus, Lufengosaurus: Young 1941; Martínez 2009). Moreover, although M. carinatus has been diagnosed on the basis of having an asymmetrically expanded distal scapula blade (with a large ventral and smaller dorsal expansion; see Barrett et al. 2019) the presence/absence of this feature cannot be assessed in No. 350. The overall appearance and size of the specimens suggests that they pertain to a sauropodomorph (Galton & Upchurch 2004), but their generalized morphology and poor preservation excludes a definitive referral to any other known upper Stormberg Group taxon (none of which possess diagnostic features of their scapulae: Knoll 2010; Chapelle et al. 2019) or to any particular sauropodo- morph clade (Yates 2007; Apaldetti et al. 2013). They are regarded as Sauropodomorpha indet. herein. Humerus The collection includes proximal (No. 354; SAM-PKC- 967; Fig. 7D–F) and distal (No. 356; SAM-PKC-968; Fig. 7G–I) parts of right humeri (registered together as NHMUK PV R3031). Owen (1854, pp. 101–102) suggested that these were parts of the same bone, a conclusion followed by Seeley (1895, p. 116, fig. 12) and Huene (1906, pl. XIV [XXI], figs 6, 7). However, although this is plausi- ble, as the two fragments match well in terms of size, this cannot be confirmed. A third specimen (No. 357) thought to be a humerus by Owen (1854, p. 101) was re-identified as an ischium by Seeley (1895, pp. 103 and 109) and as a scapula by Huene (1906; see above). Measurements are provided in Table 4. In anterior or posterior view, the proximal expansion of the humerus (No. 354) consists of a large internal tubero- sity, the humeral head and the deltopectoral crest (Fig. 7D, E). The internal tuberosity has a sub-triangular outline, with a bluntly rounded apex that extends medially. It is positioned ventral to the humeral head and the dorsal surface of the internal tuberosity merges with the latter along a smooth curve. The humeral head forms the proxi- mal-most point of the humerus but is rather poorly devel- oped and its articular surface does not extend onto the anterior surface of the humerus. Instead, it forms a low lip that overhangs it. Lateral to the humeral head, the proxi- mal margin of the humerus descends ventrally to merge with the deltopectoral crest. A shallow, ‘U’-shaped notch separates the dorsal margin of the crest from the remain- der of the proximal margin. This notch lies at a level just ventral to the ventral margin of the internal tuberosity. The deltopectoral crest is incomplete, but the preserved portion is prominent, has a narrow, sub-rectangular out- line and extends from the humerus anteriorly and slightly laterally. Its lateral margin is straight. The anterior surface of the proximal humerus is dorsoventrally flat but medio- laterally concave due to the curvature of the deltopectoral crest. The proximal humerus is strongly convex mediolaterally and slightly concave dorsoventrally in posterior view. The posterior surface of the internal tuberosity bears a low pyramidal swelling. A shallow sulcus separates this swell- ing from the articular surface of the humeral head, which forms a larger sub-elliptical boss that slightly overhangs the rest of the posterior surface. The long axis of this boss extends mediolaterally. No other features, such as intra- muscular lines, can be discerned clearly though it is unclear whether this is due to surface damage of the origi- nal bone or the introduction of casting artefacts. In proximal view, the humerus has a slightly sinuous outline (Fig. 7F). The internal tuberosity and humeral head are the most strongly expanded parts of the bone anteroposteriorly (with the humeral head showing the greater expansion) and the proximal margin tapers as it extends laterally from the region of the head, giving rise to the thin, sheet-like deltopectoral crest. None of the humeral shaft is preserved and the broken base of the proximal end has a narrow sub-triangular cross-section that tapers strongly laterally. A shallow sulcus covers most of the anterior surface of the distal humerus (No. 356; Fig. 7G), covering the area dorsal to and between the ulnar and radial condyles. The radial and ulnar condyles are similar in size and form 116 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 rounded sub-ovate bosses that project anteriorly. In anterior view, the dorsal margin of the radial condyle is confluent with a robust, low, rounded crest that extends for a short distance before merging into the surface of the bone. A prominent, triangular entepicondyle is present on the medial margin of the distal end, dorsal to the ulnar condyle, producing a strongly bevelled and medio- ventrally facing surface between it and the distal humeral surface. The posterior surface of the distal end is flat and featureless (Fig. 7H). In distal end view, the humerus has a dumbbell-shaped outline and the two condyles project both anteriorly and posteriorly for approximately the same distance (Fig. 7I). The broken base of the shaft has a stout, elliptical cross-section, with a long axis that trends mediolaterally. Both these specimens differ from the humeri preserved in the neotype of M. carinatus. In BP/1/4934 the humeral head is prominent and extends for a considerable distance dorsal to the internal tuberosity (Barrett et al. 2019) but in No. 354 the apex of the humeral head is substantially lower. Moreover, M. carinatus possesses a well-defined olecranon fossa (Barrett et al. 2019), as does Adeopappo- saurus (Martínez 2009), but this feature is absent in No. 356. An olecranon fossa is also absent in Arcusaurus, Ignavusaurus and Ngwevu (Knoll 2010; Yates et al. 2011; Chapelle et al. 2019). Other differences in both humeral shape or intramembral ratios (such as the relative length of the deltopectoral crest, which is diagnostic for M. carinatus: Barrett et al. 2019) cannot be assessed due to incompleteness or uncertainly over the association between Nos. 354 and 356. Given their similarity to other sauropodomorphs (e.g. Galton & Upchurch 2004) these specimens are here regarded are Sauropodomorpha indet. Metacarpal I No. 374 (NHMUK PV R3037 partim; Fig. 7J–L) was listed by Owen (1854, p. 104) as a fifth metatarsal and by Seeley (1895, pp. 114–115) as a first metatarsal; Huene (1906, p. 135, fig. 61, pl. XVI [XXIII], fig. 3) was the first to identify it correctly as a right metacarpal I. Here, it is described in life position (assuming that the hand was held in a supi- nated position, with the long axes of the central digits extending anteroposteriorly). Measurements are provided in Table 4. Metacarpal (Mc) I has a sub-trapezoidal outline in lateral view and both its proximal and distal ends are expanded dorsoventrally with respect to the short, stout shaft that connects them. The proximal expansion is the larger of the two and is developed asymmetrically, so that it expands further dorsally than ventrally. The posterior and anterior margins (articular surface margins) are very gently con- cave in lateral view, the ventral margin (ventral margin of the shaft) is slightly arched and the dorsal margin of the shaft is deeply notched. The ventral margin of Mc I is considerably longer than the dorsal margin contributing further to the asymmetry of the element. The lateral surface of Mc I between the two articular expansions is almost flat dorsoventrally, but gently concave antero- posteriorly. By contrast, the medial surface of the shaft is slightly convex dorsoventrally and deeply concave anteroposteriorly, due to the torsion of the distal end with respect to the rest of the bone (see below). The articular surface of the proximal expansion has an elongate, triangular outline in posterior view, whose rounded apex points dorsally and whose lateral, medial and ventral margins are all straight to very slightly concave. A dorsoventrally extending ridge traverses the articular surface and is slightly offset towards the lateral margin of the bone, dividing this surface into two shal- lowly concave facets of different sizes (a large triangular facet medially, and smaller, strip-like facet laterally). The long axis of the proximal end extends dorsoventrally. In anterior view, the distal expansion is divided into dorsal and ventrally positioned ginglymi, which give it a dumbbell-shaped outline with strongly concave medial and lateral margins and weakly convex dorsal and ventral margins. The articular surface is saddle-shaped (dorso- ventrally concave and strongly convex mediolaterally). In anterior view, the long axis of the distal end is oriented ventrolaterally–dorsomedially, so that Mc I exhibits subtle torsion between its distal and proximal ends with the two long axes offset by approximately 30°. The dorsal and ventral margins of the distal expansion each bear incipient collateral ligament pits. When viewed dorsally, the proximal and distal expan- sions are approximately equal in mediolateral width and are connected to each other by a narrow ridge. By contrast, in ventral view the proximal articulation is much more greatly expanded than the distal one, with most of this expansion extending medially, contributing further to the overall asymmetry of the bone. The ventral surface of the proximal end is covered by a large shallow sulcus that represents the articular surface for Mc II. Torsion between the proximal and distal ends of Mc I and its marked asymmetry are features that characterize a wide variety of sauropodomorph dinosaurs (Galton & Upchurch 2004). Moreover, in No. 374 the ratio between the length of the metacarpal and the maximum height of its proximal articular surface is approximately 0.85, which is regarded as synapomorphy of Massospondylidae + more derived sauropodomorphs (e.g. Yates 2007; Apaldetti et al. 2013). No. 374 possesses no other features that distinguish it from the first metacarpals of many other sauropodomorph taxa (e.g. Lufengosaurus, M. carinatus, Ngwevu: Young 1941; Barrett et al. 2019; Chapelle et al. 2019), so it is identified as that of an indeterminate massopodan herein. Metacarpal V Owen (1854, p. 103) described No. 366 (NHMUK PV R3037 partim; Fig. 7M) as a tarsal, noting similarities to the calcaneum of crocodiles, but as noted by Huene (1906, pl. XVI [XXIII], fig. 6) this element is an Mc V. As with the preceding description, this element is described as though held in life position. Measurements are provided in Table 4. In both lateral and dorsal views, Mc V consists of strongly expanded proximal and distal articular regions, which are connected by a short, stout shaft. Both the ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 117 articular regions are expanded mediolaterally and dorso- ventrally with respect to the shaft, but the proximal expansion is the larger of the two. The proximal articular region has a pyramidal shape with a sub-quadrate outline and four flat, obliquely inclined surfaces facing dorsally, ventrally, posteromedially and posterolaterally. These surfaces converge posteriorly to form a low ridge that is displaced towards the medial side of the bone. The proxi- mal articular surface is mediolaterally widest along its dorsal margin and narrow ventrally; the medial and lateral margins are subequal in length. The shaft has a sub-circular to elliptical transverse cross-section. The distal articular region forms a rounded, globose surface that is almost equally expanded dorsoventrally and mediolaterally. In dorsal or ventral view, Mc V is slightly asymmetrical with the long axis of the distal end set at a small angle with respect to the proximal end. Although No. 366 does not possess any features consid- ered synapomorphic for any sauropodomorph clade, it closely resembles the Mc V in a wide variety of sauro- podomorphs (Galton & Upchurch 2004) and is referred to Sauropodomorpha indet. herein. Manual unguals Casts of three ungual phalanges of varying sizes are available (Nos. 382, 383, 385; NHMUK PV R3026, R3037 partim; Owen 1854, p. 105). Owen (1854) did not attribute these to either the manus or pes but Seeley (1895, pp. 115–116) regarded them as part of the foot and figured No. 383 (ibid., fig. 11). Huene (1906, figs 63, 64, 68) consid- ered Nos. 383 and 385 as manual unguals, whereas he regarded No. 382 as part of the pes. For ease of compari- son, the unguals are described as though held horizon- tally (i.e. with their recurved tips directed ventrally). Measurements are provided in Table 4. In lateral view, the unguals are tallest proximally (Fig. 7N–P). The dorsal margin is strongly convex, whereas the ventral margin is strongly concave and the ungual tapers as these margins converge distally, to form a narrow, strongly recurved point (the tip is missing in all three examples). The proximal articular surface is strongly concave dorsoventrally, due in part to a large lappet that overhangs this surface dorsally. The articular surface has a sub-ovate outline, with its long axis is oriented dorso- ventrally. In dorsal view, the ungual is compressed mediolaterally and tapers anteriorly. The lateral and medial surfaces each bear grooves for claw sheaths positioned either at or slightly ventral to ungual mid- height. Seeley (1895) regarded these as pedal unguals but all three are identified as manual unguals herein. In sauro- podomorphs, manual unguals exhibit strong recurvature and are mediolaterally compressed, whereas pedal unguals are proportionally wider and lack marked recurvature (Galton & Upchurch 2004). Although these unguals bear no synapomorphies of any particular clade they are indistinguishable from those of many sauro- podomorphs (Galton & Upchurch 2004) and are referred to Sauropodomorpha indet. Pelvic girdle Ilium A left ilium (No. 358; NHMUK PV R3032 partim; SAM-PKC-969; Fig. 8A) was included among the material described by Owen (1854, 102; see also Seeley 1895, p. 108, fig. 4; Huene 1906, p. 137, pl. XV [XXII]). It is almost complete but is lacking the pubic peduncle. Measure- ments are provided in Table 5. In lateral view, the dorsal margin of the ilium is gently convex anteriorly, but this becomes straight posteriorly. The preacetabular process is anteroposteriorly short and has an almost equitriangular outline that tapers to a blunt, rounded tip in lateral view. It is mediolaterally compressed with a very narrow, sub-triangular transverse cross-section. This process is separated from the base of the pubic peduncle by a broad, ‘U’-shaped sulcus. Although the pubic peduncle is missing, the shape and orientation of its broken surface indicates that it would have extended further anteriorly than the preacetabular process. The main blade of the ilium dorsal to the acetabulum is divided into two surfaces in lateral view, by a subtle break-in-slope that occurs approximately level with the ventral margin of the preacetabular process (Fig. 8A). Consequently, the dorsal part of the iliac blade is strongly concave anteroposteriorly and dorsoventrally, whereas its ventral part is weakly convex in both directions. A poorly developed supraacetabular crest separates the lateral surface of the ilium from the acetabular articular surface. The postacetabular process is well-developed and has a sub-quadrate outline with a flat to convex dorsal margin, a short, flat posterior margin and an anteroventrally sloping, straight ventral margin (Fig. 8A). The latter is very slightly offset from the otherwise straight posterior margin of the iliac blade by a very subtle inflection. The postacetabular process is mediolaterally expanded with respect to the rest of the iliac blade and has a broad, sub-triangular transverse cross-section with the apex of this triangle oriented laterally. In lateral view, the acetabulum is almost fully open, with only its dorsal-most part slightly overhung laterally by a ventral extension of the iliac blade and its medial part formed from a shallow sheet of bone (Fig. 8A). Viewed ventrally, the acetabulum is strongly expanded medio- laterally with respect to the dorsal part of the ilium. It is widest anteriorly and tapers slightly posteriorly. The acetabular surface is strongly concave both mediolaterally and anteroposteriorly. In medial view, the ventral margin of the ilium forms a mediolaterally expanded buttress that lies dorsal to the acetabulum. Anteriorly, this buttress probably forms the ventral margin of the area that would have been occupied by the sacral rib attachments. A short, but well-defined ridge that extends for a short distance anteriorly along the base of the postacetabular process would also have contributed to the ventral margin of the sacrum attach- ment area. Sadly, the cast is not detailed enough to reveal details of either the number or shape of the sacral rib facets, although Seeley (1895) estimated that three or four sacrals would have been likely. The dorsal part of the 118 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 119 Figure 8. Casts of pelvic girdle bones from the Orpen Collection. A, NHMUK PV R3032 (partim), catalogue number 358, left ilium in lateral view. B–C, NHMUK PV R3030 (partim), catalogue number 352, ?right proximal pubis in lateral (B) and proximal views (C). D–E, NHMUK PV R3036 (partim), catalogue number 351, right proximal pubis in lateral (D) and proximal views (E). F–G, NHMUK PV R3036 (partim), catalogue number 353, ?left ischial shaft in lateral (F) and posterior views (G). 120 ISSN 2410-4418 Palaeont. afr. (2024) 58: 97–131 Table 5. Pelvic girdle, hindlimb and pes measurements in millimetres. Asterisks (*) indicate incomplete bones; dashes ('–') indicate inapplicable or substantially incomplete measurements. Ilium NHMUK catalogue number: PV R3032 (partim) Owen (1854) catalogue number: 358 Length, anteroposterior 224 Blade, height dorsal to acetabulum 89 Preacetabular process, anteroposterior length of 34 Postacetabular process, anteroposterior length of 39 Ischiac peduncle, anteroposterior length of 33 Ischiac peduncle, mediolateral width 39 Acetabulum, maximum mediolateral width 55 Pubis NHMUK catalogue number PV R3036 (partim) PV R3030 Owen (1854) catalogue number 351 352 Proximal end, anteroposterior length 101 83 Proximal end, maximum mediolateral width 32 24 Ischium NHMUK catalogue number: PV R3036 (partim) Owen (1854) catalogue number: 353 Length (as preserved) 95* Maximum anteroposterior length (as preserved) 58* Femora NHMUK catalogue number: PV R3033 (partim) PV R3033 (partim) PV R3033 (partim) Owen (1854) catalogue number: 360 361 362 Length as preserved 159* 181 245 Proximal end, maximum mediolateral width 83* – – Proximal end, maximum anteroposterior length 52 – – Distal end, maximum mediolateral width – 102 104 Distal end, maximum anteroposterior length – 84 83 Shaft, mediolateral diameter – 53 50 Shaft, anterolateral diameter – 51 51 Tibiae NHMUK catalogue number: PV R3034 (partim) PV R3034 (partim) Owen (1854) catalogue number: 363 364 Length (as preserved) 142* 135* Proximal end, maximum anteroposterior length 109 – Proximal end, maximum mediolateral width 77 – Cnemial crest, dorsoventral length 73 – Distal end, maximum anteroposterior length – 66 Distal end, maximum mediolateral width – 42 Shaft, anteroposterior diameter 48 42 Shaft, mediolateral diameter 38 35 Metatarsal II NHMUK catalogue number: PV R3037 (partim) PV R3037 (partim) Owen (1854) catalogue number: 367 369 (left Mt II, proximal) (right Mt II, proximal) Proximal end, maximum anteroposterior length 59 56 Proximal end, maximum mediolateral width 43 36 Distal shaft, maximum anteroposterior length 21 27 Distal shaft, maximum mediolateral width 21 19 Maximum length (as preserved) 77* 77* Metatarsal III NHMUK catalogue number: PV R3037 (partim) PV R3037 (partim) Owen (1854) catalogue number: 368 373 (right Mt III, distal) (Mt?, distal) Distal end, maximum anteroposterior length 29 18 Distal end, maximum mediolateral width 42 31 Shaft, maximum anteroposterior length 19 – Shaft, maximum mediolateral width 24 – Length (as preserved) 86* – Phalanges NHMUK catalogue number: PV R3037 (partim) PV R3037 (partim) PV R3037 (partim) PV R1312a Owen (1854) catalogue number: 375 378 379 380 Length 60 40 42 32 Proximal end, maximum width 40 33 25 24 Proximal end, maximum height 30 27 23 22 Distal end, maximum width 38 30 23 22 Distal end, maximum height 23 18 15 17 ilium’s medial surface is strongly convex anteroposterior- ly (so that in dorsal view, the ilium is bowed medially), as is the area immediately adjacent to the ventral buttress. The triangular area posterior to the ventral buttress and ventral to the postaceabular process is smooth and shallowly concave. In either lateral or medial view, the ischiac peduncle forms a short, posteroventrally directed process with a sub-quadrate outline. In ventral view, it has a sub-trian- gular outline with a long anterior margin (forming the posterior border of the acetabulum), an almost equally long posterolateral margin and a short medial margin. Various features of the ilium allow its identification as that of a sauropodomorph, such as the absence of a buttress between the preacetabular process and the acetabular margin and the presence of an almost fully open acetabulum (Galton & Upchurch 2004). The rounded to slightly pointed overall outline of the postacetabular process (rather than a more angular or square-ended process) is consistent with referral to Massospondylidae. Overall, the ilium is very similar to that of Adeopapposaurus (Martínez 2009), M. carinatus (Barrett et al. 2019) and Ignavusaurus (Knoll 2010), suggest- ing that referral to Massospondylidae is likely, but attribu- tion to a particular taxon is not possible given the absence of clear autapomorphies. Pubis The proximal parts of two ?right pubes are preserved (Nos. 351 and 352). Owen (1854, p. 101) originally identi- fied these as parts of a coracoid (No. 352; NHMUK PV R3030; SAM-PKC-966; Fig. 8B, C) and scapula (No. 351; NHMUK PV R3036 partim; Fig. 8D, E), respectively, but this error was corrected by Seeley (1895, pp. 103 and 110) and Huene (1906, p. 138, fig. 51). Seeley (1895, fig. 6) provided a reconstruction of the entire pubis based on No. 351 and ‘… several specimens, some of which have hith- erto escaped attention, and remain … without numbers’