Centro de Geologia, Faculdade de Ciências da Universidade de Lisboa (FCUL), Edificio C6, Campo Grande, Lisbon, Portugal 3

Transcription

1 bs_bs_banner Zoological Journal of the Linnean Society, 2014, 172, With 28 figures New generic proposal for the European Neogene large testudinids (Cryptodira) and the first phylogenetic hypothesis for the medium and large representatives of the European Cenozoic record ADÁN PÉREZ-GARCÍA 1,2,3 * and EVANGELOS VLACHOS 4 1 Departamento de Paleontología, Facultad de Ciencias Geológicas, Universidad Complutense de Madrid, C/ José Antonio Novais, 2, Ciudad Universitaria, Madrid, Spain 2 Centro de Geologia, Faculdade de Ciências da Universidade de Lisboa (FCUL), Edificio C6, Campo Grande, Lisbon, Portugal 3 Grupo de Biología Evolutiva, Facultad de Ciencias, UNED, C/ Senda del Rey, 9, Madrid, Spain 4 School of Geology, Aristotle University of Thessaloniki, University Campus, 54124, Thessaloniki, Greece Received 11 May 2014; revised 16 June 2014; accepted for publication 20 June 2014 Fossil testudinids are known in Europe since the Eocene, with several taxa of medium size (from more than 0.3 m to less than 0.7 m) recognized in the Palaeogene record, most of them being poorly known. The size of several European Neogene taxa was larger (between 1 and 2 m). These large testudinids were relatively abundant and diverse, ranging from the early Miocene to the Pleistocene. However, there is a nomenclatural gap at the generic level for the Neogene forms, as their generally used assignment to the more primitive Eocene Cheirogaster cannot be sustained. This is because relatively little material has been assigned to the described species, and also because of the absence of a detailed study comparing all of the European taxa. Here, the European Cenozoic taxa are incorporated for the first time in a data matrix, so that a hypothesis on their phylogenetic relationships is justified. This study identified the large testudinids from the Neogene of Europe as belonging to a monophyletic clade, assigned to the new genus Titanochelon. The hitherto poorly understood Testudo bolivari, proposed nearly a century ago but lacking diagnosis, is analysed in detail. It is recognized as the best-represented large testudinid from the European record, and is identified as the type species of Titanochelon gen. nov. Its comparison with the other Neogene species allowed a detailed study of the new genus and an analysis of its phylogenetic relationships with the other European taxa.. doi: /zoj ADDITIONAL KEYWORDS: Cheirogaster Europe Palaeogene phylogenetic relationships Testudinidae Testudo bolivari Titanochelon gen. nov. *Corresponding author. paleontologo@gmail.com INTRODUCTION Fossil testudinids are known in Europe since the Palaeogene. Several taxa of medium size (from more than 30 cm to less than 70 cm long) have been recognized in the Eocene Oligocene record of this continent. Most of them, such as the species attributed to Achilemys Hay, 1908, Hadrianus Cope, 1872, and Ergilemys Chkhikvadze, 1972, require revision and comparison with the type specimens from North America and Asia, as they are probably not attributable to these genera (not originally defined in Europe), but to undefined new European taxa. Other European Palaeogene forms are the Eocene Cheirogaster Bergounioux, 1935 (Cheirogaster maurini Bergounioux, 1935) and the poorly 653

2 654 A. PÉREZ-GARCÍA AND E. VLACHOS known Oligocene Testudo gigas Bravard, 1844 (de Lapparent de Broin, 2001; Gmira et al., 2013). Testudo gigas is a species usually assigned to Cheirogaster but is recognized here as not belonging to this genus (see Discussion). The available information on these Palaeogene forms is generally restricted to the shell, with the exception of very few and generally poorly preserved appendicular elements. In fact, the holotype and only known specimen of the type species of Cheirogaster is a shell, and all of the other Palaeogene specimens that have been attributed to this genus are exclusively represented by shells or isolated plates (see de Broin, 1977; de Lapparent de Broin, 2002). The revision of Cheirogaster is of paramount importance, as the putative presence of this genus has been expanded by several studies to the Neogene, including nearly all of the large tortoises of the Miocene and Pliocene of Europe (de Lapparent de Broin, 2001, 2002). Large testudinids were abundant and diverse also in the European Neogene record, ranging from the early Miocene of France to the Pleistocene of Greece and some Mediterranean islands. More than ten potentially valid species are generally recognized (see de Lapparent de Broin, 2001, 2002). In contrast to the taxa found in the Palaeogene of Europe, the size of these Neogene forms was greater than 1 m, with specimens reaching truly gigantic lengths close to 2 m long (e.g. the indeterminate species from Las Higueruelas, in Spain, and from Lesvos and Samos, in Greece; see Jiménez Fuentes, 1994, and de Lapparent de Broin, 2002). Despite the abundant postcranial remains of these tortoises, many of the nominated taxa are relatively poorly known (Gmira et al., 2013). Indeed, as several authors have pointed out (e.g. Jiménez Fuentes, 1988; Gmira et al., 2013), there is a nomenclatural gap at the generic level for the large tortoises from Europe because their putative assignment to the more primitive and medium-sized Cheirogaster (known from the late Eocene of France) cannot be sustained. For this reason, it has been indicated that the comparison of these Neogene forms with those of the Palaeogene, and better knowledge of the species involved, are required to understand the evolution of the European large tortoises (Gmira et al., 2013; Vlachos, Tsoukala & Corsini, 2014). The incompleteness of the fossil record of the large tortoises only allows comparisons of the known material from the Neogene of Europe with those of medium-sized tortoises from the Palaeogene based on detailed analysis of their shells. In this study we present relatively complete material that improves our knowledge of the Neogene fossil record of the European large tortoises. The currently available information on the shells of these taxa is limited. The preservation of the shell material available from the two species so far represented by the most complete shells (i.e. the late Miocene Testudo leberonensis Depéret, 1890, and the late Pliocene Testudo perpiniana Depéret, 1885) does not allow recognition of the morphology of most plates and scutes in the carapace. Most of the available shell information for the early Miocene Testudo ginsburgi de Broin, 1977, is based on the plastron of its holotype, corresponding to a juvenile individual. The available data on the early Miocene Testudo vitodurana Biedermann, 1862, and the middle Miocene Cheirogaster steinbacheri Karl, 1996, are very limited. Besides the recently described Pliocene Cheirogaster bacharidisi Vlachos et al., 2014, which is represented by few individuals, the relatively scarce information on the shells of all of these species has precluded both detailed comparisons amongst them, and the proposal of hypotheses concerning their individual, ontogenetic, and sexual variability. In addition, there are several Neogene species without known elements of the shell, such as the late Miocene Testudo schafferi Szalai, 1931, and the taxa named from several Mediterranean islands (see Leith-Adams, 1877; Bate, 1914). The present study is an attempt to solve the problems identified above. All the taxa involved have been compared here, and a detailed study of the Neogene forms has been performed. Therefore, in addition to reviewing specimens corresponding to all taxa of large testudinids so far described in the Neogene of Europe, we have included detailed information on a much larger group of comparative taxa than previous studies, including all of the medium-sized European fossil testudinids. No hypothesis on the phylogenetic relationships amongst the European Cenozoic forms had been hitherto proposed. Here, the European Cenozoic taxa are incorporated for the first time in a data matrix, so that a hypothesis on their phylogenetic relationships is justified. Clarification of the morphology and status of the so-far problematic genus Cheirogaster has been performed here. The study identified the large testudinids from the Neogene of Europe as belonging to a monophyletic clade, assigned to the new genus Titanochelon. Nine species have been attributed to it. The so-far poorly known and problematic Spanish Miocene Testudo bolivari Hernández-Pacheco, 1917a, has been recognized here as the best-represented large testudinid from the Neogene of Europe, its information also being much more complete than that of medium-sized forms from the Palaeogene. Therefore, we identified it as the type species of Titanochelon. In order to accurately describe the skeleton of Titanochelon, its type species, identified here from very abundant and well-preserved material, was analysed in detail. Testudo bolivari is recognized as the only European large tortoise represented by numerous specimens, from several localities. Discoveries of this taxon are relatively common in many Spanish regions, where several well-preserved specimens have been

3 A GENUS FOR THE EUROPEAN LARGE TORTOISES 655 Figure 1. Geographical and stratigraphical position of the Spanish localities where the classic material of large tortoises was discovered, and of other relevant Spanish areas cited in this paper. A, map of the geographical positions of the localities, showing the Neogene outcrops (grey colour). 1, Villalcón (Palencia); 2, Cerro del Otero (Palencia); 3, Fuensaldaña (Valladolid), La Cistérniga (Valladolid); 4, Coca (Segovia); 5, Arévalo (Ávila); 6, several sites in Madrid or very close to this city, including: Meaques stream (Casa de Campo), Barajas, Carabanchel, Cerro de los Batallones, Ciudad Universitaria, El Pardo, Parla, Puente de los Franceses, Puente de la Princesa, Vallecas; 7, Alcalá de Henares (including sites such as those of the type locality of Titanochelon bolivari: Cerro del Viso) and Los Santos de la Humosa (Madrid); 8, Villaluenga de la Sagra (Toledo), Illescas (Toledo); 9, Els Hostalets de Pierola (Barcelona), Ecoparc de Can Mata (Catalonia). B, Stratigraphical position of these localities. I, Meaques stream (Casa de Campo), Barajas, Puente de los Franceses, Vallecas; II, Cerro del Viso (Alcalá de Henares), Parla, Coca, Henares 1 (Los Santos de la Humosa); III, Carabanchel, Ciudad Universitaria, El Pardo, Puente de la Princesa; IV, Cerro del Otero, Fuensaldaña, La Cistérniga, Villaluenga de la Sagra, Illescas; V, Arévalo, Els Hostalets de Pierola, Ecoparc de Can Mata; VI, Cerro de los Batallones. MN, European Mammal Neogene biozones. identified (see García & Alberdi, 1968; Jiménez Fuentes & Carbajosa Tamargo, 1982; Jiménez Fuentes et al., 1986). It was named from abundant material from various localities (see Hernández-Pacheco, 1917a; Fig. 1), but without a diagnosis or detailed descriptions and comparisons. In addition, a discovery of abundant and well-preserved material was made after the publication of the paper of Hernández-Pacheco (1917a) and prior to the Spanish Civil War ( ) (see Royo y Gómez, 1935a and references therein). All the material found prior to the Spanish Civil War, and commonly referred as the classic material of Te. bolivari, was identified as lost (Jiménez Fuentes et al., 1988; Jiménez Fuentes & Martín de Jesús, 1991; Vlachos et al., 2014; Luján et al., in press). The Spanish researcher Eduardo Hernández-Pacheco ( ) indicated his intention to publish a detailed monograph on Te. bolivari (see Hernández-Pacheco, 1917a, b, 1921a, 1930). Unfortunately, that work has never been published. Owing to the relatively limited information on the classic material, it was generally ignored and several of its characters and data were misinterpreted in subsequent works focused on the study of large testudinids from the Neogene of Europe (e.g. de Lapparent de Broin, 2002; Vlachos et al., 2014). We rediscovered most of the classic material of Te. bolivari, including numerous unpublished specimens. In fact, very few of these specimens had been preliminarily described, including some of those coming from Alcalá de Henares and Ciudad Universitaria (Madrid), Cerro del Otero (Palencia) and Arévalo (Ávila). This detailed study provides new data on these specimens and has allowed us to amend several descriptive and interpretative aspects. We also describe unpublished material from these localities. Furthermore, we describe and figure, for the first time, classic material from other localities (e.g. Vallecas, Puente de la Princesa, Puente de los Franceses and Parla in Madrid; Fuensaldaña and

4 656 A. PÉREZ-GARCÍA AND E. VLACHOS La Cistérniga in Valladolid; Villalcón in Palencia; Illescas and Villaluenga de la Sagra in Toledo) (see Royo y Gómez, 1935a). This study allows the proposal of a lectotype and several paralectotypes. In addition to study of the classic material, detailed revision of other previously analysed specimens, coming from several Spanish Aragonian (Orleanian + Astaracian) and Vallesian sites (e.g. those published by García & Alberdi, 1968; Jiménez Fuentes & Carbajosa Tamargo, 1982; Jiménez Fuentes et al., 1986), and study of other well-preserved new specimens (e.g. material from Barajas, Madrid), have allowed us to describe for the first time, in detail, the taxon Te. bolivari. The need for such a detailed review of this species has been noted in several papers (e.g. Jiménez Fuentes, 1971; Jiménez Fuentes & Carbajosa Tamargo, 1982). As a result, most of the anatomical osteology of this taxon is described here. Owing to the abundance of the specimens analysed here, this study provides information on the individual, ontogenetic, and sexual variability of this taxon. Thus, the synonymy between Te. bolivari and Testudo richardi (Bergounioux, 1938), proposed by Jiménez Fuentes (2000), is confirmed. New data on the stratigraphical and geographical distribution of the Spanish Te. bolivari are supplied here. Its detailed comparison with other species of large tortoises from the Neogene of Europe allows us to propose, for the first time, a diagnosis for Te. bolivari. The abundance of elements of the appendicular skeleton of Te. bolivari has allowed the detailed description and recognition of several diagnostic characters, which have not only been poorly analysed in the European material, but have also often received little attention in most studies on medium and large testudinids. This study allows the recognition of Te. bolivari, and the new genus to which it belongs, Titanochelon, as one of the better known species and genera of fossil large testudinids worldwide. INSTITUTIONAL ABBREVIATIONS ACN, Archivo del Museo Nacional de Ciencias Naturales, Madrid, Spain; BSPG, Bayerische Staatssammlung für Paläontologie und Geologie, Munich, Germany; ICP, Institut Català de Paleontologia, Sabadell, Spain; LGPUT, Laboratory of Geology and Paleontology, School of Geology, Aristotle University of Thessaloniki, Thessaloniki, Greece; MAR, Museo Arqueológico Regional, Alcalá de Henares, Madrid, Spain; MNCN, Museo Nacional de Ciencias Naturales, Madrid, Spain; MNHN, Muséum national d Histoire naturelle, Paris, France; NWS, Naturmuseum Winterthur, Switzerland; PIMUZ, Paläontologische Institut und Museum, Universität Zurich, Switzerland; SMNS, Staatliches Museum für Naturkunde Stuttgart, Germany; STUS, Sala de las Tortugas de la Universidad de Salamanca, Salamanca, Spain. OTHER ABBREVIATIONS MN, European Mammal Neogene biozones. MATERIAL AND METHODS As indicated above, all the classic material of the type species of the new genus Titanochelon, Te. bolivari, was erroneously recognized as lost, having been considered as possibly destroyed as a result of the Spanish Civil War ( ) (Jiménez Fuentes et al., 1988; Jiménez Fuentes & Martín de Jesús, 1991; Vlachos et al., 2014; Luján et al., in press). The senior author of this paper, in collaboration with the previous curator of the Collection of Vertebrate Paleontology of the MNCN (Dr. Begoña Sánchez-Chillón), began the task of locating these remains in Abundant graphical material and manuscripts, corresponding to the collections of Eduardo Hernández-Pacheco ( ) and José Royo y Gómez ( ), were found and used for the identification of that generally unpublished fossil material. The unpublished photographs corresponding to the discovery, extraction, and preparation, and to different views of the specimens prepared for exhibition, deposited in the ACN were of critical importance for the identification of the material presented here. In addition, the handwritten labels accompanying several specimens also provided relevant information. A detailed inventory of the preserved classic material is given in Appendix 1. Additionally we provide the specimen numbers and identifications of other published material that we revised during this study (e.g. Coca in Segovia) and new specimens presented here (e.g. Barajas in Madrid). For additional specimens for which direct observation was not possible (e.g. the skull from Ciudad Universitaria, in Madrid) we used the published information. Many of the appendicular elements are relatively well preserved. The currently available information allows us to know that several of the most complete shells of the classic specimens were relatively well preserved and displayed in the MNCN until the late 1980s. However, some of them have dramatically deteriorated since then, as is the case of a nearly complete and articulated shell from Cerro del Otero (Palencia), which is currently represented by fragments of poorly preserved plates, and a few axial (a caudal vertebra) and appendicular elements (partial humerus). In fact, all shells had been damaged before our work. Many of the reconstructed specimens were fractured, and mixed in boxes containing tens to hundreds of fragments corresponding to several individuals from different localities. Most of the preparation of the fragmented materials took place in May and June of 2013 by the MNCN preparators with the assistance of the authors

5 A GENUS FOR THE EUROPEAN LARGE TORTOISES 657 of this paper. Many of these specimens, especially those corresponding to shells and partial shells, lacked collection numbers. All of them were inventoried after preparation. Thus, a high percentage of the classic material of Te. bolivari that was deposited in the MNCN from 1906 to 1935 was recovered. It comes from most of the localities cited by Eduardo Hernández-Pacheco and José Royo y Gómez (see Supporting Information Appendix S1). Most of this material has remained unpublished until now, because Hernández-Pacheco planned to include its detailed study in a monograph, which he never published (see Hernández-Pacheco, 1917a, b, 1921a, 1930). We also reviewed some specimens from an area with important fossil findings known from the second half of the 20 th century: Coca, in Segovia. This is because of both their good preservation and their systematic and historical importance as one of these specimens was identified by Jiménez Fuentes (1984) as the neotype of Te. bolivari. The material found in the second half of the 20 th century in the classic locality of Arévalo was also reviewed. Indeed, Jiménez Fuentes et al. (1986) identified there the neotype of Te. richardi, a taxon recognized here as a junior synonym of Te. bolivari, as was proposed by Jiménez Fuentes (2000). In light of the specimens presented here, we also discuss the lost holotype below, providing clear evidence that confirms the proposed synonymy. Finally, an unpublished specimen recently found in Barajas (Madrid) was studied here. Its good preservation and the recovery position of the appendicular elements (most in anatomical position) improve our knowledge of the morphology of the skeleton of Te. bolivari. The study of several specimens not only allows description of the detailed anatomy of Te. bolivari but also provides data on its variability. We performed comparative studies of the Spanish material with other relevant European and African taxa. Most of the compared characters are based on direct observation of these species. In order to establish the systematic position of Titanochelon, this taxon was coded in a new data matrix in which all of the European Palaeogene testudinid species of medium size whose specific validity is supported here are included (i.e. Achilemys cassouleti, Hadrianus castrensis, Ch. maurini, Te. gigas and Ergilemys bruneti). As indicated below, all of these species were recognized only by postcranial material, generally corresponding to the shell. The only one previously incorporated in any cladistic analysis is Achilemys. It was coded by the inclusion of characters from both the type species of this taxon (the North American Achilemys allabiata) and the European A. cassouleti (see Claude & Tong, 2004). Therefore, no hypothesis about the phylogenetic relationships between the medium and large European testudinids was so far justified. Four species herein included in Titanochelon were incorporated into the analysis performed in the present study: the Miocene Te. bolivari and Te. vitodurana, and the Pliocene Te. perpiniana and Cheirogaster bacharidisi. Owing to the absence of information on the shell (in the case of Te. schafferi) and the very limited information on several taxa ( Te. ginsburgi, Testudo eurysternum, Te. steinbacheri, Te. leberonensis) either because of their extremely limited known record (e.g. Te. eurysternum) or their poor preservation (e.g. Te. leberonensis), as well as the absence of information on subadult and adult specimens ( Te. ginsburgi), no other European Neogene testudinid was incorporated into this analysis. A total of 36 characters was analysed. The list of characters can be found in Appendix 2 and the matrix in Appendix 3. All characters are compared amongst the analysed taxa in the Discussion. The analysis was performed with TNT v. 1.0 (Goloboff, Farris & Nixon, 2008) using a traditional search with a tree-bisection algorithm with replicates. All characters were considered unordered and equally weighted. Bootstrap frequencies (also with replicates) and Bremer support were also calculated using the same software. SYSTEMATIC PALAEONTOLOGY CHELONII LATREILLE, 1800 CRYPTODIRA COPE, 1868 TESTUDINOIDEA BATSCH, 1788 TESTUDINIDAE BATSCH, 1788 TESTUDININAE BATSCH, 1788 TITANOCHELON GEN. NOV. Type species: Titanochelon bolivari (Hernández-Pacheco, 1917a) Included species: Titanochelon bolivari comb. nov., Titanochelon eurysternum (Gervais, ) comb. nov., Titanochelon ginsburgi (de Broin, 1977) comb. nov., Titanochelon vitodurana (Biedermann, 1862) comb. nov., Titanochelon steinbacheri (Karl, 1996) comb. nov., Titanochelon leberonensis (Depéret, 1890) comb. nov., Titanochelon schafferi (Szalai, 1931) comb. nov., Titanochelon perpiniana (Depéret, 1885) comb. nov., Titanochelon bacharidisi (Vlachos et al., 2014) comb. nov. Probably included species: cf. Titanochelon gymnesicus (Bate, 1914) comb. nov. Diagnosis: Western Euroasiatic (from Western Europe to Minor Asia) Neogene large testudinid (shell from more than 1 m up to 2 m long) based on several

6 658 A. PÉREZ-GARCÍA AND E. VLACHOS exclusive characters such as the alternating pattern of the costals, reduced proximal region of the ribs, presence of a thickened epiplastral lip, and fused trochanters of the femur. It is diagnosed by the presence of frontals narrower than prefrontals, causing a gentle constriction of the frontals in dorsal view, and by the following exclusive character combination: long crista supraoccipitalis (not in Ce. sulcata); presence of an elliptical to circular premaxillary pit (not in Ce. sulcata); relatively low carapace (shared with Ce. sulcata, but not with Ch. maurini or Chelonoidis spp.); wide shell (not in Ch. maurini); nuchal plate as wide as long or wider than long (shared with A. cassouleti, but not with Ch. maurini or Ce. sulcata); eight neural plates, the first one being rectangular; longer than wide first neural, the second and fourth neurals octagonal, with the greatest width of the second one situated in its rear half, the third and fifth neurals rectangular, wider than long, and the sixth to eighth neurals hexagonal, the length of their lateroanterior margins being similar to that of the lateroposterior ones (this neural pattern is shared with Ergilemys bruneti, but not with A. cassouleti, Hadrianus castrensis, Ch. maurini, or Ce. sulcata); absence of pointed tips in the carapacial rim in the region of contact of the sulci between the marginal scutes and the lateral border of the anterior and posterior peripherals (shared with H. castrensis and Ch. maurini, but not with A. cassouleti); two suprapygals, the first dorso-laterally surrounding the lenticular second one (shared with Ergilemys bruneti, but not with A. cassouleti, H. castrensis, and Ch. maurini); longer than wide pygal (shared with Ce. sulcata, but not with H. castrensis or A. cassouleti); absence of cervical scute (not in A. cassouleti, H. castrensis, or Ergilemys bruneti); coincidence between the medial margins of all the marginals and the suture between the costal and peripheral plates (not in A. cassouleti, H. castrensis, or Ch. maurini); significant distance between the second pair of marginals and the nuchal plate; fourth to sixth marginals in contact with the second pleural scute (shared with A. cassouleti, H. castrensis, and Ce. sulcata, but not with Ch. maurini); presence of a supracaudal scute (not in A. cassouleti or H. castrensis); second suprapygal crossed by the vertebrosupracaudal sulcus (shared with Ch. maurini, but not with A. cassouleti or H. castrensis); absence of plastral hinges (shared with A. cassouleti, H. castrensis, Ch. maurini, and Ce. sulcata, but not with Ergilemys bruneti); wide plastral lobes, at least in the region of contact with the plastral bridge (not in Ch. maurini); anterior plastral lobe longer than the posterior lobe (not in Ch. maurini); convex to flat dorsal lip of the epiplastra (not in A. cassouleti, H. castrensis, or Ce. sulcata); absence of ventral relief in the area covered by the gular scutes (present in H. castrensis, but not in Ergilemys bruneti); angle between the sagittal axis and the gular humeral sulcus of 45 or more acute (shared with Ce. sulcata, but not with A. cassouleti, H. castrensis, or Ch. maurini); humeropectoral sulcus perpendicular to the axial plane in the medial region, but with a well-developed lateral change of curvature (shared with Ce. sulcata, but not with A. cassouleti, H. castrensis, or Ch. maurini); very short pectoral scutes in the sagittal plane, but markedly expanded, towards the anterior region, laterally (not in A. cassouleti, H. castrensis, or Ch. maurini); presence of a single pair of inguinal scutes (not in Manouria emys); males with a well-developed ventral thickening in the area covered by the anal scutes (not shared with Ce. sulcata); curved diaphysis of the humerus and femur (not in A. cassouleti); fused astragalus and calcaneum (not shared with Chelonoidis spp., but shared with Ce. sulcata). Etymology: Titano-, from the singular form of the ancient Greek word (titan), referring to the large size of this taxon; -chelon from the ancient Greek word (cheloni), which means turtle or tortoise. Distribution: Neogene (early Miocene to early Pleistocene) of Europe (Portugal, Spain, France, Switzerland, Germany, Austria, Greece, Bulgaria), European Turkey and Minor Asia. This taxon may also be present in the Pleistocene deposits of some Mediterranean Islands (Baleares Islands and Malta). TITANOCHELON BOLIVARI (HERNÁNDEZ-PACHECO, 1917A) COMB. NOV. (FIGS 2 22, 23B E, 24 27) Testudo elecphantopus Vidal y Careta, 1906 Testudo sp. Fernández Navarro & Carandell Pericay, 1914; Hernández-Pacheco, 1917b Testudo aff. perpiniana Hernández-Pacheco, 1914 Testudo sp. perpiniana in size Hernández-Pacheco & Dantín Cereceda, 1915 Testudo bolivari Hernández-Pacheco, 1917a, 1921a, b, c, 1926, 1930; Royo y Gómez, 1921, 1922, 1926, 1928a, b, 1929, 1933, 1934a, b, 1935a, b; Royo y Gómez & Menéndez Puget, 1928, 1929; Bergounioux, 1938, 1957, 1958; Peyer, 1942; Bataller, 1956; Crusafont Pairó, Aguirre & García, 1968; García & Alberdi, 1968; Jiménez Fuentes, 1971, 1978a Testudo richardi Bergounioux, 1938, 1957, 1958; Bataller, 1956 Testudo bolivari Jiménez Fuentes, 1978b; Jiménez & García Marcos, 1978 Ergilemys bolivari Chkhikvadze, 1972 Ergilemys richardi Chkhikvadze, 1972 Geochelone (Geochelone) bolivari Auffenberg, 1974 Geochelone (Geochelone) richardi Auffenberg, 1974 Geochelone (Ergilemys) bolivari Mlynarski, 1976 Geochelone (Ergilemys) richardi Mlynarski, 1976 Geochelone s.l. bolivari de Broin, 1977 Geochelone s.l. richardi de Broin, 1977

7 A GENUS FOR THE EUROPEAN LARGE TORTOISES 659 Geochelone (Ergilemys) bolivari Mlynarski, 1980 Geochelone (Ergilemys) richardi Mlynarski, 1980 Geochelone bolivari Alberdi et al., 1981; Jiménez Fuentes, 1982, 1984, 1985b; Jiménez Fuentes & Carbajosa Tamargo, 1982; Cuesta et al., 1983 Geochelone bolivari Jiménez Fuentes, 1981 Cheirogaster richardi Bourgat & Bour, 1983; Bour, 1985; Jiménez Fuentes, 2000; de Lapparent de Broin, 2002; Luján et al., in press. Cheirogaster bolivari Bourgat & Bour, 1983; Jiménez Fuentes, 2000, 2003; Herráez et al., 2006; de Lapparent de Broin, 2002; Pérez-García, 2013; Luján et al., in press. Cheirogaster bolivari Jiménez Fuentes et al., 1984 Cheirogaster richardi Jiménez Fuentes et al., 1984 Geochelone richardi Jiménez Fuentes, 1984, 1985b Geochelone (s.l.) bolivari Jiménez Fuentes, 1985a Geochelone (s.l.) richardi Jiménez Fuentes, 1985a Cheirogaster (s.l.) richardi Jiménez Fuentes, 1985b, 1986a, b, 1988; Jiménez Fuentes et al., 1986, 1988 Cheirogaster (s.l.) bolivari Jiménez Fuentes et al., 1986, 1988; Jiménez Fuentes, 1988; Jiménez Fuentes & Montoya, 2002 Centrochelys bolivari Chkhikvadze, 1989a, b Centrochelys richardi Chkhikvadze, 1989a, b?cheirogaster bolivari Jiménez Fuentes & Martín de Jesús, 1991; Jiménez Fuentes, 1992, 1994; Jiménez Fuentes et al., 1993?Cheirogaster richardi Jiménez Fuentes & Martín de Jesús, 1991; Jiménez Fuentes, 1992, 1994; Jiménez Fuentes et al., 1993 Cheirogaster bolivari Vlachos et al., 2014 Lectotype: The specimen Alcalá 2, a partial skeleton including most of its shell and numerous appendicular elements, found in 1917 (see Appendix 1 and Figs 2 4). Paralectotypes: Numerous appendicular elements and some caudal vertebrae, found in 1917 in the type locality and horizon (see Appendix 1 and Figs 5 8). Other specimens attributed to this taxon: The classic material from several localities (Los Santos de la Humosa, Vallecas, Puente de la Princesa, Puente de los Franceses, Parla, Carabanchel, and Ciudad Universitaria, all of them in Madrid; Cerro del Otero and Villalcón in Palencia; Arévalo in Ávila; Fuensaldaña and La Cistérniga in Valladolid; Villaluenga de la Sagra in Toledo), as well as numerous specimens from other Spanish regions (e.g. lots of sites in Madrid such as Barajas; Coca in Segovia; and the Vallès Penedès Basin in Catalonia) (see Appendix 1 and Figs 9 22, 23B E, 24 27). Diagnosis: Member of Titanochelon differing from the other species of this genus because of the following features: presence of medially short and laterally long prefrontals, with medially long and laterally short frontals; crista supraoccipitalis elevated upwards; omegashaped femoro-anal sulcus and deep anal notch of the females; males with trapezoidal posterior lobe, showing a wide and very shallow anal notch, and with a pair of lateroposterior protrusions, well developed in adult specimens; asymmetric shape of the distal end of humerus; elongated and narrow glenoid fossa of the scapula; long neck of the coracoid; wide and short femoral head with a well-developed neck; narrow and elongated fused astragalocalcaneum. It further differs from all other species of Titanochelon in the following exclusive combination of characters: thin zygomatic arch; very shallow nuchal notch; as wide as long nuchal; first neural lacking a dorsal bump; absence of subhorizontal anterior peripherals; highly variable morphology of the anterior plastral lobe (e.g. subrounded, trilobed, trapezoidal); absent or poorly developed gular pocket; gulars covering the anterior part of the entoplastron; narrow entoplastron in relation to the width of the anterior plastral lobe; posterior border of the entoplastron located anterior to the humeropectoral sulcus or in contact with it; low angle (< 120 ) between the scapula and the acromion; medial articular surface of the intermedium contacting the radius; femoral head at an angle to the diaphysis. Etymology: In honour of Ignacio Bolivar y Urrutia, director of the MNCN and discoverer of one of the first specimens corresponding to this species (after Hernández-Pacheco, 1917a). Type locality and horizon: Cerro del Viso, in Alcalá de Henares (Madrid, Spain). MN6 zone (late Aragonian, Serravallian) of the Intermediate Unit of the Madrid Basin, belonging to the Tagus Basin (Peláez-Campomanes et al., 2003; Sesé, 2006; Eronen et al., 2010; Domingo et al., 2012). Distribution: Aragonian to early Vallesian (Langhian to early Tortonian, MN5 to MN10) of Spain. Remarks on the type material of Titanochelon bolivari: As indicated, the taxonomic status and history of the material of Te. bolivari have been problematic. Considering Articles 75.1 and of the International Code of Zoological Nomenclature, the finding of the syntypes of Te. bolivari allows us to refute the validity of the specimen STUS 352, from Coca, as the neotype of this species (see Jiménez Fuentes et al., 1988). We follow one of the introduced changes in the current edition (accessed online in March 2014; iczn/index.jsp) of the International Code of Zoological Nomenclature (change number 7): If a previously lost holotype, syntype or lectotype of a species subsequently typified by a neotype is rediscovered, the original type specimen(s) will automatically displace the neotype and become the name-bearing type. Taking into account Article 73, the specimens from the Meaques stream (Casa de Campo, Madrid), Vallecas (Madrid), Los Santos de la Humosa (Madrid), Cerro del Viso (Alcalá de Henares, Madrid), Cerro del Otero (Palencia), and Fuensaldana (Valladolid), which were found prior to the date of publication of the paper in

8 660 A. PÉREZ-GARCÍA AND E. VLACHOS which Hernández-Pacheco (1917a) defined Te. bolivari, and of which this author was aware, are identified here as syntypes. One of them, the specimen retaining the greater number of bone elements, and in addition, that was figured (see Royo y Gómez, 1928b: plates XXXIV, XXXV; Royo y Gómez, 1935a: fig. 1), is here designated as the lectotype (see Figs 2 4; Appendix 1). This specimen came from the MN6 zone (late Aragonian, middle Miocene) of Cerro del Viso, in Alcalá de Henares (Madrid, Spain). Therefore, the other syntypes that also came from this horizon and locality are recognized as paralectotypes (see Figs 5 8; Appendix 1). The type locality could be identified as the area grouping Madrid, Palencia, and Valladolid, the type horizon that groups the MN5, MN6, and MN7/8 zones, because the syntypes came from various locations and horizons. However, considering Article 73, as well as Article 76, the designation of a lectotype implies that the type locality and horizon should be more precise, corresponding to those in which it has been found: here, the MN6 zone of Cerro del Viso, in Alcalá de Henares (Madrid). Remarks on the geological and geographical context of the classic material of Titanochelon bolivari: The largest part of the outcrops in the Madrid Basin corresponds to Miocene sediments (Fig. 1A). These continental sediments have been divided into three depositional units in this basin: the Lower Unit (Ramblian middle Aragonian), the Intermediate Unit (middle Aragonian Vallesian), and the Upper Unit (Vallesian Turolian). The classic material and most of the recent discoveries of large tortoises from the Madrid Basin come from various localities in the middle and late Aragonian levels (middle Miocene) of the Lower and Intermediate Units (Fig. 1). Most of the information on the chronology of these localities found in older publications is now outdated. Therefore, here we provide updated information on this matter by grouping the various localities into six (I VI) age groups (Fig. 1B). Based also on information from the mammalian fossils of the accompanying fauna, some sites, such as those corresponding to the Meaques stream (Casa de Campo), Puente de los Franceses, and those in the Barajas area, are situated in the MN5 zone (middle Aragonian) of the Lower Unit (Fig 1B, group I). The Vallecas site is situated in the Intermediate Unit, also in the MN5 zone. Cerro del Viso (Alcalá de Henares), Henares 1 (Los Santos de la Humosa), and Parla are in the MN6 zone (late Aragonian) of the Intermediate Unit (Peláez-Campomanes et al., 2003; Domingo et al., 2012: Fig. 1B, group II). Other outcrops that have provided classic material of large tortoises in the Madrid Basin (e.g. those of Puente de la Princesa and Ciudad Universitaria) also come from Aragonian levels of the MN5 MN6 interval (Fig. 1B, group III). In addition to the specimens from the Madrid Basin, Hernández-Pacheco (1917a) attributed to Te. bolivari other Aragonian specimens from two localities in the Duero Basin: Cerro del Otero (Palencia) and Fuensaldaña (Valladolid). They came from the levels currently identified as corresponding to the late Aragonian (MN7/8) (Mazo & Jordá Pardo, 1994; Sánchez, Salesa & Morales, 1998; Mazo & Montoya, 2003). Moreover, Hernández-Pacheco (1914) recognized the presence of Testudo sp. in La Cistérniga (Valladolid). This material also comes from the late Aragonian (MN7/8) (Sánchez et al., 1998). All of these localities are chronologically grouped together in group IV (Fig. 1B). Royo y Gómez (1935a) cited the presence of some remains belonging to this group of tortoises in a town in the Tagus Basin: Villaluenga de la Sagra (Toledo). These levels are currently recognized as belonging to the late Aragonian (Beamud et al., 2006) (Fig. 1B, group IV). Although the classic material does not include specimens from Coca (Segovia, Duero basin), Jiménez Fuentes (1984) defined one of the specimens identified in that locality from the late Aragonian (MN6) (Cuesta et al., 1983; Scherler et al., 2013) as the neotype of Cheirogaster bolivari. This was motivated because of the erroneous consideration of the specimens from the classic sites as destroyed. Royo y Gómez (1935a) considered that all of the specimens of the large tortoises from the Tagus and Duero Basins had the same geological age. However, the material from one of the regions that he analysed actually came from the early Vallesian, probably from the MN9 zone (Cuesta et al., 1983; Jiménez Fuentes et al., 1993). These are the specimens from Arévalo (Ávila, Duero Basin; Fig. 1B, group V). New material was found in that locality from the 1960s onwards, allowing Jiménez Fuentes et al. (1986) to define a neotype for Cheirogaster richardi (Bergounioux, 1938). This species had been described based on a specimen from the early Vallesian (MN9 sensu Jiménez Fuentes & Martín de Jesús, 1991) of Els Hostalets de Pierola, in the Vallès-Penedès Basin (Barcelona, Catalonia; Fig. 1B, group V). The presence of specimens of this group of tortoises in the late Vallesian levels (MN10 zone, late Miocene) of the Intermediate Unit of the Madrid Basin, at the site of Cerro de los Batallones, has recently been documented (Pérez-García & Murelaga, 2013; Fig. 1B, group VI). The number of Spanish sites with remains of large tortoises has dramatically increased since the second half of the 20 th century. They are located in both the above-mentioned basins (i.e. Tagus, Duero, and Vallès- Penedès Basins) and in other regions (e.g. Ebro, Murcia- Carrascoy, and Campo de Calatrava basins) (Jiménez Fuentes, 1984; Mancheño Jiménez et al., 2001; Murelaga, Azanza & Astibia, 2006; Pérez-García et al., 2011). This has also expanded the range of stratigraphical distri-

9 A GENUS FOR THE EUROPEAN LARGE TORTOISES 661 bution of this group of tortoises in the Spanish record, the indeterminate material from Murcia-Carrascoy Basin corresponding to levels near the limit between the Miocene and the Pliocene, and the indeterminate material from Campo de Calatrava Basin being from the Pliocene (Mazo, 1996; Mancheño et al., 2013). DESCRIPTION OF THE TYPE SERIES OF THE TYPE SPECIES OF TITANOCHELON AND OF OTHER CLASSIC AND SUBSEQUENTLY FOUND RELEVANT SPECIMENS All the classic fossils attributed to Te. bolivari lack any proper description or explanations justifying their inclusion in this taxon (for a detailed historical overview see Appendix S1). Furthermore, little work has been carried out to document its morphological variation, despite the abundance of individuals. Considering the classic papers, the only exception is that of Royo y Gómez (1935a), who described the variation in the shape of the anterior lobe of the plastron, erroneously recognized by Bergounioux (1938) as representing two different species (see Discussion). Finally, as indicated above, there was no indication of a holotype or type locality of this taxon. Therefore, the descriptions of the classic material presented here are given per locality. The information below is given for the first time here. Although more extensive, this arrangement allows the proper documentation of the morphological variation within the recovered material and points out the characters that may properly define Titanochelon bolivari. Besides the recovered material of the classic localities (Alcalá de Henares, Vallecas, Puente de la Princesa, Arévalo, amongst others), we present here revised material that was found in the second half of the 20 th century (e.g. Coca, including the specimen previously proposed as the neotype of Cheirogaster bolivari) and material from a new, relevant site (i.e. Barajas). This allowed us to perform the much-needed revision of the problematic Te. bolivari, and to discuss the status of the other named taxon from the Miocene of Spain, Te. richardi. THE MATERIAL FROM THE TYPE LOCALITY: ALCALÁ DE HENARES (MADRID) The locality of Alcalá de Henares is probably the most important amongst those initially mentioned by Hernández-Pacheco (1917a), as a high percentage of the material of Te. bolivari known at that time came from there, and also because of the preservation of some of the specimens, including relatively complete postcranial skeletons. In fact, this is the classic locality; we have identified a larger number of specimens preserved today, all of them deposited in the MNCN (see Appendix 1). For these reasons, we recognize it here as the type locality of the type species of Titanochelon. All of these specimens were collected during the surveys and excavations carried out in 1917 under the supervision of Eduardo Hernández-Pacheco, in which José Royo y Gómez participated (see Appendix S1). As indicated, very little material from this locality has been hitherto described or figured. In fact, all appendicular bones remain unpublished and very little information on the shell is currently available. The preserved material consists of an individual that retains many elements of its postcranial skeleton (identified by Hernández-Pacheco as Alcalá 2, as indicated on the handwritten labels glued to several of its elements), recognized here as the lectotype of Ti. bolivari (Figs 2 4), and several appendicular elements of another individual (identified as Alcalá 3 ) (Fig. 5), as well as numerous bones corresponding to various individuals (Figs 6 8), all of them constituting the paralectotypes. The material from Alcalá de Henares allows the description of the morphology of most of the shell and of the other elements of the postcranial skeleton of Titanochelon bolivari. DESCRIPTION OF THE SPECIMEN ALCALÁ 2, LECTOTYPE OF TITANOCHELON BOLIVARI Analysed material The specimen Alcalá 2 (see Figs 2 4) corresponds to the one of about 125 cm in length cited by Hernández-Pacheco (1917a, 1917b), recognized by him as the largest of those found in 1917 in Alcalá de Henares. In fact, the maximum length of the preserved region of the plastron is about 110 cm. In the past, its disjointed and relatively complete carapace and its plastron were mounted for display at the MNCN, using an iron support (Fig. 2). We identified and reconstructed many of the elements of the carapace, as well as most of the plastron (Fig. 3). Information on this carapace was used, together with that coming from several specimens found on Cerro del Otero, for the drawing of the carapace of Te. bolivari published by Royo y Gómez (1935a) and shown here in Figure 1. Compared with the previously mounted shell (Fig. 2), the neural series was recovered almost entirely complete. A few parts of the costals are missing and only a few parts of the peripheral region could be located. Our new reconstruction (Fig. 3) differs in some respects in the placement of some of the plates, but overall is in good accordance with the reconstruction published by Royo y Gómez (1935a). As indicated by Hernández-Pacheco (1917b), abundant postcranial elements of this large specimen are also preserved. We have identified many of these bones (see Appendix 1 and Fig. 4).

10 662 A. PÉREZ-GARCÍA AND E. VLACHOS Figure 2. Historical photographs of the shell of the lectotype of Titanochelon bolivari ( Alcalá 2 ), from the MN6 zone (late Aragonian, middle Miocene) of Cerro del Viso, in Alcalá de Henares (Madrid, Spain). Specimen collected in A, right lateral view. Modified version of ACN 6553, photograph corresponding to plate XXXIV in Royo y Gómez (1928b). B, left lateral view. Modified version of the unpublished photograph ACN C, anteroventral view. Modified version of the unpublished photograph ACN D, dorsoposterior view. Modified version of ACN 6556, photograph corresponding to plate XXXV in Royo y Gómez (1928b). Original photographs taken by Eduardo Hernández-Pacheco. Shell remains The only elements of shells from Alcalá de Henares recognized by us in the current collection of the MNCN are those corresponding to the specimen Alcalá 2 (Figs 2, 3). This shell was relatively low. Four neurals are preserved. One of them, identified as the second neural, is octagonal. This plate is wider than long. Its maximum width is situated in its rear half. Two of the preserved neurals are subrectangular, wider than long. A sulcus delimiting two vertebral scutes is situated on one of them. The presence of a sulcus on the other subrectangular neural cannot be ruled out because of its preservation. The other preserved neural is the largest. Owing to this fact, and to its octagonal morphology, it is recognized as the fourth neural. Neither of these octagonal neurals is crossed by any sulcus. Although the first neural is missing, the long medial margin of the first costal plate, and the narrow anterior margin of the second neural, allow us to interpret that this subrectangular plate was noticeably longer than wide. Several costal plates are preserved, mostly fragmented. On the ventral side, all neurals preserve remains of the attachments of the dorsal vertebrae. Medially, the first pair of costals is the longest. The costals show the typical alternating pattern of the testudinids, having costals that are medially short and laterally long alternated with costals that are medially long and laterally short. Remains of the two suprapygals have been preserved. The first suprapygal embraces the second one, which is lenticular. The distal margins of the most anterior peripherals are dorsally directed. The posterior peripherals are also dorsally directed, but not the pygal. This character indicates that Alcalá 2 corresponds to a male. The sulci situated between the pleural and the marginal scutes are situated in the same regions as the sutures between the costal and the peripheral plates. The presence of a concavity in the middle posterior region of the plastron confirms that the analysed specimen is an adult male. Furthermore, this specimen has a wide and very shallow anal notch, the posterior margin of the xiphiplastra being subperpendicular to the axial plane. Other characters allow the confirmation of its sex and ontogenetic stage: the presence of a pair of bumps on the lateroposterior margins of the xiphiplastra and the presence of a well-developed ventral thickening in the area covered by the anal scutes. The morphology of the anterior plastral lobe is not known. The relatively thick dorsal epiplastral lip is wide and long, with a convex surface. It has a very poorly developed gular pocket. The entoplastron is subhexagonal, the anterior angle being much more acute than the posterior. Compared with the maximum width of the anterior lobe, the entoplastron is narrow, its maximum width being less than half of the maximum width of the lobe. The anterior region of the entoplastron is covered by the gular scutes. The posterior margin of the entoplastron ends anterior to the humero-pectoral sulcus. The hyoplastra are covered by the humeropectoral sulcus and the straight pectoro-abdominal sulcus, which are located very close to each other in the medial part. The humero-pectoral sulcus is perpendicular to the axial plane in the medial region, but it has a welldeveloped lateral change of curvature, being markedly directed towards the anterior region. The pectoroabdominal sulcus is straight. For this reason, the pectoral scutes are very short, the medial length of the humerals being about eight times greater than that of

11 A GENUS FOR THE EUROPEAN LARGE TORTOISES 663 Figure 3. The recovered shell elements of the lectotype of Titanochelon bolivari ( Alcalá 2 ) from the MN6 zone (late Aragonian, middle Miocene) of Cerro del Viso, in Alcalá de Henares (Madrid, Spain). Specimen collected in A, B, dorsal view of the carapace (MNCN 32773). C, D, ventral view of the plastron (MNCN 32771). Scale bar = 10 cm. The drawing of the carapace is modified from figure 1 of Royo y Gómez (1935a), which was based on this specimen and others from Cerro del Otero (Palencia). the pair of scutes. The hypoplastra are covered by the femorals in the posterior part. The anals cover the posterior half of the xiphiplastra. This specimen lacks a plastral hinge. Appendicular skeleton The material described in this section corresponds to that labelled by Hernández-Pacheco as Alcalá 2. It comprises non-overlapping appendicular elements, of the same ontogenetic stage, whose size indicates one individual, and all of a similar colour and fossilization type. Therefore, we can safely assume that they belong to the same individual. All the available historical information on the shell of the Alcalá 2 individual suggests the presence of appendicular elements. Therefore, based on all available information, we can safely attribute the following elements as belonging to the Alcalá 2 individual.

12 664 A. PÉREZ-GARCÍA AND E. VLACHOS Shoulder girdle Most of the shoulder girdle of the Alcalá 2 specimen is preserved; only a few parts of the scapulae and coracoids are missing (Fig. 4A D). The shoulder girdle has the typical triradiate shape of the testudinids. The scapula is long, with an elliptical-shaped crosssection (Fig. 4A, D). It is straight and is developed perpendicular to the acromion. The acromion is also straight and short. The glenoid fossa is narrow, elongated, and shallow. The angle between the scapula and the acromion is about 120. The coracoid has a very long and narrow neck and a wide, fan-shaped end (Fig. 4B, C). The acromion has a small contribution to the formation of the articular surface for the humerus. Anterior limb Both humeri from the Alcalá 2 individual are preserved complete (Fig. 4E, F). The proximal and distal ends are wide. The diaphysis shows a significant sigmoid curvature. The major trochanter is much longer than the minor, extending beyond the height of the humeral head, whereas the minor ends before the top of the humeral head. The trochanters are developed almost parallel to each other. The head is spherical in shape. The distal part of the humerus shows a weak ectepicondylar groove in dorsal view, whereas the distal end of the bone is very asymmetric because the ulnar condyle is larger than the radial. No radius is preserved but both ulnas were identified (Fig. 4G, H), the left one being complete. The lateral border of the right ulna is missing. The ulna is high and narrow. The proximal part is very narrow and with a slightly concave articular surface. The mid-shaft is triangular in crosssection. The distal part has a narrow articular surface for the radius. Several carpal elements are preserved from this individual. The ulnare (Fig. 4I) is small and narrow. Proximally, it contacts the ulna with an elongated concave surface that is slightly expanded medially and elongated laterally. Medially, it makes contact with the intermedium with two small articular surfaces. Distally, it has two slightly concave surfaces (with an obtuse angle between them) to articulate with the lateral centrale and with the fifth distal carpal. Both medial centrale are preserved and are not fused with the lateral centrale (Fig. 4J, K). They make contact with a straight symphyseal area medially. The medial centrale is thicker medially and becomes much thinner laterally. The proximal articular surface for the radius is flat to slightly concave. Dorsomedially there is a small, flat, articular surface for the intermedium. The distal articular surface for the distal carpals is slightly concave medially and slightly convex laterally. The left medial centrale is slightly longer than the right one. The distal carpal (Fig. 4L) shows the typical morphology, with a protuberance proximally and a discoid part distally. Ventrally there is a convex surface for the metacarpals. Several metacarpals have been preserved. They show the typical morphology, having a slightly convex surface proximally for the distal carpals and a distal convex surface for the first phalanx. The phalanges are long and narrow (Fig. 4M), having a concave proximal surface for the metacarpals and a convex distal surface for the unguals. The unguals are all straight in shape and wide (Fig. 4N) and therefore are from the anterior limb. They show the typical morphology, with a concave proximal surface for the phalanx. Pelvic girdle From the pelvis only the right ilium (Fig. 4P) is preserved, being straight in shape. The distal part is wide, containing an elongated concave articular surface for the femoral head. Posterior limb From the posterior limb, only a right femur (Fig. 4Q) is preserved. The proximal part of the trochanters is Figure 4. Selection of some elements of the appendicular skeleton of the lectotype of Titanochelon bolivari ( Alcalá 2 ), from the MN6 zone (late Aragonian, middle Miocene) of Cerro del Viso, in Alcalá de Henares (Madrid, Spain). Specimen collected in A, right scapula (MNCN 50229) in anterior (A1) and ventrolateral (A2) views. B, right coracoid (MNCN 50225), in dorsal (B1) and ventral (B2) views. C, left coracoid (MNCN 50212) in dorsal (C1) and ventral (C2) views. D, left scapula (MNCN 50240) in posterior (D1), anterior (D2), and ventrolateral (D3) views. E, right humerus (MNCN 50266) in posterior (E1), dorsal (E2), anterior (E3), and ventral (E4) views. F, left humerus (MNCN 50213) in posterior (F1), dorsal (F2), anterior (F3), and ventral (F4) views. G, left ulna (MNCN 50232) in dorsal (G1), lateral (G2), ventral (G3), and medial (G4) views. H, right ulna (MNCN 50226) in medial (H1), ventral (H2), lateral (H3), and dorsal (H4) views. I, right ulnare (MNCN 50256) in dorsal (I1), lateral (I2), ventral (I3), and medial (I4) views. J, right medial centrale (MNCN 50257) in dorsal (J1), proximal (J2), ventral (J3), distal (J4), and medial (J5) views. K, left medial centrale (MNCN 50258) in dorsal (K1), proximal (K2), ventral (K3), distal (K4), and medial (K5) views. L, distal carpal (MNCN 50242) in ventral (L1), medial (L2), dorsal (L3), and lateral (L4) views. M, phalanx (MNCN 50249) in dorsal (M1), medial (M2), lateral (M3), and ventral (M4) views. N, ungual (MNCN 50217) in dorsal (N1), lateral (N2), and ventral (N3) views. O, osteoderm (MNCN 50236) in dorsal (O1) and ventral (O2) views. P, right ilium (MNCN 50231) in posterior (P1), anterior (P2), medial (P3), and ventral (P4) views. Q, right femur (MNCN 50228) in dorsal (Q1), anterior (Q2), ventral (Q3), and posterior (Q4) views. Scale bars= 5 cm.

13 A GENUS FOR THE EUROPEAN LARGE TORTOISES 665

14 666 A. PÉREZ-GARCÍA AND E. VLACHOS missing. The femoral head is elliptical and developed perpendicular to the diaphysis, and shows a welldeveloped neck. The diaphysis is almost straight throughout its length. Only in the distal part is a curvature observed. The tibial condyle is more developed than the smaller fibular. These elements are connected to each other by a ridge. Behind the ridge, a shallow and elongated concavity is formed. Osteoderms Several osteoderms are preserved, being elliptical in shape, rather flat, and without any pointed ends (Fig. 4O). DESCRIPTION OF THE SPECIMEN ALCALÁ 3, PARALECTOTYPE OF TITANOCHELON BOLIVARI Analysed material All elements recognized as belonging to the specimen Alcalá 3 are appendicular bones. Eight appendicular elements are labelled as Alcalá 3 (Fig. 5). As these bones do not present overlapping parts on the same skeleton and they are similar overall, we support their attribution to the same individual. Shoulder girdle The right scapula (Fig. 5A) is preserved almost complete. The scapular process and the acromion are elliptical in cross-section and developed perpendicular to each other. The scapula shows a small degree of curvature. The articular surface for the humerus is shallowly concave and with an elongated shape. The left acromion is also preserved (Fig. 5B) and is similar to the right acromion of this individual. Anterior limb Few elements of the anterior limb have been preserved. The left humerus (Fig. 5C) is almost complete, missing only the proximal part. The shaft of the bone shows a significant curvature. The distal part is wider, and has an asymmetric shape as a result of the size of the condyles. A humeral head is also preserved disarticulated, but is not possible to identify whether it is from the left or right humerus. It has a spherical shape. Both radii are preserved. The proximal part is broad and elliptical in cross-section (Fig. 5D). The shaft of the radius shows an almost circular crosssection and is not entirely straight, showing a small curvature in the distal part. The distal part is very wide and shows a relatively flat surface for the articulation of the carpals. DESCRIPTION OF THE REMAINING PARALECTOTYPES OF TITANOCHELON BOLIVARI Analysed material Other specimens from Alcalá de Henares, also collected in 1917 (see Appendix S1), are deposited in the MNCN. Most of them correspond to appendicular elements (see Appendix 1). It cannot be ruled out that some of these bones may belong to the above-described individuals but have lost their original labels. The currently disjointed elements analysed here correspond to at least three different individuals of large tortoises, as indicated by the astragalocalcanei. Figure 5. Some elements of the appendicular skeleton of an individual corresponding to a paralectotype of Titanochelon bolivari ( Alcalá 3 ) from the MN6 zone (late Aragonian, middle Miocene) of Cerro del Viso, in Alcalá de Henares (Madrid, Spain). Specimen collected in A, right scapula (MNCN 50259) in anterior (A1), posterior (A2), and ventrolateral (A3) views. B, left acromion process (MNCN 50233) in posterior (B1) and anterior (B2) views. C, left humerus (MNCN 50230) in dorsal (C1), anterior (C2), ventral (C3), and posterior (C4) views. D, right radius (MNCN 50211) in ventral (D1), lateral (D2), dorsal (D3), and medial (D4) views. Scale bar = 5 cm.

15 A GENUS FOR THE EUROPEAN LARGE TORTOISES 667 Vertebral column Vertebral remains are scarce in the material from Alcalá de Henares, represented only by two caudal vertebrae (Fig. 6A, B). They are both procoelous, having short centra. In one specimen (Fig. 6B) small transversal processes have been preserved. Shoulder girdle Only one specimen from the shoulder girdle is preserved from Alcalá de Henares (MNCN 50275), but it is only a fragment of the scapula, not allowing further description. Humerus Few humeri remains have been found, including an incomplete right humerus (MNCN 50273, Fig. 6C) and two fragments (MNCN 50270, humeral head; MNCN 50272, humerus diaphysis). The head is spherical in shape. The diaphysis shows a sigmoid curvature. The distal part of the humerus shows a shallow ectepicondylar groove in dorsal view, whereas the distal end of the bone is very asymmetric, with the ulnar condyle much larger than the radial. Radius and ulna Several radius and ulna remains have been identified, allowing observations to be made on their articulation. The proximal part of the radius is broad and elliptical in cross-section (Fig. 6E, F). The shaft of the radius shows an almost circular cross-section and is not entirely straight, showing a small curvature in the distal part. The distal part is very wide and shows a relatively flat surface for the articulation of the carpals. The ulna is high and narrow (Fig. 6D). The proximal part is very narrow and with a slightly concave articular surface. The mid-shaft is triangular in crosssection. The distal part has a short articular surface for the radius and the remaining part is asymmetric. When articulated, the radius and ulna are not at the same level, and as a result the proximal part is asymmetric, in order to articulate with the distal part of the humerus. Medially, they articulate with a very short surface, with a long and narrow fenestra between them. Distally, they articulate in an asymmetric way as well, leaving lots of space for the intermedium distally. Intermedium The intermedium (MNCN 50005, Fig. 7A) is a carpal element with a triangular cross-section, significantly different from the remaining carpals. Proximally, it shows a concave articular surface for the ulna. Medially, it shows a slightly concave surface for the radius. Distally, it has a slightly convex surface for the medial centrale, and a concave surface for the lateral centrale. MNCN exhibits a similar morphology but is much thinner than MNCN Ulnare MNCN (Fig. 7B) is thicker and much more robust than the other two preserved specimens (MNCN and MNCN 50008), which are also smaller. Proximally, the ulnare makes contact with the ulna with an elongated concave surface that is slightly expanded medially and elongated laterally. Medially, it makes contact with the intermedium with two small articular surfaces. Distally, it has two slightly concave surfaces (with an obtuse angle between them) to articulate with the lateral centrale and with the fifth distal carpal. Lateral centrale The lateral centrale is a rounded element with a concave proximal articular surface for the intermedium and a slightly convex distal articular surface for the distal carpal (Fig. 7C). It makes contact with the medial centrale with a vertical symphyseal area. Laterally, it is rounded. All specimens (MNCN 50037, MNCN 50038, MNCN 50039) are fairly similar to each other. Medial centrale In all specimens from Alcalá de Henares that we examined, the medial centrale is separated from the lateral centrale. MNCN (Fig. 7D) is much larger, more robust, and more elongated compared with the other two specimens. MNCN is not complete (Fig. 7E). The medial centrale contacts the lateral centrale with a vertical symphysis medially. It is thicker medially and becomes much thinner laterally. The proximal articular surface for the radius is flat to slightly concave. Dorsomedially, there is a small flat articular surface for the intermedium. The distal articular surface for the distal carpals is slightly concave medially and slightly convex laterally. Distal carpals Numerous distal carpals have been preserved from Alcalá de Henares (Fig. 7F). Most of the preserved elements are most probably from the right manus, but they are too few in number to allow confident identification. MNCN is not complete; MNCN is much smaller judging from the remaining distal carpals. Generally they show the typical morphology, with a protuberance proximally and a discoid part distally. The distal discoid part is highly asymmetric with a longer medial and much shorter lateral part. Ventrally, there is a convex surface for the metacarpals. However, we observed some variation in the shape of the distal carpals. In MNCN the dorsal surface is flat to convex because the parts of this element are connected in such a way that there is a higher angle

16 668 A. PÉREZ-GARCÍA AND E. VLACHOS

17 A GENUS FOR THE EUROPEAN LARGE TORTOISES 669 Figure 6. Caudal and appendicular elements corresponding to several individuals identified as paralectotypes of Titanochelon bolivari from the MN6 zone (late Aragonian, middle Miocene) of Cerro del Viso, in Alcalá de Henares (Madrid, Spain). Specimens collected in A, caudal vertebra (MNCN 47749) in anterior (A1), posterior (A2), dorsal (A3), ventral (A4), right lateral (A5), and left lateral (A6) views. B, caudal vertebra (MNCN 50286) in anterior (B1), posterior (B2), dorsal (B3), ventral (B4), right lateral (B5), and left lateral (B6) views. C, right humerus (MNCN 50273) in anterior (C1), dorsal (C2), ventral (C3), and posterior (C4) views. D, right ulna (MNCN 50271) in dorsal (D1), lateral (D2), ventral (D3), and medial (D4) views. E, left radius (MNCN 50277) in ventral (E1), lateral (E2), medial (E3), and dorsal (E4) views. F, right radius (MNCN 50278) in ventral (F1), lateral (F2), medial (F3), and dorsal (F4) views. G, right femur (MNCN 50268) in anterior (G1), dorsal (G2), posterior (G3), and ventral (G4) views. H, right tibia (MNCN 50276) in dorsal (H1) and ventral (H2) views. I, left tibia (MNCN 50279) in dorsal (I1) and ventral (I2) views. J, right fibula (MNCN 47748) in medial (J1), lateral (J2), ventral (J3), and dorsal (J4) views. K, right fibula (MNCN 50223) in medial (K1), lateral (K2), ventral (K3), and dorsal (K4) views. L, osteoderm (MNCN 50083) in dorsal (L1) and ventral (L2) views. M, osteoderm (MNCN 50090) in dorsal (M1) and ventral (M2) views. N, osteoderm (MNCN 50102) in dorsal (N1) and ventral (N2) views. Scale bars: A K = 5 cm; L N = 1 cm. between them than in the other specimens. In MNCN the protuberance is very small. Metacarpals When disarticulated and the material belongs to individuals of varied sizes, it is sometimes difficult to distinguish the metacarpals from the distal tarsals because they are somewhat similar in morphology. Our observations on the limbs of the extant Chelonoidis nigra show that generally the metacarpals are taller and narrower than the flattened distal tarsals. The metacarpals from Alcalá de Henares generally belong to individuals of two size-groups, one of large size and one of medium size. They have a slightly convex surface proximally for the distal carpals and a distal convex surface for the first phalanx (Fig. 7G). We observed significant variation in the proportions of the metacarpals, some of them being wider and thinner than the others. Anterior phalanges Generally, the phalanges of the anterior limb are shorter than the phalanges of the posterior limb. In the material from Alcalá de Henares there are several phalanges of the anterior limb, in varied sizes. They all exhibit the typical morphology with a concave proximal surface for the metacarpals and a convex distal surface for the unguals (Fig. 7H). Anterior unguals Generally, the unguals of the anterior limb are shorter and wider than the unguals of the posterior limb. In the material from Alcalá de Henares there are plenty of anterior unguals, in various sizes, which all show the typical morphology, with a concave articulation surface proximally for the phalanx. Carpal associations There are also some associations of carpals preserved. MNCN is an association of two distal carpals, a metacarpal, and a phalanx. The metacarpal and the phalanx are not in exact articulation with the distal carpal. MNCN is a part of a distal carpal preserved together with the associated metacarpal in articulation. Pelvic girdle Pelvic fragments are very scarce in the material from Alcalá de Henares. Only one fragment (MNCN 50284) is preserved, from the ischiatic area, but it is too small to allow further description. Femur Remains of femurs are rare in the material from Alcalá de Henares. A partial right femur (MNCN 50268, Fig. 6G) and a midshaft fragment of another specimen (MNCN 50285) are preserved. The proximal part of the first one is missing. The diaphysis is almost straight throughout its length, with a small curvature in the distal part. Although not entirely preserved, the trochanters appear to be fused. Tibia Only two tibia specimens are preserved. The first (Fig. 6H) consists of the diaphysis with the distal part. In the distal part there is a deep concavity for articulation with the astragalocalcaneum, corresponding to the articulation surfaces of the proximal part of the astragalus. The second (Fig. 6I) is the proximal part with the diaphysis. This specimen is fragmented, but the cnemial crest is pronounced with an irregular surface. The proximal part is wide and triangular. Fibula The fibula is represented by at least two complete specimens in the Alcalá de Henares material (Fig. 6J, K). This bone is almost straight with only a small curvature in the midshaft. Proximally, there is a small concave surface for the femur. Distally, the articular end is

18 670 A. PÉREZ-GARCÍA AND E. VLACHOS Figure 7. Carpal elements corresponding to several specimens identified as paralectotypes of Titanochelon bolivari from the MN6 zone (late Aragonian, middle Miocene) of Cerro del Viso, in Alcalá de Henares (Madrid, Spain). Specimens collected in A, right intermedium (MNCN 50005) in anterior (A1), medial (A2), and posterior (A3) views. B, right ulnare (MNCN 50007) in anterior (B1), medial (B2), and posterior (B3) views. C, right lateral centrale (MNCN 50037) in lateral (C1), proximal (C2), and medial (C3) views. D, right medial centrale (MNCN 50004) in anterior (D1), proximal (D2), and posterior (D3) views. E, left medial centrale (MNCN 50003) in anterior (E1), proximal (E2), and posterior (E3) views. F, distal carpal (MNCN 50017) in lateral (F1), anterior (F2), and proximal (F3) views. G, metacarpal (MNCN 50031) in posterior (G1), anterior (G2), and proximal (G3) views. H, anterior phalanx (MNCN 49954) in lateral (H1) and anterior (H2) views. Scale bar = 1 cm.

19 A GENUS FOR THE EUROPEAN LARGE TORTOISES 671 enlarged, with a rounded convex articular surface to match the deep concave surface present on the fused astragalus and calcaneum. Astragalocalcaneum The astragalus and calcaneum are fused in all specimens. In MNCN (Fig. 8C), the suture between the two elements is visible, whereas in the rest is not visible (Fig. 8A, B). This may indicate that MNCN was a subadult individual. The astragalus is much larger than the calcaneum and there is no constriction in the symphysis of the two elements. The tibia articulates entirely on the astragalus in a double concave articulation surface. The fibula articulates with the symphyseal area of the astragalus and calcaneum, which is a broad, slightly concave area. Distally, the astragalocalcaneum is complex in morphology. In the ventral part of the astragalus the second and third distal tarsals are articulated in a concave and a convex surface, respectively. The fourth distal tarsal articulates in a concave surface, right below the symphysis. Distal tarsals Plenty of distal tarsal elements are preserved in the material of Alcalá de Henares. The distal tarsal from the Alcalá de Henares is flattened and wide, with a flat to concave proximal surface for the astragalocalcaneum and a convex distal articular surface for the metatarsals. MNCN is the fourth distal tarsal. It shows a rounded proximal part with a convex articular surface for the astragalocalcaneum proximally. Distally, it has two articular surfaces, one concave for the fourth metatarsal and one convex for the fifth metatarsal. The ventral part of the bone shows a protuberance. Metatarsals Plenty of disarticulated metatarsals are preserved in the studied material (Fig. 8D, E). As they belong to individuals of various sizes, it is rather difficult to use the size of the metatarsal for identification. They have a wide proximal part, with a concave articular surface for the distal tarsal and a convex distal part with a convex articular surface for the phalanx. Posterior phalanges Generally, the phalanges of the posterior limb are longer than the phalanges of the anterior limb. In the material from Alcalá de Henares there are several phalanges of the posterior limb, in various sizes. They all exhibit the typical morphology with a concave proximal surface for the metatarsals and a convex distal surface for the unguals (Fig. 8F). Posterior unguals Generally, the unguals of the posterior limb are longer and narrower than the unguals of the anterior limb. In the material from Alcalá de Henares plenty of posterior unguals are preserved, in various sizes, which all show the typical morphology, with a concave proximal surface for the phalanx (Fig. 8G). Osteoderms Numerous osteoderms of varied sizes are preserved from Alcalá de Henares (Fig. 6L N). They are elliptical, flat in most cases, and without any pointed ends. Some of them are small and rounded. THE MATERIAL FROM LOS SANTOS DE LA HUMOSA (MADRID) Analysed material The notification of the presence of large testudinids in Los Santos de la Humosa took place in 1914 (Fernández Navarro & Carandell Pericay, 1914; see Appendix S1). This town was one of those cited by Hernández-Pacheco in the paper in which he defined Te. bolivari (Hernández-Pacheco, 1917a). However, neither the specimen discovered in 1914, nor any of those found in the 1920s (see Royo y Gómez & Menéndez Puget, 1928; Royo y Gómez, 1928a, b) were collected, and no descriptions or photographs of them are known. Several excavations in Los Santos de la Humosa were carried out in the second half of the 20 th century. Relatively abundant material of the clade analysed here was collected from the site of Henares I (see Jiménez Fuentes, 1985a). These specimens are part of the MNCN collection (see numbers of the relevant specimens in Appendix 1). The specimens analysed here correspond to two epiplastra and some appendicular elements (humeri, radius, ulna, unguals, osteoderms) in a good state of preservation (Fig. 9), and other poorly preserved postcranial bones. The recovered material corresponds to several individuals, at different stages of ontogenetic development. Their characters are compatible with those of a single taxon. Epiplastra The epiplastra are relatively thick. They have a raised convex epiplastral lip without any gular pocket in the specimen MNCN (Fig. 9B); this region is not preserved in MNCN (Fig. 9A). The shape of the anterolateral margin of MNCN corresponds to a subrounded anterior plastral lobe, with a small constriction at the level of the gularo-humeral sulcus that causes a small gular protrusion (Fig. 9B). Ventrally, the gulars are narrow and long, covering the anterior part of the entoplastron (Fig. 9B3). Shoulder girdle A fragment of the scapula is preserved (MNCN 49370), showing a shallow and elliptical glenoid fossa.

20 672 A. PÉREZ-GARCÍA AND E. VLACHOS

21 A GENUS FOR THE EUROPEAN LARGE TORTOISES 673 Figure 8. Tarsal elements corresponding to several specimens identified as paralectotypes of Titanochelon bolivari from the MN6 zone (late Aragonian, middle Miocene) of Cerro del Viso, in Alcalá de Henares (Madrid, Spain). Specimens collected in A, right astragalocalcaneum (MNCN 49999) in proximal (A1), distal (A2), anterior (A3), and posterior (A4) views. B, left astragalocalcaneum (MNCN 50001) in proximal (B1), distal (B2), anterior (B3), and posterior (B4) views. C, right astragalocalcaneum (MNCN 50002) in proximal (C1), distal (C2), anterior (C3), and posterior (C4) views. D, left metatarsal (MNCN 49967) in posterior (D1), lateral (D2), and anterior (D3) views. E, right metatarsal (MNCN 49969) in anterior (E1), lateral (E2), and posterior (E3) views. F, posterior phalanx (MNCN 50029) in lateral (F1) and anterior (F2) views. G, ungual (MNCN 49974) in anterior (G1) and posterior (G2) views. Scale bar = 1cm. Humerus The humerus is one of the best-preserved elements from Los Santos de la Humosa, with a complete specimen (MNCN 73733, Fig. 9D) that belongs to a young individual; relatively complete adult specimens (e.g. MNCN 49366, Fig. 9C); and other fragments. In the proximal part, the two trochanters are developed parallel to slightly diverging from each other. The major trochanter is much larger than the minor, extending well beyond the humeral head. On the contrary, the minor trochanter reaches below the height of the humeral head. The head is slightly elliptical in shape, with the longer axis not perpendicular with the shaft, forming an angle of about 50 with the shaft. The humerus shows a significant curvature along the diaphysis. The distal part is asymmetric because the ulnar condyle is shorter than the radial one. Pelvic girdle Only a fragment of the ischium is preserved from the pelvic girdle (MNCN 49381). Its state of preservation does not allow further description. Femur Only one femur diaphysis is preserved (MNCN 49374). The shaft of the bone is straight, and only the distal part seems to develop a slight curvature. Unguals Several unguals are preserved. This element shows a shallowly concave articular surface for the phalanx (Fig. 9O). Some specimens are from the anterior limb, being wide and having a rounded tip. Others are from the posterior limb, being longer and having a pointed tip. Radius and ulna A radius (MNCN 49367, Fig. 9J N) and an ulna (MNCN 49371, Fig. 9E I) are preserved from the Los Santos de la Humosa locality, most probably from the same individual as they articulate tightly with each other. Because of their size, we can estimate that they belong to a young individual that could be the same one to which the humerus mentioned above (MNCN 73733, Fig. 9D) belongs. The radius is an element with a broad proximal part and an almost straight shaft. The wide distal region shows a slightly convex surface for the carpal articulation. The ulna is much taller than the radius. The proximal part is very narrow and constitutes a concave articular surface. The midshaft is triangular in cross-section. The distal part has a short articular surface for the radius and the remaining part is convex. The bone is strongly curved in lateral view. The proximal articulation of the two parts leaves a triangular concave surface for the distal part of the humerus (Fig. 9F). The proximal articulation surface of the radius is short and wide, being subelliptical (Fig. 9K). The ventral side of the articulation of the ulna is concave in the proximal part. The radius and ulna delimit a long, narrow, tear-shaped fenestra between them. Osteoderms Several osteoderms from this locality have been preserved (Fig. 9P, Q). They are circular to elliptical in shape, without any pointed tip. Their thickness varies, from relatively flat elements to thicker osteoderms. THE MATERIAL FROM VALLECAS (MADRID) Analysed material This specimen was found in 1906 near Cerro de Almodóvar (Vallecas). Its preserved carapace and plastron were mounted for display in the second half of the 1910s using an iron structure (Fig. 10; see Appendix S1). This specimen was identified in the MNCN and has been restored (Figs 11, 12). Compared with the original specimen (Fig. 10), only some parts of the right anterior carapace rim and some parts of the costals and fragments of the right mediolateral margin of the plastron are missing. The remaining parts have been preserved intact. Additionally, although not visible in the old photographs, some elements corresponding to its shoulder girdle, and also two cervical vertebrae, were identified (note their outline on the visceral side of the anterior lobe in Fig. 12A). They were separated from

22 674 A. PÉREZ-GARCÍA AND E. VLACHOS

23 A GENUS FOR THE EUROPEAN LARGE TORTOISES 675 Figure 9. Elements of Titanochelon bolivari from the MN6 zone (late Aragonian, middle Miocene) of Henares I, in Los Santos de la Humosa (Madrid, Spain). Specimens collected during the second half of the 20 th century. A, right epiplastron (MNCN 49385) in ventral (A1, A2) view. B, right epiplastron (MNCN 73734) in visceral (B1) and ventral (B2, B3) views. C, left humerus (MNCN 49366) in dorsal (C1) and ventral (C2) views. D, left humerus (MNCN 73733) in anterior (D1), dorsal (D2), posterior (D3), and ventral (D4) views. E I, right ulna (MNCN 49371) in ventral (E), proximal (F), dorsal (G), medial (H), and lateral (I) views. J N, right radius (MNCN 49367) in ventral (J), proximal (K), dorsal (L), lateral (M), and medial (N) views. O, ungual (MNCN 73767) in ventral (O1) and lateral (O2) views. P, Q, osteoderms (MNCN 73750, 73740) in dorsal view. Scale bars: A O = 5 cm; P, Q = 1 cm. Figure 10. Unpublished historical photographs of the shell of Titanochelon bolivari from the MN5 zone (Aragonian, middle Miocene) of Vallecas (Madrid, Spain), whose plastron corresponds to the collection number MNCN 73213, and the carapace to MNCN Specimen collected in A, right lateral view. Modified version of ACN B, left lateral view. Modified version of ACN C, anteroventral view. Modified version of ACN D, dorsoposterior view. Modified version of ACN Original photographs taken by Francisco Molina. Figure 11. MNCN 73214, the recovered carapace elements of the individual of Titanochelon bolivari from the MN5 zone (Aragonian, middle Miocene) of Vallecas (Madrid, Spain). Specimen collected in A, anterior part of the carapace in anterior (A1, A2) view. B, third or fifth neural in dorsal (B1) and visceral (B2) views. C, fourth neural in dorsal view. Scale bar = 5 cm.

24 676 A. PÉREZ-GARCÍA AND E. VLACHOS Figure 12. MNCN 73213, the recovered plastron of the individual of Titanochelon bolivari from the MN5 zone (Aragonian, middle Miocene) of Vallecas (Madrid, Spain). Specimen collected in A, visceral view. B, right viscerolateral view of the anterior lobe. C, D, ventral view. Scale bars =5cm. the visceral region of the anterior lobe of the plastron, and prepared for study (Fig. 13). Shell remains On the anterior part of the carapace, a shallow wide notch, affecting the nuchal and the first pair of peripherals, is present (Fig. 11A). The carapace is relatively low. The nuchal plate is as wide as long. Two neurals are preserved. One of them is subelliptical and wider than long. Based on its shape, it may correspond to either the third or fifth neural (Fig. 11B). Owing to the poor state of preservation, it is not possible to know whether or not this plate was traversed by a sulcus delimiting two vertebral scutes. The other neural plate is octagonal. Its width is greater than its length (Fig. 11C). Its anterior border is slightly concave, whereas the posterior border is nearly straight. It is interpreted here as a fourth neural. The distal margins of the anterior peripherals are dorsally directed. This specimen lacks a cervical scute. The sulcus between the pleural and the marginal scutes overlaps the suture between the costal and the peripheral plates. The maximum length of the preserved region of the plastron is about 95 cm (Fig. 12), whereas the total estimated length of the plastron would barely exceed

25 A GENUS FOR THE EUROPEAN LARGE TORTOISES cm. The morphology of the rim of the anterior lobe is not known. However, the presence of a welldeveloped gular protrusion, developed from the region where the sulci between the gulars and the humerals contact the lateral margin, is observed. This specimen has a wider than long anal notch. The lateral margins of the posterior plastral lobe are strongly convergent. This morphology of the posterior lobe allows us to recognize the specimen from Vallecas as corresponding to a female. This is further supported by the flat middle posterior (hyo- and hypoplastra) area of the plastron and by the absence of a ventral thickening in the region covered by the anal scutes. This specimen has a dorsally convex epiplastral lip. It lacks a gular pocket. The entoplastron is hexagonal. Its width is greater than its length. The hyoplastra are longer than the hypoplastra, and both are longer than the xiphiplastra. The gular scutes overlap the anterior region of the entoplastron. The humero-pectoral sulcus does not contact the entoplastron, being situated posterior to that plate. This sulcus is perpendicular to the axial plane in the medial region, but has a well-developed lateral change of curvature, being markedly directed towards the anterior region. The pectoral scutes are very short in the medial plane. Only the posterior region of the hypoplastra is covered by the femoral scutes. The anal scutes are relatively short and have an omegashape, being medially longer. Vertebral column Only two cervical vertebrae are preserved, almost entirely complete but slightly deformed. MNCN (Fig. 13A) has a longer body than MNCN (Fig. 13B). Both vertebra preserved well-developed anterior and posterior zygapophyses. A sagittal keel is observed in the anterior half of the dorsal region of both vertebrae. MNCN is procoelous, with a single concavity in the anterior articular surface but with a double convexity in the posterior one. MNCN is amphicoelous, having two concavities in each Figure 13. The recovered vertebrae and appendicular elements of the individual of Titanochelon bolivari from the MN5 zone (Aragonian, middle Miocene) of Vallecas (Madrid, Spain). Specimen collected in A, sixth cervical vertebra (MNCN 73210) in anterior (A1), posterior (A2), dorsal (A3), ventral (A4), right lateral (A5), and left lateral (A6) views. B, seventh cervical vertebra (MNCN 73209) in anterior (B1), posterior (B2), dorsal (B3), ventral (B4), right lateral (B5), and left lateral (B6) views. C, acromion process (MNCN 73211) in anterior (C1) and ventral (C2) views. D, coracoid (MNCN 73212) in dorsal (D1) and ventral (D2) views. Scale bars = 5 cm.

26 678 A. PÉREZ-GARCÍA AND E. VLACHOS articular surface. Based on the size, the elongated and narrow shape, and morphology of the neural spine, in those of the zygapophyses, and especially in the articular facets, MNCN is recognized as the sixth cervical and MNCN as the seventh. Shoulder girdle From the shoulder girdle only the acromion process (Fig. 13C) and the right coracoid (Fig. 13D) are preserved. The acromion is straight and elongated with a wide, rounded end. Laterally, it shows a shallow concave articular surface for the humerus. The coracoid is preserved complete (Fig. 13D), being long, with a narrow neck and a wide fan-shaped end. the axillary notch. This part is crossed by the humeropectoral and pectoro-abdominal sulci. The humeropectoral sulcus is perpendicular to the axial plane in the medial region, but has a well-developed lateral change of curvature, being markedly directed towards the anterior region. The pectoral scutes are short. THE MATERIAL FROM CERRO DEL OTERO (PALENCIA) The analysis of the material of large testudinids from Cerro del Otero allowed us to identify both some specimens corresponding to the first remains recognized in this locality in 1911, and some of those collected in 1921 (see Appendix S1). THE MATERIAL FROM PUENTE DE LA PRINCESA (MADRID) Analysed material Most of the only extracted specimen of the three found in Puente de la Princesa in June 1921 has been reconstructed (see Appendix S1). The finding of unpublished photographs corresponding to this specimen (Fig. 14) allowed us to identify it. Although the anterior margin of the plastron is recognized in the old photographs (see Fig. 14), it is no longer preserved. Shell remains This specimen corresponds to the right anterior half of a shell (Fig. 15). The distal margin of the anterior peripherals is dorsally directed. The first two right costals, and much of the third one, are preserved, as well as the complete or partial associated peripherals. The second costal is medially short, but not the third one. The third neural, with rounded lateral margins, is overlapped by the sulcus between the second and the third vertebral scutes. From the plastron, only a part of the right hyoplastron is preserved, close to Analysed material collected in 1911 Although the tortoise material found in 1911 was very scarce, some fragments of plates and of appendicular elements were recognized, corresponding to one or more specimens (see Fig. 16 and Appendix 1). Shell remains The proximal region of two costal plates was recognized (Fig. 16A, B). Both plates correspond to odd costals, so they are narrower in the medial region than in the lateral, and they are not crossed by sulci delimiting vertebral scutes, but by those delimiting pleurals. Humerus Two humeri are preserved (Fig. 16C, D). From the right humerus (Fig. 16C, also figured in plate 27 of Hernández-Pacheco & Dantín Cereceda, 1915) the distal part is missing, whereas only the midshaft is preserved from the left (Fig. 16D). The proximal part is wide. The major trochanter extends beyond the height of the humeral head, whereas the minor is much smaller. The trochanters are developed almost paral- Figure 14. Unpublished historical photographs of the shell of Titanochelon bolivari from the MN5 MN6 interval (Aragonian, middle Miocene) of Puente de la Princesa (Madrid, Spain), corresponding to the collection number MNCN Specimen collected in A, right lateral view. Modified version of ACN B, anterior view. Modified version of ACN C, posterior view. Modified version of ACN Original photographs taken by Francisco Molina.

27 A GENUS FOR THE EUROPEAN LARGE TORTOISES 679 Figure 15. The recovered shell elements of the individual of Titanochelon bolivari from the MN5 MN6 interval (Aragonian, middle Miocene) of Puente de la Princesa (Madrid, Spain), corresponding to the collection number MNCN Specimen collected in A, B, carapace in dorsal view. C, D, ventral view of the shell. Scale bar =5cm. lel to each other. The head of the humerus is spherical. The diaphysis shows a significant sigmoid curvature. Osteoderm The preserved osteoderm is flat and rounded, without a pointed tip (Fig. 16E). Additional flat osteoderms, not identified in the MNCN collection, were figured in Hernández-Pacheco & Dantín Cereceda (1915: plate 26), being mistakenly identified as altered fragments of mammalian bones. Analysed material collected in 1921 As indicated by Hernández-Pacheco (1921a), and confirmed by observation of the photographs taken by Hernández-Pacheco and Royo y Gómez in 1921 (see Fig. 17), about ten well-preserved shells, whose length ranged from 1 to 1.5 m, were found in Cerro del Otero in However, relatively scarce material from this locality is recognized in the MNCN, including shell remains (Fig. 18) and appendicular elements (Fig. 19) (see Appendix 1). Additionally, several unidentified shell

28 680 A. PÉREZ-GARCÍA AND E. VLACHOS Figure 16. Recovered postcranial elements of Titanochelon bolivari from the MN7/8 zone (late Aragonian, middle Miocene) of Cerro del Otero (Palencia, Spain). Specimens collected in A, costal fragment (MNCN 39043) in dorsal view. B, costal fragment (MNCN 39041) in dorsal view. C, right humerus (MNCN 39040) in dorsal (C1), posterior (C2), anterior (C3), and ventral (C4) views. D, left humerus (MNCN 39042) in (posterior D1) and dorsal (D2) views. E, osteoderm (MNCN 39193) in dorsal (E1) and ventral (E2) views. Scale bar =3cm. fragments with accompanying information that originate from Palencia are stored in the MNCN, but their current state of preservation does not allow further description. Shell remains One of the identified specimens consists of the anterior margin of a carapace (Fig. 18A). A very shallow nuchal notch is observed. The preserved region of the nuchal plate shows that this plate lacked a cervical scute. From the right side, only the first peripheral is preserved, whereas from the left side the first two peripherals are found. All of them are crossed by the marginal sulci, and there is good coincidence between the costoperipheral sutures and the pleuromarginal sulci. It probably corresponds to the same individual from which both the partial anterior plastral lobe (Fig. 18B) and the right bridge, attached to the axillary region of the right hyoplastron (Fig. 18C), are preserved. The morphology of the anterior plastral margin is not known (Fig. 18B). There is no gular pocket, only a step-like morphology behind the thick epiplastral lip. The preserved part of the dorsal surface of the epiplastral lip is convex. The entoplastron is relatively long but narrow,

29 A GENUS FOR THE EUROPEAN LARGE TORTOISES 681 Figure 17. Historical photographs of the some specimens of Titanochelon bolivari from the MN7/8 zone (late Aragonian, middle Miocene) of Cerro del Otero (Palencia, Spain). A, two shells in the fossil site. Modified version of ACN 6666, photograph published by Hernández-Pacheco (1921a). Photograph taken by Eduardo Hernández-Pacheco in April B, a shell and several appendicular elements in the fossil site. Modified version of the unpublished photograph ACN Photograph taken by Eduardo Hernández-Pacheco in April C, a restored plastron collected in 1921, corresponding to the recovered partial anterior plastral lobe in Figure 18D and to figure 4 in Royo y Gómez (1935a). Modified version of the unpublished photograph ACN Photograph taken by Eduardo Hernández-Pacheco. D, several shells in the fossil site. Unpublished photograph taken by Royo y Gómez in April with a probably subrhombic morphology. The gular scutes overlap the anterior region of the entoplastron. The peripherals of the bridge correspond to a relatively low shell (Fig. 18C). The humero-pectoral sulcus is perpendicular to the axial plane in the medial region, but has a well-developed lateral change of curvature, being markedly directed towards the anterior region. Another partial plastral anterior lobe is recognized (Fig. 18D). Its long epiplastral lip is also convex. There is no gular pocket, only a step-like morphology. The gulars are very wide, both dorsally and ventrally. The available information on the ventral part is very scarce, but we can observe a wide entoplastron with an angular anterior part covered by the gulars. The width/ length ratio of this plate is much larger in visceral view than in dorsal view. The humero-pectoral sulcus has the same morphology as in the previously described plastron. This fragmentary specimen corresponds to the complete plastron photographed by Hernández- Pacheco (see Fig. 17C) and of which Royo y Gómez (1935a: fig. 4) provided a drawing and indicated that its maximum length was 123 cm. Therefore, we know that its anterior lobe had rounded lateral margins, being coined towards the anterior region. Medially, the pectoral scutes were very short. This specimen belonged to a male, as deduced by the presence of a concavity in the middle posterior region of the plastron, the morphology of the posterior lobe, with lateral margins straight and slightly convergent and posterior margin subperpendicular to the axial plane, and the presence of a ventral thickening in the region occupied by the anal shields. These male characters are also present, for example, in the best-preserved specimen in the photograph in Figure 17D. However, the morphology of its anterior lobe is more coined than that of the analysed plastron. The posterior region of the shell of the cited specimen in Figure 17D has a short distance between the posterior margin of the plastron and the distal margin of the pygal, with the posterior peripherals being dorsally directed. These characters are also typical of the males. The individual situated on the left side in Figure 17A shows that, as with the specimen shown in Figure 18A, the first pair of marginal scutes is in contact in the sagittal plane. However, Royo y Gómez (1935a: fig. 3) published a schematic drawing of one of the specimens collected in Cerro del Otero in which the first vertebral scute contacted the anterior margin of the carapace. This probably corresponds to an anomaly. Royo y Gómez (1935a: fig. 5) indicated that, comparing the plastra of the specimens coming from Cerro del Otero, a relatively high range of variability in the morphology of the anterior margin was observed. He recognized specimens similar to that previously cited in Figure 17D (Royo y Gómez, 1935a: fig. 5a); specimens with sinuous anterior margins (Royo y Gómez,

30 682 A. PÉREZ-GARCÍA AND E. VLACHOS Figure 18. Recovered shell material of Titanochelon bolivari from the MN7/8 zone (late Aragonian, middle Miocene) of Cerro del Otero (Palencia, Spain). Specimens collected in A, anterior margin of a carapace (MNCN 74023), probably corresponding to the same specimen as MNCN 74024, in anterior (A1, A2) view. B, C, specimen MNCN B, anterior plastral lobe in visceral (B1) and ventral (B2, B3) views. C, right bridge in ventral (C1, C2) and ventrolateral (C3, C4) views. D, partial anterior lobe of a plastron (MNCN 32762), corresponding to the complete plastron in the sofar unpublished photograph ACN 6665 (Fig. 17C) and to figure 4 in Royo y Gómez (1935a) in visceral (D1, D2), left viscerolateral (D3), and ventral (D4, D5) views. Scale bar =5cm.

31 A GENUS FOR THE EUROPEAN LARGE TORTOISES a: fig. 5c); and individuals with a strong gular protrusion and anterior margins perpendicular to the axial plane (Royo y Gómez, 1935a: fig. 5D). Vertebral column From the vertebral column only some remains from the tail were identified (e.g. Fig. 19A). The caudal vertebra is procoelous, having short centra and relatively long transversal processes. Anterior limb Several disarticulated elements from the anterior limb were found. Besides those described below, we observed several complete elements in the classic photographs. For example, in Figure 17B, a scapula, coracoid, and a complete humerus are clearly distinguishable. The scapula and the acromion are developed perpendicular to each other and they are straight in shape. The coracoid has a narrow neck and a developed fanshaped end. The humerus shows the typical diaphysis with the sigmoidal curvature. Shoulder girdle Only one scapula (MNCN 23858) was identified, missing the acromion. The scapula is straight, with an elliptical cross-section. The articular surface for the humerus is narrow, elongated, and concave. Humerus Few remains of humeri are preserved (Fig. 19B, C). MNCN is missing only the distal part (Fig. 19C). MNCN is missing part of the trochanters and the distal part (Fig. 19B). In MNCN only the midshaft is preserved. The proximal part is wide, with a spherical head that dorsally extends well beyond the shaft of the bone with a long neck (Fig. 19B1). The trochanters are different in size, and the major trochanter extends beyond the height of the head. They are developed almost parallel to each other (note that MNCN 69682, Fig. 19C, is compressed). The diaphysis shows a significant sigmoid curvature. Carpals Some carpals were recognized (Fig. 19D F, J). Two complete intermediums are preserved (Fig. 19D). They show a triangular cross-section, with a proximal concave articular surface for the ulna. Medially, they show a slightly concave surface for the radius. Distally, a slightly convex surface for the medial centrale is present. MNCN (Fig. 19J) is a fused medial and lateral centrale, preserving the suture. The intermedium is articulated in the symphysis of the two elements in a deep complex articular surface. Similar in complexity is the surface of the medial centrale for the radius. Distally, we observed two distal carpals in articulation; they articulate firmly in the distal part of the fused centrale, with the concave surface created by the discoid part and the protuberance. The protuberance extends ventrally. Several metacarpals were found (e.g. MNCN 23827, Fig. 19E), with a proximal concave and a distal convex surface. Femur Only one femur was found (MNCN 38935, Fig. 19G), preserving the distal part. The tibial condyle is more developed than the smaller fibular; they are connected by a ridge, forming a concavity. Unguals Several unguals are preserved. MNCN is from the anterior limb, being wider than those from the posterior limb (MNCN 23834, Fig. 19H), which are long and have a pointed tip. Osteoderms Many osteoderms are preserved, being relatively flat. Their shape varies from rounded to rounded with some angular sides (e.g. Fig. 19I). No pointed osteoderms were identified. THE MATERIAL FROM ARÉVALO (ÁVILA) The large testudinid material from Arévalo comes from two different excavation periods. The first discoveries were made in In October of that year, José Royo y Gómez directed a campaign of exploration and excavation there. Some of the recovered specimens were analysed by this author (Royo y Gómez, 1935a; see Appendix S1). New discoveries were made in the second half of the 20 th century, generally corresponding to partial or complete and well-preserved shells (see García & Alberdi, 1968; Jiménez Fuentes & Carbajosa Tamargo, 1982; Jiménez Fuentes et al., 1986, 1993). Analysed classic material As indicated by Royo y Gómez (1935a), three of the four specimens found in 1933 were prepared (specimens Arévalo I, Arévalo II, and Arévalo IV ). Information published by Royo y Gómez (1935a), together with the analysis of unpublished photographs and the identification of some of the specimens in the MNCN (including some thus far unpublished elements), have allowed the description of some characters of these three specimens (Fig. 20). Some of the elements cited by Royo y Gómez (1935a), such as vertebrae of several of the specimens and a skull fragment corresponding to the specimen Arévalo IV, could not be found for the present, and they are not known from published figures or from unpublished photographs. Shell of Arévalo I Royo y Gómez collected some of the elements corresponding to the posterior region of the carapace of the

32 684 A. PÉREZ-GARCÍA AND E. VLACHOS

33 A GENUS FOR THE EUROPEAN LARGE TORTOISES 685 Figure 19. Recovered postcranial material of Titanochelon bolivari from the MN7/8 zone (late Aragonian, middle Miocene) of Cerro del Otero (Palencia, Spain). Specimens collected in A, caudal vertebra (MNCN 69668) in anterior (A1), posterior (A2), dorsal (A3), ventral (A4), right lateral (A5), and left lateral (A6) views. It corresponds to the same individual as the humerus MNCN B, right humerus (MNCN 23866) in anterior (B1), ventral (B2), and dorsal (B3) views. C, left humerus (MNCN 69682) in posterior (C1), ventral (C2), and dorsal (C3) views. D, right intermedium (MNCN 23836) in lateral (D1), anterior (D2), medial (D3), and posterior (D4) views. E, metacarpal (MNCN 23827) in anterior (E1) and posterior (E2) views. F, metatarsal (MNCN 23829) in anterior (F1) and posterior (F2) views. G, distal part of a left femur (MNCN 38935) in dorsal (G1), posterior (G2), and ventral (G3) views. H, posterior ungual (MNCN 23834) in posterior (H1) and anterior (H2) views. I, osteoderm (MNCN 23830) in dorsal (I1) and ventral (I2) views. J, carpal complex (MNCN 23860) with a fused medial and lateral centrale in articulation with two distal carpals, in anterior (J1, J2), posterior (J3), proximal (J4, J5), and distal (J6) views. Abbreviations: dc, distal carpal; int art, articulation for the intermedium; lat, lateral centrale; med, medial centrale; ra art, articulation for the radius. Scale bars: B, C, G = 5 cm; A, D F, I, J=1cm. specimen Arévalo I (see Royo y Gómez, 1935a). However, none of them could be identified in the MNCN. The recognized photographs do not allow observation of the relevant information in the poorly preserved region of its carapace (see plate 51 in Royo y Gómez, 1935a). The preservation of the recovered region of its plastron is much better than that of the carapace. Much of the region of the plastron figured by Royo y Gómez (see plate 52 in Royo y Gómez, 1935a) was recovered (Fig. 20A). Its posterior lobe is short and wide, with convergent lateral margins, and with a wider than long, and angular, anal notch. The surface of the preserved region of the plastron is flat. Anteriorly, a rather straight pectoro-abdominal sulcus is preserved in the medial part. The abdominals cover the posterior part of the hyoplastra and most of the hypoplastra. Their posterior part is covered by the femorals. The anals are medially expanded and the femoro-anal sulcus is medially straight and laterally concave on both sides. This specimen lacks ventral thickening in the region covered by the anal scutes. In visceral view, the anal scutes do not contact medially. Pelvis of Arévalo I We did not identify any appendicular elements from the Arévalo I individual. However, a complete pelvis was figured by Royo y Gómez (1935a: plate 54; corresponding to the photographs ACN 4296 and ACN 4297). The pubis has long and curved pectineal processes that are projected anterolaterally. The metischial processes are much shorter. The puboischiatic plate is rather flat and most probably did not make contact with the plastron. In the medial parts of the plate, circular to subangular puboischiatic foramina are formed. The ilium is thin and long, with a wide, straight edge at the proximal end. The three pelvic bones form a deep, concave, and almost spherical acetabulum for the femur. Shell of Arévalo II Although the posterior region of the carapace of the specimen Arévalo II was not identified in the MNCN, its preservation was better than that of the carapaces of Arévalo I and Arévalo IV, as shown in plate 49 and figure 2 of Royo y Gómez (1935a). It had two suprapygals, the first of them embracing the second one, lenticular in shape. The vertebromarginal sulcus crossed the second suprapygal. It had a supracaudal scute. The boundaries between the pleural and the marginal scutes overlapped the sutures between the costal and the peripheral plates. Shell of Arévalo IV Some elements of the posterior region of the carapace of the specimen Arévalo IV were also found (see plate 50 in Royo y Gómez, 1935a). They correspond to poorly preserved peripherals, pygal, suprapygals, and distal regions of some costals. This specimen was identified in the MNCN and has been reconstructed (MNCN 74026). Its carapace lacks a well-developed curvature of the pygal plate (thus consistent with its assignment as a female). The second suprapygal is lenticular and is embraced by the first one. It has a supracaudal scute, overlying the posterior region of the second suprapygal. Most of the fragment of the plastron figured by Royo y Gómez (1935a: plate 53) was identified in the MNCN, and has been reconstructed (Fig. 20B). It is also assignable as a female, sharing all the characters described for the plastron of Arévalo I. The overall morphology is similar to Arévalo I, showing only some small differences. These are the shape of the femoroanal sulcus, which is straighter compared with Arévalo I, and the shape of the anal notch, which is more rounded. Appendicular elements From the specimen Arévalo IV, few appendicular elements are preserved. The right scapula is fragmented (Fig. 20C), but preserves the complete acromion. Laterally, we can see an elongated shallow articular surface. In both coracoids (MNCN and MNCN 40412), the part that articulates with the scapula is preserved, showing a shallow flat to concave

34 686 A. PÉREZ-GARCÍA AND E. VLACHOS Figure 20. Some of the recovered elements of Titanochelon bolivari from the MN9 zone (early Vallesian, late Miocene) of Arévalo (Ávila, Spain). Specimens collected in A, MNCN ( Arévalo I in Royo y Gómez, 1935a) in visceral (A1) and ventral (A2, A3) views. B, MNCN ( Arévalo IV in Royo y Gómez, 1935a) in visceral (B1) and ventral (B2, B3) views. C, right scapula (MNCN 40418) of the specimen Arévalo IV in ventrolateral (C1) and anterior (C2) views. D, osteoderm (MNCN 40409) of Arévalo IV in ventral (D1) and dorsal (D2) views. Scale bars = 5 cm. Arévalo I comes from an outcrop next to the Arevalillo River. Arévalo IV was found next to the Adaja River. articular surface that contributes to the formation of the articular surface for the humerus. The crosssection of the bone at this point is triangular, and forms a narrow neck. The end of the element is not preserved. One osteoderm was also identified (Fig. 20D), which is circular to angular in shape, rather flat, and with a half-pointed end. Analysed material found in the second half of the 20 th century Several well-preserved specimens were found in Arévalo between the 1960s and the 1980s (Fig. 21; see figures and previous descriptions in García & Alberdi, 1968; Jiménez Fuentes & Carbajosa Tamargo, 1982; Jiménez Fuentes et al., 1986, 1993). One of them, the

35 A GENUS FOR THE EUROPEAN LARGE TORTOISES 687 Figure 21. Additional material of Titanochelon bolivari from the MN9 zone (early Vallesian, late Miocene) of Arévalo (Ávila, Spain). Specimens collected during the second half of the 20 th century: A, in 1980; B, in 1985; C, in 1989; D, in A, anterior half of a plastron and some left peripheral plates (STUS 2414) in visceral (A1), lateral (A2), and ventral (A3, A4) views. This specimen was identified as the neotype of Testudo richardi by Jiménez Fuentes et al. (1988). B, right half of a partial shell (STUS 5137) in ventral (B1, B2) view. C, relatively complete shell (STUS 8373), in ventral (C1, C2) view. D, relatively complete shell (unnumbered specimen preliminarily studied by García & Alberdi, 1968, and housed in ICP) in dorsal (D1), ventral (D2), left lateral (D3), anterior (D4), and posterior (D5) views. Scale bar = 10 cm.

36 688 A. PÉREZ-GARCÍA AND E. VLACHOS specimen STUS 2414 (Fig. 21A), was identified as the neotype of Te. richardi by Jiménez Fuentes et al. (1988). Shell The shells correspond to three males (Fig. 21A, B, C) and a female (Fig. 21D), as inferred by characters such as the presence of a plastral concavity in the hyohypoplastra and the morphology of the posterior lobe. The most anterior peripherals and the posterior ones are dorsally directed. The neural series, which can be observed in the ICP specimen (Fig. 21D), is composed of eight plates. The first is subrectangular, noticeably longer than wide. The second and fourth neurals are octagonal. The maximum width of the second neural is located in its posterior region. However, the fourth neural is as wide in the anterior half as in the posterior. The third and fifth neurals are subrectangular, slightly wider than long. Sulci are located between the vertebral scutes on both of these neurals. The last three neurals are hexagonal (Fig. 21D1). A shallow nuchal notch was identified. The nuchal plate is almost as wide as long (Fig. 21D4) and lacks a cervical scute. A sagittal contact between the two scutes of the first pair of marginals is present. This specimen has two suprapygal plates, the first one embracing the lenticular second one (Fig. 21D5). The morphology of the anterior plastral lobe of the Arévalo material is subject to variability. A welldeveloped change of its curvature is observed in the area of contact with the gularo-humeral sulcus in some of the specimens (Fig. 21A3, A4, D2). However, this constriction is less developed in other individuals (Fig. 21B1, B2, C1, C2). Although it is not complete, the anterior lobe of the specimen of Figure 21C1, C2 is more rounded, wider, and shorter than those of the other specimens, whereas in some other specimens (e.g. Fig. 21A) the shape of the anterior lobe is trilobed. The morphology of the entoplastron is also variable, with that of some specimens being identified as subrounded (Fig. 21C1, C2), hexagonal (Fig. 21A3, A4), or subrhombic (Fig. 21D2), but always narrow compared with the maximum width of the lobe. The gular scutes, wider in some specimens (Fig. 21B1, B2, C1, C2) than in others (Fig. 21A3, A4, D2), overlap the anterior region of the entoplastron in all of them. The humero-pectoral sulcus is perpendicular to the axial plane in the medial region, but has a well-developed lateral change of curvature, being markedly directed towards the anterior region. Although the pectoral scutes are always very short medially, their distance from the posterior end of the entoplastron may be slightly less than their medial length (Fig. 21B1, B2, D2), markedly less than that length (Fig. 21C1, C2), or a contact between the posterior margin of entoplastron and the pectoral may be present (Fig. 21A3, A4). The dorsal epiplastral lip is long and has a convex surface. A poorly developed gular pocket is present in one of the specimens (Fig. 21A1, A2). In STUS 8373, the epiplastral lip could not be observed. In the ICP specimen, there is no pocket, but the dorsal thickened part of the epiplastra is gently elevated. The shape of the anal notch is also quite variable, being very wide and shallow in the male individuals (Fig. 21B, C), whereas in the female ones (Fig. 21D2) it is narrower, deeper, and slightly rounded. The femoro-anal sulcus of the female shows the typical omega-shape (Fig. 21D2). THE HOLOTYPE OF TESTUDO RICHARDI The inclusion of the proposed neotype of Te. richardi (Fig. 21A) and the other specimens from Arévalo allow the documentation of the morphological variation of the plastron. At this point, it seems appropriate to provide some brief information on Te. richardi, the second taxon of a large tortoise that was named from Spain. Testudo richardi was defined by Bergounioux (1938), based on a single specimen from Els Hostalets of Pierola (Barcelona), which was destroyed in July 1936, during the Spanish Civil War (Jiménez Fuentes, 1986b; Jiménez Fuentes et al., 1988; Jiménez Fuentes & Martín de Jesús, 1991; Fig. 22). Bergounioux (1938) believed that the specimen came from the early Oligocene. This author published a photograph of its plastron in dorsal view (Bergounioux, 1938: fig. 5), as well as an interpretation in ventral view (Bergounioux, 1938: fig. 6). However, 20 years later he published a photograph of the plastron in ventral view (Bergounioux, 1958: plate 34), and a new interpretation of this view (Bergounioux, 1958: fig. 17). Some wrong interpretations that were made in 1938 and amended in 1958 were those concerning the elliptical morphology of the entoplastron; the absence of overlapping of the gular scutes onto the anterior region of the entoplastron; the relatively long overlap of the pectoral scutes onto the entoplastron; and the layout of the humero-pectoral sulcus; as well as their age and locality, having been found in late Miocene levels (early Vallesian) of Tarrega (Catalonia). Therefore, the actual plastral morphology of Te. richardi was totally different than previously thought. In fact, the new interpretation of these characters shows that the morphology of the plastron of Els Hostalets de Pierola (Fig. 22A) is compatible with that of the other Spanish large testudinids described here. Description Observation of the photographs mentioned above of the holotype of Te. richardi, in dorsal and ventral views, allowed us to describe its plastron (Fig. 22B). As Bergounioux (1938) indicated, its maximum length was 105 cm. Although the lateral margins of the anterior plastral lobe are subrounded, the anterior border is trilobate, with a well-developed medial protrusion an-

37 A GENUS FOR THE EUROPEAN LARGE TORTOISES 689 Figure 22. Ventral view (A, B) of the lost plastron of Titanochelon bolivari identified by Bergounioux (1938) as the holotype of Cheirogaster richardi. It comes from the MN9 zone (early Vallesian, late Miocene) of Els Hostalets de Pierola (Barcelona, Spain). Photograph taken of plate 34 of Bergounioux (1938). Scale bar = 10 cm. teriorly, and another pair of protrusions resulting from the contact of the gular scutes with the humerals. This morphology is similar to that of some of the analysed specimens (see Fig. 21A). Its posterior plastral lobe is wide, with the posterior margin subperpendicular to the axial plane. As in the other specimens with similar morphology of the posterior lobe, the middle posterior region of the plastron is concave, and that covered by the anal scutes is thickened. A pair of small and laterally directed protuberances is present at the most posterior end of the lateral margins of this lobe. All of these characters allow us to conclude that the lost holotype of Te. richardi belonged to a male individual. The epiplastral lip is long and convex dorsally. This specimen lacks a gular pocket. The gular scutes overlap the anterior region of the entoplastron. Medially, the pectoral scutes are very short. The humero-pectoral sulcus does not overlap the entoplastron. This sulcus is perpendicular to the axial plane in the medial region, but has a well-developed lateral change in curvature, being markedly directed towards the anterior region. THE MATERIAL FROM EL PARDO (MADRID) Analysed material A problematic classic specimen is the xiphiplastron found in El Pardo in 1929 (see Appendix S1). It was recognized as corresponding to a probable juvenile individual of Te. bolivari (Royo y Gómez, 1929; Royo y Gómez & Menéndez Puget, 1929). It was found in the MNCN (Fig. 23A) and its study allowed us to recognize that it does not belong to a large tortoise, but to another taxon (see Discussion). Description It is a complete but small xiphiplastron. Its maximum width is 46 mm and the maximum length is 39 mm. It is well ossified and the lateral margin is rounded. It has a well-developed anal notch and the femoroanal sulcus does not show omega morphology. THE MATERIAL FROM PUENTE DE LOS FRANCESES (MADRID) Analysed material Very few skeletal elements are preserved from Puente de los Franceses (Fig. 23B). They correspond to material collected in 1926 (see Appendix S1). Description The identified elements of the appendicular skeleton are very fragmentary, containing a part of the acromion (Fig. 23B1), which is straight throughout its length; a fragment of an ilium (Fig. 23B2), which is very massive, with a shallowly concave articular surface; two osteoderms (Fig. 23B3, 4), which are small and rather flat, without any pointed tip. All of this material may belong to the same individual. THE MATERIAL FROM FUENSALDAÑA (VALLADOLID) Analysed material The large testudinid material from Fuensaldaña recognized in the MNCN corresponds to a single appendicular bone, identified as a right femur (Fig. 23C; see Appendix S1). Description The cited bone coming from Fuensaldaña is preserved in two parts, and because a part of the shaft

38 690 A. PÉREZ-GARCÍA AND E. VLACHOS is missing there is no clear connection between the proximal femur and the midshaft (Fig. 23C). The femur is generally straight, showing a slight curvature in the distal part. The proximal part is wide, having an elongated, convex-shaped head, which is developed without any clear neck. The angle between the longer axis of the head and the diaphysis is approximately 140. The distal part is wide, with short tibial and fibular condyles that form a weak ridge. Behind the condyles, in ventral view, a shallow concavity is formed.

39 A GENUS FOR THE EUROPEAN LARGE TORTOISES 691 Figure 23. Specimens from several classic localities, corresponding to cf. Paleotestudo (A), Titanochelon bolivari (B E), and cf. Titanochelon bolivari (F). A, left xiphiplastron (MNCN 48470) of Paleotestudo sp. from the middle Aragonian of El Pardo (Madrid) in visceral (A1) and ventral (A2, A3) views. Specimen collected in Scale bars: A = 1 cm; B E = 5 cm; F = 1 cm. Note that this specimen was previously identified as a juvenile Testudo bolivari by Royo y Gómez (1929) and by Royo y Gómez & Menéndez Puget (1929). See main text for further details. B, material from the middle Aragonian (MN5) of Puente de los Franceses (Madrid). B1, acromion process (MNCN 48296) in anterior view. B2, pubic fragment (MNCN 48285). B3, B4, osteoderms (MNCN and MNCN 48298) in dorsal view. Specimens collected in C, right femur (MNCN 40249), from the late Aragonian (MN7/8) of Fuensaldaña (Valladolid) in dorsal (C1), anterior (C2), and ventral (C3) views. This discovery was cited by Hernández-Pacheco (1917b). D, right humerus (MNCN 38917) from the Aragonian of Villalcón (Palencia) in ventral (D1) and anterior (D2) views. This specimen has been housed in the MNCN since E, carapace fragment (MNCN 40726) from the late Aragonian (MN 7/8) of La Cistérniga (Valladolid) in dorsal view. This discovery was referred by Hernández-Pacheco (1914). F, anterior plastral lobe of a juvenile specimen (MNCN 47969) from the Aragonian of Illescas (Toledo) in visceral (F1), ventral (F2, F3), and left lateral (F4) views. This discovery was cited by Hernández-Pacheco (1914). Scale bars: A E = 5cm; F = 1cm. THE MATERIAL FROM VILLALCÓN (PALENCIA) Analysed material Several fragmentary and poorly preserved elements, corresponding to the large specimen of Te. bolivari from Villalcón cited by Royo y Gómez (1934a), were identified. Shell remains Only a fragment of an indeterminate plate, with all the edges broken, was recognized (MNCN 38919). Its thickness is about 3 cm. Appendicular elements Few appendicular elements are preserved, mostly fragmented. From the shoulder girdle, only a fragment of the scapula is preserved (MNCN 38915), showing a concave articular surface. From the anterior limb, the right humerus is preserved (Fig. 23D). The proximal and distal parts are missing, leaving a massive diaphysis with a small degree of curvature. Owing to the preservation further description is not possible. The morphology of the proximal end is known only by a fragment of a humeral head that is preserved (MNCN 38914). The recovered fragment does not allow the identification of the side of the body, but we can see that the shape of the head is rounded and elliptical. Only one specimen is recovered from the posterior limb. This is a diaphysis of the femur (MNCN 38916), almost straight with only a small degree of curvature. From the preserved part we can see that the trochanters are fused ventrally. THE MATERIAL FROM LA CISTÉRNIGA (VALLADOLID) Analysed material Only a specimen corresponding to a large testudinid from La Cistérniga was found in the MNCN. It is a fragment of carapace (Fig. 23E). Description The preserved fragment consists of two partial neurals and the medial region of a costal plate (Fig. 23E). Its maximum width is about 14 cm. One of the neurals, subrectangular in morphology, is traversed by a sulcus located between two vertebral scutes. It is the third or the fifth neural. Anteroposteriorly, the proximal region of the dorsal ribs is very short, as in all testudinids. THE MATERIAL FROM ILLESCAS (TOLEDO) Analysed material As Royo y Gómez (1928a) indicated, the material from Illescas, which was identified in the MNCN, corresponds to a small tortoise (Fig. 23F). A peripheral plate and an anterior plastral lobe were found. Their morphology, seen in the context of morphological variation, indicates that they may correspond to a juvenile individual of a large tortoise (see Discussion). Description The peripheral plate, probably corresponding to the anterior region, has its lateral side dorsally directed. The pleuromarginal sulcus is located next to the suture between the peripheral with the costal series. The dimensions of the anterior plastral lobe (e.g. the length of the entoplastron is 41 mm) are compatible with those of that plate, probably corresponding to the same specimen. The margin of this lobe is rounded. The entoplastron, as wide as long, is subpentagonal. Its entoplastral lip is long and dorsally convex. The gular scutes overlap the anterior region of the entoplastron. The pectoral scutes do not contact the entoplastron. The humero-pectoral sulcus is perpendicular to the axial plane in the medial region, but has a well-developed lateral change in curvature, being markedly directed towards the anterior region. Medially, the pectoral scutes are very short. THE MATERIAL FROM PARLA (MADRID) Analysed material Handwritten labels corresponding to the specimens MNCN 48482, MNCN 48483, and MNCN 48484

40 692 A. PÉREZ-GARCÍA AND E. VLACHOS allowed us to recognize the material found by Fernández Navarro in Cerro de la Coronilla (Parla), identified by Hernández-Pacheco (1914) as Testudo sp. (see Appendix S1). Description The mentioned specimens correspond to small fragments of plates having a thickness compatible with that of the large testudinids identified in many of the above-cited locations. THE MATERIAL FROM VILLALUENGA DE LA SAGRA (TOLEDO) Analysed material The handwritten labels corresponding to the specimens MNCN and MNCN confirm that they correspond to the material from Villaluenga de la Sagra referred by Royo y Gómez (1928a) to Te. bolivari. The specimens catalogued as MNCN were found by Eduardo Hernández-Pacheco, José Royo y Gómez, and Francisco Hernández-Pacheco in January The others were found by José Royo y Gómez in 1922 (see Appendix S1). Description As in the case of the material from Parla, the specimens from Villaluenga de la Sagra correspond to small fragments of plates with a thickness similar to those of most of the large testudinids described in this paper. THE MATERIAL FROM CARABANCHEL (MADRID) Analysed material A handwritten label indicates that fragments of plates catalogued as MNCN correspond to those found by José Royo y Gómez in 1929 in Gaquillo (Carabanchel), and attributed by him to Te. bolivari (Royo y Gómez, 1929; see Appendix S1). Description MNCN corresponds to fragments of plates with thicknesses that support their attribution to large testudinids. SPECIMENS FROM INDETERMINATE CLASSIC LOCATIONS Analysed material Some thus far unpublished photographs allowed us to observe partial shells of specimens not preserved today, and whose precise location is not known (Fig. 24A D). Description The specimen corresponding to Figure 24A, B is composed of the anteriormost margin of a carapace and the anterior half of a plastron, probably from Cerro del Viso (Alcalá de Henares). The carapace shows an anterior notch extending across the anterior margin of both the nuchal plate and the first pairs of peripherals. The anterior plastral lobe is rounded. However, the anterior lobe of the partial shell of Figure 24C, D, from Cerro del Viso or Cerro del Otero, has a well-developed sagittal protrusion. Both the morphology of the posterior margin of the plastron, almost straight and subperpendicular to the axial axis, and the presence of a very curved pygal, allow its identification as a male individual. THE MATERIAL FROM CIUDAD UNIVERSITARIA (MADRID) Analysed material Although Royo y Gómez (1934b, 1935a, b) referred to the discovery of several specimens of Te. bolivari in Ciudad Universitaria, none of them could be found during the present study, and the available information is limited. Only some postcranial remains could be recognized in some hitherto unpublished photographs (e.g. Fig. 24E, F). The only element figured after its preparation, and accompanied by interpretive drawings, is the partial skull of one of the individuals (see plates 47 and 48 in Royo y Gómez, 1935a). Skull Royo y Gómez (1935a) published three photographs (accompanied by simplified drawings) of a partial skull from Ciudad Universitaria. The first two (Royo y Gómez, 1935a: plate 47) are dorsoposterior and left lateral views, whereas the third (Royo y Gómez, 1935a: plate 48) is an internal view of the skull roof, with some remains of the maxilla. The third photograph does not allow any clear morphological description. In the first two photographs we could not identify the cranial sutures with confidence (most of them also indicated by Royo y Gómez with dotted lines), but we were able to infer the general topology of the cranial bones. The cranium appears to have a wide skull roof with a long frontal/ prefrontal area, showing a small constriction in the level of the frontals. Laterally, we observed a wide fossa orbitalis. The elements of the zygomatic arch are thin and tall. The ventrolateral emargination expands to the jugal and quadratojugal, whereas the posterodorsal emargination appears to be less extensive than usual (based on the preserved parts). Furthermore, the skull appears to be relatively tall, compared with its hypothesized length. Appendicular elements From Ciudad Universitaria, no postcranial elements are preserved, but in the historical photographs ACN 3133 ( specimen 1 ) and ACN 3135 ( specimen 2 ) we noticed several appendicular elements. In the former

41 A GENUS FOR THE EUROPEAN LARGE TORTOISES 693 Figure 24. Unpublished historical photographs of some specimens of Titanochelon bolivari from several Spanish classic localities. A, B, partial shell (most anterior region of the carapace and anterior half of the plastron) prepared for exhibition, probably from the MN6 zone (late Aragonian, middle Miocene) of Cerro del Viso, in Alcalá de Henares (Madrid), in anterior (A, modified version of ACN 6455) and right lateral (B, modified version of ACN 6456) views. Original photographs taken by Francisco Molina. C, D, reconstruction and preparation for exhibition of a partial shell (preserving the most anterior and most posterior regions of the plastron and the partial carapace). It may come from the MN6 zone (late Aragonian, middle Miocene) of Cerro del Viso, in Alcalá de Henares (Madrid), or from the MN7/8 zone (late Aragonian, middle Miocene) of Cerro del Otero (Palencia). José Royo y Gómez is on the far left of photograph C and Eduardo Hernández- Pacheco on the right. Original photographs belong to the Eduardo Hernández-Pacheco collection. C, modified version of ACN D, modified version of ACN E, F, two of the shells found in 1934 in Ciudad Universitaria (Madrid), in levels corresponding to the MN5 MN6 interval (Aragonian, middle Miocene). E, individual identified by Royo y Gómez (1934b) as specimen 1, showing not only the partial shell but also the pelvis. Modified version of photograph ACN 3133, taken by Royo y Gómez in November 1934 during restoration of the specimen. F, individual identified by Royo y Gómez (1934b) as specimen 2b, showing the partial shell and several appendicular elements. Modified version of photograph ACN 3135, also taken by Royo y Gómez in November 1934 during restoration of the specimen. (Fig. 24E), we could clearly see a pelvis in ventral view, preserving most of the puboischiatic plate, which is flat and has two circular foramina preserved. In the latter it was very difficult to identify any skeletal elements, but several fragments are visible (Fig. 24F). THE MATERIAL FROM COCA (SEGOVIA) Two specimens from Coca were analysed here. The first one, found in 1988, preserves an almost complete shell and scarce appendicular bones (Fig. 25). The other, found in 1969, corresponds to a partial anterior half of a shell and to several appendicular elements (Fig. 26). It was identified as the neotype of Te. bolivari (see figures and preliminary description in Jiménez Fuentes, 1971). The specimen found in 1988 As stated, this specimen retains an almost complete shell and few appendicular elements (Fig. 25). From the shell, we analysed the dorsal carapace (Fig. 25A D), the anterior lobe of the plastron (Fig. 25E), and a resin copy of its complete plastron (Fig. 25F). Owing to the mounting position of the specimen and to some problems related to its preservation, some regions of the original plastron are not available for observation. Shell This large shell is relatively low (Fig. 25A D). The anterior region of the carapace has a notch, which is very shallow and broad, affecting the nuchal and the first peripherals. The pygal plate is curved, its posterior margin being anteroventrally directed. This contrasts with the morphology of the lateral region of the last pair of peripherals, which is dorsally directed. The nuchal plate is as wide as long. This specimen has eight neural plates, of which the fourth is missing. The first one is subrectangular, longer than wide. The third and fifth neurals are also subrectangular, but wider than long. The second neural is octagonal, being wider in the posterior half than in the anterior one. The last three neurals are hexagonal. There are two suprapygal plates. The second is lenticular. The

42 694 A. PÉREZ-GARCÍA AND E. VLACHOS Figure 25. Specimen of Titanochelon bolivari found in 1988 in the MN6 zone (late Aragonian, middle Miocene) of Coca (Segovia, Spain). A D, carapace in dorsal (A1, A2), left lateral (B1, B2), anterior (C1, C2), and posterior (D1, D2) views. E, preserved anterior lobe of the plastron in visceral (E1) and ventral (E2, E3) views. F, resin copy of the currently destroyed ventral area of the shell. G, partial left scapula (STUS 8372) in anterior (G1) and ventrolateral (G2) views. H, pelvis of the same specimen (STUS 12068) in right lateral (H1), posterior (H2), and ventral (H3) views. Scale bars: A F = 10 cm; G, H = 5 cm. analysed specimen lacks a cervical scute. The boundary between the vertebral scutes is situated on the first (in the posterior part), third (in the middle, being concave posteriorly), fifth (in the middle), and eighth (in the middle) neurals. It has a single supracaudal scute, which reaches the posterior region of the second suprapygal, and the medial parts of the last peripherals. The boundary between the pleural and the marginal scutes is

43 A GENUS FOR THE EUROPEAN LARGE TORTOISES 695 situated on the suture between the costal and the peripheral plates. The anterior plastral lobe is subrounded (Fig. 25E, F). However, the posterior lobe is trapezoidal. The middle posterior region of the plastron is concave. The epiplastral lip is long, its posterior margin reaching a position near the anterior border of the entoplastron. It is convex. This specimen has a small epiplastral pocket. The entoplastron is as long as wide, being narrow compared with the maximum width of the lobe. The gular scutes overlap the anterior region of that plate. The humero-pectoral sulcus, perpendicular to the axial plane in the medial region, but anteriorly directed in the lateral area, contacts the posterior margin of the entoplastron. Medially, the pectoral scutes are very short. The region where the anal scutes are located is thickened. Appendicular elements Few appendicular elements are preserved from the individual STUS 8372 from Coca. A part of the left scapula is preserved (Fig. 25G), being straight throughout its length and with a small concave surface for the humeral head. The cross-section of the scapular bone is elliptical. Moreover, the complete pelvis is also preserved (Fig. 25H). The pectineal processes of the pubic are long and curved. The same curvature is observed in the much shorter metischial processes, meaning that the whole puboischiatic plate is elevated and does not contact the plastron. The pelvis makes contact with the plastron only with the pectineal and metischial processes. The pubic and ischium are connected medially to form circular pubo-ischiatic foramina in the middle of the plate. The ilium is massive and long, and is preserved complete until the proximal part, which is wide with a straight edge and shows many scars for muscle attachments. The three pelvic elements compose a deep, concave, and elliptical to spherical in shape acetabulum for the femoral head. The specimen found in 1969 Besides several appendicular bones, only some elements corresponding to the anterior half of both the carapace and the plastron of this specimen were found (Fig. 26). Shell The preserved elements of the carapace correspond to the almost complete nuchal plate, the complete first and second neurals, the anterior region of the third neural, the complete first right costal, the medial region of the first left costal and of the second and third pairs of costals, the two first right peripherals, and the partial third and fourth right peripherals (Fig. 26A). A broad and shallow anterior notch is present. The first neural is subrectangular, longer than wide, its posterior part being crossed by the limit between the first and second vertebrals. The second neural is octagonal, its maximum width being in the posterior half. This specimen lacks a cervical scute. The boundary between the pleural and marginal scutes coincides with the suture between the costal and the peripheral plates. A relatively poorly preserved part of its anterior plastral lobe (corresponding to the complete left entoplastron and much of the right one, the anterior half of the entoplastron and a part of the posterior half, and the anteromedial margin of the left hyoplastron) is preserved (Fig. 26B). The morphology of the anterior plastral margin is interpreted as subrounded or slightly pointed. It has a dorsally convex epiplastral lip. It is lower and shorter than that of the specimen STUS 8372, and it lacks an epiplastral pocket. The gular scutes are situated on the anterior region of the entoplastron. Shoulder girdle From the shoulder girdle both scapulae are known but are not entirely complete. The right one (Fig. 26C) preserves only the part with the articular surface, and the acromion is missing. The left one (Jiménez Fuentes, 1971: fig. 4L N) is more complete, preserving part of the acromion, but missing the proximal tip of the scapula. The articular surface is elliptical in shape and slightly concave. The left coracoid is also preserved (Fig. 26D), having a narrow neck, and a wide, triangular fan-shaped end, with most of the edge missing. Anterior limb Both humeri are preserved, almost entirely complete (Fig. 26E, F); only the minor trochanter is missing in the left one. The humerus has a narrow diaphysis with wide ends and a strong curvature over its entire length. The trochanters show a great difference in size, with the major trochanter being much longer than the minor, extending beyond the height of the humeral head. They diverge slightly from each other. The head is elliptical in shape, and is developed with an angle of approximately 120 to the diaphysis. The distal part of the humerus is marked by the presence of a shallow ectepicondylar groove in dorsal view, whereas the distal end of the bone is very asymmetric because the ulnar condyle is larger than the radial. No radius is preserved. The left ulna is complete, being high and narrow (Fig. 26G). The proximal part has a slightly concave articular surface for the humerus. The midshaft is triangular in cross-section. The distal part has a narrow articular surface for the radius. Pelvic girdle From the pelvis several parts are preserved, the most notable being the left part (Fig. 26H), preserving a deep, concave, and almost spherical acetabulum for the femoral head. The ilium is very tall, with a wide end

44 696 A. PÉREZ-GARCÍA AND E. VLACHOS Figure 26. Specimen of Titanochelon bolivari found in 1969 in the MN6 zone (late Aragonian, middle Miocene) of Coca (Segovia, Spain). It was identified as the neotype of this species by Jiménez Fuentes et al. (1988). A H, partial shell and some appendicular elements of the specimen STUS 352. A, anterior region of the carapace in dorsal (A1, A2) view. B, anterior lobe of the plastron in visceral (B1) and ventral (B2, B3) views. C, right scapula in anterior (C1) and ventrolateral (C2) views. D, coracoid in dorsal view. E, left humerus in anterior (E1), dorsal (E2), posterior (E3), and ventral (E4) views. F, right humerus in anterior (F1), dorsal (F2), posterior (F3), and ventral (F4) views. G, left ulna in dorsal (G1), medial (G2), ventral (G3), and lateral (G4) views, and view of the proximal articular surface (G5). H, pelvis in left lateral view. Scale bar = 5 cm. and a straight top border. Further elements of the pelvis were figured by Jiménez Fuentes (1971: fig. 4), from the puboischiatic plate, but they are too fragmented to allow further description. Other elements Additional elements were figured by Jiménez Fuentes (1971: fig. 4), such as posterior unguals, which are pointed and elongated, and flat, rounded osteoderms. THE MATERIAL FROM BARAJAS (MADRID) Analysed material Two new specimens from Barajas were included in our study. The first individual (MAR 2012/20/1) preserves a partial carapace and some appendicular elements in nearly anatomical position. The state of preservation of the carapace does not allow any information on sutures and sulci, but we can infer that the carapace (although deformed) was relatively low.

45 A GENUS FOR THE EUROPEAN LARGE TORTOISES 697 The appendicular elements are poorly preserved. We were only able to identify some rounded, half-pointed osteoderms covering some bones of the right forelimb. The second individual (MAR 2012/20/2) is in a very good state of preservation, and was analysed here in detail (Fig. 27). This is a relatively complete and well-preserved skeleton, prepared in ventral view, which preserves the shell and generally articulated appendicular elements. Shell The morphology of the anterior plastral lobe of the shell is not known. However, its plastral posterior lobe is well preserved. Its lateral margins are markedly divergent towards the posterior region. It has a wider than long anal notch. This morphology, together with other characters such as the lack of concavity in the middle posterior region of the plastron and the absence of a ventral thickening in the area cover by the anal scutes, allow its identification as a female individual. The entoplastron is subhexagonal, wider than long. The gular scutes overlap the anterior region of this plate. The pectoral scutes are not in contact with the entoplastron, the humero-pectoral sulcus being situated posterior to that plate. The distance between that sulcus and the entoplastron is approximately half of the medial length of the pectorals. Medially, this pair of scutes is very short. The pectoro-abdominal sulcus is perpendicular to the axial plane in the medial region, but has a well-developed lateral change of curvature, being markedly directed towards the anterior region. The femoro-anal sulcus has omega-shape morphology. Cranial remains The only preserved cranial remains are two fragments of the branchial horns or hyoid bones, found on the anterior part (Fig. 27B1, 2). One is rather small and straight in shape, found next to the dorsal vertebra. The other is longer and strongly curved, and was recovered next to the unguals of the anterior limb. Both are long and cylindrical elements. Appendicular skeleton The appendicular elements and osteoderms of the right anterior and posterior limbs are preserved in the position in which they were recovered, resting in the anterior and posterior openings of the shell (Fig. 27A1, 2). Most of the bones are also preserved in anatomical connection, with very few exceptions; the branchial horns and vertebral remains are the exceptions. Many osteoderms cover parts of the limbs, but some of them have been moved from their natural position. This is one of the rare cases of a discovery of large tortoise specimens that are preserved in anatomical position and is important for the detailed description of the anatomy of this taxon. Our analysis shows that all limb elements are preserved (although some carpal and tarsal elements cannot be observed because of the recovery position), constituting one of the most complete records of the appendicular skeleton of a fossil large tortoise in Europe. Vertebral column Few remains of the vertebrae were found. Anteriorly, the dorsal view of a posterior cervical vertebra and the posterodorsal view of the last cervical are available (Fig. 27B1, 2), not in their natural position. The first of these vertebrae is elongated, and preserves the anterior and posterior zygapophyses, as well as a low neural spine. The recovery position does not allow the morphology of the articular surfaces to be known. The eighth cervical vertebra is wider than the other, and has a simple and convex posterior articular surface. In the posterior part, behind the anal notch of the xiphiplastra, at least five remains of tail vertebrae were identified (Fig. 27D). Although not articulated, the recovery position of these elements resembles the natural tail morphology. The five vertebra show the typical structure of the caudal region of the testudininds, being procoelous and preserving (at least for the anterior part of the tail) small transverse processes. The size of the vertebra and the length of the centrum is reduced towards the end of the tail. Anterior limb Only the right anterior limb is preserved in the anterior opening of the shell, recovered in an almost anatomical position and resting on the ventral side of the anterior peripherals. The humerus, radius, and ulna are preserved in articulation (Fig. 27B1, 2). Many rounded and flat osteoderms of varied sizes cover the limb, covering most carpal elements as well (Fig. 27B2, indicated with grey colour). In the anterior part of the limb, the phalanges are preserved with their corresponding unguals. This side is the dorsal, showing that the limb of this individual was covered by a large number of osteoderms. The humerus is a massive element, with a curved diaphysis and a wide distal part. The distal part is asymmetrical, as a result of the relative sizes of ulnar and radial condyles, corresponding to the morphology of the articulated radius and ulna. The radius and ulna were not recovered articulated, as the ulna has been moved from its natural position. The distal parts of radius and ulna are covered by a number of large and flat osteoderms, which also prevent the study of the carpal elements. However, in other views (Fig. 27B3, 4) we can see that probably all of the carpals have been preserved. All phalanges are preserved, showing the typical morphology, with a concave proximal articular surface and convex distal surface for the unguals. The phalanges of the anterior limb are thick and short. The unguals are

46 698 A. PÉREZ-GARCÍA AND E. VLACHOS Figure 27. MAR 2002/20/2, specimen of Titanochelon bolivari from the MN5 zone (Aragonian, middle Miocene) of Barajas (Madrid, Spain). A1, A2, ventral view of the specimen. B, articulated right forelimb in ventral (B1, B2), lateral (B3), and ventral (B4) views. C, ventral view (C1, C2) of the articulated right hindlimb. D, caudal region, showing several vertebrae of the tail. Osteoderms are indicated in grey and bones in white. Abbreviations: as+cal, astragalocalcaneum; bh, branchial horn; cv, cervical vertebra; dt, distal tarsal; fe, femur; hu, humerus; mt, metatarsal; px, phalanx; ra, radius; ti, tibia; ul, ulna; un, ungual. Scale bars =5cm. preserved in their natural position, and are thick and wide, with a rounded tip at the end. Posterior limb The entire right posterior limb of this individual is preserved in the posterior opening of the shell (Fig. 27C1). This limb is also covered by a large number of osteoderms, but in this case there are no osteoderms covering the limb. A large amount of osteoderms was recovered around the pes of the tortoise (Fig. 27C2). The femur shows only a small curvature in the midshaft, but most of the bone is covered by the sediment, preventing further description. The tibia is massive, having a wide triangular proximal part with a smooth and nonprominent cnemial crest. Although the fibula is not visible in the recovered view, this element is probably preserved below the tibia. All the remaining tarsal elements are preserved in very good condition. The astragalocalcaneum is only visible in ventral view, being massive and rounded. Proximally, it is articulated with the tibia and fibula. Distally, it is articulated with a row of distal tarsals, which are discoid and rounded.