This is the accepted version of the journal article: Urciuoli, Alessandro; Zanolli, Clément; Almécija, Sergio; [et al.]. «Reassessment of the phylogenetic relationships of the late Miocene apes Hispanopithecus and Rudapithecus based on vestibular morphology». Proceedings of the National Academy of Sciences of the United States of America, Vol. 118, Issue 5 (January 2021), art. e2015215118. DOI 10.1073/pnas.2015215118 This version is available at https://ddd.uab.cat/record/258412 under the terms of the license 1 BIOLOGICAL SCIENCES: Anthropology 2 3 Reassessment of the phylogenetic relationships of the late Miocene apes 4 Hispanopithecus and Rudapithecus based on vestibular morphology 5 6 Alessandro Urciuolia,*, Clément Zanollib, Sergio Almécijac,d,a, Amélie Beaudete,f, Jean 7 Dumoncelg, Naoki Morimotoh, Masato Nakatsukasah, Salvador Moyà-Solàa,i,j, David 8 R. Begunk, David M. Albaa,* 9 a 10 Institut Català de Paleontologia Miquel Crusafont, Universitat Autònoma de 11 Barcelona, Edifici ICTA-ICP, c/ Columnes s/n, Campus de la UAB, 08193 12 Cerdanyola del Vallès, Barcelona, Spain; bUniv. Bordeaux, CNRS, MCC, PACEA, 13 UMR 5199, F-33600 Pessac, France; cDivision of Anthropology, American Museum 14 of Natural History, Central Park West at 79th Street, New York, NY 10024, USA; d 15 New York Consortium in Evolutionary Primatology, New York, USA; eSchool of 16 Geography, Archaeology and Environmental Studies, University of the 17 Witwatersrand, Private Bag 3, Johannesburg, WITS 2050, South Africa; fDepartment 18 of Anatomy, University of Pretoria, PO Box 2034, Pretoria, 0001, South Africa; g 19 Laboratoire AMIS, UMR 5288 CNRS, Université de Toulouse, Allées Jules Guesde 20 37, 31073, Toulouse, France; hLaboratory of Physical Anthropology, Graduate 21 School of Science, Kyoto University, Kyoto, 606 8502, Japan; iInstitució Catalana de 22 Recerca i Estudis Avançats (ICREA), Passeig de Lluís Companys 23, 08010 23 Barcelona, Spain; jUnitat d’Antropologia (Departament de Biologia Animal, Biologia 24 Vegetal i Ecologia), Universitat Autònoma de Barcelona, Campus de la UAB s/n, 25 08193 Cerdanyola del Vallès, Barcelona, Spain; kDepartment of Anthropology, 26 University of Toronto, Toronto, ON M5S 2S2, Canada 27 * 28 To whom correspondence should be addressed: 29 1 30 Alessandro Urciuoli 31 Postal address: Institut Català de Paleontologia Miquel Crusafont, Universitat 32 Autònoma de Barcelona, Edifici ICTA-ICP, c/ Columnes s/n, Campus de la UAB, 33 08193 Cerdanyola del Vallès, Barcelona, Spain 34 Phone: +34 622112509 35 Email: alessandro.urciuoli@icp.cat 36 37 David M. Alba 38 Postal address: Institut Català de Paleontologia Miquel Crusafont, Universitat 39 Autònoma de Barcelona, Edifici ICTA-ICP, c/ Columnes s/n, Campus de la UAB, 40 08193 Cerdanyola del Vallès, Barcelona, Spain 41 Phone: +34 935868604 42 Email: david.alba@icp.cat 43 44 Keywords: inner ear, semicircular canals, evolution, fossil apes, Hominidae. 45 46 ORCID: 47 Alessandro Urciuoli: 0000-0002-6265-8962 48 Clément Zanolli: 0000-0002-5617-1613 49 Sergio Almécija: 0000-0003-1373-1497 50 Amélie Beaudet: 0000-0002-9363-5966 51 Jean Dumoncel: 0000-0003-0789-7458 52 Naoki Morimoto: 0000-0002-8367-4777 53 Masato Nakatsukasa: 0000-0002-6897-8027 54 Salvador Moyà-Solà: 0000-0001-8506-1061 55 David R. Begun: 0000-0001-8916-6442 56 David M. Alba: 0000-0002-8886-5580 57 58 2 59 Abstract 60 Late Miocene great apes are key to reconstructing the ancestral morphotype 61 from which earliest hominins evolved. Despite consensus that the late 62 Miocene dryopith great apes Hispanopithecus laietanus (Spain) and 63 Rudapithecus hungaricus (Hungary) are closely related (Hominidae), ongoing 64 debate on their phylogenetic relationships with extant apes (stem hominids, 65 hominines, or pongines) complicates our understanding of great ape and 66 human evolution. To clarify this question, we rely on the morphology of the 67 inner ear semicircular canals, which has been shown to be phylogenetically 68 informative. Based on microcomputed tomography scans, we describe the 69 vestibular morphology of Hispanopithecus and Rudapithecus, and compare 70 them with extant hominoids using landmark-free deformation-based 3D 71 geometric morphometric analyses. We also provide critical evidence about the 72 evolutionary patterns of the vestibular apparatus in living and fossil hominoids 73 under different phylogenetic assumptions for dryopiths. Our results are 74 consistent with the distinction of Rudapithecus and Hispanopithecus at the 75 genus rank, and further support their allocation to the Hominidae based on 76 their derived semicircular canal volumetric proportions. Compared with extant 77 hominids, the vestibular morphology of Hispanopithecus and Rudapithecus 78 most closely resembles that of African apes, and differs from the derived 79 condition of orangutans. However, the vestibular morphologies reconstructed 80 for the last common ancestors of dryopiths, crown hominines, and crown 81 hominids are very similar, indicating that hominines are plesiomorphic in this 82 regard. Therefore, our results do not conclusively favor a hominine or stem 83 hominid status for the investigated dryopiths. 3 84 85 Significance 86 Reconstructing the phylogenetic relationships of extinct apes is challenging due to 87 their fragmentary fossil record and the recurrent independent evolution of 88 morphological features. Given the relevance of the phylogenetic signal of the bony 89 labyrinth, here we assess the phylogenetic affinities of the late Miocene great apes 90 Hispanopithecus and Rudapithecus by studying their inner ear morphology. Our 91 results are consistent with the distinct generic status of these dryopiths, which further 92 differ from the derived condition of orangutans and most closely resemble African 93 apes. However, the latter appear largely primitive (similar to the last common 94 ancestor of great apes and humans), hence our results do not conclusively favor a 95 closer relationship with African apes as opposed to great apes as a whole. 96 97 4 98 Hominoids (apes and humans) originated in Africa during the Oligocene (1) but 99 subsequently dispersed into Eurasia, giving rise to an impressive radiation during the 100 middle and late Miocene (2-3). Thus, while extant hominoids include only two 101 moderately diverse families—hylobatids (gibbons and siamangs) and hominids 102 (great apes and humans)—the panoply of extinct genera recorded during the 103 Miocene still defies classification into a coherent systematic scheme. Other than the 104 late Miocene Oreopithecus—which might be a late occurring stem hominoid (4, 5) — 105 there is consensus that most Eurasian large-bodied hominoids are members of the 106 great-ape-and-human clade (Hominidae) (2 ,3 ,6). While most Asian extinct great 107 apes such as Sivapithecus are considered to be more closely related to the 108 orangutan clade (Ponginae) than to African apes and humans (Homininae) (2, 6-8), 109 the phylogenetic affinities of European Dryopithecus and allied forms have long been 110 debated. Until a decade ago, several species of European apes from the middle and 111 late Miocene were included within this genus (9-16). However, discoveries at the 112 middle Miocene composite section of Abocador de Can Mata (6, 17-20) prompted 113 the recognition that the late Miocene species belong to one or more different genera 114 distinct from/other than Dryopithecus (2-3, 6, 7, 18, 21-26): Hispanopithecus from 115 Spain, and Rudapithecus from Hungary, the latter formerly considered a subgenus of 116 the former by some authors (6, 18, 22). 117 Together with Dryopithecus and other middle to late Miocene taxa (17, 19, 27), 118 Hispanopithecus and Rudapithecus are currently classified in a subfamily 119 (Dryopithecinae) (6, 20, 26) or tribe (Dryopithecini) (3, 7, 21) of their own, distinct 120 from pongines. Both taxa possess a hominid-like cranial morphology (6,11-13, 21, 121 25, 29-30), as shown by the high zygomatic root, reduced midfacial prognathism, 122 lack of subarcuate fossa, deep glenoid fossa, and prominent entoglenoid process. 5 123 However, there is no consensus regarding the phylogenetic position of this group— 124 being either considered stem hominids (6, 19, 28), stem hominines (2-3, 14, 16, 25), 125 or even pongines (10, 29-30) —which may be informally referred to as ‘dryopiths’. 126 Resolving the phylogenetic position of dryopiths has important implications for the 127 evolution of the great ape and human clade, since their purported hominine status 128 has led to paleobiogeographic scenarios favoring a European origin and subsequent 129 back-to-Africa dispersal for the African and human clade (2-3, 15, 24-25). 130 Disagreements and uncertainties about the phylogenetic position of extinct apes are 131 persistent, and stem from a combination of factors, including the incomplete and 132 fragmentary hominoid fossil record, the decimated current diversity of the group, and 133 pervasive homoplasy coupled with mosaic evolution (6, 23, 31-35). 134 The morphology of the semicircular canals (SCs), which partly constitute the inner 135 ear’s bony labyrinth, has been classically related to locomotion (36-42). However, 136 several studies have highlighted the possibility of inferring phylogenetic relatedness 137 based on this portion of the inner ear morphology (43-47) Recently, it has been 138 shown that this anatomical structure also embeds a strong phylogenetic signal 139 among catarrhine primates by means of 3D geometric morphometric (3DGM) 140 analyses (5, 47-49), thus being potentially useful to test phylogenetic hypotheses for 141 extinct hominoids. Previous studies relied on the SC radius of Rudapithecus 142 hungaricus and Hispanopithecus laietanus to infer slow and deliberate arboreal 143 locomotion for these species (41). However, recent analyses raised doubts about the 144 reliability of locomotor behavior predictions based on SC radius only (50-51). In 145 contrast, here we rely on µCT scans of the same specimens and a deformation- 146 based (landmark-free) 3DGM approach to assess their closest affinities in SC 147 morphology with extant hominoids and interpret them from an evolutionary viewpoint. 6 148 First, we describe the fossil remains and qualitatively compare them with extant 149 hominoids. Second, we assess if the volumetric proportions of their SCs more 150 closely resemble those of hominids than those of other anthropoids. Third, we 151 quantitatively evaluate changes in SC and vestibule morphology by means of a 152 between-group principal component analysis (bgPCA) applied to a sample of extant 153 and extinct hominoids. The affinities of the investigated fossil taxa are further 154 assessed by means of cluster analyses and group membership probabilities based 155 bgPCA results. Finally, we reconstruct the evolutionary history of the hominoid SCs 156 using a phylomorphospace approach (including reconstructed ancestral 157 morphotypes) under various phylogenetic assumptions for dryopiths. 158 159 Results 160 Descriptions and comparisons. Three-dimensional renderings of the vestibular 161 apparatus of fossil and extant hominoids investigated here are illustrated in the 162 Figure 1a-m. The vestibular apparatus of R. hungaricus is well preserved in the three 163 available specimens (Fig. 1a-c). As in extant hominids, the SCs are stout—although 164 less so than in orangutans (Fig. 1j), most humans (Fig. 1m) and gorillas (Fig. 1k)— 165 and the vestibule is large relative to the volume occupied by SCs. The anterior and 166 posterior canals are large and similar in size (Fig. 1a,b). The anterior canal is slightly 167 vertically compressed, as in extant hominoids and the fossil apes Nyanzapithecus 168 alesi (4) and Nacholapithecus kerioi (Fig. 1f), and somewhat larger in RUD 77 than 169 in RUD 200. The anterior canal is somewhat anterosuperiorly projecting, albeit much 170 less so than in Pongo (Fig. 1j) and Oreopithecus (Fig. 1e). The lateral and posterior 171 canals are slightly different between the two individuals. In RUD 77, the lateral canal 172 is noticeably smaller than the other SCs (Fig. 1a,b), slightly compressed horizontally, 7 173 and slenderer than in RUD 200. The lateral canal of RUD 200 is stout and large, 174 almost reaching the size of the vertical SCs (similar to the condition in African apes, 175 yet smaller than in gorillas; Fig. 1c), and its slender portion connects with the 176 vestibule somewhat more inferiorly than in RUD 77. The junction of the slender 177 portion of the lateral canal and the ampulla further differs between the two 178 individuals, as it protrudes anteriorly in RUD 200, while it is posterolaterally oriented 179 in RUD 77. In both individuals, the ampullary portion bends superiorly and the 180 slender segment between the connection with the vestibule and the posterolateral tip 181 of the lateral canal is straight, as in Hoolock (Fig. 1g) and in most hominids (Fig. 1j- 182 m)—except for some Gorilla and Pan specimens that show some curvature. 183 However, this section of the canal is more laterally oriented in RUD 77, while it is 184 almost parallel to the posterior canal in RUD 200. The posterior canal is elongated 185 posterolaterally in RUD 77, as in gorillas (Fig. 1k) and some humans (Fig. 1m), while 186 it is slightly more rounded in RUD 200 (Fig. 1c). In both RUD 77 and RUD 200, the 187 posterior and lateral canals approximately define a right angle (slightly more obtuse 188 in RUD 77) and the trajectory of the lateral canal does not intersect the plane 189 identified by the posterior canal. The common crus (CC) is short and slender, with 190 the slender portions of the anterior and posterior canals almost forming right angle at 191 the CC apex. The SCs are almost coplanar, with a slight amount of torsion in the 192 upper portion of the anterior one (the tip slightly bending medially), in the medial- 193 most part of the posterior one (displaced anteriorly), and in the tip of the lateral canal 194 (pointing inferiorly). 195 The vestibular apparatus of H. laietanus (Fig. 1d) differs from that of Rudapithecus 196 (especially RUD 77; Fig. 1a,b) by being more voluminous and displaying more 197 equally developed SCs. The larger volume is particularly appreciable on the 8 198 vestibular recesses (which are more voluminous than the SCs, as in orangutans; Fig. 199 1j) and in the much more inflated ampullae. The anterior canal is more vertically 200 compressed than in Rudapithecus, showing an almost rectangular shape. This canal 201 is also much slenderer than in orangutans (Fig. 1j) and gorillas (Fig. 1k), most 202 closely resembling chimpanzees (Fig. 1l). The lateral canal is stouter than the others, 203 especially in the ampullary portion. Its posterolateral-most tip slightly bends inferiorly, 204 resulting in a moderate torsion of the canal. The slender segment between the 205 connection with the vestibule and the posterolateral tip of the lateral canal is straight, 206 as in Rudapithecus (Fig. 1a-c), Hoolock (Fig. 1g) and most hominids (Fig. 1j-m), and 207 laterally oriented, as in Pongo (Fig. 1j), some humans (Fig. 1m) and RUD 77 (Fig. 208 1a-b). The ampullary portion of the lateral canal is bent superiorly, as in 209 Rudapithecus (Fig. 1a-c) and extant hominoids (Fig. 1g-m). However, unlike extant 210 great apes (Fig. 1j-l) and Rudapithecus (Fig. 1a-c), the portion between the ampulla 211 and the tip of the lateral canal is inflated. The posterior canal is small and rounded, 212 with a large ampulla. The CC is longer than in Rudapithecus (Fig. 1a-c) and in most 213 extant great apes (with Pongo showing the shortest), yet more inflated (even if much 214 less so that in orangutans; Fig. 1j), and the CC apex forms an obtuse angle. As in 215 Rudapithecus and extant hominids, the planes identified by the lateral and posterior 216 canals form a right angle and their trajectories do not intersect. 217 218 Volumetric proportions. Allometric regressions of SC volume vs. length were 219 performed separately for hominoids and the rest of anthropoids included in the 220 sample (Fig. 2a; measurements for the dryopiths are given in Supplementary Table 221 1) because it has been previously shown that the former display an allometric grade 222 shift toward relatively higher volumes at a comparable length once size-scaling 9 223 effects have been taken into account (5), with only minimal overlap. Hispanopithecus 224 falls above the hominid regression line, while Rudapithecus is situated more (RUD 225 77) or less (RUD 200) below the line, close to Nacholapithecus, but in all cases 226 within the range of extant hominids and well above the regression line of other 227 anthropoids (Fig. 2a). Gorillas are variable in this regard, while humans and 228 orangutans display stouter proportions than chimpanzees and bonobos (Fig. 2b). 229 The SCs of Hispanopithecus appear intermediate between these aforementioned 230 taxa (closer to humans and orangutans), while those of Rudapithecus, Oreopithecus, 231 and Nacholapithecus are slenderer and more comparable to those of chimpanzees 232 and bonobos. Overall, given their range of variation, all the extinct apes analyzed 233 here display extant hominid-like volumetric proportions of the vestibular apparatus. 234 235 Shape analysis. The bgPCA (Fig. 3), based on the deformation fields computed for 236 the hominoid sample, allows us to discriminate extant hominoid species, as shown 237 by classification results (99% of correctly classified individuals before and after cross 238 validation). These results closely resemble those computed using a cross-validated 239 bgPCA (Supplementary Fig.1). We also recover very significant group mean 240 differences (p<0.001) for the raw shape data (Supplementary Table 2), confirming 241 that group structure does not artifactually result from the bgPCA (52). Indeed, group 242 differences account for a substantial amount of variance (R2) in the raw shape data, 243 indicating that group separation is not spurious (52), although intergroup variance is 244 increased to a similar extent by the standard bgPCA space and the cross-validated 245 bgPCA (Supplementary Table 2). 246 bgPC1 (40.7% of the total variance) pulls apart hominids (mostly positive values) 247 from hylobatids (negative values), with no overlap. Positive values along this axis 10 248 indicate short and bulgy SCs, together with a right angle between the anterior and 249 posterior semicircular canals. Orangutans and humans display the most extreme 250 condition due to the stoutness of their SCs. Chimpanzees, bonobos, and gorillas 251 show a broad range of variation, with some individuals close to the origin due to their 252 somewhat slenderer SCs (albeit less so than in hylobatids, which display negative 253 values), and others overlapping with Pongo and Homo. Along bgPC1, 254 Hispanopithecus overlaps with australopiths, extant great apes, and humans, while 255 the Rudapithecus specimens fall within the African great ape range. Both RUD 77 256 and RUD 200 closely approach the origin, with the latter showing slightly more 257 positive values. Oreopithecus and Nacholapithecus are found on moderate negative 258 values, within the lower range of Pan and Gorilla, due to their quite slender SCs 259 (albeit clearly stouter than in hylobatids). 260 The patterns of shape variation captured by bgPC2 (33.4% of total variance; Fig. 261 3a) reflect changes in the shape of three canals as well as their relative proportions. 262 In particular, bgPC2 clearly discriminates Homo (with most negative values) from the 263 rest of the sample, due to the presence in the former of a large and rounded 264 (sometimes even slightly superiorly elongated) anterior canal, a posterolaterally 265 displaced inferior portion of the posterior canal, and a small, fairly anterolaterally 266 elongated lateral canal, whose slender portion connects to the vestibule more 267 superiorly and anterolaterally than in apes. The latter fall on intermediate and 268 positive values, with hylobatids considerably overlapping with Pan spp. (Fig. 3a). To 269 a large extent, this is due to their anterior canal shape, which appears intermediate 270 between the rounded morphology of humans and the marked vertical compression of 271 Pongo and Gorilla (the latter taxa occupying more positive values with only very 272 slight overlap with Pan and hylobatids). Both Rudapithecus and Hispanopithecus, 11 273 like Oreopithecus and Nacholapithecus, show intermediate values along this axis, 274 overlapping with hylobatids and Pan spp. (as well as the Australopithecus specimen 275 StW 573), but not with Pongo and Gorilla. Conversely, the other australopith (StW 276 578) more closely approaches humans due to its larger vertical SCs. 277 bgPC3 (11.4% of variance; Fig. 3b) is driven by the shape of the anterior canal, its 278 relative size relative compared with that of the lateral one, the length of the CC, and 279 the amount of torsion of the lateral canal. Thus, negative values reflect a large and 280 anterosuperiorly projecting anterior canal, coupled with a small lateral one, and a 281 short CC. This axis discriminates Pongo (most negative values) from the rest of the 282 sample, only minimally overlapping with some Hylobates. One individual of 283 Rudapithecus (RUD 77) and Oreopithecus overlap with the range of orangutans due 284 to their anterosuperiorly projecting anterior canal (albeit less so in RUD 77), short 285 CC, and markedly small lateral canal. A similar morphology of the anterior canal is 286 also found in some Hylobates and in one of the Australopithecus specimens (StW 287 573), resulting in moderately negative scores. Hispanopithecus and the other 288 individual of Rudapithecus (RUD 200) fall at the negative end of the gorilla and 289 human variation, due to their intermediate anterior canal morphology, longer CC (yet 290 less so than in most Pan and Gorilla individuals), and a larger lateral canal. 291 Nacholapithecus and the other australopith specimen (StW 578) fall among 292 moderate positive values, overlapping with gorillas, chimpanzees, bonobos, humans, 293 and hylobatids, due to their long CC and more vertically aligned (i.e., superiorly 294 directed) connection of the anterior canal with the CC. 295 When the inspected bgPCs are considered simultaneously to compute posterior 296 probabilities of group membership (Table 1), the Rudapithecus RUD 77 individual 297 occupies a position in the morphospace that does not fit well with most extant 12 298 hominoid genera (p < 0.05), rather approaching the position of Nacholapithecus and 299 Oreopithecus in the morphospace (Table 2). Conversely, RUD 200 shows 300 considerable similarities with Pan (p = 0.549) and Nacholapithecus. The three 301 Rudapithecus specimens fall closer to one another than either approaches the single 302 specimen of Hispanopithecus (Table 2), which is also more distant than 303 Nacholapithecus from all the considered specimens (Table 2). Hispanopithecus 304 mostly differs along bgPC1, sharing similarities in the volumetric proportions of the 305 SCs and in the vertically compressed anterior canal morphology with 306 Australopithecus individual StW 573 (Table 2). IPS18000 marginally differs from Pan 307 (p = 0.053) and is clearly an outlier compared to the remaining extant genera. 308 The cluster analyses based on the significant bgPCs (Fig. 4a) and raw shape data 309 (Fig. 4b) further support the aforementioned results, since Rudapithecus and 310 Hispanopithecus do not cluster with one another and show affinities with different 311 taxa. In particular, the cluster based on the bgPCA results (Fig. 4a) indicates that 312 Rudapithecus is most similar to both Pan and Nacholapithecus, while 313 Hispanopithecus approaches hominins. This is further supported by the raw shape 314 data cluster (Fig. 4b), which mainly differs by recovering a great ape cluster. 315 316 Phylomorphospace and reconstruction of ancestral morphologies. The shape 317 data, as captured by the bgPCA performed on the extant hominoid sample, 318 approaches the Brownian motion (BM) model of evolution, as supported by the 319 phylogenetic signal computed for the bgPCs (Kmult=0.864, p=0.019) and for the raw 320 data (i.e., the deformation fields; Kmult=0.863, p=0.017). We used phylogenetically 321 informed techniques on the shape data to visualize the direction and magnitude of 322 vestibular shape change during hominoid evolution as well as to depict the internal 13 323 nodes of the phylogeny—i.e., the inferred vestibular morphology of the last common 324 ancestors (LCAs) of major groups—as reconstructed by maximum likelihood. The 325 results are very similar irrespective of the precise phylogenetic placement of 326 dryopiths as stem hominines (Fig. 5; Supplementary Fig. 2b), stem hominids 327 (Supplementary Fig. 3a, c; Supplementary Fig. 2a), or stem pongines 328 (Supplementary Fig. 3b, d; Supplementary Fig. 2c). The crown hominoid LCA (Figs. 329 5, 6a) is reconstructed as possessing evenly-sized and moderately inflated SCs, a 330 moderately long and not inflated CC, a fairly vertically compressed, yet not 331 anterosuperiorly projecting, anterior canal, an almost rounded posterior canal, an 332 obtuse angle between the planes identified by the anterior and posterior canals 333 (close to the right angle), and a right angle among the SCs merging at the CC apex 334 (Fig. 6a). Irrespective of the phylogenetic assumptions for dryopiths, the 335 reconstructed LCA for crown hominoids is closer to hominids (especially 336 Nacholapithecus, Rudapithecus and, among extant taxa, Pan) than to hylobatids 337 (Fig. 5), in terms of their intermediate volumetric proportions, contrasting with the 338 markedly slenderer SCs of gibbons and siamangs. In turn, the LCAs of crown 339 hominines and dryopiths (Rudapithecus + Hispanopithecus) closely resemble one 340 another irrespective of the underlying phylogenetic assumptions for the fossil species 341 (Fig. 5, Supplementary Fig. 3), being extant-hominid-like in volumetric proportions 342 but otherwise showing a more plesiomorphic morphology in the evenly sized and 343 fairly rounded SCs. 344 The inferred LCA of crown hominids, in particular, closely resembles that of crown 345 hominoids, except for the stouter volumetric proportions, more derived toward the 346 extant hominid condition (Figs. 5, 6b). It displays equally-sized SCs, an obtuse to 347 right angle in the apex of a moderately long CC, and a slightly laterally elongated 14 348 posterior canal (Fig. 6b). Orangutans appear derived from the LCA by displaying 349 more inflated SCs (especially the anterior one, Figs. 1j, 6b), further diverging in the 350 opposite direction from African great apes and humans because its short and 351 extremely stout CC, as well as its anterosuperiorly projecting anterior canal and 352 marked torsion of the lateral canal (Fig. 5b). The LCA of hominines (Fig. 6c) appears 353 somewhat more derived than the LCAs of hominoids and hominids for both 354 volumetric proportions and SC shape. It displays moderately stout SCs and 355 medium/large vestibular recesses, equally developed SCs (with a slightly smaller 356 lateral one), a vertically compressed anterior canal (more so than in any other LCA), 357 a slightly laterally projecting posterior canal, and a long CC with an obtuse angle in 358 its apex (Fig. 6c). Homo and Gorilla would have evolved in opposite directions from 359 this ancestral morphology in terms of SC relative size, with humans showing the 360 largest vertical canals (Figs. 1m, 5a) and gorillas displaying a larger lateral canal 361 (Figs. 1k, 5a). Chimpanzees and bonobos, due to their equally sized SCs and fairly 362 elongated CC (Figs. 1l, 5a), are closer to the hominine LCA morphology, while 363 Australopithecus appears derived toward the human condition, due to the moderate 364 increase in the size of the anterior and posterior canals (Fig 5). The reconstructed 365 morphotype for the LCA of the investigated dryopiths (Fig. 6d) closely resembles 366 those of hominines and hominids by displaying moderately stout and evenly-sized 367 SCs (with a slightly smaller lateral one), an obtuse angle at the CC apex, and a not 368 anterosuperiorly projecting anterior canal, differing from the hominid LCA by the 369 somewhat less vertically compressed anterior canal (Fig. 6b,d). 370 Hispanopithecus and Rudapithecus appear to have diverged in opposite 371 directions from their LCA (Fig. 5). The former seems derived in the volumetric 372 proportions (similarly to, Pongo, Australopithecus, and Homo), whereas the 15 373 Rudapithecus condition in this regard is very similar to that of Pan as well as the 374 reconstructed hominid LCA, and (to a lesser extent) to those of Nacholapithecus and 375 Oreopithecus (Figs. 5, 6b). Likewise, the fairly short CC and a somewhat 376 anterosuperiorly projecting anterior canal found in Rudapithecus (less so than in 377 orangutans and Oreopithecus) contrast with the longer CC and the rectangular- 378 shaped anterior canal found in Hispanopithecus (Fig. 1d). In these regards, 379 Hispanopithecus more closely resembles the members of the African ape and 380 human clade (Fig. 1a-c). 381 In summary, each extant hominid genus is derived in a particular direction from 382 the ancestral morphology, with Pan remaining close to the hominid and hominine 383 LCAs; Nacholapithecus appears as the least derived among both extant and fossil 384 hominid taxa, together with Oreopithecus. The latter taxon also shows similarities 385 with Pongo in the anterosuperiorly projecting anterior canal (Fig. 1e,j), despite being 386 much slenderer in Oreopithecus. Overall, the dryopiths appear less derived than 387 most extant genera relative to either the crown hominid or the crown hominine LCA, 388 irrespective of their preferred phylogenetic placement. Rudapithecus appears more 389 primitive than Hispanopithecus, being closer than the latter to both Nacholapithecus 390 and Oreopithecus, and closely approaching both the reconstructed crown hominid 391 LCA and Pan (Figs. 1, 5, 6). In contrast, Hispanopithecus is in some respects more 392 derived than Rudapithecus, particularly toward orangutans, australopiths, and 393 humans in the large vestibular recesses and in the stout SC volumetric proportions, 394 and toward orangutans alone in the rounded posterior canal morphology (Fig. 5). 395 Despite Hispanopithecus sharing its CC apex morphology (intermediate between 396 African great apes and orangutans) and anterior canal shape (not anterosuperiorly 397 projecting, yet not as squared as in gorillas) with Homo and Australopithecus, this 16 398 condition could be possibly plesiomorphic for hominids as a whole, as it is also found 399 in the stem hominid Nacholapithecus (Fig 5). Overall, the two dryopiths share with 400 African great apes and humans some features (moderately stout SCs, not 401 anterosuperiorly projecting anterior canal, fairly long CC), but according to our 402 analyses these features appear primitive (being likely present in the hominid LCA 403 and, to a lesser extent, Nacholapithecus), with hominines (particularly gorillas) and 404 especially orangutans having subsequently derived in opposite directions. 405 406 Discussion 407 Our results show that the vestibular morphology of both Hispanopithecus and 408 Rudapithecus more closely resembles that of extant great apes and humans than 409 that of hylobatids, in agreement with the current consensus that they belong to the 410 great-ape-and-human clade (2-3, 6, 26). These similarities particularly concern the 411 volumetric proportions of the SCs as well as the size of the latter relative to the 412 vestibular recesses. Volumetric proportions, as reflected by the ratio between the 413 volume and the length of the SCs, appear particularly relevant given that an 414 allometric grade shift has been previously identified to characterize all extant 415 hominids, so that they display relatively more voluminous SCs than other 416 anthropoids (including hylobatids) at comparable lengths (5). The derived condition 417 of hominids has been linked with locomotion (5), but is noteworthy that chimpanzees, 418 bonobos, and gorillas are slightly more variable in SC volumetric proportions than 419 orangutans and humans. Given the relationship between SC shape variation and 420 locomotion noted by some authors (53-54), our results might reflect stronger 421 locomotor-related selection pressures in orangutans and humans. 17 422 The classification results based on the bgPCA as well as the cluster analyses 423 indicate that the two investigated dryopiths are distinguishable from one another, 424 with the three specimens (two individuals) of Rudapithecus being more similar to one 425 another than to the single specimen of Hispanopithecus. This result, together with 426 other cranial differences (e.g., morphology of the frontal squama, premaxilla, and 427 zygomatic), supports the distinction of these taxa at the genus rank (2-3, 7, 21, 25, 428 35). Rudapithecus generally displays a somewhat more primitive morphology, closer 429 to the one inferred for the crown hominid LCA. It shows some similarities with the 430 fossil hominoids Oreopithecus and Nacholapithecus. The latter taxon appears more 431 primitive than other hominids, in agreement with a previous study based on the 432 entire inner ear morphology (49). However, both the hominid-like volumetric 433 proportions of Nacholapithecus and the lack of a subarcuate fossa (55) support its 434 stem hominid status, closely resembling the morphotype reconstructed for the crown 435 hominid LCA. The morphology of Rudapithecus also resembles that of crown 436 hominids such as Pan (volumetric proportions and the relative size of the SCs) and, 437 to a lesser extent, orangutans (the somewhat anterosuperiorly projecting anterior 438 canal and the short CC). As previously noted (5), chimpanzees and bonobos appear 439 least derived than other extant hominids. This is shown by the possession of 440 similarly-sized SCs (shared with the reconstructed crown hominid and crown 441 hominine LCAs, while the dryopith LCA displays a slightly smaller lateral canal) and, 442 especially, by the fairly slender volumetric proportions (intermediate between the 443 hominine and hominid LCAs, yet closer to the latter). This is also supported by the 444 similarities between Pan species and Miocene apes, especially Nacholapithecus. 445 Nonetheless, chimpanzees and bonobos appear derived in some features (the small 446 and rounded posterior canal as well as the obtuse angle of the CC apex), just like 18 447 gorillas and humans are derived in other directions (largest lateral canal relative to 448 the other SCs and markedly enlarged vertical canals, respectively). Among 449 hominids, orangutans and humans show the most extreme condition in the 450 volumetric proportions of the SCs. Orangutans further diverge from the hominid LCA 451 by the anterosuperiorly projecting anterior canal (even more so than in hylobatids). 452 Hispanopithecus appears more derived than the other Miocene taxa, especially by 453 the stouter SCs, while it does not fit well within the variation of any extant genus. 454 More clearly than Rudapithecus, Hispanopithecus displays a mosaic of features that 455 is unknown among extant hominids, including similarities with chimpanzees and 456 bonobos (in the long CC), humans (the obtuse angle of the CC apex and the right 457 angle between the planes of the posterior and lateral canals), and orangutans (the 458 stout CC and the voluminous vestibular recesses, the latter also shared with 459 humans) coupled with some unique features (the swollen area between the ampulla 460 and the tip of the lateral canal, and the markedly inflated ampullae). 461 Interpreting the similarities of the investigated dryopiths in evolutionary terms is 462 not straightforward. The results of the phylomorphospace approach and the 463 reconstructed ancestral vestibular morphologies suggest that modern hominid-like 464 volumetric proportions of the SCs would have been present in the LCA of crown 465 hominids, while that of crown hominoids as a whole would have displayed somewhat 466 intermediate proportions between hylobatids and hominids (yet closer to the latter). 467 Differences in volumetric proportions of the SCs have been related to locomotor 468 adaptations, because they directly affect the sensitivity and steadiness of the SCs in 469 response to angular accelerations (5, 56). Hence, the moderately stout SCs of the 470 LCA of crown hominids indicate that it showed a slow type of locomotion, which was 471 present, to a large extent, also in the LCA of crown hominoids, as previously inferred 19 472 based on the size of the SC radius alone (41). Both Rudapithecus and 473 Hispanopithecus show a wide gap between the lateral and posterior canals (the 474 planes defined by them are well separated and do not intersect), caused by the 475 anterolateral location of the lateral canal. This trait has been linked to orthograde 476 behaviors (42), in agreement with the fossil evidence available for these taxa (6, 11, 477 23, 25, 35, 57-61). However, from a phylogenetic viewpoint, the presence of the 478 aforementioned feature in the investigated dryopiths is less informative than their 479 hominid-like volumetric proportions, since the former have been identified as a 480 synapomorphy of crown hominoids as a whole (5). 481 We conclude that, with differences that are consistent with their distinction at the 482 genus rank, both Hispanopithecus and Rudapithecus display a unique hominid-like 483 vestibular morphology that differs from that of any extant hominid genera but that 484 appears quite close to that ancestral for crown hominids and crown hominines— 485 mainly diverging from that of hylobatids by the stouter volumetric proportions of the 486 SCs that are uniquely characteristic of great apes and humans among anthropoids. 487 Orangutans appear most derived from such an ancestral vestibular morphology, 488 whereas the investigated dryopiths lack most orangutan-like derived features— 489 except for the slightly anterosuperiorly projecting anterior canal in Rudapithecus 490 (also found in Hylobates) and some torsion in the shape of the lateral canal (a 491 character that appears to be quite variable within hominoids). The lack of orangutan 492 derived features in dryopithecines does not completely rule out a stem pongine 493 status, as previously supported by some authors (10, 29-30), as it represents a more 494 primitive morphology that probably precedes the subsequent evolution of the 495 orangutan-like features in the pongine lineage. However, our results are more 496 consistent with a stem hominid (6, 28) or a stem hominine (2-3, 14, 16, 25) status for 20 497 the investigated dryopiths. Our results suggest that African apes and hominin genera 498 evolved in different directions from an ancestral morphology that more closely 499 resembles that of Pan among extant hominines, and which is largely plesiomorphic 500 for hominids, as further supported by similarities with the stem hominid 501 Nacholapithecus (except for the slenderer volumetric proportions of the latter). 502 Therefore, similarities between the SC morphology of the studied dryopiths and that 503 of African apes do not necessarily imply a hominine status, but overall support the 504 previous claim (5), based on extant taxa alone, that extant hominines evolved from 505 an ancestral condition quite similar to that of the crown hominid LCA, and that the 506 latter was characterized by derived volumetric proportions of the SCs. Pending the 507 analysis of other Miocene apes, Pan among the extant taxa, and Rudapithecus 508 among extinct apes, constitute the best available proxies for such ancestral 509 morphologies, being already somewhat more derived from the crown hominoid 510 condition that is best approximated by Nacholapithecus. In the future, the inclusion in 511 the analyses of additional extinct hominoids will hopefully clarify further the 512 evolutionary history of this hominoids during the Miocene. 513 514 Materials and Methods 515 Sample Composition and Acquisition. We inspected three petrosals from two 516 individuals of R. hungaricus from Rudabánya, Hungary (RUD 77, left [RUD 77L] and 517 right [RUD 77R]; and RUD 200, right) (12-13) and the single available petrosal of H. 518 laietanus from Can Llobateres 2, Spain (IPS18000, right) (10, 29-30). The 519 specimens of Rudapithecus are housed at the Geological Museum of the Mining and 520 Geological Survey of Hungary (Hungary) and were scanned with a Skyscan 1172 521 (obtaining a resolution of 0.0136 mm) at the Max Plank Institute for Evolutionary 21 522 Anthropology (Leipzig, Germany), with the following parameters: 100 kV voltage and 523 100 mA. In turn, IPS18000 is housed at the Institut Català de Paleontologia Miquel 524 Crusafont in Sabadell (Spain) and was scanned with a GE Phoenix V|Tome|X s 240 525 (obtaining a resolution of 0.0295 mm) at the Centro Nacional de Investigación sobre 526 la Evolución Humana (Burgos, Spain), with the following parameters: 125 kV voltage 527 and 120 mA. The 3D virtual models of IPS18000, RUD 200 and RUD 77R were 528 mirrored to enable the comparison with extant species. The segmented surfaces of 529 the SCs of these fossils are available from MorphoSource 530 (https://www.morphosource.org; see Supplementary Table 3). 531 The comparative sample for the volumetric proportion evaluation has been taken 532 from a previous analysis that evaluated the phylogenetic signal embedded in the 533 vestibule morphology (5), and integrated with recently published material of extant 534 hominoids (4) and humans (62), together with the stem hominid Nacholapithecus 535 kerioi (49). Overall, it consists of µCT scans of 169 dried crania and petrosals 536 belonging to 27 extant anthropoid species, including all hominid genera and a 537 selection of hylobatids, cercopithecoids, and platyrrhines, together with fossil taxa 538 (Supplementary Table 3). The 3D meshes of the inner ear bony labyrinth of StW 573 539 and StW 578 were downloaded from the Sterkfontein project of the digital repository 540 MorphoSource.org. The juvenile status of a few specimens should not affect their 541 vestibular morphology since the bony labyrinth ossifies in early prenatal stages and 542 does not change subsequently (63). The analysis of the patterns of shape variation 543 was focused on hominoids alone and was based on a subsample of 77 individuals 544 representing all extant hominoid genera (Supplementary Table 4). 545 The µCT scans (voxel size for the extant and fossil specimens added in the 546 present analysis to those originally published in ref. 5 can be found in in 22 547 Supplementary Table 5) were segmented using Avizo 9.0.1 (FEI Visualization 548 Sciences Group) to digitally extract the left bony labyrinth, when available, or that 549 from the right side (mirrored before the surface alignment). The vestibular apparatus 550 was separated from the cochlea by cutting the generated 3D surfaces right under the 551 saccule and the oval window and filling the resulting holes with Geomagic Studio 552 2014 (3D Systems) using a flat surface (5). 553 The anatomical axes used for describing semicircular canal morphology 554 corresponds to those employed in the vast majority of inner ear analysis focusing on 555 primates (36-37,42,49), which conventionally follow the same orientation as in 556 humans (i.e., superior/inferior and anterior/posterior). 557 Shape Analysis. Shape was analyzed using a deformation-based 3DGM 558 technique that does not rely on a priori defined landmarks and examines the 559 geometrical correspondences between continuous surfaces (5, 48, 64-66). This 560 method quantifies the deformation from the analyzed surfaces from a constructed 561 sample-average surface (template) (64, 66), mathematically models them as a 562 diffeomorphism, and computes a set of vectors (momenta) that describe the direction 563 and magnitude of deformation from the average template. The unscaled 3D models 564 were aligned with Avizo 9.0.1 using the ‘Align Surface’ module before running the 565 analyses. The diffeomorphisms and the momenta were computed in the Barcelona 566 Supercomputing Center (BSC) (Barcelona, Spain) with Deformetrica 4 software. The 567 3D models of the fossils were projected a posteriori in the tangent space generated 568 by means of between-group principal components analysis (bgPCA) ran on the set 569 of momenta for the hominoid-only sample using genera as grouping factor. The 570 bgPCA was computed in R Studio v.1.1.453 for R v.3.5.0 using the ade4 package 571 (67), while the cross-validated bgPCA was derived using the ‘groupPCA’ function of 23 572 the Morpho v2.6 (68) library. Group mean differences were tested by computing a 573 permutational ANOVA (1000 permutations) based on the Euclidean distance 574 between the means using the ‘adonis’ function of the Vegan package (69). The 575 amount of variance (R2) explained by group differences in the raw shape data, and in 576 the scores of both standard and cross-validated bgPCA results, was estimated with 577 the same function as for the permutation test. To further assess similarities between 578 the analyzed fossil taxa and extant hominoid genera in terms of vestibular 579 morphology, we computed Mahalanobis squared distances (D2) between the bgPC 580 scores of fossils and group centroids used in the bgPCA. The distances were also 581 used to compute the posterior probabilities of group membership for the fossil 582 specimens by means of the ‘typprobClass’ function of the Morpho v2.6 (68) package, 583 on the basis of the multivariate normal distribution of extant groups defined a priori in 584 the bgPCA analyses. The similarities between extant and fossil hominoids were 585 further investigated by means of a cluster analysis (UPGMA) on the aforementioned 586 D2 and on the Euclidean distances between pairs of species mean configurations for 587 the raw data using the ‘average’ method of the ‘hclust’ function of the ‘stats’ package 588 in R 589 Additionally, the correlation between the log-transformed cube root of SC volume 590 (ln VolSC, in mm) and log-transformed SC length (ln L, in mm) was assessed means 591 of ordinary least-squares (OLS) linear regression, as the relationship between these 592 variables has previously been shown to display an allometric grade shift between 593 hominids and other anthropoids (5). Two separate regressions were computed for 594 the non-hominid anthropoids and for great apes and humans using SPSS Statistics 595 v. 17.0 for Windows (see Fig. 7b in ref. 5). The regression for the non-hominid 596 sample was used as a baseline for computing the allometric residuals 24 597 (Supplementary Table 1; see Table 5 in ref. 5) for the extant and extinct species. 598 Comparisons between the latter and extant groups are depicted by means of box- 599 and-whisker plots. 600 601 Phylomorphospace and phylogenetic signal. Major patterns of vestibular shape 602 variation were quantified using a phylomorphospace approach (70), obtained by 603 projecting a phylogeny on to the tangent space derived from the bgPCA of a 3DGM 604 shape analysis. In this method, the tips of the phylogeny correspond to the genus 605 bgPC centroid, while the internal nodes (i.e., the ancestral states) of the tree are 606 estimated using a maximum likelihood method for continuous characters, assuming 607 that the reconstructed nodes approximate the true morphology of the ancestors. 608 Thus, when a time-calibrated phylogeny is used, its two-dimensional representation 609 enables the intuitive interpretation of the magnitude and direction of evolution, based 610 on branch length and orientation. The molecular-based phylogenetic tree for extant 611 hominoids used in this analysis was downloaded from the 10kTrees Website (ver. 3; 612 http://10ktrees.fas.harvard.edu/), while Hispanopithecus and Rudapithecus were 613 added based on the assumption that they are closely related and constitute a clade, 614 with the tips corresponding to 9.6 Ma and 10.1 Ma, respectively (20), and diverging 615 at 11.1 Ma, but considering three different phylogenetic placements for these taxa as 616 discussed in the literature during the last two decades (see above): stem hominids, 617 stem hominines, and stem pongines (Supplementary Fig. 2). Analyses were 618 repeated based on the resulting three different cladograms and their results 619 compared to evaluate the effect of phylogenetic uncertainties surrounding these 620 taxa. Oreopithecus is here considered as a stem hominoid as indicated by most 621 recent cladistic analyses (4). Nacholapithecus has been included in a stem hominid 25 622 position, 2 Myr older than the divergence between pongines and hominines (crown 623 hominids), thus always preceding the divergence of dryopiths in all the phylogenetic 624 hypotheses, and its tip corr esponds to 14.77 Ma (54, 71). The divergence between 625 crown hominoids and Oreopithecus has been placed 1 Myr older than the 626 divergence between hylobatids and hominids and its tip corresponds to its last 627 occurrence in the fossil record (7.0-6.5 Ma; ref. 72). For the South African 628 Australopithecus sp., we used the published first appearance datum for 629 Australopithecus africanus (4.02 Ma) that includes the Jacovec specimens into the 630 species (73). 631 The position in the morphospace of the internal nodes of the phylogeny (ancestral 632 morphologies) was estimated via a maximum likelihood method for continuous 633 characters (74) using the ‘fastAnc’ function of phytools v. 0.6-60 R package (75). 634 Subsequently, the bgPC scores of the ancestral states were rotated and translated 635 from the shape data back into the configuration space for interpolation and 3D 636 visualization using Deformetrica 3 software. 637 The phylogenetic signal embedded in vestibular shape, as captured by all the 638 bgPCs, was quantified by means of the multivariate phylogenetic index Kmult (76) 639 using geomorph v3.1.1 (77) R package. The Kmult statistic, like its univariate 640 counterpart (77), assesses the amount of phylogenetic signal relative to that 641 expected for character undergoing Brownian motion and reflects the accuracy with 642 which the phylogenetic tree describes the variance-covariance pattern found in the 643 shape data. It is also informative about the accumulation of the variance in the 644 phylogeny. Thus, Kmult ≈ 1 is obtained when the inspected mode of evolution can 645 adequately be described using a stochastic Brownian motion model. For Kmult < 1, 646 the majority of the variance is found within clades, thus implying that neighbor taxa 26 647 resemble one another less than expected and that the mode of evolution is not 648 aleatory, possibly as the results of homoplastic adaptations (i.e., related to function 649 rather than phylogeny). Values of Kmult > 1 indicate that variance is mostly found 650 among different clades, being obtained when close taxa are less diverse that 651 expected under Brownian motion (suggesting that phenomena of stabilizing selection 652 might have occurred). 653 654 Acknowledgements 655 This research has been funded by the Agencia Estatal de Investigación (CGL2016- 656 76431-P and CGL2017-82654-P, AEI/FEDER EU; and BES-2015-071318 to A.U.), 657 the Generalitat de Catalunya (CERC A Programme), the consolidated research 658 groups 2017 SGR 86 and 2017 SGR 116 GRC, and the French Centre National de 659 la Recherche Scientifique. Part of the analyses were performed using Barcelona 660 Supercomputing Center resources (BCV-2020-1-0008). We thank Jose Braga for 661 allowing us to use the CT scans of the human specimens; the Max Plank Institute for 662 providing access to the microCT scans of RUD 77 and RUD 200; the ESRF heritage 663 database for palaeontology, evolutionary biology and archaeology, for providing 664 access to a part of the hominoid scans used in the present analysis; and Lynn 665 Copes, Lynn Lucas, and the Museum of Comparative Zoology (Cambridge, MA) for 666 providing access to a part of the scans used in the study, originally appearing in refs. 667 79 and 80, and funded by NSF DDIG #0925793, and a Wenner-Gren Foundation 668 Dissertation Grant #8102 (both to Lynn Copes). These scans were downloaded from 669 MorphoSource.org, a web-accessible archive for 3D digital data housed by Duke 670 University. A.B. was funded by the University of the Witwatersrand. 671 27 672 References 673 1. N. J. Stevens, et al., Palaeontological evidence for an Oligocene divergence 674 between Old World monkeys and apes. Nature 497, 611–614 (2013). 675 2. D. R. Begun, “The Miocene hominoid radiation“ in A companion to 676 Paleoanthropology D. R. Begun, Ed. (Blackwell Publishing, 2013), pp. 398– 677 416. 678 3. D. R. Begun, “Fossil record of Miocene hominoids” in Handbook of 679 Paleoanthropology, 2nd edn, W. Henke, I. Tattersall, Eds. (Springer, 2015), 680 pp. 1261–1332. 681 4. I. Nengo, et al., New infant cranium from the African Miocene sheds light on 682 ape evolution. Nature 548, 169–174 (2017). 683 5. A. Urciuoli, C. Zanolli, J. Dumoncel, S. Moyà-Solà, D. M. Alba, The evolution 684 of the vestibular apparatus in apes and humans. eLife 9 (2020) 685 6. D. M. Alba, Fossil apes from the Vallès-Penedès Basin. Evol. Anthropol. 21, 686 254–269 (2012). 687 7. Begun, D. R. Miocene hominids and the origins of the African apes and 688 humans. Ann. Rev. Anthropol. 39, 67–84 (2010). 689 8. J. Kelley, “The hominoid radiation in Asia” in The Primate Fossil Record, W. 690 C. Hartwig, Ed. (Cambridge University Press, 2002), pp. 369–384. 691 9. P. Andrews, T. Harrison, E. Delson, R. L. Bernor, L. Martin, “Distribution and 692 biochronology of European and Southwest Asian Miocene catarrhines” in The 693 Evolution of Western Eurasian Neogene Mammal Faunas, R. L. Bernor, V. 694 Fahlbusch & H.-W. Mittmann, Eds. (Columbia University Press, 1996), pp. 695 168–207. 28 696 10. S. Moyà-Solà, M. Köhler, Recent discoveries of Dryopithecus shed new light 697 on evolution of great apes. Nature 365, 543 (1993). 698 11. S. Moyà-Solà, M. Köhler, A Dryopithecus skeleton and the origins of great- 699 ape locomotion. Nature 379, 156–159 (1996). 700 12. L. Kordos, Begun, R. D., A new reconstruction of RUD 77, a partial cranium of 701 Dryopithecus brancoi from Rudabánya, Hungary. Am. J. Phys. Anthropol. 702 103, 277–294 (1997). 703 13. L. Kordos, D. R. Begun, A new cranium of Dryopithecus from Rudabánya, 704 Hungary. J. Hum. Evol. 41, 689–700 (2001). 705 14. D. R. Begun, C. V. Ward, M. D. Rose, “Events in hominoid evolution” in 706 Function, Phylogeny and Fossils: Miocene Hominoid Evolution and 707 Adaptation, D. R. Begun, C. V. Ward, M. D. Rose (Plenum Press, 1997), pp. 708 389–415. 709 15. D. R. Begun, “African and Eurasian Miocene hominoids and the origins of the 710 Hominidae” in Hominoid Evolution and Environmental Change in the Neogene 711 of Europe, Vol 2. Phylogeny of the Neogene Hominoid Primates of Eurasia, L. 712 D. Bonis, G. Koufos, P. Andrews, Eds. (Cambridge University Press, 2001), 713 pp. 231–253. 714 16. D. R. Begun, “The pliopithecoidea” in: The Primate Fossil Record, W. Hartwig, 715 Ed. (Cambridge University Press., 2002), pp. 339–368. 716 17. S. Moyà-Solà, M. Köhler, D. M. Alba, I. Casanovas-Vilar, J. Galindo, 717 Pierolapithecus catalaunicus, a new Middle Miocene great ape from Spain. 718 Science 306, 1339–1344 (2004). 719 18. S. Moyà-Solà, et al., First partial face and upper dentition of the Middle 720 Miocene hominoid Dryopithecus fontani from Abocador de Can Mata (Vallès- 29 721 Penedès Basin, Catalonia, NE Spain): taxonomic and phylogenetic 722 implications. Am. J. Phys. Anthropol. 139, 126–145 (2009). 723 19. S. Moyà-Solà, et al., A unique Middle Miocene European hominoid and the 724 origins of the great ape and human clade. Proc. Natl. Acad. Sci. U.S.A. 106, 725 9601–9606 (2009). 726 20. I. Casanovas-Vilar, et al., Updated chronology for the Miocene hominoid 727 radiation in Western Eurasia. Proc. Natl. Acad. Sci U.S.A. 108, 5554–5559 728 (2011). 729 21. D. R. Begun, Dryopithecins, Darwin, de Bonis, and the European origin of the 730 African apes and human clade. Geodiversitas 31, 789–816 (2009). 731 22. D. M. Alba, et al., New dental remains of Hispanopithecus laietanus 732 (Primates: Hominidae) from Can Llobateres 1 and the taxonomy of Late 733 Miocene hominoids from the Vallès-Penedès Basin (NE Iberian Peninsula). J. 734 Hum. Evol. 63, 231–246 (2012). 735 23. D. M. Alba, S. Almécija, I. Casanovas-Vilar, J. M. Méndez, S. Moyà-Solà, A 736 partial skeleton of the fossil great ape Hispanopithecus laietanus from Can 737 Feu and the mosaic evolution of crown-hominoid positional behaviors. PLoS 738 ONE 6, e39617 (2012). 739 24. D. R. Begun, Sivapithecus is east and Dryopithecus is west, and never the 740 twain shall meet. Anthropol. Sci. 113, 53–64 (2005). 741 25. D. R. Begun, M. C. Nargolwalla, L. Kordos, European Miocene hominids and 742 the origin of the African ape and human clade. Evol. Anthropol. 21, 10–23 743 (2012). 744 26. P. Andrews, Last common ancestor of apes and humans: morphology and 745 environment. Folia Primatol. 91, 1–27 (2020). 30 746 27. M. Böhme, et al. A new Miocene ape and locomotion in the ancestor of great 747 apes and humans. Nature 575, 489–493 (2019). 748 28. D. M. Alba, et al., Miocene small-bodied ape from Eurasia sheds light on 749 hominoid evolution. Science 350, aab2625 (2015). 750 29. M. Köhler, S. Moyà-Solà, D. M. Alba, “Eurasian hominoid evolution in the light 751 of recent Dryopithecus findings” in: Hominoid Evolution and Environmental 752 Change in the Neogene of Europe, Vol 2. Phylogeny of the Neogene 753 Hominoid Primates of Eurasia, L. D. Bonis, G. Koufos, P. Andrews, Eds. 754 (Cambridge University Press, 2001), pp. 192–212. 755 30. S. Moyà-Solà, M. Köhler, New partial cranium of Dryopithecus Lartet, 1863 756 (Hominoidea, Primates) from the upper Miocene of Can Llobateres, 757 Barcelona, Spain. J. Hum. Evol. 29, 101–139 (1995). 758 31. D. R. Begun, L. Kordos, “Phyletic affinities and functional convergence in 759 Dryopithecus and other Miocene and living hominids.” in Function, Phylogeny, 760 and Fossils: Miocene Hominoid Evolution and Adaptions, D.R., Begun, C.V., 761 Ward, M.D., Rose Eds. (Plenum Press, 1997), pp. 291–316. 762 32. D. R. Begun, How to identify (as opposed to define) a homoplasy: Examples 763 from fossil and living great apes. J. Hum. Evol. 52, 559–572 (2007). 764 33. S. G. Larson, Parallel evolution in the hominoid trunk and forelimb. Evol. 765 Anthropol. 6, 87–99 (1998). 766 34. T. C. Rae, Mosaic evolution in the origin of the Hominoidea. Folia Primatol. 767 70, 125–135 (1999). 768 35. C. V. Ward, “Postcranial and locomotor adaptations of hominoids” in 769 Handbook of Paleoanthropology, 2nd edn W. Henke, I. Tattersall, Eds. 770 (Springer, 2015), pp. 1363–1386. 31 771 36. F. Spoor, B. Wood, F. Zonneveld, Implications of early hominid labyrinthine 772 morphology for evolution of human bipedal locomotion. Nature 369, 645–648 773 (1994). 774 37. F. Spoor, F. Zonneveld, Comparative review of the human bony labyrinth. 775 Yearb. Phys. Anthropol. 41, 211–251 (1998). 776 38. Spoor, F. et al., The primate semicircular canal system and locomotion. Proc. 777 Natl. Acad. Sci U.S.A. 104, 10808–10812 (2007). 778 39. David, R., et al., Motion from the past. A new method to infer vestibular 779 apparatus capacities of extinct species. C. R. Palevol 9, 397–410 (2010). 780 40. David, R., A. Stoessel, A. Berthoz, F. Spoor, D. Bennequin, Assessing 781 morphology and function of the semicircular duct system: introducing new in- 782 situ visualization and software toolbox. Sci. Rep. 6, 32772 (2016). 783 41. T. M. Ryan, et al., Evolution of locomotion in Anthropoidea: the semicircular 784 canal evidence. Proc. R. Soc. B 279, 3467–3475 (2012). 785 42. A. Le Maître, P. Schuetz, P. Vignaud, M. Brunet, New data about semicircular 786 canal morphology and locomotion in modern hominoids. J. Anat. 231, 95–109 787 (2017). 788 43. L. Rook, et al., The bony labyrinth of Oreopithecus bambolii. J. Hum. Evol. 46, 789 347–354 (2004). 790 44. M. S. Ponce de León, et al., Human bony labyrinth is an indicator of 791 population history and dispersal from Africa. Proc. Natl. Acad. Sci. U.S.A. 115, 792 4128–4133 (2018). 793 45. A. Beaudet, The inner ear of the Paranthropus specimen DNH 22 from 794 Drimolen, South Africa. Am. J. Phys. Anthropol. 170, 439–446 (2019). 32 795 46. A. Beaudet, et al., The bony labyrinth of StW 573 (“Little Foot”): Implications 796 for early hominin evolution and paleobiology. J. Hum. Evol. 127, 67–80 797 (2019). 798 47. J. del Rio, et al., Allometry, function and shape diversification in the inner ear 799 of platyrrhine primates. J. Mammal. Evol. (2020) DOI: 800 https://doi.org/10.1007/s10914-019-09490-9 801 48. A. Urciuoli, C. Zanolli, S. Almécija, S. Moyà-Solà, D. M. Alba, Analysis of the 802 primate vestibular apparatus: a comparison of landmark-based and 803 deformation-based 3D geometric morphometric approaches. Proc. Eur. Soc. 804 Stud. Hum. Evol. 7, 193 (2018). 805 49. N. Morimoto, et al., Variation of bony labyrinthine morphology in 806 Mio−Plio−Pleistocene and modern anthropoids. Am. J. Phys. Anthropol. 173, 807 276–293 (2020 808 50. T. C. Rae, P. M. Johnson, W. Yano, E. Hirasaki, Semicircular canal size and 809 locomotion in colobine monkeys: a cautionary tale. Folia Primatol. 87, 213– 810 223 (2016). 811 51. D. M. Malinzak, R. F. Kay, T. E. Hullar, Locomotor head movements and 812 semicircular canal morphology in primates. Proc. Natl. Acad. Sci. U.S.A. 109, 813 17914–17919 (2012). 814 52. A. Cardini, P. D. Polly, Cross-validated between-group PCA scatterplots: A 815 solution to spurious group separation? Evol. Biol. 47, 85–95 (2020). 816 53. A. Perier, R. Lebrun, L. Marivaux, Different level of intraspecific variation of 817 the bony labyrinth morphology in slow- versus fast-moving Primates. J. 818 Mammal. Evol. 23, 353–368 (2016). 33 819 54. L. A. Gonzales, M. D. Malinzak, R. F. Kay, Intraspecific variation in 820 semicircular canal morphology—A missing element in adaptive scenarios? 821 Am. J. Phys. Anthropol. 168, 10–24 (2019). 822 55. Y. Kunimatsu, M. Nakatsukasa, D. Shimizu, Y. Nakano, H. Ishida, Loss of the 823 subarcuate fossa and the phylogeny of Nacholapithecus. J. Hum. Evol. 131, 824 22–27 (2019). 825 56. R. D. Rabbitt, E. R. Damiano, J. W. Grant, “Biomechanics of the Semicircular 826 Canals and Otolith Organs” in The Vestibular System S. M. Highstein, R. R. 827 Fay, A. N. Popper, Eds. (Springer, 2004) pp. 153–201. 828 57. D. R. Begun, Phyletic diversity and locomotion in primitive European 829 hominids. Am. J. Phys. Anthropol. 87, 311–340 (1992). 830 58. D. R. Begun, New catarrhine phalanges from Rudabánya (Northeastern 831 Hungary) and the problem of parallelism and convergence in hominoid 832 postcranial morphology. J. Hum. Evol. 24, 373–402 (1993). 833 59. S. Almécija, D. M. Alba, S. Moyà-Solà, M. Köhler, Orang-like manual 834 adaptations in the fossil hominoid Hispanopithecus laietanus: first steps 835 towards great ape suspensory behaviours. Proc. R. Soc. B 274, 2375–2384 836 (2007). 837 60. D. M. Alba, S. Almécija, S. Moyà-Solà, Locomotor inferences in 838 Pierolapithecus and Hispanopithecus: reply to Deane and Begun (2008). J. 839 Hum. Evol. 59, 143–149 (2010). 840 61. M. Pina, D. M. Alba, S. Almécija, J. Fortuny, S. Moyà-Solà, Brief 841 communication: paleobiological inferences on the locomotor repertoire of 842 extinct hominoids based on femoral neck cortical thickness: the fossil great 34 843 ape Hispanopithecus laietanus as a test-case study. Am. J. Phys. Anthropol. 844 149, 142–148 (2012). 845 62. W. Wimmer, C. Vandersteen, N. Guevara, M. Caversaccio, H. Delingette, H., 846 “Robust cochlear modiolar axis detection in CT” in Medical Image Computing 847 and Computer Assisted Intervention – MICCAI 2019 D. Shen, et al. Eds. 848 (Springer, 2019), pp. 3–10. 849 63. N. Jeffery, F. Spoor, Prenatal growth and development of the modern human 850 labyrinth. J. Anat. 204, 71–92 (2004). 851 64. S. Durrleman, et al., “Topology preserving atlas construction from shape data 852 without correspondence using sparse parameters” in Proceedings of Medical 853 Image Computing and Computer Aided Intervention N. Ayache, H. Delingette, 854 P. Golland, K. Mori (Springer, 2012), pp. 223–230. 855 65. J. Dumoncel, S. Durrleman, J. Braga, J. P. Jessel, G. Subsol, Landmark-free 856 3D method for comparison of fossil hominins and hominids based on 857 endocranium and EDJ shapes. Am. J. Phys. Anthropol. 153, 114 (2014). 858 66. A. Beaudet, et al., Upper third molar internal structural organization and 859 semicircular canal morphology in Plio-Pleistocene South African 860 cercopithecoids. J. Hum. Evol. 95, 104–120 (2016). 861 67. S. Dray, A. Dufour, The ade4 Package: Implementing the duality diagram for 862 ecologists. J. Stat. Softw. 22, (2007). 863 68. S. Schlager, “Morpho and Rvcg - Shape Analysis in R.” in: Statistical Shape 864 and Deformation Analysis, G. Zheng, S. Li, G. Szekely Eds. (Academic 865 Press., 2017) pp. 217–256. 866 69. J. Oksanen, et al., vegan: Community Ecology Package. R package version 867 2.5-6. https://CRAN.R-project.org/package=vegan 35 868 70. B. Sidlauskas, Continuous and arrested morphological diversification in sister 869 clades of characiform fishes: A phylomorphospace approach. Evolution 62, 870 3135–3156 (2008). 871 71. M. Nakatsukasa, Y. Kunimatsu, Nacholapithecus and its importance for 872 understanding hominoid evolution. Evol. Anthropol. 18, 103–119 (2009). 873 72. L. Rook, L. Bondioli, M. Köhler, S. Moyà-Solà, R. Macchiarelli, Oreopithecus 874 was a bipedal ape after all: evidence from the iliac cancellous architecture. 875 Proc. Natl. Acad. Sci U.S.A. 96, 8795–8799 (1999). 876 73. B. Wood, E. K. Boyle, Hominin taxic diversity: Fact or fantasy? Am. J. Phys. 877 Anthropol. 159, 37–78 (2016). 878 74. D. Schluter, T. Price, A. Ø. Mooers, D. Ludwig, Likelihood of ancestor states 879 in adaptive radiation. Evolution 51, 1699–1711 (1997). 880 75. L. J. Revell, phytools: An R package for phylogenetic comparative biology 881 (and other things). Methods Ecol. Evol. 3, 217–223 (2012). 882 76. D. C. Adams, A generalized K statistic for estimating phylogenetic signal from 883 shape and other high-dimensional multivariate data. Syst. Biol. 63, 685–697 884 (2014). 885 77. D. C. Adams, M. L. Collyer, A. Kaliontzopoulou, Geomorph: Software for 886 geometric morphometric analyses. R package version 3.1.0. https://cran.r- 887 project.org/package=geomorph (2019). 888 78. S. P. Blomberg, J. T. Garland, A. R. Ives, Testing for a phylogenetic signal in 889 comparative data: behavioural traits are more labile. Evolution 57, 717–745 890 (2003). 891 79. L. E. Copes, W. H. Kimbel, Cranial vault thickness in primates: Homo erectus 892 does not have uniquely thick vault bones. J. Hum. Evol. 90, 120–134 (2016). 36 893 80. L. E. Copes, L. M. Lucas, J. O. Thostenson, H. E. Hoekstra, A collection of 894 non-human primate computed tomography scans housed in MorphoSource, a 895 repository for 3D data. Sci. Data 3, 1 (2016). 37 896 FIGURE CAPTIONS 897 Fig. 1. The vestibular apparatus morphology of Rudapithecus hungaricus (a-c), 898 Hispanopithecus laietanus (d), fossil hominoids (e-f), and individuals from extant 899 hominoid genera (g-m) as depicted by renderings of the 3D models. From left to 900 right, in posterolateral, superior, and posteromedial views: (a) R. hungaricus (RUD 901 77L); (b) R. hungaricus (RUD 77R); (c) R. hungaricus (RUD 200); (d) H. laietanus 902 (IPS18000); (e) Oreopithecus bambolii (BAC 208); (f) Nacholapithecus kerioi (BG 903 42744); (g) Hoolock hoolock (AMNH.M 83425); (h) Symphalangus syndactylus 904 (AMNH.M 106583); (i) Hylobates lar (MCZ 41424); (j) Pongo sp.(IPS10647); (k) 905 Gorilla gorilla (AMNH.M 167338); (l) Pan troglodytes (AMNH.M 51204); (m) Homo 906 sapiens (F 04). Scale bars = 5 mm 907 908 Fig. 2. (a) Allometric regressions of cube root of semicircular canal volume (ln 909 VolSC, in mm) vs. semicircular canal length (ln L, in mm) in anthropoids. Lines 910 represent ordinary least-squares (OLS) best-fit lines for extant hominids (red) and 911 other extant anthropoids (blue). Both hominids and other anthropoids show a 912 negatively allometric relationship between VolSC and L, but with a marked allometric 913 grade shift—such that hominids possess stouter canals than other anthropoid taxa at 914 comparable lengths (see Supplementary Table 1 for comparative sample 915 measurements). (b) Boxplots of allometric residuals computed using the best-fit line 916 of the non-hominid anthropoid regression as baseline. Horizontal line is the median, 917 boxes represent the interquartile range, whiskers represent maximum and minimum 918 excluding outliers (beyond 1.5 times the upper and lower quartiles), and black dots 919 are outliers. Samples for each boxplot are: Platyrrhini (n=15), Cercopithecoidea 920 (n=80), Hylobatidae (n=23), Gorilla (n=8), Pongo (n=8), Pan (n=25), and Homo 38 921 (n=10). Note that all the fossils overlap with extant hominoids. However, while 922 Australopithecus (StW 573 and StW 578) and Hispanopithecus (IPS18000) 923 approach the human and orangutan condition, Rudapithecus RUD 200 and 924 Oreopithecus (BAC 208) only overlap with African apes, while Nacholapithecus (BG 925 42744) and Rudapithecus RUD 77 only overlap with chimpanzees and bonobos, and 926 marginally also with the upper range of hylobatids and cercopithecoids. 927 928 Fig. 3. Results of the bgPCA based on vestibular shape (as reflected by deformation 929 data) in hominoids using genera as grouping factor (variance explained by each 930 bgPC is included within parentheses): (a) bgPC2 vs. bgPC1; (b) bgPC3 vs. bgPC1. 931 Extreme conformations for each bgPC are displayed: (c) bgPC1; (d) bgPC2; (e) 932 bgPC3. Convex hulls are drawn for each hominoid genus and colored as follows: 933 blue, Pongo; black, Gorilla; green, Pan; lilac, Homo; red, Hylobates; orange, 934 Symphalangus; cyan, Hoolock. 935 936 Fig. 4. Dendrograms resulting from UPGMA cluster analyses based on: (a) the 937 Mahalanobis squared distances (D2) computed for the meaningful between-group 938 principal components (bgPC1–bgPC3, 86% of variance) depicted in Figure 2 939 (cophenetic correlation: 0.75); (b) the Euclidean distances computed for the raw 940 shape data (i.e., the deformation fields) obtained from the deformation-based 3D 941 geometric morphometrics analysis (cophenetic correlation: 0.77). 942 943 Fig. 5. Phylomorphospaces of the vestibular apparatus in hominoids, obtained by 944 projecting the phylogenetic tree that considers dryopithecines a clade of stem 945 hominines (Supplementary Fig. 1b) on bivariate plots between bgPCs. The tips 39 946 correspond to genus bgPCA score centroids: (a) bgPC2 vs. bgPC1; (b) bgPC3 vs. 947 bgPC1. Key ancestral morphologies reconstructed using maximum likelihood for the 948 last common ancestors (LCAs) of various clades are depicted by means of 949 arrowheads. See Supplementary Fig. 2 for the results based on alternative 950 phylogenetic hypotheses (Supplementary Fig. 1a,c). 951 952 Fig. 6. Reconstructed vestibular shape for the last common ancestor (LCA) of the 953 main clades of interest as inferred using maximum likelihood methods for 954 deformation-based 3DGM analyses applied to the hominoid sample under the stem- 955 hominine phylogenetic hypothesis for dryopiths (Supplementary Fig. 1b), in 956 posterolateral (left), superior (middle), and posteromedial (right) views. The 957 reconstructed LCAs depicted are the following: (a) crown hominoids; (b) crown 958 hominids; (c) crown hominines; (d) dryopithecines (Hispanopithecus + 959 Rudapithecus). The results for the other phylogenetic hypotheses (not shown) are 960 virtually identical 40 a) Rudapithecus RUD 77L b) Rudapithecus RUD 77R c) Rudapithecus RUD 200 d) Hispanopithecus IPS18000 e) Oreopithecus f) Nacholapithecus g) Hoolock h) Symphalangus i) Hylobates j) Pongo k) Gorilla l) Pan m) Homo Platyrrhini Cercopithecoidea Pongo spp. Gorilla gorilla a 1.5 Oreopithecus bambolii Pan paniscus Nacholapithecus kerioi Pan troglodytes 1.4 Hylobates lar Pan sp. Symphalangus syndactylus Australopithecus sp. Hoolock hoolock Homo sapiens 1.3 Rudapithecus hungaricus Hispanopithecus laietanus 1.2 IPS18000 ln VolSC 1.1 RUD 200 RUD 77R 1.0 RUD 77L 0.9 0.8 0.7 0.6 3.4 3.5 3.6 3.7 3.8 3.9 ln L Allometric residuals Platyrrhini b Cercopithecoidea Hylobatidae Gorilla Homo Pan Pongo D 44 D 08 W 00 0 St 573 8 77 20 57 RU427 RU C 2 St 80 W S1 BA BG IP í 0.0 0.2 0.4 0.6 a c Oreopithecus bambolii Hylobates lar Symphalangus syndactylus 2 Hoolock hoolock Nacholapithecus kerioi Pongo spp. 1 Gorilla gorilla + bgPC1 bgPC2 33.4% BG 42744 0 – BAC 208 RUD 200 IPS18000 RUD 77L RUD 77R StW 573 −1 StW 578 −2 Rudapithecus hungaricus Hispanopithecus laietanus Pan paniscus d Pan troglodytes −3 Pan sp. Australopithecus sp. Homo sapiens −3 −2 −1 0 1 2 bgPC1 40.7% + bgPC2 b – 2 StW 573 1 BG 42744 e bgPC3 11.4% IPS18000 0 RUD 200 StW 578 RUD 77R RUD 77L −1 BAC 208 + −2 bgPC3 – −3 −3 −2 −1 0 1 2 bgPC1 40.7% a b Hylobates Hylobates Symphalangus Oreopithecus Symphalangus Hoolock Hoolock Nacholapithecus Nacholapithecus Oreopithecus Rudapithecus Rudapithecus Pan Pan Australopithecus Australopithecus Gorilla Hispanopithecus Gorilla Pongo Homo Homo Hispanopithecus Pongo a 2 bgPC2 33.4% Gorilla Pongo 1 Oreopithecus Pan Nacholapithecus 0 Hylobates Hoolock Rudapithecus Hispanopithecus Symphalangus −1 Australopithecus −2 −3 bgPC1 40.7% Homo −3 −2 −1 0 1 2 3 b bgPC3 11.4% Gorilla 1 Pan Symphalangus Nacholapithecus Hispanopithecus 0 Hoolock Homo Australopithecus Hylobates Rudapithecus −1 Oreopithecus −2 bgPC1 40.7% Pongo −3 −2 −1 0 1 2 3 Crown hominoid LCA Crown hominid LCA Crown hominine LCA Dryopithecine LCA a Crown hominoid LCA b Crown hominid LCA c Crown hominine LCA d Dryopithecine LCA 1 Table 1 Mahalanobis squared distances (D2) between fossil scores and extant hominoid group centroids and associated posterior 2 probabilities (p) of group membership for all fossil individuals. Note that these are probability estimates of having a particular score given 3 membership in a particular group, not the likelihood of group membership in each of a priori defined groups given a particular score. The lowest 4 D2 and the highest probability for each specimen are in bold. D2 Hoolock Symphalangus Hylobates Pongo Gorilla Pan Homo IPS18000 (Hispanopithecus laietanus) 6.086 6.861 10.532 9.229 5.329 1.407 5.076 RUD 77R (Rudapithecus hungaricus) 2.044 2.776 4.233 8.831 9.809 2.736 4.063 RUD 77L (Rudapithecus hungaricus) 2.252 3.325 4.104 6.735 10.091 3.132 5.621 RUD 200 (Rudapithecus hungaricus) 2.478 3.312 5.338 7.475 5.174 0.745 6.566 BAC 208 (Oreopithecus bambolii) 4.359 6.261 4.809 4.087 15.042 7.218 11.210 BG 42744 (Nacholapithecus kerioi) 1.200 1.662 3.546 10.016 4.942 0.683 7.673 StW 573 (Australopithecus sp.) 5.323 5.367 9.865 14.341 5.192 1.246 4.182 StW 578 (Australopithecus sp.) 9.400 10.538 13.134 10.554 16.362 7.661 2.552 p Hoolock Symphalangus Hylobates Pongo Gorilla Pan Homo IPS18000 (Hispanopithecus laietanus) < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 0.053 < 0.001 RUD 77R (Rudapithecus hungaricus) 0.006 0.003 < 0.001 < 0.001 < 0.001 0.011 < 0.001 RUD 77L (Rudapithecus hungaricus) 0.015 0.005 < 0.001 < 0.001 < 0.001 0.012 < 0.001 RUD 200 (Rudapithecus hungaricus) < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 0.549 < 0.001 BAC 208 (Oreopithecus bambolii) 0.016 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 BG 42744 (Nacholapithecus kerioi) 0.035 0.012 < 0.001 < 0.001 < 0.001 0.184 < 0.001 StW 573 (Australopithecus sp.) < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 0.026 < 0.001 StW 578 (Australopithecus sp.) < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 0.013 1 Table 2 Mahalanobis distances (D2) between dryopiths and other fossils. These distances are based on the scores of the 2 significative between-group principal components (bgPC1–bgPC3). D2 IPS18000 RUD 77R RUD 77L RUD 200 IPS18000 (Hispanopithecus laietanus) — 2.037 2.504 1.012 RUD 77R (Rudapithecus hungaricus) 2.037 — 0.179 0.772 RUD 77L (Rudapithecus hungaricus) 2.504 0.179 — 0.848 RUD 200 (Rudapithecus hungaricus) 1.012 0.772 0.848 — BAC 208 (Oreopithecus bambolii) 6.678 2.495 1.385 3.571 BG 42744 (Nacholapithecus kerioi) 2.286 1.270 1.505 0.416 StW 573 (Australopithecus sp.) 0.703 2.587 3.657 1.637 StW 578 (Australopithecus sp.) 3.080 2.745 3.087 4.479 3 4 5