This is the accepted version of the article: Bouchet, Florian; Urciuoli, Alessandro; Beaudet, Amélie; [et al.]. «Comparative anatomy of the carotid canal in the Miocene small-bodied catarrhine Pliobates cataloniae». Journal of human evolution, Vol. 161 (December 2021), art. 103073. DOI 10.1016/j.jhevol.2021.103073 This version is available at https://ddd.uab.cat/record/250776 under the terms of the license 1 Comparative anatomy of the carotid canal in the Miocene small-bodied catarrhine Pliobates 2 cataloniae 3 4 Abstract 5 The small-bodied Miocene catarrhine Pliobates cataloniae (11.6 Ma, Spain) displays a 6 mosaic of catarrhine symplesiomorphies and hominoid synapomorphies that hinders 7 deciphering its phylogenetic relationships. Based on cladistic analyses it has been 8 interpreted as a stem hominoid or as a pliopithecoid. Intriguingly, the carotid canal 9 orientation of Pliobates was originally described as hylobatid-like. The variation in carotid 10 canal morphology among anthropoid clades shown in previous studies suggests that this 11 structure might be phylogenetically informative. However, its potential for phylogenetic 12 reconstruction among extinct catarrhines remains largely unexplored. Here we quantify the 13 orientation, proportions, and course of the carotid canal in Pliobates, extant anthropoids 14 and other Miocene catarrhines (Epipliopithecus, Victoriapithecus, and Ekembo) using 3D 15 morphometric techniques. We also compute phylogenetic signal and reconstruct the 16 ancestral carotid canal course for main anthropoid clades. Our results reveal that carotid 17 canal morphology embeds strong phylogenetic signal but mostly discriminates between 18 platyrrhines and catarrhines, with an extensive overlap among extant catarrhine families. 19 The analyzed extinct taxa display a quite similar carotid canal morphology more closely 20 resembling that of extant catarrhines. Nevertheless, our results for Pliobates highlight some 21 differences compared with the pliopithecid Epipliopithecus, which displays a somewhat 22 more platyrrhine-like morphology. In contrast, Pliobates appears as derived toward the 23 modern catarrhine condition as the stem cercopithecid Victoriapithecus and the stem 24 hominoid Ekembo, which more closely resemble one another. Moreover, Pliobates appears 1 25 somewhat derived toward the reconstructed ancestral hominoid morphotype, being more 26 similar than other Miocene catarrhines to the condition of great apes and the hylobatid 27 Symphalangus. Overall, our results rule out previously noted similarities in carotid canal 28 morphology between Pliobates and hylobatids, but do not show particular similarities with 29 pliopithecoids either—as opposed to extant and other extinct catarrhines. Additional 30 analyses will be required to clarify the phylogenetic relationships of Pliobates, particularly 31 given its dental similarities with dendropithecids. 32 33 Keywords: Internal carotid artery; Petrosal; Geometric morphometrics; Fossil primates; 34 Phylogeny; Evolution. 35 36 1. Introduction 37 1.1. Pliobates and catarrhine evolution 38 Crown catarrhines include two main extant clades distinguished at the superfamily rank: 39 Old World monkeys (Cercopithecoidea) and apes and humans (Hominoidea). Earliest stem 40 catarrhines have been reported from the latest Eocene (34.5–29.5 Ma) of Afro-Arabia 41 (Seiffert, 2006, 2012; Seiffert et al., 2010). In turn, earliest crown catarrhines (both 42 cercopithecoids and hominoids) date back to the late Oligocene (~25 Ma) of East Africa 43 (Stevens et al., 2013)—in accordance with molecular estimates indicating that the 44 cercopithecoid–hominoid divergence occurred during the Oligocene (Chatterjee et al., 2009; 45 Perelman et al., 2011; Springer et al., 2012; Finstermeier et al., 2013; Pozzi et al., 2014). 46 Catarrhines did not disperse into Eurasia until later in the early Miocene, following the 47 closure of the Tethys Seaway and the establishment of intermittent land bridges at ~19 Ma 48 (Harzhauser et al., 2007; Harrison, 2013). Several catarrhine lineages eventually dispersed 2 49 from Afro-Arabia into Eurasia (Roos et al., 2019; Gilbert et al., 2020a), including: 50 pliopithecoids (considered an Eurasian clade of stem catarrhines, first recorded ~18–17 Ma; 51 Harrison and Gu, 1999; Begun, 2002, 2017; Harrison, 2013); hylobatids (first recorded ~13.8– 52 12.5 Ma; Gilbert et al., 2020b); large-bodied hominoids (presumably hominids, first recorded 53 ~16.5–14 Ma; Casanovas-Vilar et al., 2011; Harrison, 2017); and various lineages of 54 cercopithecoid primates, successively recorded at different times in Eurasia, namely 55 colobines (~8.5 Ma; Alba et al., 2015a), cercopithecins (~8–6.5 Ma; Gilbert et al., 2014), 56 macaques (~5.9–5.3 Ma; Alba et al., 2014a), and geladas (1.6–1.2 Ma; Alba et al., 2014b). 57 A panoply of small-bodied, putative stem catarrhine genera (including, but not restricted 58 to, dendropithecids) are recorded in the early to middle Miocene of Africa (Harrison, 2010, 59 2013). Some Miocene Eurasian taxa have previously been linked to some of these African 60 stem catarrhines—implying yet another catarrhine dispersal event from Africa—but such 61 claims have been mostly rejected by reinterpreting them as pliopithecoids or hylobatids 62 (Harrison, 2016 and references therein). The most likely exception is some isolated dental 63 remains from Pakistan (~17–16 Ma; Barry et al., 1987; Bernor et al., 1988), which are most 64 similar to African dendropithecids (Harrison, 2016) but might ultimately prove to be stem 65 pliopithecoids (Sankhyan et al., 2017). In turn, as remarked by Roos et al. (2019), the small- 66 bodied catarrhine Pliobates (middle/late Miocene of Spain, 11.6 Ma; Alba et al., 2015b) 67 might denote an additional out-of-Africa dispersal event of catarrhines unless it is 68 interpreted as a pliopithecoid. This taxon was originally recovered by Alba et al.’s (2015b) 69 cladistic analysis as a stem hominoid morphologically more derived than Proconsul s.l. 70 (currently split into Proconsul s.s. and Ekembo; McNulty et al., 2015) and subsequently 71 reinterpreted as a pliopithecoid by different cladistic analyses (Nengo et al., 2017; Gilbert et 72 al., 2020b), as first suggested by Benefit and McCrossin (2015). However, given the mosaic 3 73 nature of its cranial and postcranial skeleton—combining plesiomorphic, stem catarrhine- 74 like features (including dental resemblances with dendropithecids) with crown hominoid 75 synapomorphies (Alba et al., 2015b)—the possibility remains that Pliobates is a late 76 descendant of an African stem catarrhine lineage. 77 Deciphering the phylogenetic position of Pliobates would have important implications for 78 our current understanding of catarrhine evolution and paleobiogeography during the 79 Miocene. If ultimately shown to be a pliopithecoid or as a member of a different stem 80 catarrhine lineage, it would imply remarkable cranial and postcranial convergences with 81 crown hominoids; in turn, if interpreted as a non-pliopithecoid stem catarrhine or as a stem 82 hominoid, it would indicate that pliopithecoids were not the only small catarrhines to 83 disperse from Africa into Europe. Unfortunately, determining the evolutionary history and 84 phylogenetic relationships of small-bodied catarrhines is still hindered by the fragmentary 85 nature of their fossil remains and the incompleteness of their fossil record—with long ghost 86 lineages for both pliopithecoids and hylobatids. Deciphering the phylogenetic position of 87 extinct taxa ultimately requires performing cladistic analyses with all available evidence from 88 multiple anatomical regions. However, as mentioned above, previous attempts in this regard 89 have yielded contradictory results for Pliobates (compare Alba et al., 2015b with Nengo et 90 al., 2017 and Gilbert et al., 2020b). Therefore, additional research on the various characters 91 included in these analyses is required to refine their definition and scoring for both extant 92 and extinct taxa. With this aim in mind, here we focus on one of the most intriguing features 93 included in the original diagnosis of Pliobates: the orientation of the carotid canal, which was 94 originally described as most closely resembling hylobatids than either pliopithecoids or other 95 hominoids (Alba et al., 2015b). Given the potential implications of such similarities—if 96 interpreted as a hylobatid synapomorphy or hominoid symplesiomorphy—we decided to 4 97 analyze them quantitatively by means of a novel morphometric approach devised to 98 measure the size and shape of the carotid canal. To the best of our knowledge, this is the 99 first study to quantitatively assess the orientation, proportions, and course of this 100 anatomical structure in extant and extinct anthropoids. 101 102 1.2. The carotid canal in primate phylogenetics 103 All haplorrhines and non-cheirogaleid lemuriforms have an internal carotid artery 104 enclosed by a bony canal (termed ‘carotid canal’) that derives from the anterior lamina of 105 the petrosal plate (Bugge, 1974, 1980; Cartmill et al., 1981; MacPhee and Cartmill, 1986; Kay 106 et al., 2008; Boyer et al., 2016). The carotid canal runs from the posterolaterally located 107 external carotid foramen to the endocranial space (Bugge, 1980; MacPhee and Cartmill, 108 1986; Kay et al., 2008; Boyer et al., 2016). In haplorrhines, the carotid canal displays a 109 ‘perbullar pathway’: it is located in the primary medial wall of the auditory bulla and runs 110 through the middle ear cavity, passing through the transverse septum that separates the 111 tympanic cavity from the anterior accessory cavity (Cartmill et al., 1981; MacPhee and 112 Cartmill 1986; Kay et al., 2008). The canal then opens more anteromedially into the 113 endocranial surface, with an opening located either on the basisphenoid or between the 114 basisphenoid and petrosal (Boyer et al., 2016). In adult anthropoids, one of the two 115 branches of the internal carotid artery (the stapedial artery) is missing and only the 116 promontory artery remains patent (Bugge 1974, 1980; Rosenberger and Szalay, 1980; 117 MacPhee and Cartmill, 1986; Kay et al., 2008; Boyer et al., 2016). Therefore, the carotid 118 canal of anthropoid primates corresponds to a unique and well-developed bony enclosure. 119 Low homoplasy is one of the most important criteria for choosing phylogenetically 120 informative morphological characters (Lieberman, 1999). Among characters least affected by 5 121 homoplasy (convergence, parallelism, reversal, and homoiology), soft-tissue traits have 122 previously been proven more efficient in producing accurate primate phylogenies than hard- 123 tissue traits (Gibbs et al., 2000, 2002; Diogo and Wood, 2011). Given that vessel and nerve- 124 related characters are soft-tissue traits, this rationale might potentially apply to the carotid 125 canal as well. A few previous studies have included carotid canal features in character-taxon 126 matrices devised for cladistic analysis, focusing on the position of the carotid foramen in the 127 bulla as well as on the overall direction and orientation of the canal itself (Beard and 128 MacPhee, 1994; Ross, 1994; Shoshani et al., 1996; Ross et al., 1998; Kay et al., 2008; Alba et 129 al., 2015b). These studies aimed at resolving phylogenetic relationships among higher-rank 130 primate clades (Beard and MacPhee, 1994; Ross, 1994; Shoshani et al., 1996; Ross et al., 131 1998; Kay et al., 2008) and/or deciphering the phylogenetic relationships of particular fossil 132 taxa (Ross et al., 1998; Kay et al., 2008; Alba et al., 2015b). Nonetheless, some of them found 133 differences between platyrrhines and catarrhines, as well as among some catarrhine groups. 134 Based on these studies, the morphology of the carotid canal seems promising for 135 phylogenetic reconstruction in catarrhine primates. However, the exact phylogenetic 136 potential of this structure for distinguishing among catarrhine subclades and, hence, 137 clarifying the relationships of extinct taxa remains unexplored. 138 In the light of all the considerations above, the aim of this paper is threefold: (1) 139 investigate the orientation, proportions, and course of the carotid canal in a wide sample of 140 extant anthropoid primates, to test its potential for phylogenetic inference in extinct 141 catarrhines; (2) re-evaluate previously noted differences and similarities among several 142 catarrhine clades; and (3) describe the carotid canal morphology of Pliobates and compare it 143 with that of extant and fossil catarrhines, so as to gain additional insight on the controversial 144 phylogenetic relationships of this genus. To do so, we collected µCT scans of both fossil taxa 6 145 and an extant comparative sample, and developed a newly devised semiautomatic iterative 146 protocol to obtain comparable 3D carotid canal surfaces (i.e., 3D models). These are 147 analyzed by means of traditional morphometrics and three-dimensional geometric 148 morphometric (3DGM) and imaging techniques, to quantify the orientation, proportions, 149 and course of the carotid canal within the petrosal in a standardized manner. 150 151 2. Materials and methods 152 2.1. Materials 153 Studied sample This study focuses on the cranium of the holotype (IPS58443) of Pliobates 154 cataloniae, which consists of a partial skeleton (Alba et al., 2015b). The carotid canal is 155 entirely preserved in the right temporal fragment (IPS58443.1; Alba et al., 2015b: Fig. 1A). In 156 contrast, in the left temporal fragment (same specimen) the inferiormost part of the canal is 157 broken, with most of its lateral and anterior walls missing along a length of >1.5 mm. 158 IPS58443.1 was originally scanned at a resolution of 95 µm (Alba et al., 2015b). To explore 159 fine bony structures, the specimen was scanned again by X-ray microtomography at the 160 Centro Nacional de Investigación sobre la Evolución Humana (CENIEH, Burgos, Spain) using a 161 Phoenix V|Tome|X s240 μCT scanner with the following parameters: 0.35 mA current, 170 162 kV voltage, 0.2 mm Cu filter, and a magnification of 9.52. The final reconstructed volume has 163 an isometric voxel size of 21 µm. 164 Fossil comparative sample Three Miocene catarrhines are included in the comparative 165 sample: the Miocene pliopithecoid Epipliopithecus vindobonensis (see Zapfe, 1961), the stem 166 cercopithecoid (victoriapithecid) Victoriapithecus macinnesi (see review in Benefit and 167 McCrossin, 1997), and the stem hominoid (proconsulid) Ekembo heseloni (see review in 168 McNulty et al., 2015). There are four available fragmentary temporal bones of 7 169 Epipliopithecus (Zapfe, 1961): NMB OE 303 (right and left temporals, individual II), NHMW 170 1970/1397/0002 (right temporal, individual III; holotype), and NHMW 1970/1397/0003 (left 171 temporal, individual III; holotype). The carotid canal is only sufficiently preserved in the 172 latter specimen, which includes a portion of the bulla, the external acoustic meatus, and 173 most of the petrosal (Zapfe, 1961: Fig. 28). In contrast, the anterior end of the canal is 174 missing from the remaining specimens. The petrosal of the holotype was scanned at the 175 Vienna µCT lab (Vμ-CTL) using a Viscom X8060 μCT scanner with the following parameters: 176 0.24 mA current, 120 kV voltage, 0.5 mm Cu filter, and a magnification of 7.50. The final 177 reconstructed volume has an isometric voxel size of 22 µm. Ekembo heseloni is represented 178 by the right temporal specimen KNM-RU 2036al (Davis and Napier, 1963: Fig. 1; Alba et al., 179 2015b: Fig. 4C), which is part of the holotype (Walker et al., 1993), while V. macinnesi is 180 represented by the left temporal of the cranium KNM-MB 29100 (Benefit and McCrossin, 181 1997: Fig. 1; Benefit, 1999: Fig. 3; Jablonski and Frost, 2010: Fig. 23.1). The μCT scans of 182 KNM-RU 2036al and KNM-MB 29100 have isometric voxel sizes of 64 µm and 44 µm, 183 respectively. They were both kindly made available for this study by the Department of Earth 184 Sciences, National Museums of Kenya (which holds the copyright) and the Department of 185 Human Evolution, Max Plank Institute for Evolutionary Anthropology, Leipzig, Germany. 186 Extant comparative sample Our extant comparative sample consists of μCT scans of 127 187 crania belonging to 41 anthropoid species (13 platyrrhines and 28 catarrhines) from 36 188 genera, representing all extant hominoid genera as well as all cercopithecoid subtribes and 189 platyrrhine (sub)families (Table 1; Supplementary Online Material [SOM] Table S1). A 190 minimum of three individuals per genus are included. Ontogenetic changes in basicranial 191 angulation and elongation—somewhat related to each other and potentially affecting 192 petrosal morphology—continue up to adulthood as measured by M3 eruption (Lieberman 8 193 and McCarthy, 1999; Lieberman et al., 2000). Therefore, all individuals included in the 194 comparative sample are fully adult (exhibiting the M3 in occlusion, or the M2 in 195 callitrichines). Most of the μCT scans included in our extant comparative sample were 196 downloaded from MorphoSource (https://www.morphosource.org; Table 1; SOM Table S1). 197 Three μCT scans of specimens from the AMNH were kindly provided by Sergio Almécija, 198 while seven additional μCT scans of specimens housed in the Institut Català de Paleontologia 199 Miquel Crusafont (Sabadell, Spain) were scanned at the CENIEH. The µCT resolution for the 200 specimens included in the extant comparative sample ranges from 27 to 131 µm (Table 1; 201 SOM Table S1). 202 203 2.2. Morphometric methods 204 Segmentation For each specimen, the carotid canal was virtually extracted through 205 semiautomatic threshold-based segmentation in Avizo v. 7.0 (Visualization Sciences Group, 206 Mérignac). The right carotid canal was segmented for most specimens; when this was not 207 possible, the left canal was segmented and mirrored. Both extremities of the resulting 3D 208 canal surfaces (i.e., 3D models) were cut along homologous planes to allow their comparison 209 among different individuals. First, to cut the posterior end of the canal, we used a best-fit 210 plane based on landmarks placed on its external aperture (Fig. 1a, l; SOM Fig. S1a, b). Then, 211 the anterior limit of the carotid canal was identified by determining a landmark on the 212 anterior part of the petrosal (the ‘intersection ridge–groove’ [IRG]; Fig. 1b–l; SOM Fig. S1c, d; 213 see SOM S1) and cut through such a homologous point using a script coded in R v. 4.0.2 (R 214 Core Team, 2020; see SOM file S1). 215 Alignment and canal orientation The orientation of the carotid canal within the petrosal was 216 quantified by means of two 2D angles representing the canal superior orientation along the 9 217 (para)sagittal plane (2DYZA) and its medial orientation along a transverse plane (2DXYA)— 218 which can be related to canal ‘orientation’ and ‘direction’ sensu Alba et al. (2015b: Tables S5 219 and S6), respectively. Each was computed as the angle between the vector connecting the 220 canal endpoints (from posterior to anterior) and a standard reference vector. Both the canal 221 vector and the reference vector were computed after performing a Procrustes alignment 222 based on petrosal/tympanic landmarks (Fig. 2; Table 2) and applying the translation-scaling- 223 rotation parameters resulting from this alignment to the two canal endpoints (SOM S1). This 224 procedure allowed us to retain the information on the position, orientation and shape 225 variation of the canal inside the petrosal. For both angles, the standard reference vector 226 corresponds to a 2D vector parallel to the anteroposterior cranial axis (from posterior to 227 anterior), in lateral view for 2DYZA and in superior view for 2DXYA (Fig. 3). Aligning the 228 carotid canal on the basis of petrosal/tympanic landmarks instead of the cranium as a whole 229 enables the analysis of fragmentary fossils and has several advantages (see SOM S1 for 230 further details), while being consistent with previous studies (Beard and MacPhee, 1994; 231 Ross, 1994; Shoshani et al., 1996; Ross et al., 1998; Kay et al., 2008; Alba et al., 2015b) that 232 assessed carotid foramen locations relative to the bulla instead of general cranial axes or 233 planes. 234 Volume and length The volume (V, in mm3) of the 3D canal surfaces was computed using the 235 module ‘Surface Area Volume’ in Avizo. To calculate canal length (L, in mm), we used 236 equidistant slices between the two canal endpoints (which correspond to type III landmarks; 237 Bookstein, 1991). These slices were used to generate 8 additional equidistant semilandmarks 238 that follow the canal streamline using the 'digit.curve' function of 'Geomorph' v. 3.3.1 239 (Adams et al., 2019), summing a total of 10 canal landmarks, which define the canal 240 streamline. This canal landmark configuration thus corresponds to a total of 10 canal 10 241 streampoints. Canal length was taken as the sum of the length of the resulting nine canal 242 segments (see SOM S1 for further details). 243 Robusticity index Carotid canal proportions were measured by a dimensionless index of 244 canal robusticity as L/V1/3. Even though the carotid canal is a hollow structure, we refer to 245 this index as ‘robusticity’ because it measures how thick the canal is on average relative to 246 its length. Low values of the index denote robust canals, whereas high values denote slender 247 canals. The same terms have been applied to other hollow cranial structures, such as the 248 semicircular canals of the bony labyrinth (Urciuoli et al., 2020, 2021a, 2021b). The diameter 249 (and hence the cross-sectional area) of the carotid canal varies along its length, so that using 250 the cube root of the volume as the denominator is equivalent to relying on the hypothetical 251 average cross-sectional area. We used the cube root of the volume to make the index 252 dimensionless (i.e., a shape variable), and verified that the robusticity index was free from 253 allometric (size-scaling) effects by computing a phylogenetic generalized least-squares 254 (PGLS) regression of ln L against ln V1/3 on the species means of the whole sample. We 255 further computed a major axis regression of the same parameters, using the function 256 ‘lmodel2’ of the R package 'lmodel2' v. 1.7-3 (Legendre, 2018). The objective was to test 257 whether length and volume of the carotid canal covary in a linear (i.e., isometric) or non- 258 linear (i.e., allometric) fashion, and to do so under the assumptions that the independent 259 variable is measured with error (major axis) or without error (PGLS). Note that, under 260 geometric similarity, linear dimensions and volumes are expected to be correlated with an 261 allometric slope of 1/3 (Gould, 1966). However, this does not apply here because we used 262 the cube root of the canal volume, so that allometry would imply an allometric slope 263 significantly different from 1. 11 264 Canal course configuration To assess carotid canal course within the petrosal, we applied the 265 translation-scaling-rotation parameters resulting from the alignment of the 266 petrosal/tympanic landmarks described above to the 10 canal landmark configuration. As for 267 canal orientation, we were able to retain the position, orientation and shape variation of the 268 canal inside the petrosal. Therefore, the postalignment 10 canal landmark configuration 269 represents the course of the carotid canal within the petrosal (SOM S1) and will be 270 thereafter referred to as ‘carotid canal course configuration’. 271 Size scaling To better understand the influence of size-scaling (allometric) effects on carotid 272 canal shape, we computed multiple bivariate regressions using PGLS. On the one hand, we 273 relied on allometric regressions of canal measurements (as measured by L and V1/3) against 274 overall canal size (as measured by centroid size [CS]; see Section 2.3 below) and body size (as 275 measured by body mass [BM]); the allometric regression between CS and BM was also 276 computed. On the other hand, we computed regressions between carotid canal shape— 277 including not only the above-mentioned index of canal robusticity, but also the multivariate 278 shape variables derived by means of between-group principal component analysis (bgPCA; 279 see Section 2.3 below)—against log-transformed carotid canal size (as measured by V1/3 and 280 CS) and BM. To minimize the effects of body size dimorphism, for each taxon we relied 281 exclusively on average female BM taken from the literature (Smith and Jungers, 1997; 282 Delson et al., 2000; Turner et al., 2018). Natural logarithms (ln) were used in all instances. 283 For allometric regressions between metrical linear variables, the null hypothesis of isometry 284 was rejected when unity was excluded from the 95% CI of the allometric slope. In contrast, 285 for bivariate regressions of shape vs. size, isometry was rejected when the regression was 286 significant (i.e., when the slope was significantly different from zero). 287 12 288 2.3. Between-group principal component analyses 289 Carotid canal course variation among clades was examined using a bgPCA of the carotid 290 canal course configurations using the ‘groupPCA’ function of the R package 'Morpho' v. 2.8 291 (Schlager, 2017). We used the following four main crown anthropoid clades as grouping 292 factor: platyrrhines, cercopithecids, hylobatids, and hominids, with fossils projected onto the 293 morphospace a posteriori. The bgPCA analysis was repeated based on a catarrhine-only 294 subsample to see if it provided a better discrimination among extant catarrhine families 295 once the variance introduced by platyrrhines was removed. 296 As explained in the preceding section, allometric effects on canal course configuration 297 were assessed by means of PGLS regressions of between-group principal component (bgPCs) 298 against log-transformed canal size (ln CS and ln V1/3) or log-transformed body size (ln BM)— 299 see Section 2.2 for further details. 300 We verified that the bgPCA grouping structure was not spurious by comparing our bgPCA 301 results with those of a cross-validated bgPCA (Cardini et al., 2019). This procedure iteratively 302 resamples the data set by excluding one individual at a time and using the remaining (n – 1) 303 individuals to compute the model that is then used to classify the omitted individual. A 304 cross-validated bgPCA significantly reduces the distortion of mean group differences, so that 305 its comparison with the non-cross-validated bgPCA allows one to ascertain whether (or to 306 what extent) the latter is affected by spurious grouping effects. As further recommended by 307 Cardini and Polly (2020), we also relied on a permutational analysis of variance 308 (PERMANOVA) to (1) explore the significance (p) of group mean differences and (2) compute 309 the variance (R2) explained by them. The permutations consist in randomly resampling n 310 times the data set (thus modifying group composition, as the individuals are permuted), 311 each time recomputing the distance between the centroids of all pairs of newly formed 13 312 groups. Based on this set of permutations, here consisting of between-group Euclidean 313 distances, both p and R2 were computed within the full space of canal course configurations 314 as well as within the bgPCA spaces (both standard and cross-validated bgPCs). As in the case 315 of the cross-validated bgPCA scatterplot, comparing the p and R2 values of the bgPCA 316 (before and after cross-validation) with those of the raw canal course configuration 317 coordinates permits verifying if the former inflates the differences among groups. These 318 computations were done using the ‘lm.rrpp’ function of the R package 'RRPP' v. 0.6.1 319 (Collyer and Adams, 2018). 320 The discrimination among a priori defined groups was evaluated on the basis of correctly 321 classified individuals after cross-validation. Posterior probabilities of group membership 322 were computed for fossil specimens based on the squared Mahalanobis distances between 323 their bgPC scores (bgPC1 and bgPC2) and those of extant group centroids using the 324 ‘typprobClass’ function of 'Morpho'. Finally, a neighbor-joining (NJ) cluster was computed 325 using a matrix of Euclidean distances based on the raw coordinates of the canal course 326 configurations with the ‘nj’ function of the R package 'phangorn' v. 2.5.5 (Schliep, 2011). 327 328 2.4. Statistical analyses 329 Box and whisker plots were employed to assess the variation in carotid canal orientation 330 (2DYZA and 2DXYA), proportions (index L/V1/3), and course (bgPC1 and bgPC2). To test for 331 significant differences among extant groups, we used Kruskal-Wallis non-parametric tests 332 and Bonferroni-corrected Mann-Whitney post hoc pairwise comparisons performed in R. 333 When significant differences were found, Pliobates and other extinct genera were compared 334 with the distributions of extant groups using z-scores—computed as z = (individual value – 14 335 group mean) / SD—to assess if the fossil specimens fall within the variation of the latter (|z| 336 ≤ 1.96). 337 338 2.5. Phylogenetically informed methods 339 Phylogenetic tree To compute PGLS regressions and the phylomorphospaces, as well as to 340 quantify phylogenetic signal, we used a time-calibrated phylogenetic tree among the 341 investigated taxa. The tree for the extant taxa was primarily derived from the 10kTrees v.3 342 website (Arnold et al., 2010), although some species that were not included there were 343 added a posteriori following Meyer et al. (2011) and Springer et al. (2012)—see SOM S1 for 344 further details. Two different hypotheses for the extinct genera were considered, except for 345 Victoriapithecus, which was invariably considered a stem cercopithecoid. Under hypothesis 1 346 (SOM Fig. S2a), Epipliopithecus, Ekembo, and Pliobates were considered successive stem 347 hominoids, following the results of Alba et al.’s (2015b) cladistic analysis. In contrast, under 348 hypothesis 2 (SOM Fig. S2b), only Ekembo was considered a stem hominoid, while both 349 Pliobates and Epipliopithecus were considered stem catarrhines (pliopithecoids), following 350 the results of other cladistic analyses (Nengo et al., 2017; Gilbert et al., 2020b). The tip and 351 divergence dates for the extinct taxa are detailed in SOM S1. 352 Phylogenetic signal To assess the phylogenetic signal embedded in carotid canal course, we 353 computed Pagel’s λ (Pagel, 1999) and Blomberg’s K (Blomberg et al., 2003) statistics using 354 the R package 'phytools' v. 0.7-70 (Revell, 2012). These metrics compare the variance in the 355 phylogenetic tree tips relative to that expected under a Brownian motion evolutionary 356 model. Pagel’s λ measures the covariance among related species and ranges from 0 (no 357 phylogenetic signal) to 1 (high phylogenetic signal). In turn, Blomberg’s K reflects to what 358 extent the phylogenetic tree matches the variance in the data (and where variance is 15 359 concentrated): K ≈ 1 means that trait evolution approximates the expectations under a 360 Brownian motion model; K ≪ 1 means that closely related taxa resemble each other less 361 than expected under a Brownian motion model (such that variance is mostly found within 362 clades); finally, K ≫ 1 means that closely related units resemble each other more than 363 expected under a Brownian motion model (such that variance is mostly found among 364 clades). 365 Phylomorphospace and reconstruction of ancestral morphotypes To quantify and visualize 366 patterns of carotid canal course variation along the branches of the phylogeny, we produced 367 a phylomorphospace by projecting the phylogenetic tree onto the two first bgPCs 368 (Sidlauskas, 2008). In a phylomorphospace, the taxon centroids are connected to one 369 another in the morphospace following the phylogenetic relationships given by the chosen 370 phylogeny. In the case of fossil taxa, because there are competitive hypotheses regarding 371 their phylogenetic relationships, various time-calibrated cladograms were devised a priori— 372 resulting in a different phylomorphospace for each phylogenetic hypothesis. Scores of the 373 internal nodes—last common ancestors (LCAs)—were estimated based on a maximum 374 likelihood method for continuous characters using the ‘fastAnc’ function of 'phytools'. To 375 explore the extent to which extant and extinct taxa differ from the various LCAs, we rotated 376 and translated the bgPC scores estimated for the LCAs back into the Kendall’s shape space 377 (i.e., configuration space) to obtain their landmarks coordinates and thus visualize the 378 estimated ancestral carotid canal course conformations. 379 380 3. Results 381 The 3D carotid canal surface of Pliobates (SOM File S2) is compared with those of 382 Ekembo, Epipliopithecus, and Victoriapithecus, as well as selected taxa from the extant 16 383 comparative sample in Figure 4. Individual data for the variables describing the orientation, 384 proportions, and course of the carotid canal in the extant comparative sample are reported 385 in SOM Table S2. 386 387 3.1. Carotid canal orientation 388 For 2DYZA, the angle values range from ca. 30° to 90° (Fig. 5a; Tables 3 and 4). On this 389 basis, and taking into account the mean extant clade differences (SOM Tables S3 and S4), we 390 defined three different states (potentially usable in future cladistic analyses): oblique (<55°); 391 subvertical (55–65°); and vertical (>65º). Hominoids have smaller angles than platyrrhines 392 and cercopithecids, i.e., a less vertically oriented canal, whereas cercopithecids have the 393 highest angles, corresponding to vertical canals (Fig. 5a; Table 3; SOM Tables S3 and S4). 394 Pliobates has the lowest value among Miocene catarrhines and does not significantly differ 395 from the hylobatid range of variation (Fig. 5a; Tables 4 and 5). Epipliopithecus and Ekembo 396 display very similar values to one another and most closely fit with the hominoid variation, 397 although they only significantly differ from cercopithecids (Fig.5a; Tables 4 and 5). In 398 contrast, Victoriapithecus has a higher angle, more similar to the monkey condition (i.e., a 399 more vertical canal; Fig.5a; Table 4), although it does not significantly differ from any extant 400 group (Table 5). 401 In the case of 2DXYA, the values range from ca. -4° to 45° (Fig. 5a; Tables 3 and 4; see 402 SOM S1 regarding the interpretation of negative angles). On this basis, and taking into 403 account the mean differences among extant clades (SOM Tables S3 and S4), we defined two 404 discrete states: anteriorly oriented (≤30°); and medially oriented (>30°). Despite some 405 overlap, platyrrhines display a significantly higher 2DXYA—i.e., a more medially (less 406 anteriorly) directed canal—than the three catarrhine families, which do not differ 17 407 significantly from one another (Fig. 5b; Table 3; SOM Tables S3 and S4). Among the fossil 408 specimens, Ekembo shows the highest 2DXYA whereas Victoriapithecus shows the lowest 409 (Fig.5b; Table 4). Based on the z-scores (Table 5), Ekembo displays a higher angle than extant 410 hominoids, Victoriapithecus and Pliobates show a lower angle than extant platyrrhines, and 411 Epipliopithecus displays a higher angle than hylobatids. 412 413 3.2. Carotid canal proportions and body size scaling 414 Robusticity index The two allometric regressions between L and V1/3 are significant (p < 415 0.001; Table 6), but the PGLS regression indicates slight negative allometry, whereas the 416 major axis regression does not allow us to exclude isometry. This suggests that the canal 417 robusticity index L/V1/3 is only slightly affected by intrinsic size-scaling effects, i.e., that 418 carotid canal length and volume covary in an almost linear fashion when geometric similarity 419 is considered. This is further confirmed by the lack of significant correlation between L/V1/3 420 against CS (Table 6). When separate allometric regressions of L and V1/3 against CS are 421 computed, it emerges that V1/3 and CS display an isometric relationship, whereas L and CS 422 display a minimally positive allometric relationship (with a slope very close to 1). Overall, this 423 suggests that these three variables are suitable proxies for the size of the carotid canal and 424 that the robusticity index is only very slightly affected, if at all, by intrinsic size-scaling effects 425 related to the overall size of the canal. 426 Based on the robusticity index, and despite some overlap, platyrrhines display 427 significantly slenderer canals than catarrhine families, while hominids also display slenderer 428 canals than both cercopithecids and hylobatids, with the latter showing the stoutest canals 429 (Fig. 5c; Table 3; SOM Tables S3 and S4). Ekembo and Epipliopithecus exhibit similar indices, 430 while Victoriapithecus shows a somewhat higher value but similarly overlaps with all 18 431 catarrhine families (Fig. 5c; Table 4), and Pliobates shows an even higher value and falls 432 outside the hylobatid range (Fig. 5c; Table 4). However, none of the fossil specimens 433 significantly differs from any extant clade (Table 5). 434 Body-size allometry The fact that the robusticity index is not markedly affected by the size of 435 the canal itself notwithstanding, either carotid canal size and/or proportions could still be 436 affected by body size-scaling effects as measured by regressions against BM (Table 6). Both L 437 and CS display a negative allometric relationship relative to BM—slightly below the 438 theoretical isometric slope of 1/3 predicted by geometric similarity—whereas, in contrast, 439 V1/3 fits well with the prediction of geometric isometry. This suggests that the volume of the 440 carotid canal is a better proxy for BM than either L or CS if size-scaling effects are not 441 corrected for, because the former variable is directly proportional to BM in a linear fashion, 442 whereas both L and CS display negative allometry relative to BM. The similar allometric 443 pattern of L and CS is not surprising, because the latter was computed from the carotid canal 444 configurations—which are sets of canal streampoints—and, hence, it only contains 445 information about carotid canal length (but not volume). 446 Differences in size scaling between L and V are reflected in the existence of a mild 447 negatively allometric relationship when L is considered the dependent variable, although 448 such a relationship is not significant when both variables are considered to display error (i.e., 449 isometry cannot be excluded in the major axis regression; Table 6). Differences in scaling 450 between L and V relative BM, in any case, are reflected in the regression between the 451 robusticity index and BM (Table 6), which is of inverse proportionality (i.e., a negative slope): 452 since the numerator (L) of the index increases slightly slower than the denominator (V1/3) 453 relative to BM, the index tends to decrease as the body size increases. However, it is 19 454 noteworthy that despite being significant, the correlation between the robusticity index and 455 body size only explains 14% of the variance (Table 6). 456 457 3.3. Carotid canal course within the petrosal 458 Principal components analysis The bgPCA discriminates between platyrrhines and 459 catarrhines but not among catarrhine families (Fig. 6). The cross-validated bgPCA (SOM Fig. 460 S3) is virtually identical to the standard bgPCA, and the permutational ANOVA results (SOM 461 Table S5) show that mean group differences are significant in all instances. The percentage 462 of variance explained by bgPCAs only increases slightly as compared with the raw data— 463 indicating that grouping structure is not spurious. The percentage of correctly classified 464 individuals by the bgPCA (Table 7) is moderately high (79%), even after cross-validation 465 (77%), although much higher in the case of platyrrhines (92%) and hominids (91%) than for 466 cercopithecoids (63%) and hylobatids (64%). 467 bgPC1 (Fig. 6a) explains most of the variance (94%), is not significantly correlated with CS 468 (p = 0.21; Table 6), and shows a strong phylogenetic signal at p < 0.001 (λ = 0.95; K = 3.28). 469 Although bgPC1 is not significantly correlated to ln CS, it displays significant allometry 470 relative to both V1/3 and BM, but only explains 19% and 25% of the shape variance, 471 respectively (Table 6). This axis discriminates platyrrhines (negative scores) from catarrhines 472 (positive scores) at p < 0.001 (Fig. 5d; Table 3; SOM Tables S3 and S4) because of longer 473 canals that originate more inferolaterally in the petrosal of the former (Figs. 6b, 7; SOM Figs. 474 S4–S6). In contrast, catarrhines display shorter canals that originate more superiorly and are 475 overall located more medially in the petrosal (Figs. 6b and 7; SOM Figs. S4–S6). Extant 476 catarrhine families extensively overlap along bgPC1 and do not significantly differ in terms of 477 canal length and foramen/canal superoinferior and mediolateral position (Fig. 5d; Table 3; 20 478 SOM Tables S3 and S4). Pliobates and the remaining fossil taxa display similar (moderately 479 negative) bgPC1 scores that are intermediate between those of extant catarrhines and 480 extant platyrrhines (Figs. 5d and 6a; Table 4) and only slightly overlap with cercopithecids 481 and hominids or, in the case of Epipliopithecus, also marginally with platyrrhines. Their canal 482 configuration (SOM Figs. S4–S6) is thus somewhat shorter and more medially located than in 483 platyrrhines, although Epipliopithecus more closely resembles the platyrrhine condition than 484 Pliobates, Victoriapithecus, and Ekembo do. 485 bgPC2 (Fig. 6a) explains only 5% of variance, is significantly correlated with CS, V1/3, and 486 BM (explaining respectively 58%, 46%, and 35% of the shape variance; Table 6), and shows 487 significant phylogenetic signal at p < 0.001 (λ = 0.99; K = 0.62). Variation along bgPC2 (Figs. 488 6b, 7) reflects the canal position on the petrosal (more anterior toward negative values, and 489 more posterior toward positive scores). Platyrrhines and catarrhines extensively overlap 490 along this axis, but hominids display significantly more positive values than other groups 491 despite considerable overlap (Fig. 5e; SOM Figs. S4–S6; SOM Tables S3 and S4). Along bgPC2, 492 Epipliopithecus, Victoriapithecus and Ekembo are found on slightly negative scores and 493 overlap with platyrrhines, cercopithecids and hylobatids, whereas Pliobates has a slightly 494 positive score that further overlaps with hominids (Figs. 5e and 6a; Table 4). The canal 495 configuration of Pliobates is thus somewhat more posteriorly located than in the other 496 extinct taxa (SOM Figs. S4–S6), particularly Epipliopithecus. 497 Z-scores (Table 5) indicate that, for bgPC1, Epipliopithecus significantly differs from all 498 extant groups, while other extinct taxa match the hominid distribution only. In contrast, for 499 bgPC2, Epipliopithecus, Victoriapithecus and Ekembo only significantly differ from hominids, 500 whereas Pliobates matches the variation of the four extant groups. When both bgPCs are 501 considered simultaneously, the analyzed extinct taxa differ from the four extant anthropoid 21 502 groups except for Pliobates, which matches the hominid variation. This is confirmed by 503 posterior probabilities of group membership (Table 8), according to which Epipliopithecus, 504 Victoriapithecus, and Ekembo are classified as cercopithecoids—although Epipliopithecus 505 falls outside their range of variation (p = 0.018)—whereas Pliobates is classified as a hominid 506 and only significantly differs from platyrrhines. According to these results, the canal course 507 of both Ekembo and Victoriapithecus fit better the cercopithecoid variation, whereas 508 Pliobates fits better with hominids, and Epipliopithecus does not fit well with any of the 509 extant groups analyzed. This notwithstanding, the results of the NJ cluster analysis (Fig. 8) 510 indicate that Pliobates and Epipliopithecus are most similar to one another. Indeed, the four 511 analyzed extinct taxa are grouped together in a cluster of their own, distinct from the 512 subcluster of great apes and that of platyrrhines. 513 The discrimination among the catarrhine families in the catarrhine-only bgPCA is slightly 514 better than in the analysis based on the whole sample but the three groups still overlap to 515 an important extent (SOM Fig. S7). The first axis explains most of the variance (85.33%) and 516 is the only one that discriminates between hominids and the remaining groups—as in the 517 whole sample analysis, in which only one axis (bgPC2) discriminates between them. The 518 results for the fossils in the catarrhine-only analysis are not more conclusive than for the 519 analysis including platyrrhines, with only minor differences. In particular, Epipliopithecus falls 520 closer to the hylobatid centroid and is classified as a hylobatid (p = 0.54), although it displays 521 no significant differences with either cercopithecids (p = 0.30) or hominids (p = 0.29); 522 Victoriapithecus still matches better the cercopithecid range of variation (p = 0.22), as it is 523 the case for Ekembo (p = 0.73), while Pliobates is classified first as a hominid (p = 0.39) but 524 also falls within the range of hylobatids (p = 0.14)—apparently owing to the more hominid- 525 like condition of Symphalangus as compared to other hylobatids. 22 526 Phylomorphospace The results of the phylomorphospace (Fig. 9; SOM Fig. S8) and the 527 reconstructed ancestral morphotypes (Fig. 10; SOM Fig. S9) are virtually identical 528 irrespective of the hypothesis considered. The crown platyrrhine LCA displays a very long 529 canal that originates inferiorly and is located very laterally in the petrosal (Fig. 10e; SOM Fig. 530 S9e), whereas the crown catarrhine LCA displays a condition almost indistinguishable from 531 the mean anthropoid configuration (Fig. 10f; SOM Fig. S9f). The inferred crown 532 cercopithecoid (Fig. 10g; SOM Fig. S9g) and crown hominoid (Fig. 10h; SOM Fig. S9h) 533 ancestral morphotypes are not very divergent from one another or from the catarrhine LCA, 534 but both are more derived than the latter by displaying a somewhat shorter and more 535 medially and posteriorly located canal. The crown hylobatid (Fig. 10i; SOM Fig. S9i) and 536 crown hominid (Fig. 10j; SOM Fig. S9j) LCAs more clearly diverge from the crown hominoid 537 LCA in different directions. Thus, the hylobatid LCA displays a higher bgPC1 score 538 overlapping with extant hylobatids and cercopithecoids, due to a shorter and more medially 539 located canal. In contrast, the hominid LCA displays a higher bgPC2 than the crown hominoid 540 LCA (like the extant hylobatid Symphalangus) due to a more posteriorly located canal. 541 The four analyzed extinct taxa are broadly similar to the crown catarrhine LCA’s condition 542 (Fig. 9). Indeed, both Victoriapithecus (Fig. 10d) and Ekembo (Fig. 10c) most closely resemble 543 the crown catarrhine LCA rather than the crown ceropithecoid (Fig. 10g) or hominoid (Fig. 544 10h) LCAs. Epipliopithecus (Fig. 8b) displays a slightly more platyrrhine-like condition in 545 bgPC1 but is nevertheless very distinct from the crown platyrrhine LCA (Fig. 10e) and closer 546 to the inferred crown catarrhine ancestral condition. Pliobates (Fig. 10a) closely resembles 547 the crown catarrhine LCA in canal length and carotid foramen/canal superoinferior and 548 mediolateral position (as reflected in bgPC1), but unlike Victoriapithecus and Ekembo it 549 appears derived along bgPC2 toward the reconstructed ancestral hominid condition—albeit 23 550 being more similar along this axis to Symphalangus than to extant great apes—due to its 551 more posteriorly located canal. 552 553 4. Discussion 554 4.1. Carotid canal orientation 555 Two different characters were defined by Alba et al. (2015b: Tables S5 and S6) to code 556 the carotid canal morphology, one describing the canal ‘orientation’ (more or less 557 horizontal) and another (following Shoshani et al., 1996) describing the canal ‘direction’ 558 (more or less anterior). In Alba et al. (2015b), canal ‘orientation’ was defined as the 559 inclination between the canal and the basioccipital, whereas canal ‘direction’ was coded 560 based on the spatial relationship between the imaginary line emerging from the canal and 561 the foramen magnum. Canal ‘direction’ sensu Alba et al. (2015b) could be anterior or 562 posterior relative to the foramen magnum, i.e., more laterally or more anteriorly directed, 563 respectively. To evaluate this character, previous authors (Shoshani et al., 1996) placed a 564 straight wire inside the carotid canal and noted the point of intersection between the 565 imaginary line along the wire and the foramen magnum. In Alba et al. (2015b) on the other 566 hand, this character was evaluated conducting visual inspections on μCT scans, drawing the 567 course of the canal on basicranium inferior view and noting the point of intersection 568 between the imaginary line in continuation of the canal course and the foramen magnum. 569 2DYZA (canal superior orientation in the present study) is comparable with Alba et al.’s 570 (2015b) canal ‘orientation’ because this angle is measured along the parasagittal plan, 571 although based on values obtained we prefer to categorize them as 572 oblique/subvertical/vertical (as opposed to horizontal/oblique/subvertical as in Alba et al., 573 2015b). In contrast, 2DXYA (canal medial orientation in the present study) is not directly 24 574 comparable with Alba et al.’s (2015b) canal ‘direction’, because while both studies rely 575 primarily on the line passing through the two canal endpoints (i.e., whole carotid canal), in 576 the present study the alignment process and the quantitative assessment of this character 577 (i.e., possibility of having more than two states vs. two qualitative states only in Alba et al. 578 [2015b]) make that this measure of canal orientation, for a same canal specimen, might 579 differ between the two studies. 580 Pliobates was coded by Alba et al. (2015b) as displaying a horizontal orientation 581 parallel to the basioccipital, like hylobatids, whereas the stem catarrhine Saadanius, 582 cercopithecoids (including Victoriapithecus) and hominids were coded with a subvertical 583 orientation, and Epipliopithecus and Ekembo with an intermediate (oblique) condition. Our 584 results for superior canal orientation (2DYZA) indicate that, in this regard, Pliobates does not 585 differ from hylobatids and display a lower value than Ekembo and Epipliopithecus. However, 586 our results further show that both hylobatids and hominids display a rather oblique 587 orientation (i.e., a horizontal orientation parallel to the basioccipital using the terminology of 588 Alba et al., 2015b), with Ekembo and Epipliopithecus resembling hominoids as a whole. 589 Nonetheless, all these taxa possess a less vertical orientation than platyrrhines (i.e., a 590 subvertical canal) and, especially, cercopithecoids (including Victoriapithecus), in which the 591 canal is vertical. To sum up, according to Alba et al. (2015b) hylobatids and Pliobates had a 592 horizontal canal, Ekembo and Epipliopithecus an oblique canal, and cercopithecoids 593 (including Victoriapithecus) and hominids a subvertical canal, whereas according to our 594 results all hominoids (including Ekembo), Epipliopithecus, and Pliobates have an oblique 595 canal, platyrrhines a subvertical canal, and cercopithecoids (including Victoriapithecus) an 596 even more vertical canal. Therefore, the superior orientation of the carotid canal is 25 597 consistent with the cercopithecoid status of Victoriapithecus, but it is not informative 598 regarding the stem catarrhine vs. stem hominoid status of Pliobates. 599 In turn, Pliobates was coded by Alba et al. (2015b) as resembling Epipliopithecus and 600 hylobatids in the ‘direction’ of the canal. The condition of all these taxa was coded as 601 anteriorly directed, the imaginary line emerging from the carotid canal crossing the foramen 602 magnum more posteriorly than in Saadanius, cercopithecids, and hominids, whose canal 603 would be more laterally directed, as in platyrrhines (see also Shoshani et al., 1996). In 604 contrast, our results indicate that hylobatids extensively overlap in canal medial orientation 605 with other extant catarrhines, which only significantly differ from platyrrhines. Furthermore, 606 Pliobates does not specifically resemble hylobatids or Epipliopithecus, but rather crown 607 catarrhines as a whole, including Victoriapithecus. On the other hand, Epipliopithecus and, 608 especially, Ekembo, exhibit a more platyrrhine-like condition, i.e., a medially directed canal. 609 To sum up, our results indicate that cercopithecoids (including Victoriapithecus), crown 610 hominoids, and Pliobates have an anteriorly oriented canal, whereas platyrrhines, Ekembo, 611 and Epipliopithecus have a medially oriented canal. 612 613 4.2. Carotid canal proportions 614 Volumetric proportions—which were not considered by Alba et al. (2015b)—as 615 measured by the robusticity index of the carotid canal, further confirm the differences in 616 carotid canal morphology between platyrrhines and catarrhines. Our allometric analyses 617 indicate that these differences cannot be explained by body size-scaling effects alone. In 618 particular, the volume of the carotid canal displays an isometric relationships with body size, 619 whereas length displays a slight negative allometric relationship with the latter. As a result, 620 larger-bodied taxa tend to display lower values of the robusticity index (i.e., slightly stouter 26 621 carotid canals). However, correlation with body mass only explains 14% of the variance in 622 carotid canal robusticity. 623 Based on our results, platyrrhines display on average slenderer canals (i.e., lesser 624 volume relative to length) than catarrhines. However, as for medial canal orientation, 625 catarrhine families largely overlap in canal proportions, although great apes display less 626 robust canals than cercopithecids and hylobatids, and are somewhat intermediate between 627 them and platyrrhines. The analyzed fossil catarrhines more closely resemble each other in 628 canal proportions than in canal orientation—being catarrhine-like but further overlapping 629 with the platyrrhine range of variation. Pliobates displays somewhat slenderer canals than 630 the other extinct taxa and, unlike them, it does not overlap with the hylobatids, even more 631 clearly than in the case of medial canal orientation thus further reinforcing the rejection of 632 possible closer similarities with hylobatids the latter family. 633 Our results for the extant taxa are in broad agreement with those previously obtained 634 by Boyer et al. (2016) on relative internal carotid artery area. Boyer et al. (2016) found 635 internal carotid area to scale isometrically with cranial area (measured as the square root of 636 the product between prosthion-inion length and bizygomatic breadth). Among catarrhines, 637 these authors found a pattern of decreasing robusticity in relative carotid artery area from 638 hylobatids to hominids, and to cercopithecids, which is the same as we found for the carotid 639 canal as a whole. The main difference between our results and those of Boyer et al. (2016) is 640 that, in the latter study, the carotid artery robusticity of platyrrhines was found to be 641 intermediate between hylobatids and hominids, whereas our results show that platyrrhines 642 have slenderer canals than catarrhines. This discrepancy is easily explained by the fact that 643 Boyer et al. (2016) considered internal carotid artery area relative to cranial area, whereas 644 we relied on the ratio between the length and the volume of the canal. Our results (see 27 645 below for further discussion) indicate that platyrrhines display relatively longer carotid 646 canals than catarrhines, thereby resulting in less robust volumetric proportions for the canal 647 as a whole. Given that Boyer et al. (2016) did not took canal length into account, this 648 explained the above-mentioned discrepancy for platyrrhines (but not the other groups) as 649 compared to our results. 650 651 4.3. Carotid canal course 652 Our bgPCA results indicate that the morphology of the canal embeds strong phylogenetic 653 signal, although most of the variance merely discriminates between platyrrhines and 654 catarrhines, due to differences in canal length and location reflected in bgPC1. In particular, 655 extant catarrhines display shorter canals that originate more superiorly than in platyrrhines. 656 It is noteworthy to stress beforehand that Pliobates and hylobatids do not show particular 657 similarities in bgPC1 scores. 658 Previous studies suggested that the carotid foramen is located ventrally (Ross, 1994; Ross 659 et al., 1998) and posteriorly (Ross, 1994; Ross et al., 1998; Kay et al., 2008) in both 660 platyrrhines and catarrhines, whereas Alba et al. (2015b) noted some differences in the 661 anteroposterior carotid foramen location between hylobatids and other catarrhines (with 662 the former displaying a more anteriorly located carotid foramen). Previous studies also 663 noted that platyrrhines display a more medially located foramen than catarrhines (Ross et 664 al., 1998; Kay et al., 2008). Our bgPC1 results indicate that extant platyrrhines and 665 catarrhines differ not only regarding the length of the canal, but also the location of the 666 (external) carotid foramen, such that in catarrhines the entire carotid canal (i.e., not only the 667 foramen but the whole structure) is more superiorly and medially located within the 668 petrosal. Our results therefore point to differences between platyrrhines and catarrhines in 28 669 terms of dorsoventral location of the carotid foramen, contrary to previous studies (Ross, 670 1994; Ross et al., 1998). This might be attributable to the different methodologies employed 671 (i.e., because of having performed an alignment on the specimens in the present study). 672 Similarly, because of the alignment, our conclusions for the location along the mediolateral 673 axis might be the opposite as those noted by previous authors (Ross et al., 1998; Kay et al., 674 2008), who simply recorded foramen position relative to the bulla. Since catarrhines possess 675 a tubular ectotympanic, catarrhine landmark configurations are shifted more medially. This 676 is the consequence of alignment optimization of their tympanic landmarks with those of 677 platyrrhines, whose annular ectotympanic causes their tympanic landmarks to be more 678 medially positioned. Regardless, we observe an extensive overlap among extant catarrhine 679 families in the position of the carotid foramen, at least based on bgPC1. 680 Our results for bgPC2 lend some support that hylobatids display on average a more 681 anteriorly located foramen in the petrosal than hominids. However, rather than indicating a 682 particular condition for hylobatids, our bgPC2 results suggest that only hominids differ in this 683 regard from other catarrhines by displaying a more posteriorly located foramen. This 684 conclusion should be taken with caution, given that bgPC2 shows an extensive overlap 685 between hominids and the upper range of other crown catarrhines. Furthermore, bgPC2 686 reflects a smaller amount of variance, embeds lesser phylogenetic signal than bgPC1, and is 687 more strongly correlated with size—although the percentage of variance explained by 688 allometry varies depending on whether canal size or body size is considered, our results 689 clearly indicate that this axis is more strongly influenced by size-scaling than bgPC1. 690 Therefore, one should be cautious when interpreting similarities among taxa based on 691 bgPC2 scores. 29 692 Our 3DGM results for the analyzed fossil taxa do not support particular similarities 693 between Pliobates and hylobatids in terms of overall carotid canal course (contra Alba et al., 694 2015b), and indicate instead that Pliobates is most similar in this regard to Epipliopithecus, 695 Victoriapithecus and Ekembo. All these fossil catarrhines appear overall closer to extant 696 catarrhines in terms of canal length and location in the petrosal, albeit being somewhat 697 intermediate between platyrrhines and catarrhines. Epipliopithecus appears slightly more 698 platyrrhine-like, whereas the other extinct genera only overlap with the lowest range of 699 cercopithecids and hominids (but not hylobatids). Pliobates only differs slightly from the 700 remaining fossil catarrhines by more closely resembling hominids and Symphalangus along 701 bgPC2 (reflecting a more posteriorly located foramen/canal). As noted above, bgPC2 is quite 702 strongly correlated with CS, suggesting that the differences between hominids and the rest 703 of the sample could merely result from the larger body size of the former. However, this 704 cannot explain the differences along this axis between Pliobates and Epipliopithecus, as the 705 former more closely resembles the large-bodied hominids and the largest-bodied hylobatid 706 (Symphalangus) despite a much lower estimated body size than Epipliopithecus (4–5 kg vs. 707 11–12 kg, respectively; Alba et al., 2015b). 708 709 4.4. Evolutionary implications 710 Alba et al. (2015b) concluded that Pliobates displays a mosaic of plesiomorphic (stem 711 catarrhine-like) cranial and postcranial features and more derived (crown hominoid) cranial 712 and postcranial synapomorphies, coupled with a dendropithecid-like dentition and some 713 cranial similarities with hylobatids. A stem hominoid status more derived than that of 714 Ekembo was supported for Pliobates by Alba et al. (2015b) based on the results of their 30 715 cladistic analysis. Such a phylogenetic placement would support the view that some 716 purported synapomorphies of crown catarrhines, best exemplified by the completely 717 ossified tubular ectotympanic, would have been independently acquired independently in 718 cercopithecoids and hominoids—a possibility taken into account by some previous authors 719 (Begun, 2002). A stem hominoid status for Pliobates would further support that the last 720 common ancestor of crown hominoids would have been more hylobatid-like than previously 721 assumed (Alba et al., 2015b). Interestingly, one of the resemblances between Pliobates and 722 hylobatids found by Alba et al. (2015b) was the carotid canal morphology, which is at odds 723 with the results of our more elaborate analyses of this anatomical structure, which denote a 724 generalized catarrhine morphology with some closer similarities with extant hominids and 725 uniquely Symphalangus among hylobatids. 726 Contrary to Alba et al. (2015b), subsequent cladistic analyses–based on different taxon- 727 character matrix–have recovered Pliobates as a stem catarrhine more closely related to 728 pliopithecoids (Nengo et al., 2017; Gilbert et al., 2020b). Based on these results, Gilbert et al. 729 (2020a: 399) concluded that “Pliobates is most likely a derived, late-occurring pliopithecoid 730 or other stem catarrhine taxon that has converged on some interesting but minor details of 731 extant hominoid elbows and wrists”. In fact, the similarities in elbow and wrist morphology 732 between Pliobates and hominoids are not minor but quite extensive, more so than those 733 displayed by Ekembo and certainly than those convergently acquired by atelids (Alba et al., 734 2015b). However, it is not unconceivable that these similarities—being functionally related 735 to enhanced forearm rotation and ulnar deviation capabilities—might have independently 736 evolved between Pliobates and crown hominoids. Under such scenario, in which postcranial 737 similarities between Pliobates and crown hominoids would be dismissed as convergences, it 738 should be taken into account that the dental morphology of Pliobates suggests much closer 31 739 affinities to dendropithecids (particularly Micropithecus) than to pliopithecoids (Alba et al., 740 2015b). A dendropithecid status for Pliobates would not imply an independent acquisition of 741 a fully-ossified ectotympanic, but would not automatically resolve the phylogenetic 742 placement of the former, given the controversies surrounding the phylogenetic relationships 743 of dendropithecids—considered either stem catarrhines more derived than pliopithecoids 744 (Harrison, 2010, 2013, 2017; Nengo et al., 2017; Gilbert et al., 2020b) or stem hominoids 745 more basal than proconsulids (Rae, 1999; Zalmout et al., 2010; Begun, 2015; Alba et al., 746 2015b; Rossie and Hill, 2018). 747 Unfortunately, the carotid canal morphology of dendropithecids is currently unknown, 748 thereby precluding a direct comparison with Pliobates. Furthermore, despite substantial 749 differences between platyrrhines and catarrhines in terms of carotid canal orientation and 750 proportions as well as carotid canal length and (foramen) location in the petrosal, our results 751 indicate a considerable overlap among extant catarrhines. As a result, carotid canal 752 morphology is of limited utility for investigating the phylogenetic position of fossil 753 catarrhines, although some insights can be drawn for the investigated taxa. Before 754 discussing the evolutionary implications of our results, however, it should be noted that the 755 small sample sizes available for extinct taxa make it necessary interpret the differences 756 among the analyzed fossil specimens by assuming that they are more or less representative 757 of their respective taxa. Although this caveat is not restricted to the carotid canal 758 morphology, it should not be forgotten that we cannot know a priori how similar in carotid 759 canal morphology a particular fossil specimen is to the (unknown) average condition of the 760 taxon to which it belongs. For example, given the considerable variation that can be 761 observed in some extant taxa for carotid canal course (e.g., Aotus, Macaca, Mandrillus), the 762 possibility remains that some of the differences inferred among extinct taxa, or between 32 763 them and extant catarrhine clades, might vanish (or become more accentuated) if additional 764 specimens became available. In the meantime, caution is required by bearing in mind that 765 the reliability of inferences for fossil taxa is limited by the number of specimens available to 766 them, which often preclude an adequate assessment of intraspecific variability. 767 Based on the currently available specimens, our results indicate that the stem hominoid 768 Ekembo and the stem cercopithecoid Victoriapithecus are virtually identical in carotid canal 769 volumetric proportions and course morphology, although Ekembo displays a more 770 platyrrhine-like medial orientation (like Epipliopithecus) and a more hominoid-like superior 771 orientation (like Epipliopithecus and Pliobates to some extent). The less medially/more 772 anteriorly oriented carotid canal in Pliobates than in Ekembo would be consistent with a 773 more derived stem hominoid status, but given the modern cercopithecoid-like carotid canal 774 orientation of Victoriapithecus, the condition of extant cercopithecoids and hominoids might 775 have evolved independently to some extent. In turn, based on the condition of 776 Epipliopithecus, the possession of an obliquely oriented canal in hominoids might represent 777 a catarrhine symplesiomorphy; in that case, the condition of extant cercopithecoids would 778 have independently evolved from that of extant platyrrhines. Alternatively, the oblique 779 orientation of the canal might be synapomorphic for hominoids, in which case the similar 780 condition of Epipliopithecus would be a convergence. Either way, this feature is not 781 informative regarding the phylogenetic position of Pliobates. 782 In terms of carotid canal course, as reflected by our 3DGM analysis, Pliobates does not 783 show particular affinities with hylobatids, and rather resembles the other analyzed extinct 784 taxa, which are overall more similar to catarrhines but somewhat intermediate between 785 platyrrhines and both cercopithecids and hominids. The phylomorphospace shows that both 33 786 Victoriapithecus and Ekembo apparently reflect well the inferred ancestral crown catarrhine 787 morphotype, despite their generally accepted crown catarrhine (respectively, stem 788 cercopithecoid and stem hominoid) status (e.g., Benefit and McCrossin, 1997; Benefit, 1999; 789 Alba et al., 2015b; Nengo et al., 2017; Gilbert et al., 2020b). In contrast, Epipliopithecus 790 appears somewhat more plesiomorphic, whereas Pliobates appears slightly derived (in the 791 anteroposterior position of the carotid foramen/canal) toward the great ape condition, 792 being closer to the inferred ancestral hominid morphotype as well as the hylobatid 793 Symphalangus. 794 In the case of Epipliopithecus, our results are in agreement with the widely held opinion 795 that pliopithecoids are a clade of stem catarrhines, predating the divergence of 796 cercopithecoids and hominoids (Harrison, 2013; Begun, 2002, 2017; Nengo et al., 2017; 797 Gilbert et al., 2020a, b; Sankhyan et al., 2017; Harrison et al., 2020; Urciuoli et al., 2021; 798 contra Alba et al., 2015b). For Pliobates, our results are more ambiguous and subject to 799 interpretation, although it is noteworthy that they do not support a closer relationship with 800 pliopithecoids–at least, based on the information provided by Epipliopithecus. They suggest 801 instead that, in carotid canal morphology, Pliobates is more derived than Epipliopithecus 802 toward the inferred ancestral crown catarrhine morphotype, like Ekembo and 803 Victoriapithecus. Furthermore, our results discount very close similarities in carotid canal 804 morphology between Pliobates and hylobatids as a whole, but indicate some resemblances 805 with the hylobatid Symphalangus and great apes. In the latter regard, Pliobates is 806 approximately equidistant from the crown catarrhine and crown hominoid reconstructed 807 morphotypes. It is uncertain whether this reflects a more derived status of Pliobates or an 808 independent development, although the latter is plausible in light of the Symphalangus 809 condition—if the hylobatid LCA reconstructed in the phylomorphospace is broadly correct, 34 810 Symphalangus must have independently acquired a more great ape-like condition; 811 alternatively, the condition shared by the hominid LCA and Symphalangus might be 812 plesiomorphic for crown hominoids and the reconstructed hylobatid LCA would reflect a 813 secondary reversal of the remaining hylobatids. Of course, the morphology of a single 814 anatomical region cannot provide much insight on the phylogenetic position of a given taxon 815 among catarrhines, particularly if—as it is the case of the carotid canal—it does not 816 adequately discriminate among catarrhine clades. However, our results for this structure do 817 not show any particular resemblances between pliopithecoids and Pliobates (beyond those 818 displayed by all the extinct taxa analyzed), and further fail to contradict the alternative 819 phylogenetic status of Pliobates as a stem hominoid. The phylogenetic status of Pliobates 820 thus seems far from being settled. 821 822 5. Summary and conclusions 823 Previous research on the primate carotid canal within a phylogenetic context has been 824 limited to the inclusion of qualitative features related to carotid foramen position in the 825 bulla and overall direction and orientation of the canal itself in character-taxon matrices 826 devised for cladistic analysis (Beard and MacPhee, 1994; Ross, 1994; Shoshani et al., 1996; 827 Ross et al., 1998; Kay et al., 2008; Alba et al., 2015b). From a quantitative viewpoint, only 828 Boyer et al. (2016) investigated the allometric relationship between carotid canal area and 829 BM (Boyer et al., 2016), whereas some authors relied on carotid canal area for inferring 830 encephalic blood flow rates and blood flow metabolism by further taking vertebral canal size 831 into account (Boyer and Harrington, 2018, 2019; Beaudet et al., 2020). Thus, our study is the 832 first to rigorously quantify the orientation, direction, overall course and volumetric 35 833 proportions of the carotid canal in 3D among a large sample of anthropoids. The allometric 834 regressions performed also represent an advancement in the understanding of body size- 835 scaling effects on carotid canal morphology. The methodological protocols devised here 836 enable repeatability among different researchers and thus will enable a refinement of our 837 conclusions by adding additional extant and fossil specimens in the future. The analyzed 838 variables are no panacea for clarifying the phylogeny of extinct catarrhines, because various 839 extant groups overlap to a large extent and also because intraspecific variation will be 840 difficult to evaluate due to small samples sizes. However, the analyzed variables offer the 841 prospect to provide taxonomically and phylogenetically relevant information for extinct taxa, 842 particularly if they are added to character-taxon matrices devised for cladistic analysis 843 including information from other anatomical areas. Other future directions of work on the 844 carotid canal should deepen our current understanding about the influence of function, as 845 well as allometric and phylogenetic constraints, on the evolution of this anatomical 846 structure. This would be required not only to better interpret carotid canal morphology from 847 an adaptive viewpoint, but also to derive from it paleobiological inferences for extinct taxa. 848 In this study, we rely on 3D morphometric techniques to quantify the orientation, 849 proportions, and course of the carotid canal in the small-bodied catarrhine Pliobates, and 850 compare it with extant anthropoids and other Miocene catarrhines (Epipliopithecus, 851 Victoriapithecus, and Ekembo). Our results reveal that Pliobates and the other analyzed 852 Miocene catarrhines are broadly similar to one another in carotid canal morphology and 853 more closely resemble extant catarrhines than platyrrhines. Unfortunately, carotid canal 854 morphology does not adequately discriminate among catarrhine clades, thereby hindering a 855 more detailed phylogenetic assessment of Pliobates on this basis. The difficulty of assessing 856 intraspecific variation in carotid canal morphology in the extinct taxa, coupled with the lack 36 857 of marked differences among extant catarrhine families, make it impossible to reach 858 definitive conclusions about the systematic position of Pliobates on this basis alone. 859 However, our results rule out the previously noted similarities in carotid canal morphology 860 between Pliobates and hylobatids and indicate instead that this extinct catarrhine is broadly 861 similar to catarrhines as a whole and, in particular, to other Miocene catarrhines (both stem 862 and crown). 863 It is however noteworthy that the carotid canal of Pliobates is not particularly similar to 864 that of pliopithecoids (as represented by Epiplioithecus, which appears somewhat more 865 platyrrhine-like), thereby not supporting (or contradicting) the hypothesis that Pliobates is a 866 pliopithecoid. Like the stem hominoid Ekembo and the stem cercopithecoid Victoriapithecus, 867 Pliobates appears somewhat more derived than Epipliopithecus toward the modern 868 catarrhine condition, and indeed, in some respects, Pliobates appears even more derived 869 toward the ancestral hominoid morphotype than the remaining Miocene catarrhines 870 analyzed, more closely resembling extant hominids and the hylobatid Symphalangus. 871 Therefore, our results do not enable us to discount the hypothesis that Pliobates is a stem 872 hominoid rather than a stem catarrhine. Given the dental similarities between Pliobates and 873 African dendropithecids, the former might be a late offshot of the latter clade that dispersed 874 into Eurasia independently from pliopithecoids. Unfortunately, this possibility could not be 875 directly assessed in this study because the carotid canal morphology of dendropithecids is 876 unknown. Additional analyses focused on other anatomical (both cranial and postcranial) 877 areas will be required to further clarify the phylogenetic relationships of Pliobates, as well as 878 to clarify whether its postcranial similarities with crown hominoids might have been 879 independently acquired. 37 880 881 Acknowledgments 882 This work has been funded by the Agencia Estatal de Investigación (CGL2016-76431-P and 883 CGL2017-82654-P, AEI/FEDER, EU), the Generalitat de Catalunya (CERCA Programme and 884 Consolidated Research Groups 2017 SGR 086 and 2017 SGR 116), the Ministerio de Ciencia, 885 Innovación y Universidades (PRE2018-083299 to F.B. and BES-2015–071318 to A.U.), and the 886 European Commission under the Marie Skłodowska-Curie Individual Fellowship Programme 887 (H2020-MSCA-IF-2018-837966 to M.P.). A.B. was funded by the Centre of Excellence in 888 Palaeosciences (CoE-Pal), Institut français d’Afrique du Sud (IFAS), National Research 889 Foundation (NRF), Palaeontological Scientific Trust (PAST), and Wits. Regarding the CT 890 scanning of Epipliopithecus, this research also received support from the Synthesys Project 891 (http://synthesys3.myspecies.info/), which is financed by the European Community 892 Research Infrastructure Action under the FP7 (ref. AT-TAF-4689). We thank Fred Spoor, the 893 Department of Earth Sciences of the National Museums of Kenya, and the Department of 894 Human Evolution of the Max Planck Institute for Evolutionary Anthropology in Lepzig for 895 granting access to the CT scans of Ekembo and Victoriapithecus, as well as Ursula Göhlich for 896 access to the Epipliopithecus material housed in the Naturhistorisches Museum Wien and 897 Martin Dockner for assistance during the CT scanning process of this material at the µCT lab 898 of Universität Wien. We also thank the following people and institutions for providing access 899 to scans used in this study: Eric Delson and the American Museum of Natural History 900 (AMNH), funded by the AMNH and the New York Consortium in Evolutionary Primatology 901 (NYCEP); Lauren A. Gonzales, the Museum of Comparative Zoology (MCZ), the Southern 902 Illinois University (SIU), and the Smithsonian National Museum of Natural History (USNM), 903 funded by NSF grants BCS 1440742 & 1552848; Kari L. Allen and the Duke University, 38 904 Department of Evolutionary Anthropology (DUEA), regarding scans that originally appeared 905 in Allen’s PhD dissertation, funded by an NSF DDIG and the Leakey Foundation; Lynn Copes, 906 Lynn Lucas, and the MCZ, funded by an NSF DDIG #0925793 and the Wenner Gren 907 Foundation; and Sergio Almécija, funded by NSF grant BCS 1316947 and Stony Brook 908 University. These scans were downloaded from MorphoSource.org, a web-accessible archive 909 for 3D digital data housed by Duke University. We further thank the Editor (Clément Zanolli), 910 the Associate Editor (Eric Delson), and three anonymous reviewers for helpful comments 911 that helped us to improve a previous version of this paper. 912 913 References 914 Adams, D.C., Collyer, M.L., Kaliontzopoulou, A., 2019. Geomorph: Software for geometric 915 morphometric analyses. R Package Version 3.3.1. https://cran.r- 916 project.org/web/packages/geomorph/index.html. 917 Alba, D.M., Delson, E., Carnevale, G., Colombero, S., Delfino, M., Giuntelli, P., Pavia, M., 918 Pavia, G., 2014a. First joint record of Mesopithecus and cf. Macaca in the Miocene of 919 Europe. J. Hum. Evol. 67, 1–18. 920 Alba, D.M., Colombero, S., Delfino, M., Martínez-Navarro, B., Pavia, M., Rook, L., 2014b. A 921 thorny question: The taxonomic identity of the Pirro Nord cervical vertebrae revisited. J. 922 Hum. Evol. 76, 92–106. 923 Alba, D.M., Montoya, P., Pina, M., Rook, L., Abella, J., Morales, J., Delson, E., 2015a. First 924 record of Mesopithecus (Cercopithecidae, Colobinae) from the Miocene of the Iberian 925 Peninsula. J. Hum. Evol. 88, 1–14. 39 926 Alba, D.M., Almécija, S., DeMiguel, D., Fortuny, J., Pérez de los Ríos, M., Pina, M., Robles, 927 J.M., Moyà-Solà, S., 2015b. Miocene small-bodied ape from Eurasia sheds light on 928 hominoid evolution. Science 350, aab2625. 929 Arnold, C., Matthews, L.J., Nunn, C.L., 2010. The 10kTrees website: A new online resource for 930 primate phylogeny. Evol. Anthropol. 19, 114–118. 931 Barry, J.C., Jacobs, L.L., Kelley, J., 1987. An early middle Miocene catarrhine from Pakistan 932 with comments on the dispersal of catarrhines into Eurasia. J. Hum. Evol. 15, 501–508. 933 Beard, K.C., MacPhee, R.D.E. 1994. Cranial anatomy of Shoshonius and the antiquity of 934 Anthropoidea. In: Fleagle, J.G., Kay, R.F. (Eds.), Anthropoid Origins. Plenum Press, New 935 York, pp. 55–97. 936 Beaudet, A., Clarke, R.J., Heaton, J.L., Pickering, T.R., Carlson, K.J., Crompton, R.H., 937 Jashashvili, T., Bruxelles, L., Jakata, K., Bam, L., Van Hoorebeke, L., Kuman, K., Stratford, 938 D., 2020. The atlas of StW 573 and the late emergence of human-like head mobility and 939 brain metabolism. Sci. Rep. 10, 4285. 940 Begun, D.R., 2002. The Pliopithecoidea. In: Hartwig, W.C. (Ed.), The Primate Fossil Record. 941 Cambridge University Press, Cambridge, pp. 221–240. 942 Begun, D.R., 2015. Fossil record of Miocene hominoids. In: Henke, W., Tattersall, I. (Eds.), 943 Handbook of Paleoanthropology, 2nd ed. Springer, Heidelberg, pp. 1261–1332. 944 Begun, D.R., 2017. Evolution of the Pliopithecoidea. In: Fuentes, A. (Ed.), The International 945 Encyclopedia of Primatology. John Wiley & Sons, Hoboken. 946 https://doi.org/10.1002/9781119179313.wbprim0165. 947 Benefit, B.R., 1999. Victoriapithecus: the key to Old World monkey and catarrhine origins. 948 Evol. Anthropol. 7, 155–174. 40 949 Benefit, B.R., McCrossin, M.L., 1997. Earliest known Old World monkey skull. Nature 388, 950 368–371. 951 Benefit, B.R., McCrossin, M.L., 2015. A window into ape evolution. Science 350, 515–516. 952 Bernor, R.L., Flynn, L.J., Harrison, T., Hussain, S.T., Kelley, J., 1988. Dionysopithecus from 953 southern Pakistan and the biochronology and biogeography of early Eurasian catarrhines. 954 J. Hum. Evol. 17, 339–358. 955 Blomberg, S.P., Garland, T., Ives, A.R., 2003. Testing for phylogenetic signal in comparative 956 data: behavioral traits are more labile. Evolution, 57, 717–745. 957 Bookstein, F.L., 1991. Morphometric Tools for Landmark Data: Geometry and Biology. 958 Cambridge University Press, Cambridge. 959 Boyer, D.M., Kirk, E.C., Silcox, M.T., Gunnell, G.F., Gilbert, C.C., Yapuncich, G.S., Allen, K.L., 960 Welch, E., Bloch, J.I., Gonzales, L.A., Kay, R.F., Seiffert, E.R., 2016. Internal carotid arterial 961 canal size and scaling in Euarchonta: Re-assessing implications for arterial patency and 962 phylogenetic relationships in early fossil primates. J. Hum. Evol. 97, 123–144. 963 Boyer, D.M., Harrington, A.R., 2018. Scaling of bony canals for encephalic vessels in 964 euarchontans: Implications for the role of the vertebral artery and brain metabolism. J. 965 Hum. Evol. 114, 85–101. 966 Boyer, D.M., Harrington, A.R., 2019. New estimates of blood flow rates in the vertebral 967 artery of euarchontans and their implications for encephalic blood flow scaling: A 968 response to Seymour and Snelling (2018). J. Hum. Evol. 128, 93–98. 969 Bugge, J., 1974. The cephalic arterial system in insectivores, primates, rodents, and 970 lagomorphs, with special reference to the systematic classification. Acta Anat. 87, 1–160. 41 971 Bugge, J., 1980. Comparative anatomical study of the carotid circulation in New and Old 972 World Primates. In: Ciochon, R.L., Chiarelli, A.B. (Eds.), Evolutionary Biology of the New 973 World Monkeys and Continental Drift. Plenum Press, New York, pp. 293–316. 974 Cardini, A., O'Higgins, P., Rohlf, F.J., 2019. Seeing distinct groups where there are none: 975 spurious patterns from between-group PCA. Evol. Biol. 46, 303–316. 976 Cardini, A., Polly, P.D., 2020. Cross-validated between group PCA scatterplots: A solution to 977 spurious group separation? Evol. Biol. 47, 85–95. 978 Cartmill, M., MacPhee, R.D.E., Simons, E.L., 1981. Anatomy of the temporal bone in early 979 anthropoids, with remarks on the problem of anthropoid origins. Am. J. Phys. Anthropol. 980 56, 1–21. 981 Casanovas-Vilar, I., Alba, D.M., Garcés, M., Robles, J.M., Moyà-Solà, S., 2011. Updated 982 chronology for the Miocene hominoid radiation in Western Eurasia. Proc. Natl. Acad. Sci. 983 USA 108, 5554–5559. 984 Chatterjee, H.J., Ho, S.Y.W., Barnes, I., Groves, C. 2009. Estimating the phylogeny and 985 divergence times of primates using a supermatrix approach. BMC Evol. Biol. 13, 259. 986 Collyer, M.L., Adams, D.C., 2018. RRPP: An R package for fitting linear models to high‐ 987 dimensional data using residual randomization. Methods Ecol. Evol. 9, 1772–1779. 988 Delson, E., Terranova, C.J., Sargis, E.J., Jungers, W.L., Jablonski, N.G.C., Dechow, P., 2000. 989 Body mass in Cercopithecidae (Primates, Mammalia): estimation and scaling in extinct 990 and extant taxa. Anthropol. Pap. Am. Mus. Nat. Hist. 83, 1–159. 991 Diogo, R., Wood, B., 2011. Soft-tissue anatomy of the primates: Phylogenetic analyses based 992 on the muscles of the head, neck, pectoral region and upper limb, with notes on the 993 evolution of these muscles. J. Anat. 219, 272–359. 42 994 Finstermeier, K., Zinner, D., Brameier, M., Meyer, M., Kreuz, E., Hofreiter, M., Roos, C., 2013. 995 A mitogenomic phylogeny of living primates. PLoS One 8, e69504. 996 Gibbs, S., Collard, M., Wood, B., 2000. Soft-tissue characters in higher primate phylogenetics. 997 Proc. Natl. Acad. Sci. USA 97, 11130–11132. 998 Gibbs, S., Collard, M., Wood, B., 2002. Soft-tissue anatomy of the extant hominoids: A review 999 and phylogenetic analysis. J. Anat. 200, 3–49. 1000 Gilbert, C.C., Bibi, F., Hill, A., Beech, M.J., 2014. Early guenon from the late Miocene 1001 Baynunah Formation, Abu Dhabi, with implications for cercopithecoid biogeography and 1002 evolution. Proc. Natl. Acad. Sci. USA 111, 10119–10124. 1003 Gilbert, C.C., Pugh, K.D., Fleagle, J.G., 2020a. Dispersal of Miocene hominoids (and 1004 pliopithecoids) from Africa to Eurasia in light of changing tectonics and climate. In: 1005 Prasad, G.V., Patnaik, R. (Eds.), Biological Consequences of Plate Tectonics. New 1006 Perspectives on post-Gondwana Break-up—A Tribute to Ashok Sahni. Springer, Cham, pp. 1007 393-412. 1008 Gilbert, C.C., Ortiz, A., Pugh, K.D., Campisano, C.J., Patel, B.A., Premjit, N., Fleagle, J.G., 1009 Patnaik, R., 2020b. New Middle Miocene ape (Primates: Hylobatidae) from Ramnagar, 1010 India fills major gaps in the hominoid fossil record. Proc. R. Soc. B. 287, 20201655. 1011 Gould, S.J., 1966. Allometry and size in ontogeny and phylogeny. Biol. Rev., 41, 587–640. 1012 Harrison, T., 2010. Dendropithecoidea, Proconsuloidea and Hominoidea (Catarrhini, 1013 Primates). In: Werdelin, L., Sanders, W.J. (Eds.), Cenozoic Mammals of Africa. University of 1014 California Press, Berkeley, pp. 429–469. 1015 Harrison, T., 2013. Catarrhine origins. In: Begun, D.R. (Ed.), A Companion to 1016 Paleoanthropology. Blackwell Publishing, Oxford, pp. 376–396. 43 1017 Harrison, T., 2016. The fossil record and evolutionary history of hylobatids. In: Reichard, 1018 U.H., Hirai, H., Barelli, C. (Eds.), Evolution of Gibbons and Siamang. Phylogeny, 1019 Morphology, and Cognition. Springer, New York, pp. 90–110. 1020 Harrison, T., 2017. Miocene primates. In: Fuentes, A. (Ed.), The International Encyclopedia of 1021 Primatology. John Wiley & Sons, Hoboken. 1022 https://doi.org/10.1002/9781119179313.wbprim0227. 1023 Harrison, T., Gu, Y., 1999. Taxonomy and phylogenetic relationships of early Miocene 1024 catarrhines from Sihong, China. J. Hum. Evol. 37, 225–277. 1025 Harrison, T., Zhang, Y., Wei, G., Sun, C., Wang, Y., Liu, J., Tong, H., Huang, B., Xu, F., 2020. A 1026 new genus of pliopithecoid from the late Early Miocene of China and its implications for 1027 understanding the paleozoogeography of the Pliopithecoidea. J. Hum. Evol. 145, 102838. 1028 Harzhauser, M., Kroh, A., Mandic, O., Piller, W. E., Göhlich, U., Reuter, M., Berning, B., 2007. 1029 Biogeographic responses to geodynamics: A key study all around the Oligo–Miocene 1030 Tethyan Seaway. Zool. Anz. 246, 241–256. 1031 Kay, R.F., Simons, E., Ross, J.L., 2008. The basicranial anatomy of African Eocene/Oligocene 1032 anthropoids. Are there any clues for platyrrhine origins? In: Fleagle, J.G., Gilbert, C.C. 1033 (Eds.), Elwyn Simons: A Search for Origins. Springer, New York, pp. 125–158. 1034 Legendre, P. 2018. lmodel2: Model II Regression. https://cran.r- 1035 project.org/web/packages/lmodel2/index.html. 1036 Lieberman, D.E., 1999. Homology and hominid phylogeny: Problems and potential solutions. 1037 Evol. Anthropol. 7, 142–151. 1038 Lieberman, D.E., McCarthy, R.C., 1999. The ontogeny of cranial base angulation in humans 1039 and chimpanzees and its implications for reconstructing pharyngeal dimensions. J. Hum. 1040 Evol. 36, 487–517. 44 1041 Lieberman, D.E., Ross, C.F., Ravosa, M.J., 2000. The primate cranial base: Ontogeny, 1042 function, and integration. Yearbk. Phys. Anthropol. 43, 117–169. 1043 MacPhee, R.D.E., Cartmill, M., 1986. Basicranial structures and primate systematics. In: 1044 Swisher, D.R., Erwin, J. (Eds.), Comparative Primate Biology: Systematics, Evolution, and 1045 Anatomy. Alan R. Liss, New York, pp. 219–275. 1046 McNulty, K.P., Begun, D.R., Kelley, J., Manthi, F.K., Mbua, E.N., 2015. A systematic revision of 1047 Proconsul with the description of a new genus of early Miocene hominoid. J. Hum. Evol. 1048 84, 42–61. 1049 Meyer, D., Rinaldi, Ir.D., Ramlee, H., Perwitasari-Farajallah, D., Hodges, J.K., Roos, C., 2011. 1050 Mitochondrial phylogeny of leaf monkeys (genus Presbytis, Eschscholtz, 1821) with 1051 implications for taxonomy and conservation. Mol. Phylogenet. Evol. 59, 311–319. 1052 Nengo, I., Tafforeau, P., Gilbert, C.C., Fleagle, J.G., Miller, E.R., Feibel, C., Fox, D.L., Feinberg, 1053 J., Pugh, K.D., Berruyer, C., Mana, S., Engle, Z., Spoor, F., 2017. New infant cranium from 1054 the African Miocene sheds light on ape evolution. Nature 548, 169–174. 1055 Pagel, M., 1999. Inferring the historical patterns of biological evolution. Nature 401, 877– 1056 884. 1057 Perelman, P., Johnson, W.E., Roos, C., Seuánez, H.N., Horvath, J.E., Moreira, M.A.M., Kessing, 1058 B., Pontius, J., Roelke, M., Rumpler, Y., Schneider, M.P.C., Silva, A., O’Brien, S.J., Pecon- 1059 Slattery, J., 2011. A molecular phylogeny of living primates. PLoS Genetics 7, e1001342. 1060 Pozzi, L., Hodgson, J.A., Burrell, A.S., Sterner, K.N., Raaum, R.L., Disotell, T.R., 2014. Primate 1061 phylogenetic relationships and divergence dates inferred from complete mitochondrial 1062 genomes. Mol. Phylogenet. Evol. 75, 165–183. 1063 R Core Team, 2020. R: A language and environment for statistical computing. R Foundation 1064 for Statistical Computing, Vienna. 45 1065 Rae, T.C., 1999. Mosaic evolution in the origin of the Hominoidea. Folia Primatol. 70, 125– 1066 135. 1067 Revell, L.J., 2012. Phytools: an R package for phylogenetic comparative biology (and other 1068 things). Methods Ecol. Evol. 3, 217–223. 1069 Roos, C., Kothe, M., Alba, D.M., Delson, E., Zinner, D., 2019. The radiation of macaques out 1070 of Africa: Evidence from mitogenome divergence times and the fossil record. J. Hum. Evol. 1071 133, 114–132. 1072 Rosenberger, A.L., Szalay, F., 1980. On the tarsiiform origins of the Anthropoidea. In: 1073 Ciochon, R.L., Chiarelli, A.B. (Eds.), Evolutionary Biology of the New World Monkeys and 1074 Contnental Drift. Plenum Press, New York, pp. 139–157. 1075 Ross, C., 1994. Evidence for anthropoid and tarsier relationships. In: Fleagle, J.G., Kay, R.F. 1076 (Eds.), Anthropoid Origins. Plenum Press, New York, pp. 469–547. 1077 Ross, C., Williams, B., Kay, R.F., 1998. Phylogenetic analysis of anthropoid relationships. J. 1078 Hum. Evol. 35, 221–307. 1079 Rossie, J.B., Hill, A., 2018. A new species of Simiolus from the middle Miocene of the Tugen 1080 Hills, Kenya. J. Hum. Evol. 125, 50–58. 1081 Sankhyan, A.R., Kelley, J., Harrison, T., 2017. A highly derived pliopithecoid from the Late 1082 Miocene of Haritalyangar, India. J. Hum. Evol 105, 1–12. 1083 Schlager, S., 2017. Morpho and Rvcg e shape analysis in R: R-packages for geometric 1084 morphometrics, shape analysis and surface manipulations. In: Zheng, G., Li, S., Székely, G. 1085 (Eds.), Statistical Shape and Deformation Analysis. Methods, Implementation and 1086 Applications. Academic Press, London, pp. 217–256. https://cran.r- 1087 project.org/web/packages/Morpho/index.html. 1088 Schliep, K.P., 2011. phangorn: phylogenetic analysis in R. Bioinformatics 27, 592–593. 46 1089 Seiffert, E.R., 2006. Revised age estimates for the later Paleogene mammal faunas of Egypt 1090 and Oman. Proc. Natl. Acad. Sci. USA 103, 5000–5005. 1091 Seiffert, E.R., 2012. Early primate evolution in Afro-Arabia. Evol. Anthropol. 21, 239–253. 1092 Seiffert, E.R., Simons, E.L., Fleagle, J.G., Godinot, M., 2010. Paleogene anthropoids. In: 1093 Werdelin, L., Sanders, W.J. (Eds.), Cenozoic Mammals of Africa. University of California 1094 Press, Berkeley, pp. 369–391. 1095 Shoshani, J., Groves, C.P., Simons, E.L., Gunnell, G.F., 1996. Primate phylogeny: 1096 Morphological vs molecular results. Mol. Phylogenet. Evol. 5, 102–154. 1097 Sidlauskas, B., 2008. Continuous and arrested morphological diversification in sister clades of 1098 characiform fishes: aphylomorphospace approach. Evolution 62, 3135–3156. 1099 Smith, R.J., Jungers, W.L., 1997. Body mass in comparative primatology. J. Hum. Evol. 32, 1100 523–559. 1101 Springer, M.S., Meredith, R.W., Gatesy, J., Emerling, C.A., Park, J., Rabosky, D.L., Stadler, T., 1102 Steiner, C., Ryder, O.A., Janečka, J.E., Fisher, C.A., Murphy, W.J., 2012. Macroevolutionary 1103 dynamics and historical biogeography of primate diversification inferred from a species 1104 supermatrix. PLoS One 7, e49521. 1105 Stevens, N.J., Seiffert, E.R., O’Connor, P.M., Roberts, E.M., Schmitz, M.D., Krause, C., 1106 Gorscak, E., Ngasala, S., Hieronymus, T.L., Temu, J., 2013. Palaeontological evidence for 1107 an Oligocene divergence between Old World monkeys and apes. Nature 497, 611–614. 1108 Turner, T.R., Schmitt, C.A., Cramer, J.D., Lorenz, J., Grobler, J.P., Jolly, C.J., Freimer, N.B., 1109 2018. Morphological variation in the genus Chlorocebus: Ecogeographic and 1110 anthropogenically mediated variation in body mass, postcranial morphology, and growth. 1111 Am. J. Phys. Anthropol. 166 (S3), 682–707. 47 1112 Urciuoli, A., Zanolli, C., Beaudet, A., Dumoncel, J., Santos, F., Moyà-Solà, S., Alba, D.M., 2020. 1113 The evolution of the vestibular apparatus in apes and humans. eLife 9, e51261. 1114 Urciuoli, A., Zanolli, C., Beaudet, A., Pina, M., Almécija, S., Moyà-Solà, S., Alba, D.M., 2021a. A 1115 comparative analysis of the vestibular apparatus in Epipliopithecus vindobonensis: 1116 Phylogenetic implications. J. Hum. Evol. 151, 102930. 1117 Urciuoli, A., Zanolli, C., Almécija, S., Beaudet, A., Dumoncel, J., Morimoto, N., Nakatsukasa, 1118 M., Moyà-Solà, S., Begun, D.R., Alba, D.M., 2021b. Reassessment of the phylogenetic 1119 relationships of the late Miocene apes Hispanopithecus and Rudapithecus based on 1120 vestibular morphology. Proc. Natl. Acad. Sci. USA 118, e2015215118. 1121 Walker, A., Teaford, M.F., Martin, L., Andrews, P., 1993. A new species of Proconsul from the 1122 early Miocene of Rusinga/Mfangano Islands, Kenya. J. Hum. Evol. 25, 43–53. 1123 Zalmout, I.S., Sanders, W.J., MacLatchy, L., Gunnell, G., Al-Mufarreh, Y.A., Ali, M.A., Nasser, 1124 A.-A.H., Al-Masary, A.M., Al-Sobhi, S.A., Nadhra, A.O., Matari, A.H., Wilson, J.A., Gingerich, 1125 P.D., 2010. New Oligocene primate from Saudi Arabia and the divergence of apes and Old 1126 World monkeys. Nature 466, 360–365. 1127 Zapfe, H., 1961. Die Primatenfunde aus der miozänen Spaltenfüllung von Neudorf an der 1128 March (Děvínská Nová Ves), Tschechoslowakei. Schweizer. Palaeontol. Abh. 78, 1–293. 1129 1130 Figure captions 1131 1132 Figure 1. Schematic illustration of the protocol used in this study for cutting the extremities 1133 of the carotid canal as exemplified in a Macaca fascicularis individual. a) To cut the posterior 1134 end, a best-fit plane defined for the landmarks placed at the carotid foramen is computed. b) 1135 To cut the anterior end, the 3D canal surface is first imported as a cloud of points in R. c) The 48 1136 projection of IRG on the cloud of points (IRG’) is computed. d) 360 points are created around 1137 the axis IRG–IRG’. e, f) 360 possible cutting planes passing through IRG are created (each 1138 being based on IRG, IRG’ and one of the points rotating around the axis), here depicted in 1139 anterior (e) and superior (f) views. The cross-sectional area obtained by cutting the canal 1140 using each possible plane is calculated as follows: g–i) the points of the cloud belonging to 1141 the plane are recorded (g), the most external points (in red) are kept to create a convex hull 1142 that approximates the shape of the canal contour (h), and a set of triangles is drawn, each of 1143 them having the same point of the contour as the most acute vertex, and the two 1144 subsequent points on the contour as the other vertices (i); the total area is the obtained by 1145 summing those of each individual triangle. j, k) The plane with the smallest cross-sectional 1146 area (that is, orthogonal to the 3D canal surface and therefore to the canal streamline) (j) is 1147 used as a reference for cutting the canal at its anterior end (k). l) The centroids of the 1148 (posterior) carotid foramen landmark set and of the anterior cross-section (i.e., obtained by 1149 cutting with the orthogonal plane), representing the posterior and anterior endpoints of the 1150 carotid canal, respectively, are calculated. IRG = Intersection ridge–groove; IRG’ = orthogonal 1151 projection on the 3D carotid canal surface of IRG. 1152 1153 Figure 2. Petrosal landmarks used for aligning the specimens as exemplified in a Macaca 1154 fascicularis individual in different views: a) right lateral; b) inferior; c) sagittal cross section 1155 (right-mirrored for visualization purpose); d) superior transverse cross section. Landmarks: 1 1156 = porion; 2 = inferior external auditory meatus; 3 = posterior external auditory meatus; 4 = 1157 posterior internal auditory meatus; 5 = anterior internal auditory meatus; 6 = medial 1158 anterior pyramidal ridge; 7 = intersection ridge-groove; 8 = greater superficial petrosal nerve 1159 hiatus; 9 = stylomastoid foramen. 49 1160 1161 Figure 3. Graphic representation of the 2D angles defining the carotid canal orientation 1162 relative to the cranium. a) Virtual external reconstructions of the cranium, in superior (left) 1163 and lateral (right) views. b) Virtual cross sections of the cranium to visualize the petrosal 1164 bone, in transverse (left) and sagittal (right) views. c) Representations of the carotid canal 1165 medial orientation (2DXYA; left) and superior orientation (2DYZA; right) within the cranium, 1166 rendered in semitransparency; red lines represent the 2D canal vector and its direction, 1167 orange squares represent the plane along which the angle can vary, and dotted black lines 1168 represent the standard reference vector. For 2DXYA, the plane of angle variation is a 1169 transverse plane, whereas for 2DYZA the plane of angle variation is a (para)sagittal plane. 1170 For both angles, the standard reference vector is a 2D vector parallel to the anteroposterior 1171 cranial axis (from posterior to anterior). Both angles are calculated between the 2D canal 1172 vector and the 2D standard reference vector. Because the anthropoid mean configuration is 1173 extremely similar to the canal course configuration of Presbytis (i.e., before alignment), the 1174 cranium of this taxon is used as a reference for orienting the canal following the cranial axes 1175 of reference. 1176 1177 Figure 4. Virtual reconstructions of carotid canal comparing Pliobates with extant 1178 anthropoids and some Miocene catarrhines: a) Pliobates (IPS58443.1); b) Epipliopithecus 1179 (NHMW 1970/1397/0003); c) Ekembo (KNM-RU 2036al); d) Victoriapithecus (KNM-MB 1180 29100); e) Alouatta; f) Aotus; g) Cebus; h) Mico; i) Plecturocebus; j) Cercopithecus; k) 1181 Colobus; l) Mandrillus; m) Presbytis; n) Hylobates; o) Hoolock; p) Nomascus; q) 1182 Symphalangus; r) Gorilla; s) Pan; t) Pongo. The canals are shown in medial view with their 1183 posterior ends horizontally aligned. Scale bars = 5 mm. 50 1184 1185 Figure 5. Box-and-whisker plots of the variables describing the orientation, proportions and 1186 course of the carotid canal in the fossil specimens and main extant clades analyzed in this 1187 study: a) 2D YZ angle; b) 2D XY angle; c) index L/V(1/3); d) bgPC1; e) bgPC2. Horizontal lines 1188 denote medians, boxes depict the interquartile range, whiskers the minimum-maximum 1189 values excluding outliers, dots denote outliers (beyond 1.5 times above or below the 1190 interquartile range). Abbreviations: L = length (in mm); V = volume (in mm3); bgPC = 1191 between-group principal component; 2DYZA = 2D angle in the YZ plane (canal superior 1192 orientation; in degrees); 2DXYA = 2D angle in the XY plane (canal medial orientation; in 1193 degrees). 1194 1195 Figure 6. Results of the between-group principal component analysis (bgPCA) based on 1196 carotid canal course variation among main extant anthropoid clades. a) Bivariate plot of 1197 bgPC2 vs. bgPC1 (the variance explained by each bgPC is denoted in parentheses). b) 1198 Minimum (blue) and maximum (red) landmark conformations compared with the 1199 anthropoid mean configuration (black) for bgPC1 and bgPC2, in superior (top), lateral 1200 (middle), and anterior (bottom) views (anatomical axes of reference are specified next to 1201 each view; see Fig. 7 for a representation of the carotid canal within the cranium). Convex 1202 hulls correspond to: platyrrhines (blue), cercopithecids (red), hylobatids (green), and 1203 hominids (gray). Abbreviation: bgPC = between-group principal component. 1204 1205 Figure 7. Representations of the carotid canal relative to the cranium in superior (left), 1206 lateral (middle), and anterior (right) views. a) Virtual external reconstruction of the cranium. 1207 b) Virtual cross sections of the cranium to visualize the petrosal bone (from left to right, 51 1208 transverse, sagittal, and coronal sections are depicted). c) Virtual reconstructions of the 1209 carotid canal within the cranium rendered in semitransparency. d, e) Minimum (blue) and 1210 maximum (red) landmark conformations compared with the anthropoid mean configuration 1211 (black) for bgPC1 (d) and bgPC2 (e) within the cranium rendered in semitransparency. 1212 Because the anthropoid mean configuration is extremely similar to the canal course 1213 configuration of Presbytis (i.e., before alignment), the cranium of this taxon is used as a 1214 reference for orienting the canal following the cranial axes of reference. 1215 1216 Figure 8. Neighbor-joining cluster (unrooted) based on the Euclidean distances calculated 1217 from raw coordinates of the canal course configurations in Pliobates, extant anthropoids and 1218 other Miocene catarrhines. Cophenetic = 0.94. 1219 1220 Figure 9. Phylomorphospace of carotid canal course variation in extant anthropoids and 1221 extinct Miocene catarrhines including Pliobates. Epipliopithecus and Pliobates are considered 1222 sister taxa within a stem catarrhine pliopithecoid clade and Ekembo is considered a stem 1223 hominoid while Victoriapithecus is considered a stem cercopithecoid (see SOM Fig. S2b for a 1224 representation of this phylogenetic hypothesis). Ancestral nodes: 1 = crown platyrrhines; 2 = 1225 crown catarrhines; 3 = crown cercopithecoids; 4 = crown hominoids; 5 = crown hominids; 6 = 1226 crown hylobatids. See SOM Figure S8 for a phylomorphospace derived from the alternative 1227 phylogenetic hypothesis (SOM Fig. S2b). 1228 1229 Figure 10. Configurations of the carotid canal course in Pliobates compared with other 1230 Miocene catarrhines and the reconstructed carotid canal course for the last common 1231 ancestors (LCAs) of various anthropoid clades: a) Pliobates (IPS58443.1); b) Epipliopithecus 52 1232 (NHMW 1970/1397/0003); c) Ekembo (KNM-RU 2036al); d) Victoriapithecus (KNM-MB 1233 29100); e) crown platyrrhines; f) crown catarrhines; g) crown cercopithecoids; h) crown 1234 hominoids; i) crown hylobatids; j) crown hominids. The canal courses of the LCAs have been 1235 inferred based on the phylomorphospace displayed in Figure 9 (see SOM Fig. S9 for the 1236 inferred LCA configurations derived from the alternative phylogenetic hypothesis). For each 1237 specimen/LCA, the configurations are displayed in superior (left), lateral (middle), and 1238 anterior (right) views. The mean configuration computed for the whole anthropoid sample is 1239 represented by black dots, while configurations for Miocene catarrhines and LCAs are 1240 represented in beige. The anthropoid mean configuration was used as a reference for 1241 visualizing all the configurations at the same relative scale. 1242 53 Table 1 Extant comparative sample used in this study based on µCT scans. Sample size and sex composition, institutional repository for the specimens, digital repository source, and voxel size range are indicated for each species. See SOM Table S1 for further details on the specimens included. Taxon Species n M F ? Repository Source Voxel size (µm) Platyrrhini Alouatta palliata 5 2 3 0 DUEA MS 52–67 Platyrrhini Aotus griseimembra 5 2 3 0 USNM MS 34–38 Platyrrhini Ateles geoffroyi 3 0 3 0 AMNH, MCZ MS 62–63 Platyrrhini Cebus imitator 3 0 3 0 MCZ MS 80 Platyrrhini Chiropotes chiropotes 3 1 2 0 USNM MS 47–49 Platyrrhini Lagothrix lagotricha 3 1 2 0 USNM MS 58–60 Platyrrhini Mico argentatus 3 2 1 0 MCZ MS 40 Platyrrhini Pithecia monachus 1 0 0 1 MCZ MS 50 Platyrrhini Pithecia pithecia 2 1 1 0 MCZ MS 50 Platyrrhini Plecturocebus cupreus 5 3 2 0 AMNH, MCZ MS 47–51 Platyrrhini Saguinus fuscicollis 1 0 1 0 AMNH MS 27 Platyrrhini Saguinus midas 2 2 0 0 MCZ MS 40 Platyrrhini Saimiri sciureus 3 1 1 1 MCZ MS 47 Cercopithecidae Cercocebus torquatus 2 2 0 0 AMNH MS 90–120 Cercopithecidae Cercocebus galeritus 1 1 0 0 AMNH, MCZ MS 70 Cercopithecidae Cercopithecus mitis 3 0 3 0 MCZ MS 80 Cercopithecidae Chlorocebus pygerythrus 3 1 2 0 SIU MS 46–54 Cercopithecidae Colobus guereza 3 1 2 0 AMNH MS 66–73 Cercopithecidae Erythrocebus patas 3 2 1 0 MCZ MS 80 Cercopithecidae Lophocebus albigena 3 1 2 0 MCZ MS 90 Cercopithecidae Macaca fascicularis 5 2 3 0 MCZ MS 60–91 Cercopithecidae Mandrillus sphinx 3 2 1 0 AMNH, MCZ MS 75–126 Cercopithecidae Miopithecus talapoin 3 3 0 0 MCZ MS 50 Cercopithecidae Nasalis larvatus 3 0 3 0 MCZ MS 71 Cercopithecidae Papio anubis 3 3 0 0 MCZ MS 108–118 Cercopithecidae Piliocolobus badius 3 2 1 0 MCZ MS 80–90 Cercopithecidae Presbytis rubicunda 3 1 2 0 MCZ MS 80 Cercopithecidae Procolobus verus 2 1 1 0 AMNH MS 76–80 Cercopithecidae Semnopithecus entellus 3 2 1 0 AMNH MS 52–108 Cercopithecidae Simias concolor 2 1 1 0 AMNH MS 56 Cercopithecidae Theropithecus gelada 3 2 1 0 AMNH MS 84–120 Cercopithecidae Trachypithecus cristatus 3 0 3 0 MCZ MS 50 Hominidae Gorilla gorilla 6 1 5 0 AMNH, MCZ MS 106–131 Hominidae Pan paniscus 5 1 3 1 AMNH, ICP, MCZ MS, own data 66–116 Hominidae Pan troglodytes 6 3 3 0 AMNH, ICP, MCZ MS, own data 77–111 Hominidae Pongo sp. 6 0 6 0 ICP, MCZ MS, own data 77–125 Hylobatidae Hoolock hoolock 3 1 2 0 AMNH MS 52–84 Hylobatidae Hylobates agilis 1 0 0 1 ICP MS, own data 52 Hylobatidae Hylobates lar 4 0 4 0 MCZ MS 67 Hylobatidae Nomascus leucogenys 1 1 0 0 AMNH MS 105 Hylobatidae Symphalangus syndactylus 2 0 2 0 AMNH MS 62–79 Abbreviations: M = male; F = female; ? = unknown sex; AMNH = American Museum of Natural History, New York, USA; DUEA = Duke University, Evolutionary Anthropology, Durham, USA; ICP = acronym of the collections of the Institut Català de Paleontologia Miquel Crusafont, Sabadell, Spain; MCZ = Museum of Comparative Zoology, Harvard University, Cambridge, USA; MS = MorphoSource (https://www.morphosource.org); SIU = Southern Illinois University, Carbondale, USA; USNM = Smithsonian National Museum of Natural History, Washington, D.C., USA. Table 2 Definition of the petrosal/tympanic landmarks used in this paper for aligning the specimens. No. Landmark Definition 1 Porion Most superior point on the margin of the external auditory meatus 2 Inferior external auditory meatus Most inferior point on the margin of the external auditory meatus 3 Posterior external auditory meatus Most posterior point on the margin of the external auditory meatus 4 Posterior internal auditory meatus Most posterior point on the margin of the internal auditory meatus 5 Anterior internal auditory meatus Most anterior point on the margin of the internal auditory meatus 6 Medial anterior pyramidal ridge Most medial point where anterior pyramidal ridge starts running laterally above the fossa for the semilunar ganglion 7 Intersection ridge-groove Point of intersection between the laterally running anterior pyramidal ridge and the groove of the greater petrosal nerve 8 Greater petrosal nerve hiatus Most superior, anterior point on the margin of the hiatus for the greater petrosal nerve 9 Stylomastoid foramen Center of the stylomastoid foramen Table 3 Descriptive statistics of the variables describing the orientation, proportions and course of the carotid canal in the main extant anthropoid clades. See SOM Table S2 for the individual values of extant anthropoid specimens. 2DYZA 2DXYA L/V(1/3) bgPC1 bgPC2 Clade n Mean SD Range Mean SD Range Mean SD Range Mean SD Range Mean SD Range Platyrrhini 39 64.06 7.33 50.66–79.36 31.92 7.57 8.25–44.26 3.68 0.39 2.90–4.42 -0.36 0.11 0.50–0.07 -0.01 0.11 -0.26–0.17 Cercopithecidae 54 69.71 7.42 57.2–85 16.69 10 -3.54–33.87 2.86 0.42 1.40–3.87 0.17 0.08 -0.11–0.34 -0.04 0.09 -0.29–0.14 Hominidae 23 54.90 7.83 37.27–69.11 13.74 7.38 -0.82–27.78 3.23 0.31 2.57–3.82 0.12 0.08 -0.10–0.20 0.12 0.05 -0.01–0.20 Hylobatidae 11 54.51 12.39 38.27–78.27 10.27 7.44 1.00–23.27 2.68 0.47 1.97–3.37 0.18 0.08 0.06–0.32 -0.04 0.10 -0.25–0.12 Abbreviations: 2DYZA = 2D angle in the YZ plane (canal superior orientation; in degrees); 2DXYA = 2D angle in the XY plane (canal medial orientation; in degrees); L = length (in mm); V = volume (in mm3); bgPC = between-group principal component. Table 4 Variables describing the orientation, proportions and course of the carotid canal in the fossil specimens. Species Catalog No. 2DYZA 2DXYA L/V(1/3) bgPC1 bgPC2 Ekembo heseloni KNM-RU 2036al 53.58 34.73 3.12 -0.02 -0.03 Epipliopithecus vindobonensis NHMW 1970/1397/0003 53.59 27.62 3.05 -0.08 -0.06 Pliobates cataloniae IPS58443.1 35.58 16.86 3.58 -0.03 0.05 Victoriapithecus macinessi KNM-MB 29100 66.73 10.36 3.25 -0.02 -0.04 Abbreviations: 2DYZA = 2D angle in the YZ plane (canal superior orientation; in degrees); 2DXYA = 2D angle in the XY plane (canal medial orientation; in degrees); L = length (in mm); V = volume (in mm3); bgPC = between-group principal component. Table 5 Z-score analysis of the three variables describing the orientation, proportions and course of the carotid canal in the fossil specimens and compared to the variation expressed by the main extant anthropoid clades.a Variable and species Catalog No. Platyrrhini Cercopithecidae Hominidae Hylobatidae 2DYZA Ekembo heseloni KNM-RU 2036al -1.43 -2.17 -0.17 -0.08 Epipliopithecus vindobonensis NHMW 1970/1397/0003 -1.43 -2.17 -0.17 -0.07 Pliobates cataloniae IPS58443.1 -3.88 -4.60 -2.46 -1.52 Victoriapithecus macinnesi KNM-MB 29100 0.36 -0.40 1.51 0.98 2DXYA Ekembo heseloni KNM-RU 2036al 0.37 1.80 2.84 3.29 Epipliopithecus vindobonensis NHMW 1970/1397/0003 -0.56 1.09 1.88 2.33 Pliobates cataloniae IPS58443.1 -1.99 0.02 0.42 0.89 Victoriapithecus macinnesi KNM-MB 29100 -2.85 -0.63 -0.46 -0.02 L/V(1/3) Ekembo heseloni KNM-RU 2036al -1.46 0.61 -0.36 0.93 Epipliopithecus vindobonensis NHMW 1970/1397/0003 -1.64 0.44 -0.58 0.78 Pliobates cataloniae IPS58443.1 -0.27 1.69 1.09 1.89 Victoriapithecus macinnesi KNM-MB 29100 -1.13 0.91 0.04 1.20 bgPC1 Ekembo heseloni KNM-RU 2036al 3.23 -2.30 -1.78 -2.39 Epipliopithecus vindobonensis NHMW 1970/1397/0003 2.70 -2.97 -2.50 -3.08 Pliobates cataloniae IPS58443.1 3.14 -2.41 -1.90 -2.51 Victoriapithecus macinnesi KNM-MB 29100 3.19 -2.35 -1.83 -2.45 bgPC2 Ekembo heseloni KNM-RU 2036al -0.25 0.02 -3.16 0.04 Epipliopithecus vindobonensis NHMW 1970/1397/0003 -0.51 -0.31 -3.78 -0.27 Pliobates cataloniae IPS58443.1 0.48 0.93 -1.47 0.87 Victoriapithecus macinnesi KNM-MB 29100 -0.25 0.02 -3.17 0.03 Abbreviations: 2DYZA = 2D angle in the YZ plane (canal superior orientation; in degrees); 2DXYA = 2D angle in the XY plane (canal medial orientation; in degrees); L = length (in mm); V = volume (in mm3); bgPC = between-group principal component. a Significant z-scores at p < 0.05 are bolded. Table 6 Phylogenetic generalized least-squares (PGLS) allometric regressions between carotid canal size variables (L, V1/3 or CS) or between carotid canal size and BM, together with PGLS bivariate regressions between carotid canal shape (as represented by the canal robusticity index or bgPCs) vs. log-transformed carotid canal size or BM. Regression pa R2 Slope Slope 95% CI ln L vs. ln V1/3 <0.001 0.81 0.86 0.73, 0.99 ln L vs. ln V1/3 b <0.001 0.81 0.95 0.81, 1.10 L/V1/3 vs. ln CS >0.05 -0.02 0.11 –0.34, 0.57 ln L vs. ln CS <0.001 0.99 1.04 1.02, 1.06 ln V1/3 vs. ln CS <0.001 0.81 0.99 0.84, 1.13 ln L vs. ln BM <0.001 0.68 0.25 0.20, 0.31 ln V1/3 vs. ln BM <0.001 0.90 0.30 0.27, 0.34 ln CS vs. ln BM <0.001 0.69 0.25 0.19, 0.30 L/V1/3 vs. ln BM <0.01 0.14 -0.17 –0.29, –0.05 bgPC1 vs. ln CS >0.05 0.01 0.12 –0.07, 0.31 bgPC1 vs. ln V1/3 <0.01 0.19 0.26 0.10, 0.41 bgPC1 vs. ln BM <0.001 0.25 0.10 0.05, 0.15 bgPC2 vs. ln CS <0.001 0.58 0.21 0.16, 0.27 bgPC2 vs. ln V1/3 <0.001 0.46 0.17 0.11, 0.23 bgPC2 vs. ln BM <0.001 0.35 0.05 0.04, 0.08 Abbreviations: L = carotid canal length (mm); V1/3 = carotid canal volume cube root (mm); L/V1/3 = shape index of canal robusticity; CS = carotid canal centroid size (mm); BM = body mass (g); bgPC = between-group principal component (representing carotid canal course). a Regressions are significant when p < 0.05. b For ln L vs. ln V1/3, a major axis regression was also computed to test whether the two variables are isometric when deviation from the best-fit line is minimized simultaneously for both variables (instead of only the dependent variable, in this case ln L). Our results show that, unlike for the PGLS regression (which displays slight negative allometry), isometry cannot be discounted for the major axis regression. Table 7 Number (n) and percentages (%) of correctly and incorrectly classified individuals obtained by the between-group principal components analysis after cross-validation. In total, 98 out of 127 (77.2%) individuals were correctly classified (in bold). Platyrrhini Cercopithecidae Hominidae Hylobatidae Platyrrhini 36 (92.3%) 0 (0%) 0 (0%) 3 (7.7%) Cercopithecidae 1 (1.9%) 34 (63.0%) 9 (16.7%) 10 (18.5%) Hominidae 0 (0%) 2 (8.7%) 21 (91.3%) 0 (0.0%) Hylobatidae 0 (0%) 3 (27.3%) 1 (9.1%) 7 (63.6%) Table 8 Posterior probabilities of group membership based on the between-group principal component scores of fossil specimens.a Species Catalog No. Platyrrhini Cercopithecidae Hominidae Hylobatidae Ekembo heseloni KNM-RU 2036al <0.001 0.098 0.059 0.087 Epipliopithecus vindobonensis NHMW 1970/1397/0003 <0.001 0.018 0.009 0.016 Pliobates cataloniae IPS58443.1 <0.001 0.064 0.169 0.056 Victoriapithecus macinessi KNM-MB 29100 <0.001 0.089 0.054 0.079 a Each value corresponds to the probability of having the given score if the specimen was a member of the group considered, not the likelihood of group membership in each of the a priori defined groups given a particular score (the greater the number, the higher the probability). For each specimen the highest probability is bolded.