1. Evolutionary Biology
Download icon

The evolution of the vestibular apparatus in apes and humans

  1. Alessandro Urciuoli  Is a corresponding author
  2. Clément Zanolli
  3. Amélie Beaudet
  4. Jean Dumoncel
  5. Frédéric Santos
  6. Salvador Moyà-Solà
  7. David M Alba  Is a corresponding author
  1. Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain
  2. Laboratoire PACEA, UMR 5199 CNRS, Université de Bordeaux, France
  3. University of the Witwatersrand, South Africa
  4. University of Pretoria, South Africa
  5. Laboratoire AMIS, UMR 5288 CNRS, Université de Toulouse, France
  6. Institució Catalana de Recerca i Estudis Avançats (ICREA), Spain
Research Article
  • Cited 2
  • Views 2,042
  • Annotations
Cite this article as: eLife 2020;9:e51261 doi: 10.7554/eLife.51261

Abstract

Phylogenetic relationships among extinct hominoids (apes and humans) are controversial due to pervasive homoplasy and the incompleteness of the fossil record. The bony labyrinth might contribute to this debate, as it displays strong phylogenetic signal among other mammals. However, the potential of the vestibular apparatus for phylogenetic reconstruction among fossil apes remains understudied. Here we test and quantify the phylogenetic signal embedded in the vestibular morphology of extant anthropoids (monkeys, apes and humans) and two extinct apes (Oreopithecus and Australopithecus) as captured by a deformation-based 3D geometric morphometric analysis. We also reconstruct the ancestral morphology of various hominoid clades based on phylogenetically-informed maximum likelihood methods. Besides revealing strong phylogenetic signal in the vestibule and enabling the proposal of potential synapomorphies for various hominoid clades, our results confirm the relevance of vestibular morphology for addressing the controversial phylogenetic relationships of fossil apes.

eLife digest

Humans, gorillas, chimpanzees, orangutans and gibbons all belong to a group known as the hominoids. This ‘superfamily’ also includes the immediate ancestors and close relatives of these species, however in many instances the evolutionary relationships between these extinct ape species remain controversial.

While DNA can help evolutionary biologists to work out how living species are related to one another, fossils are typically the principle source of information for extinct species. Inferring evolutionary relationships from fossils must be done with caution, but the bony cavity that houses the inner ear – which is involved in balance and hearing and fairly common in the fossil record – has proven useful for tracing the evolution of certain groups of mammals. However, no one had previously looked to see if this structure could give insights into the evolutionary relatedness among living and extinct hominoids.

Urciuoli et al. have now used a 3D imaging technique to capture the complex shapes of the inner ear cavities of 27 species of monkeys and apes, including humans and two extinct apes (Oreopithecus and Australopithecus). The results confirmed that the shape of these structures most closely reflected the evolutionary relationships between the species and not, for example, how the animals moved.

Urciuoli et al. went on to identify features of these bony chambers that were shared within several hominoid groups, and to estimate what the inner ears of the ancestors of these groups might have looked like. The results for Australopithecus, for example, were consistent with it being most closely related to modern humans than other apes, while those for the enigmatic Oreopithecus supported the view that it was a much older species of ape that converged in some respects with other apes still alive today.

The findings highlight the potential of the inner ear for reconstructing the early branches of our family tree. They also offer the prospect of refining the controversial evolutionary relationships within the impressive diversity of extinct ape species.

Introduction

Catarrhine primates (Old World anthropoids) include two extant subclades: cercopithecoids (Old World monkeys) and hominoids (apes and humans). Based on molecular (e.g., Springer et al., 2012) and paleontological (e.g., Harrison, 2013; Stevens et al., 2013) data, both groups diverged during the late Oligocene (≥25 Ma), but experienced very different evolutionary histories. Hominoids first radiated in the early Miocene of Africa (Harrison, 2010; Begun, 2013; Begun, 2015) and subsequently diversified into Eurasia during the middle and late Miocene (Alba, 2012; Begun, 2015). Their diversity and geographic distribution (humans excluded) was much greater during the Miocene than at present, being currently restricted to a few genera in southeastern Asia and Africa. In contrast, extant cercopithecoid lineages started to diversify later and experienced a major radiation during the late Miocene (Jablonski and Frost, 2010), being currently much more diverse and widely distributed than hominoids in both Africa and Asia.

The decimated current diversity of hominoids, coupled with the fragmentary nature of their fossil record, abundant homoplasy (e.g., Larson, 1998), and the lack of known fossil hylobatids prior to the latest Miocene (Harrison, 2016) make it difficult to confidently infer the phylogenetic relationships of extinct hominoids and thus reliably infer the morphotype of the last common ancestor (LCA) of various hominoid subclades. This is required not only to adequately understand the evolutionary history of the group as a whole, but also to reconstruct the LCA of chimpanzees and humans, from which the earliest hominins evolved during the late Miocene. The contribution of Miocene apes to debates about hominoid evolution is thus diminished by the numerous controversies about the phylogenetic position of the former.

For example, putative stem hominoids from the early Miocene of Africa (Begun, 2013; Begun, 2015), such as the proconsulid Ekembo, lack most of the synapomorphies of crown hominoids, such that some authors still contend that they might represent stem catarrhines instead (Harrison, 2010; Harrison, 2013). Even more uncertain is the position of dendropithecids (e.g., Micropithecus) and other small-bodied catarrhines from the early to middle Miocene of Africa, which are generally interpreted as stem catarrhines (Harrison, 2010; Harrison, 2013) but might include stem hominoids as well (Alba et al., 2015; Begun, 2015). The same controversy applies to the European middle to late Miocene Pliobates, recovered as a stem hominoid more derived than proconsulids (Alba et al., 2015) or alternatively as a member of the stem catarrhine pliopithecoid radiation (Nengo et al., 2017). Regarding undoubted hominoids, the distinction between stem and crown taxa is by no means less controversial, in part due to the virtual lack of fossil hylobatids since their origin in the early Miocene until the latest Miocene (Harrison, 2016). This is best exemplified by the late Miocene Oreopithecus from Italy, variously considered a hominid (Moyà Solà and Köhler, 1997; Harrison and Rook, 1997) or a stem hominoid (Nengo et al., 2017). Finally, the phylogenetic placement of Miocene apes from Eurasia is also controversial. For example, most Asian forms have been classically considered pongines (Begun, 2013), but most recently the late Miocene Lufengpithecus from China has been reinterpreted as a stem hominid (Kelley and Gao, 2012; Begun, 2015). Even more controversial is the position of the middle to late Miocene European dryopithecines, interpreted as either stem hominids (Alba, 2012; Alba et al., 2015) or hominines (Begun, 2013; Begun, 2015), and further controversies apply when trying to decipher the phylogenetic relationships among various members of this extinct group.

Deciding among phylogenetic hypotheses for Miocene apes has consequences for our current understanding of hominin origins, from calibrating molecular data to estimate their divergence time to the reconstruction of the ancestral locomotor repertoire from which the earliest bipeds arose. For example, the recently described dryopithecine Danuvius from Germany has been used to argue that bipedal and suspensory adaptations characterized the last common ancestor of crown hominids (Böhme et al., 2019). However, without a phylogenetic analysis supporting a more basal branching of Danuvius compared with the older dryopithecine Pierolapithecus (Moyà-Solà et al., 2004), which was orthograde but lacked adaptations to both bipedalism and suspension (Alba et al., 2010; Alba, 2012), the implications for the ancestral condition of the group remain moot.

The recent recovery of enamel proteome sequences from the early Pleistocene (1.9 Ma) fossil pongine Gigantopithecus (Welker et al., 2019) offers some hope that molecular data will become available for Miocene apes sometime in the future. In the meantime, to better resolve the phylogeny of Miocene hominoids, the search for morphological features not very prone to homoplasy is crucial. Anatomical structures that develop early during development and are not remodeled thereafter (such as the enamel-dentine junction of teeth) represent the best candidates (Corruccini, 1987; Skinner et al., 2008). In this regard, the inner ear is also a very promising anatomical area (Spoor and Zonneveld, 1998; Spoor et al., 2003), even if thus far its phylogenetic implications have been mostly explored for fossil hominins only (Quam et al., 2016; Conde-Valverde et al., 2018; Ponce de León et al., 2018; Beaudet, 2019a; Beaudet et al., 2019b), while its application to fossil apes has been mainly devoted to locomotor inferences (David et al., 2010; Malinzak et al., 2012; Rook et al., 2004; Ryan et al., 2012).

Housed in the highly mineralized petrosal bone, which is frequently preserved in the fossil record, the inner ear is composed of a series of endolymph-filled membranes (encased in the corresponding bony labyrinths), namely the cochlea or cochlear duct (involved in hearing) and the vestibular apparatus (devoted to balance and vision stability). The vestibule consists of three (anterior, posterior, and lateral) semicircular canals (SCs) and two macular organs (utricle and saccule). The approximately orthogonal SCs sense angular accelerations and decelerations of the head, while the maculae perceive linear accelerations and thus provide gravity reference (Spoor and Zonneveld, 1998; Rabbitt et al., 2004; Johnson Chacko et al., 2018; Cheung and Ercoline, 2018).

Differences in the relative size and morphology of the SCs have been correlated with locomotor agility (Spoor et al., 1994; Spoor et al., 2007; Walker et al., 2008; Silcox et al., 2009; Ryan et al., 2012; Perier et al., 2016) and positional behavior (Le Maître et al., 2017), albeit not without criticism (Rae et al., 2016; Benson et al., 2017; Coutier et al., 2017). On the other hand, the bony labyrinth morphology has been considered of great importance for phylogenetic reconstruction in various mammals including hominins (Quam et al., 2016; Conde-Valverde et al., 2018; Ponce de León et al., 2018; Beaudet, 2019a; Beaudet et al., 2019b) and nonhuman primates (Lebrun et al., 2010; Lebrun et al., 2012; Grohé et al., 2016; Mennecart et al., 2017; Costeur et al., 2018; Schwab et al., 2019). While previous research in hominoids has yielded encouraging results for phylogenetic reconstruction (Spoor and Zonneveld, 1998; Spoor et al., 2003; Rook et al., 2004; Gunz et al., 2012), according to some authors phylogeny may not be a major component of ape vestibular morphology (Le Maître et al., 2017).

Determining if and to what extent inner ear anatomy reflects phylogeny among extant hominoids is central for assessing the potential of this anatomical area to more confidently resolve the controversial phylogenetic relationships of fossil apes. To provide insight into this question, we use phylogenetically-informed statistical analyses to test the significance and quantify the amount of phylogenetic signal of vestibular shape captured by three-dimensional geometric morphometrics (3DGM) in living hominoids and a broader sample of extant anthropoids. To capture vestibular shape, we mostly rely on a landmark-free, deformation-based 3DGM approach that takes the whole surface into account (Glaunés and Joshi, 2006; Durrleman et al., 2012b; Durrleman et al., 2012a) and enables integrating the spatial trajectory of the semicircular canals with their thickness and volumetric variations, the latter two being more difficult to assess based on mainstream landmark-based 3DGM. Since our results reveal the presence of strong phylogenetic signal in the vestibular morphology, we also employ maximum likelihood methods (Felsenstein, 1988; Schluter et al., 1997) to reconstruct the ancestral vestibular morphology for the LCA of main hominoid subclades (crown hominoids, hominids, and hominines), with the aim to identify phylogenetically informative characters that can be used in formal cladistic analysis. To test the reliability and illustrate the usefulness of our approach from a phylogenetic viewpoint, we also include two extinct hominoid taxa: the early hominin Australopithecus and the aforementioned controversial late Miocene ape Oreopithecus. The well-known phylogenetic placement of Australopithecus as the sister-taxon of humans predicts that its vestibular morphology will be somewhat derived towards the modern human condition. On the other hand, our analysis will enable testing the competing phylogenetic hypotheses for Oreopithecus (stem hominoid vs. hominid), thereby illustrating the potential of our method for clarifying the controversial affinities of extinct apes.

Results

Semicircular canal shape variation among anthropoids

In spite of a similar spatial configuration of the three SCs, their trajectory, stoutness and relative proportions are quite variable among anthropoids (Figure 1). A bgPCA performed among major anthropoid groups (platyrrhines, cercopithecoids, hylobatids, and hominids) enables their accurate distinction (Figure 2, Figure 2—figure supplement 1a, Figure 2—source data 1), as shown by classification results (more than 95% individuals correctly classified after cross-validation; Table 1). In particular, bgPC1 discriminates hylobatids from hominids and all the remaining taxa (Figure 2, Figure 2—figure supplement 1c). A landmark-based analysis applied to the same sample yields very similar results except for hylobatids, due to the reasons explained in the next section. Shape variation in the analyzed sample accounts for a strong phylogenetic signal (Kmult = 1.248, p<0.001), and this also holds for the first three bgPCs separately (see below).

The vestibular apparatus of extant hominoids, other anthropoids, and some fossil hominoids, in lateral view.

(a) Homo sapiens (EMBR 121); (b) Pan paniscus (MCZ 38019); (c) Pan troglodytes (AMNH.M 51204); (d) Gorilla gorilla (AMNH.M 167338); (e) Pongo pygmaeus (IPS 10647); (f) Hylobates lar (MCZ 41424); (g) Hoolock hoolock (AMNH.M 83425); (h) Symphalangus syndactylus (AMNH.M 106583); (i) Ateles geoffroyi (MCZ 29628); (j) Theropithecus gelada (AMNH.M 60568); (k) Erythrocebus patas (MCZ 47017); (l) Trachypithecus cristatus (MCZ 35597); (m) Piliocolobus badius (MCZ 24793); (n) Oreopithecus bambolii (BAC 208); (o) Australopithecus sp. (StW 573); (p) Australopithecus sp. (StW 578). Scale bars equal 5 mm.

Figure 2 with 1 supplement see all
Main patterns of vestibular shape variation among the analyzed anthropoid sample as shown by bivariate plots of principal components from the between-group principal components analysis (bgPCA) using a few major clades (i.e., platyrrhines, cercopithecoids, hylobatids and hominids) as grouping factor.

(a) bgPC2 vs. bgPC1. (b) bgPC3 vs. bgPC1. Variance explained by each bgPC is included within parentheses. Color code: dark green, platyrrhines; orange, hominids; red, hylobatids; brown, papionins; green, cercopithecins; blue, colobines. Colored stars correspond to: green, Oreopithecus; black, Australopithecus sp. (StW 573); red Australopithecus sp. (StW 578). Lateral (top/left), superior (middle), and posterior (bottom/right) views of deformation maximum (red) and minimum (blue) conformations for each bgPC are shown along each axis. Hominids are distinguished from hylobatids and cercopithecoids along bgPC1 (positive vs. mostly negative values, respectively), which is mainly driven by the volumetric proportions of the SCs and by their size relative to that of the vestibular recesses. bgPC2, driven by the size of the anterior and posterior SCs relative to the lateral one, distinguishes platyrrhines (more negative values) from catarrhines. Hylobatids (on positive values) differ from all other anthropoids along bgPC3 due to the reduced and posteriorly tilted posterior SC, as well as by the relative orientation among the canals. Oreopithecus does not match the variability of extant anthropoids as its morphology shows a mosaic of primitive and derived features. The two Australopithecus specimens match instead the range of extant hominids, with StW 573 being most similar to Pan, while StW 578 to Homo.

Table 1
Percentage of correctly classified individuals with cross-validation according to the groups (hominids, hylobatids, cercopithecoids, platyrrhines) used in the between-group principal components analysis based on group-centroid distances.
CercopithecoideaHominidaeHylobatidaePlatyrrhini
Cercopithecoidea96.3%3.8%0.0%0.0%
Hominidae3.3%96.7%0.0%0.0%
Hylobatidae5.9%0.0%94.1%0.0%
Platyrrhini0.0%0.0%0.0%100.0%

When analyzed individually, we recover a similarly strong phylogenetic signal (λ = 1 and K = 1.15; Table 2) for bgPC1 (68.8% of total variance). bgPC1 captures differences in thickness and cross-section of the SCs, and is also driven by the development of the macular organs relative to the canals. Great apes fall on positive values (Figure 2, Figure 2—figure supplement 1a) due to their stout and flattened SCs combined with an extensive vertical compression of the anterior canal, a more anterosuperior insertion of the lateral canal into the vestibule, and a greater volume of the vestibular recesses relative to the canals (Figures 3a–e, 4a–e and 5a–e). Hylobatids, as well as colobine and papionin cercopithecoids, showing slender and elongated canals but maintaining well developed ampullae, largely overlap on negative values, while cercopithecins and platyrrhines display intermediate values due to their slightly inflated SCs (Figures 2, 3f–h, 4f–h and 5f–h). bgPC2 (19.6% of total variance) also bears strong phylogenetic signal (λ = 0.91; K = 1.51; Table 2), with variance accumulating among rather than within clades (as indicated by K > 1). This axis separates platyrrhines—especially Ateles (Figure 1i)—from other anthropoids due to the more reduced lateral canal in the former, which is inversely proportional to anterior canal development and vertical elongation (Figure 2a). In contrast, Gorilla (Figures 1a, 3a, 4a and 5a) occupies the opposite end of the distribution due to its large lateral canal and reduced anterior one, the latter being also vertically compressed, whereas the remaining hominoids show intermediate values along bgPC2. bgPC3 (11.6% of variance), which is driven by both trajectory and relative size of the SCs (Figure 2b), still displays a strong and significant phylogenetic signal (λ = 1 and K = 1.16; Table 2). Hylobatids (Figures 1f–h, 3f–h, 4f–h and 5f–h) display the highest scores for bgPC3, as a result of the right to acute angle formed by the apex of the common crus (CC), the latter being also shorter, a more anterosuperiorly projecting and long anterior canal, an obtuse angle between the planes identified by the anterior and lateral canals, and a more developed lateral canal relative to the posterior one, which is also posteriorly oriented. Most of the taxa fall within moderate positive and moderate negative values, with some cercopithecoids (Cercopithecus, Macaca, Papio and Nasalis, among others) and Cebus occupying the negative end of the distribution. African apes, cercopithecines, and Nasalis occupy an intermediate position, being characterized by a well-developed and anteriorly inclined posterior canal, an obtuse to right angle of the CC apex, and an obtuse angle between the vertical canals, combined with a larger lateral canal. Orangutans fall on moderately positive values and are distinguished from other great apes and humans by an anterosuperiorly projecting anterior canal (less than in hylobatids).

Results of the 3DGM deformation-based analysis superimposed on the vestibular apparatus of hominoids in lateral view.

Cumulative displacement variations are rendered by pseudocolor scale ranging from dark blue (5.1 µm) to dark red (0.42 mm). Black arrows correspond to the vectors identifying the direction and amount of displacement. (a) Homo sapiens (EMBR 121); (b) Pan paniscus (MCZ 38019); (c) Pan troglodytes (AMNH.M 51204); (d) Gorilla gorilla (AMNH.M 167338); (e) Pongo pygmaeus (IPS10647); (f) Hoolock hoolock (AMNH.M 83425); (g) Symphalangus syndactylus (AMNH.M 106583); (h) Hylobates lar (MCZ 41424). Scale bars equal 5 mm. Hominids (a–e) differ from other anthropoids in the stouter semicircular canals and the vertically compressed (more eccentric) anterior semicircular canal, while hylobatids (f–h) possess slender canals more similar to those of cercopithecoids (j–m).

Results of the 3DGM deformation-based analysis superimposed on the vestibular apparatus of hominoids, in superior view.

Cumulative displacement variations are rendered by pseudocolor scale ranging from dark blue (<5.1 µm) to dark red (0.42 mm). Black arrows correspond to the vectors identifying the direction and amount of displacement. (a) Homo sapiens (EMBR 121); (b) Pan paniscus (MCZ 38019); (c) Pan troglodytes (AMNH.M 51204); (d) Gorilla gorilla (AMNH.M 167338); (e) Pongo pygmaeus (IPS10647); (f) Hoolock hoolock (AMNH.M 83425); (g) Symphalangus syndactylus (AMNH.M 106583); (h) Hylobates lar (MCZ 41424). Scale bars equal 5 mm. Hominids (a–e) differ from other anthropoids in the stouter semicircular canals, except for the lateral canal of Ateles. Among great apes, Pongo (e) displays the most inflated canals and a uniquely displaced lateral semicircular canal, Pan (b, c) has the least derived morphology, and Gorilla (e) possesses a laterally protruding lateral canal. Homo (a) combines a slight (variable) reduction of the lateral canal with an enlarged posterior canal. Hylobatids (f–h) show an obtuse angle between the posterior and the anterior SCs.

Results of the 3DGM deformation-based analysis superimposed on the vestibular apparatus of hominoids in posterior view.

Cumulative displacement variations are rendered by pseudocolor scale ranging from dark blue (<5.1 µm) to dark red (0.42 mm). Black arrows correspond to the vectors identifying the direction and amount of displacement. (a) Homo sapiens (EMBR 121); (b) Pan paniscus (MCZ 38019); (c) Pan troglodytes (AMNH.M 51204); (d) Gorilla gorilla (AMNH.M 167338); (e) Pongo pygmaeus (IPS10647); (f) Hoolock hoolock (AMNH.M 83425); (g) Symphalangus syndactylus (AMNH.M 106583); (h) Hylobates lar (MCZ 41424). Scale bars equal 5 mm. Homo (a), together with Gorilla (d), displays a markedly posterolaterally protruding posterior canal. The two species of Pan (b, c) can be distinguished from one another for the orientation of the anterior canal (more medially inclined in P. troglodytes, (c). Hylobatids (f–h) display a small posterior canal relative to the size of the anterior and lateral ones.

Table 2
Phylogenetic signal results for a between-group principal components analysis (bgPCA) applied to vestibular shape deformation fields in the analyzed sample of extant anthropoids.
bgPC1bgPC2bgPC3
Variance68.79%19.60%11.61%
Eigenvalue0.8210.2340.138
Pagel’s λ1.000 (p<0.001)0.921 (p<0.001)1.000 (p<0.001)
Blomberg’s K1.152 (p<0.001)1.514 (p<0.001)1.163 (p<0.001)

When fossil specimens are plotted a posteriori onto the tangent space identified by extant taxa, Oreopithecus (BAC 208; Figures 1n and 6a) falls on moderately positive scores for bgPC1 (where the distributions of hominids, cercopithecins, and platyrrhines overlap), while the two Australopithecus individuals (StW 573 and StW 578; Figures 1o,p and 6b–c) fall within the range of living great apes and humans (Figure 2, Figure 2—figure supplement 1a). This is due to the volumetric proportions of their SCs and the possession of voluminous vestibular recesses (although the latter character is less pronounced in Oreopithecus). In BAC 208, SC volume is greater on the lateral canal, while the two Australopithecus specimens possess stouter vertical canals. When the relative size of the SCs is taken into account, StW 573 shows more evenly developed canals than StW 578 and BAC 208, which both display a smaller lateral canal. Their position along the bgPC2 axis reflects these differences, with StW 573 falling on positive scores (close to the mean value for hominids and cercopithecoids; Figure 2a, Figure 2—figure supplement 1b) and the other two specimens occupying moderately negative values (within the range of extant catarrhines and approaching that of platyrrhines; Figure 2a, Figure 2—figure supplement 1b). This is caused by the comparatively smaller lateral canal and by the large vertical canal in StW 578 and in BAC 208. Due to the acute angles between the planes identified by the anterior and lateral canals and that between the planes of the posterior and anterior canals, Oreopithecus falls at the negative end of the extant anthropoid distribution for the bgPC3 (Figure 2b, Figure 2—figure supplement 1c). On the other hand, the two Australopithecus specimens occupy more intermediate values because of the possession of a right angle between the planes of the aforementioned canals.

Results of the 3DGM deformation-based analysis superimposed on the vestibular apparatus of the fossil specimens included in the present study.

Each vestibule is displayed in lateral (left), superior (middle), and posterior (right) views. Cumulative displacement variations are rendered by pseudocolor scale ranging from dark blue (<5.1 µm) to dark red (0.42 mm). Black arrows correspond to the vectors identifying the direction and amount of displacement. (a) Oreopithecus bambolii (BAC 208); (b) Australopithecus sp. (StW 573); (c) Australopithecus sp. (StW 578). Scale bars equal 5 mm. Oreopithecus (a) displays large vertical canals and a small, flat lateral one that intersect the plane of the posterior SC. The angle between the anterior and posterior SCs is one of the narrowest among the studied sample. StW 573 (b) and StW 578 (c) differ by means of a larger lateral canal in the previous and by the more rounded and developed anterior SC in the latter.

When the bgPCs are considered at the same time (Figure 2), the australopith specimens fall well within the great ape and human range. This is further supported by their posterior probabilities of group membership based on the proximity of fossil specimen scores to groups centroids, with StW 573 and 578 being classified as hominids with p=0.678 and p=0.190, respectively (Table 3). StW 573 falls close to Pan and Homo, whereas StW 578 occupies an intermediate position between humans and orangutans due to its stouter volumetric proportions. When the posterior probabilities are computed using the centroids of the hominoid genera, StW 573 is classified as Pan as first option (p=0.368) and as Homo as second (p=0.264), while StW 578 is more clearly classified as Homo with p=0.727 (Table 4). These results suggest that both australopith specimens show vestibular similarities with extant humans, but that StW 573 display a more plesiomorphic (chimpanzee-like) morphology. In turn, Oreopithecus shows a mosaic of vestibular features (Pan-like volumetric proportions, small lateral canal, and acute angles between the anterior and both the posterior and the lateral canals) that does not match the condition of any extant taxon (Figure 2). The posterior probabilities indicate closest similarities with cercopithecoids, followed by hominids and platyrrhines, although in all instances it falls outside the variability of the extant members of these groups (p<0.05; Table 3). When comparisons are restricted to hominoid genera, Oreopithecus appears more similar to humans than to any ape genus, but again with a posterior probability that indicates significant differences with all of them (p<0.05; Table 4).

Table 3
Posterior probabilities of group membership based on the bgPC scores for fossil specimens in the analysis based on the anthropoid sample.

Note that these are probability estimates of having a particular score given membership in a particular group, not the likelihood of group membership in each of a priori defined groups given a particular score. Highest probability for each specimen in bold.

CercopithecoideaHominidaeHylobatidaePlatyrrhini
BAC 208 (Oreopithecus)p=0.046p=0.013p<0.001p=0.012
StW 573 (Australopithecus)p=0.005p=0.678p<0.001p<0.001
StW 578 (Australopithecus)p<0.001p=0.190p<0.001p<0.001
Table 4
Posterior probabilities of group membership based on the bgPC scores for fossil specimens in the analysis restricted to hominoid genera.

Note that these are probability estimates of having a particular score given membership in a particular group, not the likelihood of group membership in each of a priori defined groups given a particular score. The highest probability for each specimen in bold.

HoolockHylobatesSymphalangusPongoGorillaPanHomo
BAC 208 (Oreopithecus)p<0.001p<0.001p<0.001p<0.001p<0.001p=0.002p=0.026
StW 573 (Australopithecus)p<0.001p<0.001p<0.001p=0.019p<0.001p=0.368p=0.264
StW 578 (Australopithecus)p<0.001p<0.001p<0.001p=0.062p=0.077p=0.030p=0.727

The multivariate regression between shape (deformation fields) and size (log-transformed volume of the vestibule) shows a significant correlation (i.e., allometry) at p<0.001, but nevertheless explains only a limited portion of the variance (R2 = 0.192). Bivariate regressions of the bgPCs against log-transformed cube root of vestibular volume reveal a significant correlation only for bgPC1 (R2 = 0.635, p<0.001, Table 2). A visual inspection of the scatter of points (Figure 7a) suggests that allometry for bgPC1 is more marked in hominids. This is confirmed when separate regressions are performed for hominids (R2 = 0.480, p<0.001) and the rest of the sample (R2 = 0.058, p<0.01), with the former displaying also a higher slope (Table 5). When phylogeny is considered by means of PGLS regression (Table 5), the correlation for the whole sample is still significant but explains much less variance (R2 = 0.261, p<0.01), becoming non-significant for hominids and the rest of the sample separately. bgPC3 shows a low yet significant correlation with volume for non-hominids, which becomes non-significant after PGLS correction (Table 5). Both Australopithecus and Oreopithecus overlap with the hominid scatter of points, with the two australopith specimens falling above the hominid regression line, whereas BAC 208 falls slightly below (although well above that of non-hominid anthropoids).

Bivariate regressions between (a) vestibular shape (as represented by bgPC1) and log-transformed cube root of vestibular volume (ln Vol), and (b) semicircular canal log-transformed cube root of volume (ln VolSC) vs.log-transformed length (ln L).

Lines represent OLS best-fit lines for the whole anthropoid sample (black), hominids (red), and other anthropoid taxa (blue). Note that there is a significant correlation between bgPC1 (which captures differences in SC thickness and macular organ size) and vestibular volume, more marked in hominids than in the rest of the sample (see Table 3 for further details). Note as well that hominids and other anthropoids show a similar negatively allometric relationship between the cube root of volume and length of the SCs, but with a marked allometric grade shift—such hominids possess stouter canals than other anthropoid taxa at comparable lengths once size-scaling effects have been taken into account (see Table 3 for further details). Both australopiths fall above the regression line for hominids, while Oreopithecus is slightly below, yet well above the regression line for non-hominid taxa. Color code as in Figure 2.

Table 5
Bivariate regressions of vestibular shape vs. volume and of semicular canal volume vs. length.

Both ordinary least-square linear regressions (OLS) and phylogenetic generalized least-square regressions (PGLS) are provided for the whole anthropoid sample, as well as hominids and non-hominids separately. Vestibular shape is represented by the first three principal components (bgPC), while vestibular volume (Vol) is represented by its log-transformed cube root. Semicular canal volume (VolSC) and length (L) are represented by the log-transformed cube root and the log-transformed length, respectively, of the three semicircular canals together. For each regression, the coefficient of determination (R2), the significance of the slope (p), and the slope and intercept values with their corresponding standard error (SE) and 95% confidence intervals (CI) are included. Regressions bolded when significant at p<0.05. For bgPCs vs. VOL regressions, a significant correlation denotes allometry, while for VOLSC vs. L regressions there is allometry when the correlation is significant and the 95% CI for the slope excludes unity.

R2pslopeSE95% CIinterceptSE95% CI
OLS
Anthropoids (n = 142)
 bgPC1 vs. ln Vol0.635<0.0015.2570.3354.6005.914−6.1910.399−6.973−5.409
 bgPC2 vs. ln Vol0.0080.1460.4530.309−0.1541.060−0.5330.368−1.2540.188
 bgPC3 vs. ln Vol0.0000.3850.1920.220−0.2390.622−0.2260.261−0.7380.287
 ln VolSC vs. ln L0.288<0.0010.8970.1180.6661.128−2.3280.429−3.169−1.487
Hominids (n = 30)
 bgPC1 vs. ln Vol0.480<0.0016.4961.2334.0798.913−7.7811.797−11.303−4.259
 bgPC2 vs. ln Vol0.0260.195−1.4001.054−3.4660.6662.3071.538−0.7085.322
 bgPC3 vs. ln Vol0.0390.153−0.7200.490−1.6810.2401.0360.714−0.3652.436
 ln VolSC vs. ln L0.2510.0030.6210.1900.2490.992−1.0840.705−2.4650.297
Non-hominids (n = 112)
 bgPC1 vs. ln Vol0.0580.0061.9900.7090.6013.380−2.6450.785−4.183−1.108
 bgPC2 vs. ln Vol0.0100.152−1.0270.712−2.4220.3681.0610.788−0.4822.605
 bgPC3 vs. ln Vol0.0460.0131.3320.5290.2952.368−1.4660.585−2.613−0.319
 ln VolSC vs. ln L0.219<0.0010.4540.0800.2970.611−0.7830.291−1.352−0.213
PGLS
Anthropoids (n = 27)
 bgPC1 vs. ln Vol0.2610.0043.4011.0651.3145.488−3.7191.376−6.416−1.022
 bgPC2 vs. ln Vol0.0030.4160.6250.756−0.8582.107−1.270.962−3.1550.615
 bgPC3 vs. ln Vol0.0630.111−0.7270.440−1.5890.1350.8790.570−0.2381.996
 ln VolSC vs. ln L0.437<0.0010.5020.1090.2880.7163.1440.1242.9013.388
Hominids (n = 5)
 bgPC1 vs. ln Vol0.2210.2404.4403.042−1.52110.402−4.6274.499−13.4454.192
 bgPC2 vs. ln Vol0.0360.760−1.1443.421−7.8505.5611.8924.990−7.88811.672
 bgPC3 vs. ln Vol0.0620.687−0.3920.883−2.1221.3380.5961.309−1.9703.162
 ln VolSC vs. ln L0.5530.0930.6310.2590.1241.1382.9200.3272.2793.561
Non-hominids (n = 22)
 bgPC1 vs. ln Vol0.0080.2941.3761.276−1.1253.877−1.6741.512−4.6391.290
 bgPC2 vs. ln Vol0.0300.4450.7630.978−1.1542.679−1.4441.152−3.7010.814
 bgPC3 vs. ln Vol0.0200.500−0.4190.610−1.6160.7770.5770.730−0.8542.008
 ln VolSC vs. ln L0.504<0.0010.6340.1340.3710.8963.0520.1372.7843.320

The bivariate regression between the log-transformed cube root of the SC volume and SC length shows in all instances a significant correlation that nevertheless only explains a limited (ca. 20–30%) amount of variance (Table 5, Figure 7b). Isometry cannot be rejected for anthropoids as a whole, but a negatively allometric relationship emerges (revealing that length increases faster than volume) when hominids and other taxa are analyzed separately (Table 5). The latter is confirmed by PGLS regressions for the whole sample and the two groups separately, which further explain a higher proportion of variance (Table 5), although the hominid regression is not significant with all probability due to small sample size. The bivariate plot (Figure 7b) shows an allometric grade shift between hominids and the remaining anthropoid taxa, which is confirmed by ANCOVA results—indicating no significant differences (F = 0.705, p=0.403) between the allometric slopes but significantly different intercepts (F = 263.26, p<0.001) between the two groups. This indicates that hominids possess more voluminous (i.e., stouter) canals than other anthropoids at equal lengths, with only minimal overlap. All the fossil specimens display hominid-like volumetric proportions (Figure 7b): StW 573 and 578 fall slightly above the hominid regression line, whereas BAC 208, due to its slenderer semicircular canals, falls below (although much closer than to the non-hominid regression line).

Exploration of a preexisting group structure in the tangent space of the vestibular shape

Recently, caution has been advised regarding the use of between-group PCA (bgPCA) applied to 3D geometric morphometric (GM) data, as it could produce spurious grouping when there are fewer groups than variables (Cardini et al., 2019). However, the same study also highlighted that the presence of a strong covariance among the variables (as found in many biological structures) largely reduces the magnitude of the problem. Interestingly, due to the properties of diffeomorphisms, the set of momenta is expected to be highly correlated, as close momenta tend to covary.

Prior to computing the bgPCA, we explored the principal components resulting from the vestibular shape GM analyses to investigate the presence of a preexisting group structure, which was found to be similar to that showed by bgPCA for both extant and fossil taxa (Figure 8). We used hierarchical clustering analysis (HCA) on the deformation fields for assessing the probability of correct classification of individuals according to the groups used in the bgPCA. The confusion matrix resulting from the HCA shows that most individuals are correctly identified in the corresponding groups (Table 6). Only in the deformation-based analyses hylobatids show a low percentage of classification (24% of the individuals), mostly due to the great similarity in the volumetric proportions and surface shape of the vestibule between this group and cercopithecoids.

Tangent space of vestibular shape among the analyzed anthropoid sample as shown by bivariate plots of principal components from the PCA of deformation-based analysis.

Variance explained by each PC is included within parentheses. Convex hulls correspond to: hominids (orange), hylobatids (red), cercopithecoids (blue), platyrrhines (green). Color code: dark green, platyrrhines; orange, hominids; red, hylobatids; brown, papionins; green, cercopithecins; blue, colobines. Hominids are distinguished from hylobatids and other anthropoids along PC1 (mostly negative vs. positive values, respectively), while PC2 tends to distinguish platyrrhines (more negative values) from catarrhines. Even if slightly more overlapping, the groups identified when observing the morphospace obtained with bgPCA are already present in the PCA.

Table 6
Probability of correct classification of individuals from the hierarchical clustering analyses of deformation fields according to the groups (hominids, hylobatids, cercopithecoids, platyrrhines) used in the bgPCA.
HominidaeHylobatidaeCercopithecoidaePlatyrrhini
90.0%23.5%65.0%66.7%

Another way to ascertain that the group separation observed in our bgPCA is not the result of any bias is to compare different kind of analyses with a same data set and the same grouping factor. We thus compared our deformation-based results with those obtained from a configuration of 3D semilandmarks commonly used to investigate the vestibular shape (Gunz et al., 2012). Both analyses, based on the full primate sample, yielded a similar group separation (compare Figure 2 with Figure 9), as shown by the components resulting from the bgPCA. These results are also coherent with the biological reality, enabling the discrimination of major anthropoid clades in agreement with their phylogenetic relationships. In the landmark-based approach, hylobatids largely overlap with great apes in both bgPC1 (occupying an intermediate position) and bgPC2 (Figure 9a), but can be distinguished from them (and other anthropoids) to a large extent based on bgPC3 (Figure 9b). Shape variation in the analyzed sample, as captured by landmark-based 3DGM, accounts for a strong phylogenetic signal (Kmult = 0.973, p<0.001), and this also holds for the first three bgPCs separately (see below).

Main patterns of vestibular shape variation among the analyzed anthropoid sample as shown by bivariate plots of principal components from the landmark-based between-group principal components analysis (bgPCA) using major taxa as in Figure 2.

(a) bgPC2 vs. bgPC1. (b) bgPC3 vs. bgPC1. Variance explained by each bgPC is included within parentheses. Lateral (top), superior (middle), and anterior (bottom) views of landmarks and semilandmark maximum (red) and minimum (blue) conformations for (c) bgPC1, (d) bgPC2, and (e) bgPC3. Color code as in Figure 2. Hominoids and cercopithecoids are distinguished along bgPC1 (mostly negative vs. positive values, respectively), which is mainly driven by the vertical compression of the anterior SC, as well as the position of the lateral (more anteriorly connecting with the vestibule) and the posterior (displaced posteromedially) canals. bgPC2, driven by the size of the anterior and posterior SCs relative to the lateral one, tends to distinguish platyrrhines (more negative values) from catarrhines, while bgPC3, driven by the size of the anterior and lateral SCs relative to that of the posterior canal, as well as by the relative orientation among the canals, tends to distinguish most hylobatids (more negative values) from great apes, humans, and cercopithecoids (more positive values). The two Australopithecus specimens fall within the hominid range, flanking that of hylobatids. Similarly to the deformation-based approach, StW 578 is more similar to humans (matching the distribution of Homo), while StW 573 is found within the scatter of points of Pan species. In the landmark-based analysis, Oreopithecus overlaps within cercopithecoids and also falls close to platyrrhines.

Shape differences along bgPC1 (53.3% of total variance) in the landmark-based approach embed a strong phylogenetic signal (λ = 1 and K = 1.26; Table 7). This component correlates with the insertion of the lateral canal on the vestibule, the size and shape of the posterior canal, and the roundness of the SCs. Great apes and humans fall on negative values for the first axis (Figure 9c), as they are characterized by smaller canals compared to the size of the vestibular recesses and less rounded SCs (particularly the anterior one, which is vertically compressed). Hylobatids stand on intermediate scores that largely overlap with the hominid range (Figure 9) due to a combination of long SCs, a vertically compressed anterior canal, and well separated lateral and posterior canals (since posterior canal is posteriorly displaced and the lateral canal inserts anteriorly in the vestibule). In contrast, Old world monkeys tend to be located on positive values of bgPC1 and display a protruding lateral canal that intersects the plane defined by the posterior canal.

Table 7
Phylogenetic signal results for a between-group principal components analysis (bgPCA) applied to vestibular shape procrustes residuals in the analyzed sample of extant anthropoids.
bgPC1bgPC2bgPC3
Variance53.34%29.29%17.37%
Eigenvalue2.4621.3520.802
Pagel’s λ1.000 (p<0.001)0.902 (p<0.001)0.932 (p=0.006)
Blomberg’s K1.226 (p=0.001)1.081 (p=0.001)0.528 (p=0.01)

bgPC2 (29.3% of variance) separates platyrrhines—especially Ateles found in the most negative scores (Figure 1i)—from other anthropoids due to the more reduced lateral canal in the former, which is inversely proportional to anterior canal development and vertical elongation (Figure 9d). This pattern is shared, although to a lesser extent, by humans (Figure 1d) and Theropithecus (Figure 1j), which possess more developed anterior and posterior canals relative to the lateral one. In contrast, Gorilla (Figure 1a) occupies the positive end of the distribution due to its large lateral canal and reduced anterior one, the latter being also vertically compressed, whereas the remaining hominoids show intermediate values along bgPC2. The identified phylogenetic signal displayed by bgPC2 is slightly reduced as compared to bgPC1, but still significant and very high (λ = 0.9 and K = 1.08;).

The bgPC3 (17.4% of variance) is driven by both trajectory and size of the SCs, especially of the posterior one (Figure 9e), and displays less phylogenetic signal than bgPC1 and bgPC2 (λ = 0.93 and K = 0.53, Table 7). Along this axis Homo, Ateles and Theropithecus (Figure 9b) occupy the positive end, as due to their large and anteriorly inclined posterior canal that protrudes laterally, an obtuse angle of the CC apex, large posterior canal, and an angle between the anterior and posterior canals that is close to 90°. Great apes and the majority of non-hominoid taxa fall in an intermediate position (Figure 9b), differing from the aforementioned genera by the less obtuse angle in the CC apex and a larger lateral canal. Hylobatids and, to a lesser extent, Trachypithecus show the lowest scores for bgPC3, as the result of the right to acute angle formed by the apex of the CC, more developed anterior and lateral canals relative to the posterior one, and a posteriorly tilted posterior canal.

When plotted onto the tangent space of extant taxa, the two Australopithecus specimens overlap with the range of great apes and humans for all the bgPCs (Figure 9, Figure 9—source data 1). They show a vertically compressed anterior canal together with well-separated lateral and posterior canals. The lateral canal is also moderately sinuous and its ampullar portion bends upwards, thus resulting in a negative score for bgPC1. The two specimens can be distinguished from one another by means of bgPC2 (Figure 9a), with StW 573 falling on positive values and StW 578 occupying negative ones. This is explained by the smaller lateral canal relative to the vertical SCs in StW 578, which therefore overlaps with extant human variation in both bgPC1 and bgPC2, whereas StW 573 overlaps instead with chimpanzees and bonobos (Figure 9a). Along bgPC3, both specimens display intermediate values due to the large posterior canal and for the almost right angle between the planes of the anterior and posterior SCs, overlapping with all extant anthropoids except hylobatids.

In turn, along bgPC1 Oreopithecus displays more positive values than hominoids and falls well within the range of non-hominoid anthropoids (it only slightly overlaps with the positive end of the hominoid distribution; Figure 9) due to its more coplanar lateral canal that almost intersects the plane defined by the posterior canal. Furthermore, due to its small lateral canal and fairly short CC, BAC 208 displays an intermediate value for bgPC2, within the range of extant catarrhines and slightly above the positive end of platyrrhine distribution (Figure 9a). Finally, along bgPC3 Oreopithecus falls on a positive score, differing from hylobatids, as a result of the acute angles found between anterior canal plane and those defined by both the posterior and lateral canals (Figure 9b).

Overall, the two 3DGM techniques used in this paper generally yielded similar results except for hylobatids and Oreopithecus (along bgPC1 alone). This is attributable to differences in the underlying methodological assumptions of each method when computing shape variation. In particular, our 3DGM landmark protocol measures the spatial trajectory of SCs based on their midline skeleton (Gunz et al., 2012) and hence it does not capture differences in volumetric proportions. In contrast, by comparing surfaces as a whole (Durrleman et al., 2012b; Durrleman et al., 2012a), the deformation analysis is particularly sensitive to volumetric differences. In addition, the amount of identified phylogenetic signal is very similar for both techniques, affecting the entire variance. Together with the results of the HCA, we confidently show that the separation found between the groups in the bgPCA of this study already exists in the shape data and that is not a spurious effect produced by the bgPCA method itself.

Hominoid phylomorphospace of the vestibule

The phylomorphospace approach applied to vestibular shape variation in hominoids infers different branch lengths for hominids and hylobatids from the ancestral morphology estimated for crown hominoids, which falls much closer to great apes and humans than to hylobatids for both bgPC1 and bgPC3 (Figure 10). According to our reconstructions based on the extant taxa (Figure 10), the crown hominoid LCA vestibular morphology (Figure 11a) would be characterized by equally developed and slightly inflated SCs, a fairly vertically compressed anterior canal, and by the slender portion of the lateral canal connecting more anteromedially with the vestibule. The estimated morphology for the LCA of hylobatids (Figure 11b) resembles to some extent that of the crown hominoid LCA (slightly vertically compressed anterior canal and lack of intersection between the lateral and posterior canals), combined with more monkey-like features (markedly slender canals with inflated ampullae), and others exclusive to hylobatids (an obtuse angle between a slightly anteriorly protruding anterior canal and a small posterior canal relative to the others). Hylobatid genera are generally less diverging from one other than great apes and humans. Hoolock (Figures 1f, 3f, 4f and 5f) apparently displays the most primitive morphology among hylobatids (with equally developed, rounded, almost orthogonal canals, and less anteriorly protruding anterior canal), while Hylobates (Figures 1h, 3h, 4h and 5h) and, to a lesser extent, Symphalangus (Figures 1g, 3g, 4g and 5g) show the slenderest SCs and an extremely anteriorly protruding anterior canal (as noted by Le Maître et al., 2017).

Reconstructed evolutionary history of the vestibular apparatus in the sample restricted to hominoids.

The depicted phylomorphospaces are obtained by projecting the phylogeny displayed in Figure 13 on the bivariate plots between principal components: (a) bgPC2 vs. bgPC1 (see Figure 2a); (b) bgPC3 vs. bgPC1 (see Figure 2b). Color code: red, Hylobates lar; orange, Symphalangus syndactylus; light green, Hoolock hoolock; blue, Pongo pygmaeus; dark green, Gorilla gorilla; cyan, Pan troglodytes; purple, Pan paniscus; fuchsia, Homo sapiens; chartreuse, Oreopithecus bambolii; black, Australopithecus sp. Key ancestral morphologies reconstructed using maximum likelihood for the last common ancestor of various clades are depicted as follows: 1, crown hominoids (hylobatids and hominids); 2, crown hylobatids (gibbons and siamangs); 3, crown hominids (great apes and humans); 4, crown hominines (African great apes and humans); 5, Pan-Homo clade; 6 Australopithecus-Homo clade.

Reconstructed vestibular shape for the last common ancestors (LCA) of the main clades of interest as reconstructed using maximum likelihood methods for deformation-based 3DGM analyses applied to the anthropoid sample (Figure 7), in lateral (left), superior (middle), and posterior (right) views.

Cumulative displacement variations are rendered by pseudocolor scale ranging from dark blue (<5.1 µm) to dark red (0.42 mm). Black arrows correspond to the vectors identifying the direction of displacement. The reconstructed LCAs depicted are the following: (a) crown hominoids; (b) crown hylobatids; (c) crown hominids; (d) crown hominines; (e) Pan-Homo clade; (f) Australopithecus-Homo clade.

The reconstructed morphologies for LCAs of crown hominids (Figure 11c) and, to a lesser extent, crown hominines (Figure 11d) and the Pan-Homo clade (Figure 11e) are not very far from the crown hominoid ancestral condition (Figure 11a) for any of the first three bgPCs. The crown hominid LCA is characterized by a lateral insertion of the slender portion of the lateral canal on the vestibule, a moderate medioventral displacement of the posterior canal, and an increased vertical compression of the anterior canal, in combination with thick and bulgy canals and well-developed vestibular recesses. The hominine and Pan-Homo clade LCAs possess stouter canals and are very similar to one another, distinguished only by the size of the lateral and anterior canals. The LCA of the Pan-Homo clade shows a larger and less vertically compressed anterior canal, and a smaller lateral one, which also connects more anteriorly with the vestibule. The hominin (Australopithecus-Homo) LCA (Figure 11f) is closer to Homo, being characterized by the stoutest volumetric proportions, with larger vertical canals relative to the Pan-Homo clade LCA, yet smaller than those found in humans. Its anterior canal is more vertically compressed than in Homo, rather resembling the morphology of Pan, while the posterior canal is rounded, thus being intermediate between the human (laterally projecting) and the chimpanzee (laterally compressed) morphology.

Among crown hominids, Pan (Figures 1b–c, 3b–c, 4b–c and 5b–c) more closely resembles the morphology of the inferred LCAs of crown hominids and hominoids (Figure 11a,c) in the moderately inflated and equally developed SCs and in the degree of the vertical compression of the anterior canal. Pongo (Figures 1e, 3e, 4e and 5e) occupies the positive end along bgPC1 (Figure 10) due to the possession of relatively small but extremely stout canals (especially the anterior one and the common crus). It also exhibits the most vertically compressed anterior canal and a ‘triangular’ lateral canal (i.e., showing straight slender portions of the bony labyrinth close to the ampulla and to the connection with the vestibule, as previously outlined by Spoor and Zonneveld, 1998). Finally, Gorilla and Homo (Figures 1a, 3d, 4d and 5d) are derived in opposite directions along bgPC2 (Figure 10a), as the former exhibits increased lateral canal radius with a flattened cross-section, while humans retain quite rounded canals shape and cross-section, and show a reduction of the lateral canal, as opposed to more developed anterior and posterior canals.

Australopithecus is closer to humans than to any great ape, due to the possession of large vertical canals and stout volumetric proportions, being closest to the LCA of the Australopithecus-Homo clade and only slightly more derived in the same direction as Pongo due to the stouter canals (Figure 10). In contrast, Oreopithecus appears more plesiomorphic than humans and extant great apes in volumetric proportions, resembling those found in the LCA of crown hominoids. Nevertheless, Oreopithecus is clearly distinct from hylobatids in both SC stoutness and shape, being most clearly distinguished from gibbons and siamangs based on the acute angles defined the anterior canal with both the posterior and lateral canals (Figure 10b). Oreopithecus is also derived in terms of SC size (with the vertical canals much larger than the lateral one), similarly to humans albeit to a greater extent, and opposite to gorillas, due to the remarkably smaller lateral canal.

Discussion

Previous research on the morphology of the vestibular apparatus among extant mammals has focused on its relationship to positional behavior (Spoor et al., 2007; Perier et al., 2016; Le Maître et al., 2017), particularly in order to make locomotor inferences in extinct species (Spoor et al., 1994; Walker et al., 2008; Silcox et al., 2009; Ryan et al., 2012). However, the phylogenetic signal embedded in vestibular morphology has not been adequately quantified among hominoids, because previous attempts were either exploratory (Gunz et al., 2012) or based on a restricted sample (Le Maître et al., 2017). Our results indicate that main anthropoid groups can be distinguished based on vestibular shape variation, and that there are important differences not only between hylobatids, great apes, and humans, but also among extant great ape genera. A significant phylogenetic signal is found to affect the entire variance of the anthropoid sample. Thus, the shape of the SCs is overall informative from a phylogenetic viewpoint—as hypothesized for strepsirrhine primates (Lebrun et al., 2010) and carnivorans (Schwab et al., 2019), but in contrast to previous results for hominoids (Le Maître et al., 2017) and some other mammals (Grohé et al., 2016; Costeur et al., 2018).

Based on the analysis of the shape of the vestibular apparatus, we identify several potential hominoid synapomorphies (Table 8; Figure 12a–g), including among others a posteromedially displaced posterior canal and a straight segment between the lateral-most point of the lateral canal and its anteromedially situated insertion on the vestibule. These features result in an anteromedially located lateral canal (i.e., the plane defined by the posterior canal is always separated from the trajectory of the lateral canal, even when the latter is well developed, as in Gorilla and Hylobates). This would imply an increased sensitivity for angular accelerations occurring along the coronal plane, which has been correlated with orthogrady in extant hominoids (Le Maître et al., 2017).

Illustration of hominoid, hominid and hylobatid synapormophies for the vestibular apparatus.

See Table 8 for further details. character #1: (a) rounded (gray) and compressed (green) anterior canal; character #2: (b–c) non-posteriorly displaced posterior canal (gray), posteriorly displaced posterior canal (green); character #3: (b–c) lateral canal intersecting(dashed line) the posterior one (gray), lateral canal non-intersecting the posterior one (in green); character #4: (d-e) curved (gray) and straight (green) medial portion of the lateral canal; character #5: (f–g) flat (gray) and bent upwards (green) trajectory of the lateral canal ampullar portion; character #6: (h–i) anterior and posterior SCs forming an angle (dotted arc) close to the right angle (gray), anterior and posterior canals forming an obtuse angle (green); character #7: (h–i) posterior canal equal in size to the other SCs and forming a right angle with the lateral canal (gray), small posterior canal relative the other SCs and inclined posteriorly respective to the lateral canal (green); character #8: (j–k) slender SCs (gray), stout canals (green); character #9: (j–k) small vestibular recesses (gray), enlarged vestibular recesses (green).

Table 8
Phylogenetically informative discrete characters of vestibular morphology that represent potential synapomorphies of either hominoids or hominids.
Character #Character definitionCharacter statesaSynapomorphic for
#1Anterior SCb0 = rounded; 1 = vertically compressedHominoidea
#2Posterior SC0 = non posteromedially displaced; 1 = posteromedially displacedHominoidea
#3Insertion of the lateral SC slender portion on the vestibule0 = posterior (lateral SC intersecting the posterior SC plane); 1 = anterior (not intersecting)Hominoidea
#4Lateral SC medial portion0 = curved; 1 = straightHominoidea
#5Trajectory of the lateral SC ampullar portion0 = flat; 1 = bent upwardsHominoidea
#6Angle between the planes identified by the anterior and posterior SCs0 = close to right angle, 1 = obtuseHylobatidae
#7Inclination and size of the posterior SC relative to the size of the anterior and lateral SCs0 = forming a right angle and equal or larger in size, 1 = posteriorly tilted and smallerHylobatidae
#8Robusticity of SCs0 = slender; 1 = stoutHominidae
#9Extension of vestibular recesses relative to that of the SCs0 = smaller; 1 = similar in sizeHominidae
  1. Abbreviations: SC = semicircular canals.

    a Character state 0 represents the primitive condition reconstructed for the last common ancestor of crown catarrhines.

  2. b Some platyrrhines display a superior eccentricity of the anterior SC that might be apomorphic.

The most evident character shared by extant apes and humans (even if somewhat variable in the latter and in Hoolock), and further displayed by the extinct genera analyzed here, is the vertical compression of the anterior canal (Figure 12a), as noted for great apes only in a previous analysis (Spoor and Zonneveld, 1998). Since the subarcuate fossa arguably constrains the shape of the anterior canal (Jeffery et al., 2008), the hominoid morphology might be related to the absence of the fossa in great apes and siamangs (Moyà-Solà and Köhler, 1993; Gannon et al., 1988; Spoor and Leakey, 1996). However, the combination of a well-developed fossa and marked vertical compression of the anterior SC found in Hylobates argues against this hypothesis. The latter is further rejected by the cercopithecoid morphology, characterized by a rounded anterior SC even in the largest terrestrial genera (Papio, Theropithecus, and Mandrillus), which unlike other cercopithecoids display a much reduced (or even absent) subarcuate fossa (Gannon et al., 1988; Spoor and Leakey, 1996).

As a result of vertical compression, all hominoids display the anterior canal projected anterosuperiorly to some extent (less accentuated in Homo and Hoolock). Such a projection of the anterior canal in Hylobates has been interpreted as a hylobatid synapomorphy (Le Maître et al., 2017). Our results raise doubts about the latter view and indicate instead that the anterior projection of this canal is variable within both hylobatids and hominids, and that the LCA of crown hylobatids might have not shown the patent elongation of the anterior canal that is found in Hylobates and Symphalangus. On the other hand, here we identify two potential synapomorphies for hylobatids (Table 8; Figure 12h,i): an obtuse angle between the planes defined by the anterior and posterior canals, and a posteriorly-inclined posterior canal, which is smaller relative to the anterior and lateral ones.

In turn, bgPC1 clearly separates hominids (great apes and humans) from cercopithecoids and hylobatids and enables the identification of some potential hominid synapomorphies (Table 8; Figure 12j,k). In particular, hominids differ from other anthropoids, including hylobatids, by derived volumetric proportions of the SCs (stouter canals relative to their length, or shorter canals relative to their volume) even when size-scaling considerations are taken into account, as well as by the possession of more extensive vestibular recess for a similar size of the SCs (Table 8; Figure 12j,k), as reflected in bgPC1. In particular, cercopithecoids and hylobatids completely overlap due to the possession of slender SCs, while hominids as a whole (even if more markedly orangutans) differ by their swollen and relatively shorter SCs (with only few cercopithecins falling within the hominid range). This might be related to the fact that hylobatids and cercopithecoids, unlike great apes, are swift moving animals that make fast and large head movements, thus requiring a limited duct sensitivity to avoid overstimulation and a quick response to angular displacement (Spoor and Zonneveld, 1998). This hypothesis is supported by biophysical models suggesting that the length of the membranous ducts is inversely proportional to their sensitivity and that a larger lumen of the ducts correlates with a reduced steadiness of the response to an external angular stimulus (i.e., the abrupt change of the position and/or posture) (Muller, 1994; Rabbitt et al., 2004). Therefore, species with shorter and thicker ducts (such as hominids) require more time to perceive and react to sudden head displacements, while being more sensitive to fine movements. Nevertheless, caution must be used when inferring the lumen of the ducts from that of the bony canals, as the amount of the SC cross-section occupied by perilymphatic space is variable depending on the species (Ramprashad et al., 1984; Spoor and Zonneveld, 1998).

Superimposed on the aforementioned hominoid and hominid synapomorphies (Table 8; Figure 12), there are also marked differences among hominid genera. Such differences mainly relate the relative size among the SCs (which varies particularly along bgPC2 and bgPC3), while hylobatid genera are less diverse in this regard. Chimpanzees and bonobos are characterized by equally developed SCs and a moderately short CC. Similarly, orangutans possess evenly proportioned SCs and can be distinguished from Pan by a shorter CC, more inflated canals, and a greater vertical compression of the anterior SC. Gorillas display the largest intrageneric variability in the studied sample, especially with regard to SCs slenderness, coupled with some other distinctive traits (obtuse angle of the CC apex, and longer lateral SC and CC).

Humans differ from apes in the enlarged vertical canals, a laterally protruding and inferiorly displaced posterior canal, and a moderately smaller lateral canal. Relatively enlarged vertical canals are also found in Theropithecus. The human morphology has been linked to bipedalism (Spoor et al., 1994; Spoor et al., 2003), as accelerations during bipedal walking mainly occur along the vertical axis, so that the broadly similar morphology of Theropithecus might be related to the bipedal shuffling characteristic of this genus during foraging, causing them spend an extremely large amount of time with an erect trunk posture (Wrangham, 1980).

Irrespective of the functional implications of the variation in vestibular morphology among anthropoids, our analyses show that this variation bears strong phylogenetic signal and, hence, has potential for reconstructing the evolutionary history of the group—particularly hominoids, which in spite of their extant decimated diversity are more variable in this regard (particularly when size differences among the SCs are considered) (Spoor and Zonneveld, 1998; Le Maître et al., 2017) than the taxonomically more diverse cercopithecoids. Although functional demands frequently lead to the independent evolution of similar morphologies (homoplasy), often function is not decoupled from—but superimposed on—phylogeny, with many clades being characterized by synapomorphic features linked to the adaptation for new functions. Therefore, to the extent that vestibular morphology appears to be linked to positional behavior (Spoor et al., 1994; Spoor et al., 2007; Walker et al., 2008; Silcox et al., 2009; Ryan et al., 2012; Le Maître et al., 2017), the higher variation of vestibular morphology displayed by hominoids compared to cercopithecoids agrees with the more diverse and varied locomotor repertoires of the former. This is because cercopithecoids as a whole are largely pronograde terrestrial quadrupeds that mostly differ in the degree of arboreality vs. terrestriality (Fleagle, 2013; Gosselin-Ildari, 2013), while crown hominoids are characterized by a derived and versatile orthograde body plan and associated adaptations throughout the body that enable very different and very specialized antipronograde behaviors—vertical climbing (all apes), ricochetal brachiation (hylobatids), arboreal quadrumanous suspension and clambering (orangutans), below-branch arm-swinging as well as semiterrestrial knuckle-walking (African apes), and terrestrial bipedalism (humans) (Hunt, 1991; Thorpe and Crompton, 2006).

The fact that, in terms of positional behavior, extant hominoid lineages have more significantly diverged in different directions from their last common ancestor with cercopithecoids explains why extant hominoids more strongly differ in vestibular features—even if the functional link of some vestibular features remains to be better determined. Determining the order in which these features evolved is therefore required to use them for inferring the phylogenetic placement of extinct hominoids. Besides proposing various potential synapomorphies for the hominoid and hominid clades, we further reconstruct the evolution of the vestibular apparatus in this group by estimating ancestral vestibular morphotypes by means of maximum likelihood and a molecular phylogeny. From the LCA of crown hominoids, great apes and humans appear derived in the opposite direction of hylobatids with regard to their volumetric proportions (stout vs. slender SCs, respectively). According to our ancestral state reconstruction for crown hominoids, in this regard hylobatids appear secondarily convergent with cercopithecoids. However, this hypothesis (and the alternate one, that hylobatids merely reflect more closely the primitive condition for crown catarrhines as a whole) should be tested by means of adding extinct stem cercopithecoids and fossil hominoids of less controversial affinities than Oreopithecus to the analysis. It is noteworthy that Hoolock, in agreement with the basal position of this genus among extant hylobatids, apparently retains a more primitive morphology than other hominoids for various features. While the characters related to an anteromedial displacement of the lateral SC appear synapomorphic for hominoids, the only incipient vertical compression of the anterior canal SC morphology of Hoolock might be plesiomorphic for hylobatids, in which case the marked vertical compression of this canal would be a synapomorphy of hominids only, with the remaining hylobatid genera having also evolved it in parallel.

While various great ape lineages and humans further diverged from one another from the more derived condition of the reconstructed crown hominoid LCA, as noted above gibbons and siamangs might have secondarily converged to some extent with cercopithecoids by evolving slenderer SCs, presumably as a result of similar evolutionary pressures posed by fast-moving types of locomotion. Based on our reconstructed ancestral morphotypes, bonobos and (to a lesser extent) chimpanzees would be closer to the LCAs of hominids, hominines, and the Pan-Homo clade, than either humans or the remaining extant great apes (gorillas and orangutans). The latter would have diverged from the hominid LCA in markedly different directions both in terms of SC configuration and stoutness. Our results therefore support the view that not only hominins, but also gorillas and, to a large extent, the orangutan lineage diverged from ancestors with a largely Pan-like vestibular morphology. It would be tempting to interpret this pattern in locomotor terms (e.g., by suggesting a semiterrestrial ancestry not only for hominines, but also for crown hominids as a whole). However, caution is required as other selection pressures and/or non-adaptive factors could have potentially played an equally, if not more significant, role in determining vestibular shape variation in this group.

With regard to the fossil hominoids analyzed here, Australopithecus not only displays the various hominoid synapomorphies mentioned above, but also hominid-like volumetric proportions of the SCs and, as expected in a bipedal hominin, human-like vestibular features such as large anterior and posterior canals. This is in agreement with the large amount of habitual bipedal behaviors inferred for the almost complete skeleton (StW 573) to which one of the analyzed specimens belongs (Heaton et al., 2019), as well as previous analyses of the inner ear as a whole (Beaudet et al., 2019b). Following our ancestral state reconstruction, Australopithecus appears derived in the same direction as humans, although it more closely resembles the morphology of the Australopithecus-Homo clade LCA, which is derived from the reconstructed LCA of the Pan-Homo clade in the opposite direction as chimpanzees and bonobos are. The fact that the two analyzed specimens of Australopithecus are classified as humans with a moderately high probability reflects the fact that their vestibular morphology already approximates the human condition, although maintaining plesiomorphic characters, particularly in the specimen that more closely resembles chimpanzees. On the other hand, their similar classification probability with other extant great ape genera is consistent with a more primitive vestibular morphology. It is noteworthy that, although the two analyzed individuals are very similar to one another (Beaudet et al., 2019b), they display noticeable differences in vestibular morphology, with StW 578 showing more human-like (even if stouter, approximating the orangutan morphology) canals and StW 573 retaining a more African great ape-like morphology. This could be related with diachronic changes within South African Australopithecus or to its previously noted heterogeneity (Clarke, 2013; Grine, 2013), and might help to discern, coupled with other features, the number of species represented among the current samples.

Unlike Australopithecus, Oreopithecus displays a mosaic vestibular morphology that defies a simple phylogenetic interpretation, as it does not fit well among the variation displayed by any extant hominoid genus. This is due to the combination of some hominoid and hominid synapomorphies with more plesiomorphic, cercopithecoid-like, or even platyrrhine-like, features. In particular, the hominid-like volumetric proportions of the Oreopithecus SCs would support the contention that this taxon is a great ape (Begun et al., 1997), or even a member of the European dryopithecine radiation, as previously argued by some authors (Köhler and Moyà-Solà, 1997; Harrison and Rook, 1997). However, other, apparently more plesiomorphic vestibular features, are at odds with such an interpretation, and even with the previous suggestion that Oreopithecus postcranium would be consistent with a giant hylobatid that emphasized cautious climbing (Sarmiento, 1987). In particular, the vestibular morphology of Oreopithecus does not overlap with that of extant hylobatids in any respect, particularly differing by the acute angle between the anterior and posterior canal planes and by the large posterior canal relative to the others. Oreopithecus also appears more primitive than crown hominoids in the shape of the lateral canal, which is flat (rather than displaying an upwards bent ampullar portion) and posteriorly displaced (especially in the junction between the ampulla and the vestibular recesses). In these regard, the Oreopithecus morphology more closely resembles that of cercopithecoids and platyrrhines, respectively, possibly reflecting a plesiomorphic condition that would be more consistent with a stem hominoid status, as recently supported by some other authors (Nengo et al., 2017).

If our interpretation above is correct, then the stout volumetric proportions of Oreopithecus would be homoplastic with those of great apes and humans, representing an independent acquisition that might be functionally related to the evolution of a slower mode of locomotion—in agreement with previous analysis of the inner ear of this taxon (Rook et al., 2004; Ryan et al., 2012) and the possession of an orthograde body plan with adaptations related to cautious vertical climbing and forelimb-dominated suspension (Sarmiento, 1987; Harrison and Rook, 1997). This is plausible given that hylobatids appear to some extent convergent in this regard to cercopithecoids, due to their agile locomotion. This suggests that volumetric proportions are quite labile in evolutionary terms, so that other features and functional considerations must also be considered when interpreting the vestibular morphology of extinct taxa in phylogenetic terms. Interestingly, Oreopithecus resembles australopiths and humans in the possession of a larger vertical and posterior canals relative to the lateral one, which apparently represents another homoplasy that would lend some support to the controversial claim that bipedalism featured prominently among the locomotor repertoire of this taxon (Köhler and Moyà-Solà, 1997; Rook et al., 1999). However, the SCs of Oreopithecus are clearly distinguishable from those present in Homo and Australopithecus regarding volumetric proportions, orientation, and shape. This rules out a hominin-like bipedalism for Oreopithecus—in further agreement with the lack of the lower torso features than in australopiths and humans are functionally linked to committed bipedalism (Hammond et al., 2020)—but would not be at odds with the possession of more varied orthograde positional behaviors combining climbing with a different type of bipedalism (more related to a stable bipedal stance and short distance shuffling instead of fast walking or running), as previously inferred based on the foot of this taxon (Köhler and Moyà-Solà, 1997).

In conclusion, our study provides new insight into the evolution of the vestibular apparatus in hominoids and confirms the potential of SC shape for investigating further the phylogenetic affinities of fossil apes, which are still controversial due to the inherent limitations of the fossil record. This is not to say that functional considerations must not be taken into account—rather the contrary, as several of the discussed vestibular features are arguably linked with the demands of particular positional behaviors, as noted by previous authors (Spoor et al., 2007; Walker et al., 2008; Silcox et al., 2009; Ryan et al., 2012; Perier et al., 2016; Le Maître et al., 2017). However, as exemplified by the analysis of the extinct hominin Australopithecus, the various characters identified as potentially synapomorphic for either crown hominoids or hominids offer the prospect of refining the phylogenetic placement of fossil apes for which their stem vs. crown hominoid status is controversial—as these features can be easily scored from CTs of the petrosal bone and incorporated into formal cladistic analyses including information from other anatomical areas. On the other hand, our ancestral state reconstructions rely mainly on living taxa, which is potentially problematic in the case of hominoids, which were much more diverse in the Miocene and appear quite prone to homoplasy, particularly with regard to the locomotor adaptations of the few surviving lineages. Even if the quantification of phylogenetic signal based on the phylogeny of extant taxa indicates that vestibular morphology overall is not significantly affected by homoplasy, the evolutionary history of vestibular morphology presented here on the basis of ancestral morphotypes should be treated with caution as a set of working hypotheses that require further testing based on the information provided by a larger fossil sample. In particular, given the relationship between vestibular morphology and positional behavior, and the fact that the locomotor apparatus of extinct hominoids frequently displays a mosaic of primitive and derived features unknown among the surviving lineages (e.g., Moyà-Solà et al., 2004; Alba, 2012; Alba et al., 2015; Böhme et al., 2019), it may be predicted that the vestibular morphology of extinct hominoids will similarly display unique combinations of features. This is illustrated here by the condition of Oreopithecus, which is nevertheless most consistent with that of a stem hominoid somewhat convergent with hominids in terms of locomotion. In any case, our conclusions should be subjected to further scrutiny in the future by means of the inclusion of additional fossil taxa, with emphasis on Miocene hominoids as well as stem cercopithecoids.

Materials and methods

Sample composition and acquisition

Request a detailed protocol

The analyzed sample includes microcomputed tomography (µCT) scans of 142 dried anthropoid crania belonging to 27 species and 25 genera, including all extant great ape genera and a selection of hylobatids, cercopithecoids, and platyrrhines (Table 9 and Supplementary file 1). A few specimens are juveniles instead of adults, but this should not affect their vestibular morphology as the bony labyrinth ossifies in early prenatal stages, bounding its shape and size (Jeffery and Spoor, 2004; Perier et al., 2016). The hominoid subsample consists of 48 individuals belonging to 8 species and seven genera (Supplementary file 1). For each specimen, the bony labyrinth was virtually extracted (from the left side when possible, or otherwise from the right side and mirrored) by segmenting the µCT image stacks (voxel size reported in Supplementary file 1). Virtual 3D models were generated using Avizo 9.0.1 software (FEI Visualization Sciences Group). The fossil sample consists of one left bony labyrinth belonging to the late Miocene stem-hominoid Oreopithecus bambolii (Rook et al., 2004) and of two right inner ear that have been virtually extracted from the Australopithecus specimens StW 573 and StW 578 from Sterkfontein (Beaudet et al., 2019b).The vestibular apparatus was separated from the cochlea by cutting right under the oval window and the saccule, and filling the resulting hole with a flat surface in Geomagic Studio 2014 software (3D Systems).

Table 9
Sample of extant anthropoid specimens analyzed in this paper based on µCT image stacks.

See Supplementary file 1 for further details on each specimen.

FamilySpeciesnMF?
AtelidaeAlouatta palliata5320
AtelidaeAteles geoffroyi5140
CebidaeCebus apella5320
CercopithecidaeCercocebus galeritus5320
CercopithecidaeCercopithecus mitis5050
CercopithecidaeChlorocebus pygerythrus5230
CercopithecidaeColobus guereza5230
CercopithecidaeErythrocebus patas5320
CercopithecidaeLophocebus albigena5230
CercopithecidaeMacaca fascicularis5140
CercopithecidaeMandrillus sphinx5500
CercopithecidaeMiopithecus talapoin5320
CercopithecidaeNasalis larvatus5050
CercopithecidaePapio anubis5320
CercopithecidaePiliocolobus badius5410
CercopithecidaePresbytis hosei5140
CercopithecidaePresbytis rubicunda5230
CercopithecidaeTheropithecus gelada5410
CercopithecidaeTrachypithecus cristatus5050
HylobatidaeHoolock hoolock6240
HylobatidaeHylobates lar7070
HylobatidaeSymphalangus syndactylus4220
HominidaeGorilla gorilla7250
HominidaeHomo sapiens5230
HominidaePan paniscus5140
HominidaePan troglodytes7430
HominidaePongo pygmaeus6042
  1. Abbreviations: n, total number of specimens; M, males; F, females; ?, unknown sex.

Shape analysis

Request a detailed protocol

Our 3DGM approach is based on deformation methods (Durrleman et al., 2012b; Durrleman et al., 2012a), which do not rely on a priori defined landmarks but consider instead the geometrical correspondences between continuous surfaces, and are particularly convenient for comparing overall shape and complex 3D surface changes (Durrleman et al., 2012b; Dumoncel et al., 2014; Beaudet et al., 2016a). This method relies on the construction of a sample-average surface model (template) and its deformation to the investigated surfaces (Durrleman et al., 2012b; Durrleman et al., 2012a; Beaudet et al., 2016b). Unlike in classical landmark-based 3DGM analyses, the surfaces are represented by a set of oriented faces and the comparisons do not assume a point-to-point correspondence between samples (Durrleman et al., 2012b). Prior to the analysis, the unscaled vestibular surfaces were aligned and scaled using the ‘Align Surface’ module of Avizo 9.0. Then, deformations between surfaces were mathematically modeled as a diffeomorphism (i.e., a one-to-one deformation of the 3D space that is smooth, invertible, and with a smooth inverse), and a set of momenta (vectors representing the flow of deformations from the initial position of the control points on the template to the target shape) were estimated with Deformetrica 3 software. Due to its high-demanding computational power, analyses were run in the CALMIP supercomputing center (Toulouse, France).

We inspected interspecific major patterns of shape variation by means of between-group principal components analysis (bgPCA) of the deformation-based shape residuals, using major clades (i.e., platyrrhines, cercopithecoids, hylobatids and hominids) as grouping factor. The restricted platyrrhine sample included in this study is aimed to serve as an outgroup for catarrhines, since the description of the vestibular morphology variation among New World monkeys as a whole is beyond the scope of this paper. Each group has been designed to include a large number of individuals (>>10) in order to prevent spurious separation between the groups used in the analysis. We preferred bgPCA over linear discriminant analysis (LDA) because the latter produces overexaggerated separation among groups when the number of variables is close to the number of the analyzed individuals (Mitteroecker and Bookstein, 2011). Taking into account that some authors have recently recommended caution when interpreting bgPCA results, as they might present spurious grouping (Cardini et al., 2019), we compared our results with those of a landmark-based 3DGM analysis ran on the same sample and investigated the presence of a preexisting group structure (See Exploration of a preexisting group structure in the tangent space of the vestibular shape’ section for further details).

Correlation between SC shape and size (allometry) was assessed using multivariate regression of the deformation fields against log-transformed cube root of the entire vestibular volume (ln Vol, in mm3), as well as bivariate regressions computed for each bgPC against ln Vol. We further inspected the correlation between log-transformed cube root of the SC volume (ln VolSC, in mm3; including the SCs and the CC cut at their connection with the vestibular recesses) and log-transformed length (ln L, in mm; measured along the streamline of the SCs and of the CC). All these regressions were performed for the whole anthropoid sample, as well as for hominids and non-hominids separately, as we detected several differences for the former. In the case of the ln VolSC vs. ln L regression, we also checked the homogeneity of slopes and intercepts between hominoids and the rest of the sample via analysis of covariance (ANCOVA). Regressions were performed by means of ordinary least-square linear regression (OLS) as well as by taking into account the phylogenetic non-independence between data, that is by fitting our linear models via phylogenetic generalized least squares (PGLS). The data used to compute the regressions are given in Figure 7—source data 1.

The hierarchical clustering analysis (HCA) of the deformation fields (used to identify preexisting group structure embedded in our shape data) has been computed caret v6.0–84 (Kuhn, 2008) and Factominer v1.34 (Le et al., 2008) R packages. The discrimination and the amount of overlap among the groups defined a priori for the bgPCA has been quantified by computing the number of correctly classified individuals with cross-validation using the Morpho v2.6 (Schlager, 2014) R package. Finally, we estimated the posterior probabilities of group membership for the fossil specimens based on the Mahalanobis distances between their bgPC scores and group centroids using the ‘typprobClass’ function of Morpho v2.6 (Schlager, 2014) R package. These probabilities, which were computed for both bgPCA (based on main anthropoid groups as well as on extant hominoid genera only) denote the likelihood of the specimens to belong to each group without assuming that it must belong to one of them (which is required when comparing the fossils with extant genera), so that the sum of probabilities for each specimen does not equal one. Probabilities < 0.05 indicate that the specimen falls outside the variability of that particular group. Further statistical analyses were carried out using different R packages in RStudio v.1.1.453 for R v.3.5.0: ape v5.1 (Paradis, 2012), phytools v0.6–60 (Revell, 2012), Morpho v2.6 (Schlager, 2014), caper v1.0.2 (Orme et al., 2013), and geomorph v3.1.1 (Adams et al., 2019).

Phylogenetic signal analysis

Request a detailed protocol

To assess phylogenetic signal, that is the degree to which related species resemble each other (Felsenstein, 1985; Harvey and Pagel, 1991), we used a phylogenetic tree (Figure 13) derived from a time-calibrated molecular phylogeny for the extant taxa (Springer et al., 2012). Estimated mean divergence dates for the various included extant clades are indicated in Figure 13 (see Table 1 and S1 in Springer et al. (2012) for 95% composite credibility intervals). The phylogenetic placement of Oreopithecus has been controversial; for our purposes here, we followed Nengo et al. (2017) in considering this taxon as a stem-hominoid, although other possibilities are discussed in the text. The node of the Oreopithecus-crown hominoid divergence has been placed 1 Ma older than the divergence of crown apes and humans and its tip corresponds to its last occurrence in the fossil record (7.0–6.5 Ma; Rook et al., 2000). For the South African Australopithecus sp., we used the published first appearance datum for Australopithecus africanus (4.02 Ma) that includes the Jacovec specimens into the taxon (Wood and Boyle, 2016).

Time-calibrated molecular phylogeny of extant anthropoids used in the analyses of phylogenetic signal and PGLS regressions as inferred from a species supermatrix.

Fossil taxa have been added a posteriori according to their phylogenetic position and their point estimates correspond to their last occurrence in the fossil record (Rook et al., 2000; Wood and Boyle, 2016). Oreopithecus is here considered as a stem hominoid, predating the split between hominids and hylobatids, while Australopithecus has been added following the first appearance datum for A. africanus based on the Jacovec Cavern specimens. For the phylomorphospace the phylogenetic tree has been pruned to include extant and extinct hominoids only. Branches are color-coded: dark green, platyrrhines; brown, papionins; green, cercopithecins; blue, colobines; orange, hominids; red, hylobatids. Extinct taxa are denoted by a dagger.

We computed Pagel’s λ (Pagel, 1999), Blomberg’s K (Blomberg et al., 2003) and Kmult (Adams, 2014) using phytools v.0.6–44 (Revell, 2012) and geomorph v3.1.1 (Adams et al., 2019) R packages. These metrics compare the variance in the phylogenetic tree tips relative to that expected under a Brownian motion evolutionary model. Pagel’s λ is a scaling coefficient of the expected correlations between related species on the tree, and varies from 0 (no correlation due to absence of phylogenetic signal) to 1 (covariance proportional to phylogenetic distance, implying maximal phylogenetic signal). Blomberg’s K and its multivariate generalization Kmult inform on how precisely the variance-covariance patterns found in the data are matched by the phylogenetic tree and where variance accumulates: K ≈ one implies that the mode of evolution closely resembles that expected under Brownian motion; when K < 1, close relatives resemble each other less than expected (variance is accumulated within clades), implying an evolutionary pattern that departs from a purely stochastic model (which could be caused by adaptation uncorrelated with phylogeny, that is homoplasy); finally, K > 1 is found when close relatives are more similar than expected under Brownian motion (variance is among clades), which could indicate stabilizing selection.

Phylomorphospace and ancestral states analysis

Request a detailed protocol

To quantify major patterns of vestibular shape variation along the branches of the phylogeny we relied on a phylomorphospace approach (Sidlauskas, 2008), which allows us to intuitively visualize the extent and direction of the inferred shape change by means of branch length and orientation. This method projects the phylogenetic tree (Figure 13) onto the tangent space computed from the bgPCA and estimates the position in the morphospace of the internal nodes (ancestral morphologies) via a maximum likelihood (ML) method for continuous characters (Felsenstein, 1988; Schluter et al., 1997) using the ‘fastAnc’ function of phytools v0.6–60 R package (Revell, 2012). Subsequently, the bgPC scores of the ancestral states are rotated and translated from the shape data back into the configuration space for interpolation and 3D visualization using Deformetrica 3 software.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
    Function, Phylogeny and Fossils: Miocene Hominoid Evolution and Adaptation
    1. DR Begun
    2. CV Ward
    3. MD Rose
    (1997)
    389–415, Events in Hominoid Evolution, Function, Phylogeny and Fossils: Miocene Hominoid Evolution and Adaptation, New York, Plenum Press, 10.1007/978-1-4899-0075-3_18.
  11. 11
    A Companion to Paleoanthropology
    1. DR Begun
    (2013)
    398–416, The Miocene Hominoid Radiations, A Companion to Paleoanthropology, Oxford, Blackwell, 10.1002/9781118332344.ch21.
  12. 12
    Handbook of Paleoanthropology
    1. DR Begun
    (2015)
    1261–1332, Fossil Record of Miocene Hominoids, Handbook of Paleoanthropology, London, Springer, 10.1007/978-3-642-39979-4_32.
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
    The Paleobiology Of Australopithecus
    1. R Clarke
    (2013)
    Australopithecus from Sterkfontein Caves, South Africa, The Paleobiology Of Australopithecus, Dordrecht, Springer, 10.1007/978-94-007-5919-0_7.
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
    Landmark-free 3D method for comparison of fossil hominins and hominids based on endocranium and EDJ shapes
    1. J Dumoncel
    2. S Durrleman
    3. J Braga
    4. JP Jessel
    5. G Subsol
    (2014)
    American Journal of Physical Anthropology 153.
  26. 26
  27. 27
    Topology Preserving Atlas Construction from Shape Data without Correspondence Using Sparse Parameters
    1. S Durrleman
    2. M Prastawa
    3. JR Korenberg
    4. S Joshi
    5. A Trouvé
    6. G Gerig
    (2012b)
    In: N Ayache, H Delingette, P Golland, K Mori, editors. Proceedings of Medical Image Computing and Computer Aided Intervention. Berlin: Springer-Verlag. pp. 223–230.
    https://doi.org/10.1007/978-3-642-33454-2_28
  28. 28
  29. 29
  30. 30
    Primate Adaptation and Evolution.
    1. JG Fleagle
    (2013)
    San Diego: Academic press.
  31. 31
  32. 32
    Template Estimation Form Unlabeled Point Set Data and Surfaces for Computational Anatomy
    1. JA Glaunés
    2. S Joshi
    (2006)
    In: X Pennec, editors. Proceedings of the International Workshop on the Mathematical Foundations of Computational Anatomy Sophia-Antipolis. INRIA. pp. 28–39.
  33. 33
    The Evolution of Cercopithecoid Locomotion: A Morphometric, Phylogenetic, and Character Mapping Approach. Ph.D Dissertation
    1. AD Gosselin-Ildari
    (2013)
    Stony Brook University.
  34. 34
    The Alpha Taxonomy of Australopithecus africanus
    1. FE Grine
    (2013)
    In: E Delson, EJ Sargis, editors. The Paleobiology of Australopithecus. Dordrecht: Springer. pp. 73–104.
    https://doi.org/10.1007/978-94-007-5919-0_6
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
    Catarrhine Origins
    1. T Harrison
    (2013)
    In: D. R Begun, editors. A Companion to Paleoanthropology. Oxford: Blackwell Publishing. pp. 376–396.
    https://doi.org/10.1002/9781118332344
  40. 40
    The Fossil Record and Evolutionary History of Hylobatids. Evolution of Gibbons and Siamang
    1. T Harrison
    (2016)
    In: UH Reichard, H Hirai, C Barelli, editors. Phylogeny, Morphology, and Cognition. London: Springer. pp. 90–110.
    https://doi.org/10.1007/978-1-4939-5614-2_4
  41. 41
    Enigmatic Anthropoid or Misunderstood Ape?
    1. T Harrison
    2. L Rook
    (1997)
    In: DR Begun, CV Ward, editors. Function, Phylogeny, and Fossils. Boston: Springer. pp. 327–362.
    https://doi.org/10.1007/978-1-4899-0075-3_16
  42. 42
    The Comparative Method in Evolutionary Biology
    1. PH Harvey
    2. MD Pagel
    (1991)
    Oxford: Oxford University Press.
  43. 43
  44. 44
    Positional behavior in the Hominoidea
    1. KD Hunt
    (1991)
    International Journal of Primatology 12:95–118.
    https://doi.org/10.1007/BF02547576
  45. 45
    Cercopithecoidea
    1. NG Jablonski
    2. S Frost
    (2010)
    In: L Werdelin, WL Sanders, editors. Cenozoic Mammals of Africa. London:  University of California Press. pp. 393–429.
    https://doi.org/10.1525/california/9780520257214.003.0023
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
    FactoMineR: an R package for multivariate analysis
    1. S Le
    2. J Josse
    3. F Husson
    (2008)
    Journal of Statistical Software, 25, 10.18637/jss.v025.i01.
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
    The phylogenetic relationships of Oreopithecus bambolii Gervais, 1872
    1. S Moyà Solà
    2. M Köhler
    (1997)
    Comptes Rendus de l'Académie des Sciences Paris 324:141–148.
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
    Biomechanics of the Semicircular Canals and Otolith Organs
    1. RD Rabbitt
    2. ER Damiano
    3. JW Grant
    (2004)
    In: SM Highstein, RR Fay, AN Popper, editors. The Vestibular System. New York: Springer. pp. 153–201.
    https://doi.org/10.1007/0-387-21567-0_4
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
    The phylogenetic position of Oreopithecus and its significance in the origin of the Hominoidea
    1. EE Sarmiento
    (1987)
    American Museum Novitates 2881:1–44.
  80. 80
    Morpho: calculations and visualisations related to geometric morphometrics
    1. S Schlager
    (2014)
    Morpho: calculations and visualisations related to geometric morphometrics, https://github.com/zarquon42b/Morpho.
  81. 81
  82. 82
  83. 83
  84. 84
  85. 85
  86. 86
  87. 87
  88. 88
  89. 89
  90. 90
  91. 91
  92. 92
  93. 93
  94. 94
  95. 95
  96. 96
  97. 97

Decision letter

  1. Jessica C Thompson
    Reviewing Editor; Yale University, United States
  2. Diethard Tautz
    Senior Editor; Max-Planck Institute for Evolutionary Biology, Germany
  3. Jessica C Thompson
    Reviewer; Yale University, United States
  4. David R Begun
    Reviewer; University of Toronto, Canada
  5. Ignacio Martínez
    Reviewer; Universidad de Alcala, Spain

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This paper provides a comprehensive three-dimensional analysis of the inner-ear morphology of extant primates, with a focus on understanding the evolutionary history of apes and humans. As one of the hardest parts of the skull, the inner is often well-preserved in fossils and associated with diagnostic parts of the anatomy such as teeth. This makes it a useful candidate for the reconstruction of evolutionary patterns from fossil evidence, but first it is necessary to establish the degree to which it preserves a signal of these patterns. This paper provides such evidence with its analysis of living primates, and then draws in additional evidence from three fossil specimens: two putative human ancestors (Australopithecus) and one much older ape (Oreopithecus) with an unresolved evolutionary history. In addition to describing how inner-ear morphology varies along the branches of ape evolution, this paper will be of interest to researchers who are interested in the morphometric technique that was used to execute the analysis. The authors were able to describe shape variation in a more objective manner than more traditional use of set anatomical landmarks and semilandmarks, and this has potential application across any field of study that relies on the quantification and comparison of morphology.

Decision letter after peer review:

Thank you for sending your article entitled "The evolution of the vestibular apparatus in apes and humans" for peer review at eLife. Your article is being evaluated by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation is being overseen by Diethard Tautz as the Senior Editor.

Summary:

This is a dense, technical manuscript that addresses the topic of morphological variation of the vestibular apparatus, and advances knowledge of key aspects of the evolution of anthropoid primates, especially hominoids. It is successful in demonstrating the utility of the bony labyrinth in preserving a phylogenetic signal across a relatively large sample of diverse, but also closely related, taxa. Both the sample and the methodology used are adequate and novel, and allow the authors to rigorously address the problem in question. Therefore, the results and conclusions reached (the establishment of the validity of the use of the vestibular apparatus, the proposal of characters and their phylogenetic polarity, and the description of the morphology of the LCAs for the different groups considered) are solidly justified and will serve other authors in future works.

However, this is also a highly specialized issue in primate evolution. Its broader significance to a larger readership remains unclear, even though the manuscript hints at larger implications for resolving fossil hominoid phylogeny, which is both controversial and has bearing on understanding the origins of the hominin clade. Although it may seem hominin-centric to place the significance of this manuscript with our own origins, that is certainly where its broader appeal will lie. The lack of such data, indeed the lack of more explicit emphasis on this issue in the discussion, makes this manuscript appear to be better suited in its present form for a more specialized journal. Although the reviewers appreciated the authors' caution in not overstating the significance of their results, a common concern was the broader appeal of the work as it presently stands.

Because of this ambiguity, in order to be published in eLife, the authors should do the following:

1) Provide a fossil test of the method they present here, preferably with Miocene hominoids, so that its broader applicability is very clear. The Abstract hints that some of this work may already be underway; however, if that is not the case and the authors do not already have those data and analyses in hand, then they should submit this version of the manuscript to a more specialized journal and consider eLife as a venue for their fossil application. This is because a typical revision turnaround time should be no more than about two months.

2) The authors should make their data presentation clearer, so that the differences between groups are explained in a very straightforward way. They should also amend their figures so that they are more easily readable, and correct some errors in the tables. These points are summarized below:

a) Methods: The authors should justify and explain their sample selection more in that part of the methods. For example, what is the validity for exploring the platyrrhines as a whole compared to, say the hominids as a whole (for example in Table 3), when the groups are not at the same taxonomic level? Also, throughout the text, and in the Materials and methods section, the definitions and measurement method of the SC-volume or SC-length variables appear to be missing. Since these variables are used in an important way in the manuscript, both variables must be adequately defined and also explained in detail how they have been measured so that other authors can measure them in the same way. A timeline for the proposed split times between the different analytical units that went into the analysis (based on both genetic and fossil estimates would be ideal) would be most helpful.

b) Fossil application: As for application to future fossil data, there are issues applying these results to the fossil record (as the authors note). It is not clear that stem hominines, for example, will follow the same pattern as crown hominines. If they do not, is this evidence that they are not in fact hominines, or simply that the synapomorphy of crown hominines had not yet evolved in early hominines? There is also some evidence of overlap in all the variables described. How is this quantified? What are the ranges of variation? To assess the phylogenetic signal in a fossil this variability needs to be quantified. Would it not be possible to do a DFA in order to assess the degree to which membership in a particular group is robust and predictable? What percentage of the members of each group are correctly identified? Certainly I would expect such an analysis if fossils are added to the analysis.

c) Figures require adjustment:

i) Currently, Figure 3—figure supplement 1 is the same as Figure 3. According to the text, Figure 3—figure supplement 1 should include hominoids and other anthropoids in superior view, but only includes hominoids in lateral view.

ii) Include in Figure 4 the values of R2 and the corresponding p values. It can make reading easier.

iii) A key directly on some figures would be extremely helpful (especially for Figure 6). This would make for a more understandable (visual) summary of the major differences between groups. At the end of the caption of Figure 6 it is said that humans can be distinguished by their more positive values but it seems that the distribution of the only four individuals of Homo is very similar to that of P. troglodytes and, perhaps, it would be better to remove this comment on humans from the caption of Figure 6.

iv) It would be easier to understand Figure 8 if the main morphological characteristics of each LCAs were indicated by arrows (or otherwise) in each of them.

d) Tables require adjustment:

i) In Table 2 caption it should say "as well as hominids and non-hominids separately" instead of "as well as hominoids and non-hominoids separately", since Table 2 only includes the terms hominids and non-hominids.

ii) So that the characters in Table 5 can be properly used by other authors, it would be very useful if the character states were illustrated with images.

iii) In Table 6 there is an error in the number of orangutan males: it should say 2 not 0, judging by the total of individuals (6) and the number of females (4).

iv) The narrative descriptions of the overall shape differences are clearly-written, but it would be easier to conceptualize and follow if there was less text and a single figure or set of figures with descriptive arrows highlighting the major differences. Such a figure may also include proposed or possible functional reasons for the differences, so that the visual form can be immediately linked to its potential adaptive purpose (where one may plausibly exist). A table summarizing some of these differences would also be a useful way to consolidate some of these results into a more digestible (and, frankly, useable) presentation.

https://doi.org/10.7554/eLife.51261.sa1

Author response

Because of this ambiguity, in order to be published in eLife, the authors should do the following:

1) Provide a fossil test of the method they present here, preferably with Miocene hominoids, so that its broader applicability is very clear. The Abstract hints that some of this work may already be underway; however, if that is not the case and the authors do not already have those data and analyses in hand, then they should submit this version of the manuscript to a more specialized journal and consider eLife as a venue for their fossil application. This is because a typical revision turnaround time should be no more than about two months.

As suspected by the editors, we are currently working on this approach with the aim to apply to extinct primate taxa. However, these fossils cannot be included in the present manuscript because they are unpublished and their analysis must be accompanied by detailed descriptions and comparisons outside the scope of this paper. However, to address the aforementioned request we have included in our analyses three specimens from two extinct hominoid species whose inner ear morphology is available from the literature. The addition of these taxa, which include a putative stem hominoid of uncertain affinities (Oreopithecus) and two specimens of early hominins (Australopithecus), support the reliability and usefulness of our approach. On the one hand, Oreopithecus bambolii (BAC 208) is an enigmatic late Miocene ape from Italy, whose affinities have been largely discussed during the last decades. Although there is currently a consensus that it is a hominoid, in the past it was even argued to be a cercopithecoid, and it is currently uncertain whether it is more closely related to early Miocene stem hominoids from Africa or to late Miocene stem hominids (dryopithecines) from Europe. Our analyses allow us to test among these competing hypotheses, thereby illustrating the potential of our approach for clarifying the phylogenetic affinities of Miocene hominoids. On the other hand, the vestibular apparatus morphology of two recently published individuals of South African Australopithecus (StW 578 and StW 573) enable testing the reliability of our approach, because the phylogenetic position of Australopithecus is well known (more closely related to humans than to other extant hominoids) and, hence, it can be predicted a priori that Australopithecus will show some derived features of Homo (as previously noted by the landmark-based approach published by Beaudet et al., 2019). Note that we included Amélie Beaudet as a coauthor due to her contribution to the analysis and interpretation of the Australopithecus vestibular morphology. Regarding Oreopithecus, our results support the hypothesis that it is a late-surviving stem hominoid and recover greatest similarities between australopiths and humans. Overall, we therefore think that the addition of these fossils greatly strengthens our conclusions and that the revised manuscript is more appealing to a broader readership than the original version.

2) The authors should undertake the recommendations of the reviewers to make their data presentation clearer, so that the differences between groups are explained in a very straightforward way. They should also amend their figures so that they are more easily readable, and correct some errors in the tables. These points are summarized below:

This has been done following the reviewers’ recommendations; see below for further details.

a) Methods: The authors should justify and explain their sample selection more in that part of the methods. For example, what is the validity for exploring the platyrrhines as a whole compared to, say the hominids as a whole (for example in Table 3), when the groups are not at the same taxonomic level?

Sample composition and selection have been justified in greater detail in the methods section of the revised version. The platyrrhine taxa included in this study only serve as an outgroup for catarrhines, i.e., to evaluate the polarity of evolutionary change. The description and analysis of vestibular morphology variability among New World monkeys is beyond the scope of this analysis, which is focused on hominoids. A different sample selection, more extensively sampling platyrrhine species and individuals, would be required to analyze the evolution of vestibular morphology among various platyrrhine subclades.

Also, throughout the text, and in the Materials and methods section, the definitions and measurement method of the SC-volume or SC-length variables appear to be missing. Since these variables are used in an important way in the manuscript, both variables must be adequately defined and also explained in detail how they have been measured so that other authors can measure them in the same way.

We concur with the reviewer’s request. Therefore, in the revised version we have amended the Materials and methods section by adding a definition for the SC volume and length variables. Based on these definitions, other researches should be able to take the same measurements and replicate our analyses.

A timeline for the proposed split times between the different analytical units that went into the analysis (based on both genetic and fossil estimates would be ideal) would be most helpful.

In the original version of the manuscript we already provided such information below the evolutionary tree depicted in Figure 9 (currently Figure 13), to which we have now added the two extinct hominoid genera incorporated into the revised version. To be more clear-cut in this regard, in the Materials and methods section of the revised version we have explicitly noted that such information is provided in the figure, and the uncertainty ranges of such datings are provided in several tables published by Springer et al., 2012. In case the reviewer considers it necessary, we could provide a table with the mean divergence rates and 95% composite credibility intervals, but we think it is not necessary, as the information is available from the literature.

b) Fossil application: As for application to future fossil data, there are issues applying these results to the fossil record (as the authors note). It is not clear that stem hominines, for example, will follow the same pattern as crown hominines. If they do not, is this evidence that they are not in fact hominines, or simply that the synapomorphy of crown hominines had not yet evolved in early hominines?

The main problem with fossil specimens is that their phylogenetic relationships are often controversial, so that different phylogenetic placements should be inspected to assess if the implications of the fossil taxa to the reconstruction of ancestral conditions varies. This problem, however, does not apply to extinct taxa of well-known phylogenetic relationships (such as Australopithecus, see below). Furthermore, the synapomorphies recognized based on the living taxa should contribute to further clarify the position of more controversial extinct taxa (such as Oreopithecus, see also below), based on the possession or absence of shared-derives features of extant clades. Of course, extinct taxa might eventually show that some of these synapomorphies are indeed homoplasies independently evolved among various extant hominoid lineages, or that some independently evolved on some extinct taxa. However, the condition displayed by the extant forms should constitute the baseline to assess the position of the extinct forms.

As for the specific concern by the reviewer about hominines, a priori, a stem hominine would be expected to display some (but not necessarily all) of the synapomorphies of the hominine crown group, while at the same time lacking apomorphies of any hominine subclade. In contrast, a crown hominine would be expected to display all of the synapomorphies of the hominine crown group plus some derived features of one of the hominine subclades. The same applies for other groups. In the case of our analysis of inner ear morphology, the ancestral condition reconstructed for crown hominines closely resemble that for crown hominids as a whole, making it more difficult a priori to discern hominids from stem hominines than from pongines on this basis alone. But if a particular extant taxon shows some but not all synapomorphies of a particular crown group, then it must be assumed it is a member of the stem lineage (unless further analyses show some of the purported synapomorphies to have independently evolved among extant members of the group). In the revised version, we have been more explicit about the presence/absence of crown synapomorphies when interpreting extinct taxa, and have exemplified this when discussing the two extinct genera added to the revised version.

There is also some evidence of overlap in all the variables described. How is this quantified? What are the ranges of variation? To assess the phylogenetic signal in a fossil this variability needs to be quantified. Would it not be possible to do a DFA in order to assess the degree to which membership in a particular group is robust and predictable? What percentage of the members of each group are correctly identified? Certainly I would expect such an analysis if fossils are added to the analysis.

Unlike classical morphometric methods based on linear measurements or shape ratios, it makes no sense to quantify variation in the original variables as they represent mathematical moments, that could be assimilated to a kind of "data coordinates". It is only thanks to composite multivariate analyses such as PCA and bgPCA that these data can be interpreted. In this regard, the overlap among groups can be assessed (even if not quantified) from the bivariate plots of the various PCs explaining most of the variance. We concur with the reviewer that it would be useful to assess the discrimination among groups by computing the number of original correctly classified specimens. Because of this reason, the reviewer suggests to perform a discriminant analysis, which would further enable to compute the closest similarities of the fossil specimens as compared to the extant taxa. The problem is that discriminant analyses are designed to maximize the distinction among the a priori defined groups and, to do so, they distort the morphospace in a way that tends to exaggerate a lot the group distinction. Furthermore, for mathematical reasons related to the way a discriminant analysis is computed, it is required that the number of variables is lower than the number of specimens. The usual way to circumvent this issue is to use a limited number of PC scores as variables. Nevertheless, the choice of the most adequate number of PC scores is questionable. For this reason, we rely instead on a classification based on the between-group PCA, which is similarly based on intergroup variance but does not distort so much the morphospace to maximize classification accuracy. The classification results with cross-validation have been added to the new Table 1 and reported for fossils in the new Tables 3 and 4; the Results, Material and Methods, and Discussion sections have been amended accordingly. The cross-validated classification and the posterior probabilities results were computed by Frédéric Santos, which has been accordingly added as a coauthor.

c) Figures require adjustment:

i) Currently, Figure 3—figure supplement 1 is the same as Figure 3. According to the text, Figure 3—figure supplement 1 should include hominoids and other anthropoids in superior view, but only includes hominoids in lateral view.

We thank the reviewers for bringing this issue to our attention. It was our inadvertent mistake during the submission process. We have changed the file for Figure 3—figure supplement 1 with the correct one. Figure 3—figure supplement 1 now depicts the vestibular superior view in both extant and extinct hominoids. In addition, since eLife is not restrictive in the use of figures, but rather encourages it for enhancing the comprehensibility of the text, we have moved Figure 3—figure supplements 1 and 2 to Figure 4 and Figure 5, respectively.

ii) Include in Figure 4 the values of R2 and the corresponding p values. It can make reading easier.

We have added the R2 and p values for each regression to Figure 4 (now Figure 7). The text has been colored to match that of the corresponding regression line.

iii) A key directly on some figures would be extremely helpful (especially for Figure 6). This would make for a more understandable (visual) summary of the major differences between groups.

We thank the reviewer for bringing this issue to our attention. Accordingly, we have implemented the key legend for Figure 5 (now Figure 8) and added convex hulls to Figures 2 and 6 (the latter now Figure 9), in order to help readers to identify the distribution of the groups distinguished in the bgPCA plots, as well as to identify the species to which each symbol corresponds.

At the end of the caption of Figure 6 it is said that humans can be distinguished by their more positive values but it seems that the distribution of the only four individuals of Homo is very similar to that of P. troglodytes and, perhaps, it would be better to remove this comment on humans from the caption of Figure 6.

Done.

iv) It would be easier to understand Figure 8 if the main morphological characteristics of each LCAs were indicated by arrows (or otherwise) in each of them.

Following the reviewer’s suggestion, we modified Figure 8 (now Figure 11). In particular, we implemented a pseudocolor map for the 3D meshes, which visually explains the zones of maximum and minimum deformation for the surface. Furthermore, we added a set of vectors (indicated by arrows of different size) that correspond to the magnitude and direction of the deformation. We think that these additions will be very helpful for readers to identify and visualize the characters that we describe within the main text.

d) Tables require adjustment:

i) In Table 2 caption it should say "as well as hominids and non-hominids separately" instead of "as well as hominoids and non-hominoids separately", since Table 2 only includes the terms hominids and non-hominids.

Corrected (note that former Table 2 is now Table 5).

ii) So that the characters in Table 5 can be properly used by other authors, it would be very useful if the character states were illustrated with images.

Following the reviewer’s suggestion, we added a new figure (Figure 12) that summarizes and visually illustrates the proposed synapomorphies to the readers, based on the information reported in former Table 5 (now Table 8).

iii) In Table 6 there is an error in the number of orangutan males: it should say 2 not 0, judging by the total of individuals (6) and the number of females (4).

We thank the reviewer for bringing this issue to our attention. We amended Table 6 (now Table 9) by adding a column for individuals of unknown sex. This is because two of the Pongo specimens included in this work are represented by CT scans of juvenile individuals’ petrosals, thus preventing us from determining their sex.

iv) The narrative descriptions of the overall shape differences are clearly-written, but it would be easier to conceptualize and follow if there was less text and a single figure or set of figures with descriptive arrows highlighting the major differences. Such a figure may also include proposed or possible functional reasons for the differences, so that the visual form can be immediately linked to its potential adaptive purpose (where one may plausibly exist).

We appreciate the reviewer’s suggestion to improve the clarity of the morphological descriptions by means of images. Accordingly, as stated above (see point no. 2), we added a new figure that summarized the most important morphological differences among the analyzed taxa (see Figure 12). Furthermore, major differences are highlighted by the deformation maps newly reported in Figures 3, 4, 5 and 6, which illustrate the magnitude and direction of the changes by means of vectors (depicted by arrows). In contrast, we refrained from adding potential functional links in any of the figures, not only because it would complicate their readability, but also to present an oversimplified functional interpretation of the features discussed. We consider that caution must be exercised when inferring locomotor repertoires from vestibular morphology, so we prefer to refer the readers to the more detailed explanations provided in the main text. Furthermore, the main aim of our manuscript is rather to highlight the phylogenetic significance of the various features of the vestibular apparatus, irrespective of whether their functional meaning is more or less well known.

A table summarizing some of these differences would also be a useful way to consolidate some of these results into a more digestible (and, frankly, useable) presentation.

We consider that our former Table 5 (currently Table 8) already summarizes the main differences between hominoids and non-hominoids, and also between hylobatids and hominids, in a useful way. The latter is particularly true in the sense that characters are defined and scored in a cladistic manner, which greatly facilitates their inclusion in future cladistic analyses as well as their applicability for discussion the phylogenetic placement of the extinct taxa analyzed. The only information that is missing from this table are autapomorphic features of extant hominoid taxa, but we do not consider it necessary to summarize such information in a similar manner – particularly because autapomorphic features are not phylogenetically informative from the viewpoint of parsimony analysis. Therefore, we did not implement further changes to address this particular suggestion.

https://doi.org/10.7554/eLife.51261.sa2

Article and author information

Author details

  1. Alessandro Urciuoli

    Institut Català de Paleontologia Miquel Crusafont, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Barcelona, Spain
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    alessandro.urciuoli@icp.cat
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6265-8962
  2. Clément Zanolli

    Laboratoire PACEA, UMR 5199 CNRS, Université de Bordeaux, Pessac, France
    Contribution
    Conceptualization, Data curation, Software, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5617-1613
  3. Amélie Beaudet

    1. School of Geography, Archaeology and Environmental Studies, University of the Witwatersrand, Johannesburg, South Africa
    2. Department of Anatomy, University of Pretoria, Pretoria, South Africa
    Contribution
    Data curation, Software, Formal analysis, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9363-5966
  4. Jean Dumoncel

    Laboratoire AMIS, UMR 5288 CNRS, Université de Toulouse, Toulouse, France
    Contribution
    Software, Methodology
    Competing interests
    No competing interests declared
  5. Frédéric Santos

    Laboratoire PACEA, UMR 5199 CNRS, Université de Bordeaux, Pessac, France
    Contribution
    Data curation, Software, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1445-3871
  6. Salvador Moyà-Solà

    1. Institut Català de Paleontologia Miquel Crusafont, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Barcelona, Spain
    2. Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain
    3. Unitat d’Antropologia (Departament de Biologia Animal, Biologia Vegetal i Ecologia), Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Barcelona, Spain
    Contribution
    Formal analysis, Supervision, Funding acquisition, Writing - review and editing
    Competing interests
    No competing interests declared
  7. David M Alba

    Institut Català de Paleontologia Miquel Crusafont, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Barcelona, Spain
    Contribution
    Conceptualization, Data curation, Formal analysis, Supervision, Investigation, Writing - original draft, Writing - review and editing
    For correspondence
    david.alba@icp.cat
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8886-5580

Funding

Agencia Estatal de Investigación (CGL2016-76431-P)

  • David M Alba

Agencia Estatal de Investigación (CGL2017-82654-P)

  • Alessandro Urciuoli
  • Salvador Moyà-Solà

Generalitat de Catalunya (2017 SGR 86)

  • Alessandro Urciuoli
  • Salvador Moyà-Solà

Generalitat de Catalunya (2017 SGR 116)

  • David M Alba

Agencia Estatal de Investigación (BES-2015–071318)

  • Alessandro Urciuoli

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

Access to the HPC resources of CALMIP supercomputing center were granted under the allocation 2016-[P1440] attributed to the AMIS Laboratory, Toulouse (France). This work has been funded by the Agencia Estatal de Investigación and European Regional Development Fund of the European Union (projects CGL2016-76431-P and CGL2017-82654-P, AEI/FEDER, EU), by the Ministerio de Economía y Empresa (grant BES-2015–071318 to AU), the Generalitat de Catalunya (CERCA Programme, and consolidated research groups 2017 SGR 86 and 2017 SGR 116 GRC), and the French CNRS. We deeply thank Sergio Almécija for helpful advice and discussion during the elaboration of this paper and José Braga for valuable suggestions. We are grateful to Lorenzo Rook for giving access to the Oreopithecus bambolii petrosal bone originally published in Rook et al. (2004). Lynn Copes, Lynn Lucas, and the MCZ provided access to a part of the scans used in the study, originally appearing in Copes et al. (2016), funding for the collection of which was provided by NSF DDIG #0925793, and a Wenner-Gren Foundation Dissertation Grant #8102 (both to Lynn Copes). The Evolutionary Studies Institute provided access to the Australopithecus sp. data originally appearing in Beaudet (2019a), the collection of which was funded by PAST, DST-NRF, CoE-Pal, Claude Leon Foundation, IFAS, French embassy, Leakey Foundation, Wits University. The scans for fossil and extant species were downloaded from MorphoSource.org, a web-accessible archive for 3D digital data housed by Duke University.

Senior Editor

  1. Diethard Tautz, Max-Planck Institute for Evolutionary Biology, Germany

Reviewing Editor

  1. Jessica C Thompson, Yale University, United States

Reviewers

  1. Jessica C Thompson, Yale University, United States
  2. David R Begun, University of Toronto, Canada
  3. Ignacio Martínez, Universidad de Alcala, Spain

Publication history

  1. Received: August 21, 2019
  2. Accepted: January 18, 2020
  3. Version of Record published: March 3, 2020 (version 1)

Copyright

© 2020, Urciuoli et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,042
    Page views
  • 272
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

  1. Further reading

Further reading

    1. Evolutionary Biology
    Nathalie Feiner et al.
    Research Article

    Plasticity can put evolution on repeat if development causes species to generate similar morphologies in similar environments. Anolis lizards offer the opportunity to put this role of developmental plasticity to the test. Following colonization of the four Greater Antillean islands, Anolis lizards independently and repeatedly evolved six ecomorphs adapted to manoeuvring different microhabitats. By quantifying the morphology of the locomotor skeleton of 95 species, we demonstrate that ecomorphs on different islands have diverged along similar trajectories. However, microhabitat-induced morphological plasticity differed between species and did not consistently improve individual locomotor performance. Consistent with this decoupling between morphological plasticity and locomotor performance, highly plastic features did not show greater evolvability, and plastic responses to microhabitat were poorly aligned with evolutionary divergence between ecomorphs. The locomotor skeleton of Anolis may have evolved within a subset of possible morphologies that are highly accessible through genetic change, enabling adaptive convergence independently of plasticity.

    1. Evolutionary Biology
    2. Genetics and Genomics
    María Angélica Bravo Núñez et al.
    Research Article

    Killer meiotic drivers are genetic parasites that destroy ‘sibling’ gametes lacking the driver allele. The fitness costs of drive can lead to selection of unlinked suppressors. This suppression could involve evolutionary tradeoffs that compromise gametogenesis and contribute to infertility. Schizosaccharomyces pombe, an organism containing numerous gamete (spore)-killing wtf drivers, offers a tractable system to test this hypothesis. Here, we demonstrate that in scenarios analogous to outcrossing, wtf drivers generate a fitness landscape in which atypical spores, such as aneuploids and diploids, are advantageous. In this context, wtf drivers can decrease the fitness costs of mutations that disrupt meiotic fidelity and, in some circumstances, can even make such mutations beneficial. Moreover, we find that S. pombe isolates vary greatly in their ability to make haploid spores, with some isolates generating up to 46% aneuploid or diploid spores. This work empirically demonstrates the potential for meiotic drivers to shape the evolution of gametogenesis.