Available online at www.sciencedirect.com

Journal of Human Evolution 54 (2008) 615e637

Variation in guenon skulls (I): species divergence, ecological and genetic differences

Andrea Cardini a,b,*, Sarah Elton b

a Dipartimento del Museo di Paleobiologia e dell’Orto Botanico, Universita´ di Modena e Reggio Emilia, via Universita` 4, Modena 41100, Italy b Functional Morphology and Evolution Unit, Hull York Medical School, University of Hull, Cottingham Road, Hull HU6 7RX, UK Received 1 June 2007; accepted 21 September 2007

Abstract

Guenons are the most diverse clade of African monkeys. They have varied ecologies, include arboreal and terrestrial species, and can be found in nearly every region of sub-Saharan Africa. Species boundaries are often uncertain, with a variable number of species and subspecies mostly recognised on the basis of their geographic distribution and pelage. If guenon soft tissue patterns show high variability, the same does not seem to hold for skull morphology. Guenon skulls are traditionally considered relatively undifferentiated and homogeneous. However, patterns of var- iation in skulls have never been examined using a large number of specimens sampled across the breadth of species diversity. Thus, in the present study, skulls of adult guenons and two outgroup species are analysed using three-dimensional geometric morphometrics. Three-dimensional co- ordinates of 86 anatomical landmarks were measured on 1,315 adult specimens belonging to all living guenon species except Cercopithecus dryas. Species are well-discriminated using shape but the best discrimination occurs when species have either a long evolutionary history (e.g., Alleno- pithecus nigroviridis) or represent extremes of size variation (Miopithecus sp. and Erythrocebus patas). Interspecific phenetic relationships reflect size differences. Four main clusters are found that mainly correspond to four size groups: the smallest species (Miopithecus sp.), the largest species (E. patas plus the study outgroups), a group of medium-small arboreal guenons, and a group of medium-large arboreal and terrestrial guenons. Correlations between interspecific shape distances and interspecific differences in size are higher than between shape distances and genetic dis- tances. However, if only the component of interspecific shape variation which is not correlated to evolutionary allometry is used in the comparison with genetic distances, correlations are up to 1.4 times larger than those including allometric shape. The smallest correlations are those between shape and ecological distances, which is consistent with the lack of clusters clearly reflecting broad ecological specialisations (e.g., arboreality versus terrestriality). Thus, size, which is generally considered more evolutionarily labile than shape, seems to have played a major role in the evolution of the guenons. The incongruence between interspecific shape differences and phylogeny might be explained by a large proportion of shape changes having occurred along allometric trajectories that tend to be conserved within this clade. Ó 2007 Elsevier Ltd. All rights reserved.

Keywords: Allometry; Cercopithecids; Geometric morphometrics; Size and shape; Morphological distinctiveness; Phylogenetic signal

Introduction Cercopithecini primarily comprises arboreal species, many of which are distinguished on the basis of pelage and calls The highly speciose guenons (Primates, Cercopithecini), (Kingdon, 1997). Craniodental features are seen as being widespread in sub-Saharan Africa, are an ecologically and poor species discriminators, at least within the species-rich ge- behaviourally varied tribe (Butynski, 2002). Their skull mor- nus Cercopithecus (Verheyen, 1962; Wood and Richmond, phologies, in contrast, appear to be much less diverse. The 2000), and have been the subject of remarkably few studies. Given the interest in the taxonomy, ecology, behaviour, and * Corresponding author. Dipartimento del Museo di Paleobiologia e biogeography of guenons, perceptions of craniodental unifor- dell’Orto Botanico, Universita´ di Modena e Reggio Emilia, via Universita´ 4, Modena 41100, Italy. mity must surely have contributed to this lack of attention. E-mail addresses: [email protected], [email protected] (A. Cardini), The most comprehensive study of guenon craniodental mor- [email protected] (S. Elton). phology to date was conducted by Verheyen (1962), who

0047-2484/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.jhevol.2007.09.022 616 A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 analysed 221 adult specimens of 13 guenon species using linear monkeys (O’Higgins and Collard, 2001), and adult mangabeys measurements of cranial, mandibular, and dental characters. (Cercocebus and Lophocebus) may share a divergent static Along with noting their general uniformity, lack of diagnostic scaling pattern when compared to larger-bodied papionins characters, and pronounced sexual dimorphism [explored fur- (Singleton, 2002). This indicates that the relationship between ther in Cardini and Elton (in press)], he found that allometry size and shape in primates is not necessarily straightforward or had a strong influence on guenon skull morphology. Interge- evolutionarily conserved across the majority of taxa. Investi- neric comparisons of the ratio between interorbital width and gating the effects of allometry on the skulls of the behaviour- glabella-prosthion length, which best discriminated among ally and ecologically diverse guenons, which [as discussed in genera, indicated roughly parallel size-based trajectories Cardini and Elton (in press)] exhibit a wide range of body within the majority of the clade (Verheyen, 1962). The links be- masses, will therefore help to shed further light on the links tween body size and morphology in guenons were further high- between size and shape in primate morphology. lighted by Martin and MacLarnon (1988), who, after removing Examining guenon skull morphologies will also assist in the effects of body size through the use of residual values of determining whether hard tissue features of the skull map cranial and dental dimensions, found that adults of the smallest onto the species boundaries that are indicated by calls, ecol- guenon, Miopithecus talapoin, grouped with the largest, Eryth- ogy, geographic range, and soft tissue. Geometric morphomet- rocebus patas. It thus appears likely that evolutionary or inter- rics is an ideal tool for this: demonstrated to be powerful specific allometry (sensu Klingenberg, 1996, 1998) makes an enough to identify and quantify small inter- and intraspecific important contribution to the morphological divergence of gue- differences in primate morphology (e.g., Frost et al., 2003; non taxa. It has also been suggested that ontogenetic scaling Cardini et al., 2007). Nonetheless, traditional morphometric contributes to differences in skull form within the clade. In studies have identified highly divergent taxa, including Alleno- a comparison of M. talapoin and C. cephus, Shea (1992) deter- pithecus nigroviridis (Verheyen, 1962; Martin and MacLarnon, mined that skull proportions differed significantly between the 1988) and Miopithecus (Verheyen, 1962), found in molecular two species, attributing this to a decrease in growth rates and analyses to probably be basal members of the guenon clade resultant smaller adult body size in M. talapoin. (Tosi et al., 2005). The patas monkey, which is ecologically The effects of allometric scaling on primate skull morphol- and morphologically distinctive, tends also to be separated at ogy have been the focus of considerable research, reviewed the generic level, although recent molecular data suggest extensively elsewhere (Singleton, 2002). Many studies (e.g. that as E. patas forms a monophyletic group with the other ter- Cheverud and Richtsmeier, 1986; Shea, 1992; O’Higgins and restrial guenons (the Cercopithecus lhoesti group and C. ae- Collard, 2001; Leigh et al., 2003; Leigh, 2006) have investi- thiops), it would be better placed in a more inclusive gated the role of ontogenetic allometry in determining cranial terrestrial guenon genus, possibly Chlorocebus (Tosi et al., and mandibular form. Others (e.g., Marroig and Cheverud, 2004). The majority of guenon species are included in the ap- 2001, 2005; Singleton, 2002; Frost et al., 2003; McNulty, parently cranially homogeneous Cercopithecus. These hard 2004) have concentrated on assessing how size contributes tissue similarities, alongside a sparse fossil record prior to to interspecific variation within adults of closely related spe- 1Ma(Leakey, 1988), suggest rapid, recent divergence within cies. In neotropical primates, allometry was found to account Cercopithecus and possibly guenons as a whole. However, re- for 20e40% of all morphological variation within each genus, cent genetic studies have indicated that the tribe has a rela- with only small divergence (z30) of allometric vectors be- tively long evolutionary history (Tosi et al., 2005). Based on tween and within genera, despite at least 30 million years of evidence from X-chromosomal DNA, Allenopithecus split evolutionary diversification and considerable variation in abso- from the rest of the guenons at 9.3 1.0 Ma, with divergence lute size (Marroig and Cheverud, 2001, 2005). This indicates of Miopithecus, the terrestrial guenons, and the arboreal gue- that although phenotypic means changed during the evolution nons occurring at 8.1 1.0 Ma (Tosi et al., 2005). At of South American monkeys, covariance structure of their 4.8 1.2 Ma, the three ‘terrestrial’ guenon taxa (E. patas, C. skulls remained relatively constant. The diversification of skull aethiops, and the C. lhoesti group) diverged, followed by the shape occurred mainly along lines of least evolutionary resis- separation of the major arboreal lineages (the Cercopithecus tance (Schluter, 2000) defined by size variation, which, in turn, mona/C. neglectus/C. diana group and the Cercopithecus mi- was generally associated with dietary shifts. When evolution tis/C. cephus group) at 4.6 0.7 Ma. Speciation within the ar- did not occur along lines of least evolutionary resistance, the boreal guenons probably happened relatively rapidly during amount and pace of morphological changes were, respectively, the Plio-Pleistocene, with the differentiation of C. mona, C. small and slow. In Old World primates, size also appears to neglectus, and C. diana at 3.5 0.6 Ma, and C. mitis and C. contribute to shape differences between taxa (Singleton, cephus at 2.2 0.6 Ma. The Central African forest belt is 2002; Frost et al., 2003; McNulty, 2004). However, in contrast commonly seen as the region where much Plio-Pleistocene to the conserved patterns evident in neotropical primates, there guenon evolution took place (Hamilton, 1988), with the subse- appears to be significant diversity in morphological scaling, quent dispersal of some taxa, including C. aethiops and C. mi- both ontogenetic (O’Higgins and Collard, 2001) and, to a lesser tis, to other parts of sub-Saharan Africa (Elton, 2007). degree, static (Singleton, 2002), within the sister clade of the The study reported here has two main aims. First, similar- guenons, the Papionini. Specifically, ontogenetic scaling ap- ities in skull shape (phenetic analyses) are used to group spe- pears to differ between mandrills and the other ‘dog-faced’ cies, with similarity relationships of females compared to A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 617 those of males. Given the perceived uniformity of the skull form of a large triangle with distant vertices in order to mini- alongside the obvious size and ecological differences of mise the measurement error relative to the size of the triangle. many guenons, we are interested in examining the relative im- Data collected on the mandible were then aligned onto the portance of allometry and ecology in determining skull form. same coordinate system as those collected on the cranium Since the clade probably has a relatively long evolutionary by applying a least-squares superimposition (see below) of history (Tosi et al., 2004, 2005), we also assess whether ge- the three points so that the rigid rotation derived from them ap- netic distances are related to differences in skull shape. The plies to all landmark coordinates. Software written in Visual second aim is to assess the degree of morphological diver- Basic (Jones, unpublished) was used for this. The three land- gence among species, using anatomical landmarks of the skull marks used for matching the cranial and mandibular configu- in discriminant analyses. This will shed light on whether spe- rations were eventually discarded and only the 86 anatomical cies boundaries determined by soft tissue features, biogeogra- landmarks were used in the analyses. phy, and behaviour are mirrored by small but significant differences in skull shape. Geometric morphometrics

Materials and methods In landmark-based geometric morphometrics (Rohlf and Marcus, 1993; Adams et al., 2004), the form of an organism Sample and data collection (or its organs) is captured by the Cartesian coordinates of a configuration of anatomical landmarks. Landmarks are topo- For the purposes of the current study, and partially in rec- graphically corresponding points (Marcus et al., 2000) that can ognition of the fact that published classifications have not be precisely located on all individuals. Geometric morphomet- yet been revised to reflect new molecular evidence, we follow ric analyses were performed using Morpheus (Slice, 1999), the classification of Grubb et al. (2003) in recognising four Morphologika (O’Higgins and Jones, 1999), TPSSmall 1.20 guenon genera (Allenopithecus, Miopithecus, Erythrocebus, (Rohlf, 2003), NTSYS-pc 2.2c (Rohlf, 2005), and SPSS and Cercopithecus) and also accept their species-level taxon- 11.5.0 (SPSS, 2002). omy. The sample comprised 1,315 adult specimens drawn Differences in landmark coordinates due to the position of from all living guenon species (Grubb et al., 2003) except the specimens during the digitisation process were removed, C. dryas (Table 1). Only one specimen of C. solatus, a poorly and size was standardised. This was achieved by optimally studied species that inhabits a small area of Gabon and is superimposing landmark configurations using a least-squares considered vulnerable by conservationists, was found in the algorithm (Rohlf and Slice, 1990) that scales configurations collections studied. This specimen was included only in the or- to unit centroid size, centres them on their centroid, and rotates dinations with all species and pooled sexes (PCA and ssPCA, the configurations to minimise the sum of squared distances see below), and its missing landmarks were estimated using between corresponding landmarks. The process is called gen- the average of males of the closely related C. lhoesti. The Af- eralised Procrustes analysis (GPA). Centroid size is a measure rican monkeys Cercocebus atys and Colobus guereza were in- of the dispersion of landmarks around the centroid of the land- cluded in the whole study sample as outgroups, with C. atys mark configuration and is computed as the square root of the representing the Papionini (sister clade of the guenons) and sum of squared distances of all landmarks from the centroid. C. guereza the Colobinae (sister clade of guenons and papio- The new Cartesian coordinates obtained after the registration nins). The maturity of each specimen was assessed on the basis are the aligned coordinates or shape coordinates used for sta- of third molar and canine eruption. The museums from which tistical comparisons of individuals. The shape differences specimens were derived are listed in the Acknowledgements. between landmark configurations of two individuals can be Three-dimensional coordinates of anatomical landmarks summarised by their Procrustes distance (PRD), which is ap- were directly collected by the same person on crania and man- proximately the square root of the sum of squared distances dibles using a 3D-digitiser (MicroScribe 3DX, Immersion between pairs of corresponding landmarks. Henceforth, we Corporation). Landmarks were digitised only on the left side will refer to Procrustes shape distances by simply using the to avoid redundant information in symmetric structures. The term ‘shape distances’ or the abbreviation PRD. set (configuration) of 86 landmarks used for the analysis is Before statistical comparison of shape variables was under- shown in Fig. 1 and landmarks are listed in Table 2. Measure- taken, the registered landmark configurations were projected ment error and estimates of missing landmarks (on average 1.6 to an Euclidean space tangent to the shape space. This approx- landmarks are missing in 11.0% of the specimens) are de- imation is done because the shape space (Kendall, 1984)is scribed in Cardini and Elton (in press). curved while standard statistical analyses are performed in Three registration points were digitised on pieces of plasti- an Euclidean space. This is analogous to the approximation cine stuck on the two condyles and below the incisors of the of the distance relationship between points of a small region mandible of each specimen. These landmarks were recorded of the Earth’s surface on a flat map (Rohlf, 1998). This twice: first, the mandible articulated to the cranium (after approach is satisfactory when variations are small (Rohlf, the digitisation of the cranial landmarks) and, then, on the dis- 2003), as in the present data (results not shown). articulated mandible (after the digitisation of the mandibular An extensive introduction to applications of geometric mor- landmarks). The three registration points were chosen in the phometrics in biology is provided by Zelditch et al. (2004). 618 A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637

Table 1 Species samples and missing landmarks 1 1 Genus Species (abbreviation) and taxonomic authority Sex n NmissL missL Allenopithecus nigroviridis (nig) (Lang, 1923) female 7 ee male 8 ee Cercopithecus aethiops (aet) (Linnaeus, 1758) female 169 27 2.4 male 227 39 1.7 ascanius (asc) (Audebert, 1799) female 37 2 1.5 male 39 6 1.3 campbelli (cam) (Waterhouse, 1838) female 32 ee male 31 ee cephus (cep) (Linnaeus, 1758) female 29 5 1.0 male 29 1 1.0 diana (dia) (Linnaeus, 1758) female 32 2 2.0 male 32 1 1.0 erythrogaster (eryg) (Gray, 1866) female 4 1 1.0 male 5 ee erythrotis (eryt) (Waterhouse, 1838) female 4 1 2.0 male 10 ee hamlyni (ham) (Pocock, 1907) female 15 5 1.6 male 15 2 1.5 lhoesti (lhoe) (Sclater, 1899) female 16 3 1.3 male 18 2 1.5 mitis (mit) (Wolf, 1822) female 67 3 1.3 male 79 8 1.4 mona (mon) (von Schreber, 1775) female 16 2 1.0 male 19 1 1.0 neglectus (neg) (Schlegel, 1876) female 24 4 1.5 male 27 4 1.5 nictitans (nic) (Linnaeus, 1758) female 24 3 1.7 male 23 2 2.0 petaurista ( pet) (von Schreber, 1774) female 16 1 1.0 male 25 ee pogonias ( pog) (Bennett, 1833) female 38 2 1.0 male 38 6 1.7 preussi ( pre) (Matschie, 1898) female 3 ee male 5 1 1.0 sclateri (scla) (Pocock, 1904) female 5 ee male 6 2 1.0 solatus2 (sol ) (Harrison, 1988) female ee e male 1 1 7.0 Erythrocebus patas ( pat) (von Schreber, 1774) female 9 2 1.0 male 21 1 1.0 Miopithecus ogouensis (ogo)(Kingdon, 1997) female 16 3 2.3 male 11 1 1.0 talapoin (tal ) (von Schreber, 1774) female 2 ee male 3 1 1.0 Cercocebus atys (aty) (Audebert, 1797) female 23 ee male 21 ee Colobus guereza ( gue) (Ru¨ppell, 1835) female 17 ee male 17 ee Total pooled 1,315 145 1.7 1 Abbreviations: NmissL, number of specimens with missing landmarks; missL, average number of missing landmarks in specimens with missing landmarks (pooling all species and sexes: 1,170 specimens have no missing landmarks, 85 have one, 41 have two, nine have three, three have four, four have five, two have six, and one has seven).

Detailed mathematical descriptions of geometric morphomet- Statistics ric methods are available in Bookstein (1991) and Dryden and Mardia (1998). Guidelines on how to implement linear statis- Statistical analyses were performed using NTSYS-pc 2.2f tical models in geometric morphometrics can be found in (Rohlf, 2005) and SPSS 11.5.0 (2002). Principal components Rohlf (1998) and Klingenberg and Monteiro (2005). analysis (PCA) of shape coordinates was used to extract major A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 619

Fig. 1. Landmark configuration. See Table 1 for definitions. axes of shape variation. Multivariate regressions of shape the shape space and the size-shape space (see below) of all spe- coordinates onto sex or centroid size were used to compute cies with pooled sexes. A PCA of shape coordinates was per- vectors (trajectories) of shape variation predicted by the ex- formed and scatterplots of the main axes of variation were planatory variable. The natural logarithm of centroid size examined. Forms were compared in size-shape space (Dryden was used in all regressions of shape onto size following Mit- and Mardia, 1998; Mitteroecker et al., 2004). Thus, a principal teroecker et al. (2004, 2005). Matrix correlation was used to components analysis (ssPCA) was performed on a matrix that measure the similarity of two matrices of pairwise distances includes shape coordinates and log-transformed centroid size. between the same taxonomic units. Matrix correlation was cal- Log-centroid size has the largest variance in this matrix and culated as the Pearson correlation of distances between any therefore the first principal component of the analysis (ssPC1) pair of taxonomic units in the first matrix and the correspond- is closely aligned with size. Thus, ssPC1 summarises shape var- ing distances in the second matrix. Significance was tested us- iation correlated to size in the pooled sample regardless of ing 10,000 random permutations, including observed. Thus, species and can be interpreted as the ‘common’ allometric tra- the observed correlation was compared with its permutational jectory. The inspection of ssPCs other than the first one provides distribution, which was obtained by comparing the first matrix clues as to shape differences along trajectories that do not with all possible matrices in which the order of cases in the simply follow the direction dictated by the main axis of size var- other matrix were permuted (Rohlf, 2005). iation across all study species (including within-species size- UPGMA (unweighted pair-group method using arithmetic related shape differences that may not occur along this axis). average) phenograms were used to summarise phenetic rela- In this sense, an ssPCA can be seen as a simple but preliminary tionships described by matrices of shape distances. Discussion technique to explore patterns of shape variation that can be is confined to results which are consistently found in ordina- further investigated using specific analyses. tions and phenograms. Mean shapes relationships and partial disparity. Mean Analysis of the pooled sample in the shape space and size- shapes were computed for each species (separate sexes) and shape space. The variation in shape was examined by exploring used for examining phenetic relationships. One hundred 620 A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637

Table 2 Definition and numbering of landmarks (L). The terms ‘anterior’ and ‘posterior’ are used with reference to Fig. 1. Landmarks 65 to 86 are on the mandible L Definition 1 prosthion: anteroinferior point on projection of premaxilla between central incisors 2 prosthion2: anteroinferiormost point on premaxilla, equivalent to prosthion but between central and lateral incisors 3 posteriormost point of lateral incisor alveolus 4 anteriormost point of canine alveolus 5 mesial P3: most mesial point on P3 alveolus, projected onto alveolar margin 6e9 contact points between adjacent premolars/molars, projected labially onto alveolar margin 10 posterior midpoint onto alveolar margin of M3 11e14 contact points between adjacent premolars/molars, projected lingually onto alveolar margin 15 posteriormost point of incisive foramen 16 meeting point of maxilla and palatine along midline 17 greater palatine foramen 18 point of maximum curvature on the posterior edge of the palatine 19 tip of posterior nasal spine 20 meeting point between the basisphenoid and basioccipital along midline 21 meeting point between the basisphenoid, basioccipital, and petrous part of temporal bone 22 most medial point on the petrous part of temporal bone 23 most medial point of the foramen lacerum 24 meeting point of petrous part of temporal bone, alisphenoid, and base of zygomatic process of temporal bone 25e26 anterior and posterior tip of the external auditory meatus 27 stylomastoid foramen 28, 30 distal and medial extremities of jugular foramen 29 carotid foramen 31 basion: anteriormost point of foramen magnum 32, 35 anterior and posterior extremities of occipital condyle along margin of foramen magnum 33 hypoglossal canal 34 centre of condylar fossa 36 opisthion: posteriormost point of foramen magnum 37 inion: most posterior point of the cranium 38 most lateral meeting point of mastoid part of temporal bone and supraoccipital 39 nasospinale: inferiormost midline point of piriform aperture 40 point corresponding to largest width of piriform aperture 41 meeting point of nasal and premaxilla on margin of piriform aperture 42 rhinion: most anterior midline point on nasals 43 nasion: midline point on frontonasal suture 44 glabella: most forward projecting midline point of frontals at the level of the supraorbital ridges 45 supraorbital notch 46 frontomalare orbitale: where frontozygomatic suture crosses inner orbital rim 47 zygo-max superior: anterosuperior point of zygomaticomaxillary suture taken at orbit rim 48 centre of nasolacrimal foramen (fossa for lacrimal duct) 49 centre of optic foramen 50 uppermost posterior point of maxilla (visibile through pterygomaxillary fissure) 51 frontomalare temporale: where frontozygomatic suture crosses lateral edge of zygoma 52 maximum curvature of anterior upper margin of zygomatic arch 53 zygo-max inferior: anteroinferior point of zygomaticomaxillary suture 54 zygo-temp superior: superior point of zygomaticotemporal suture on lateral face of zygomatic arch 55 zygo-temp inferior: inferolateral point of zygomaticotemporal suture on lateral face of zygomatic arch 56 posteriormost point on curvature of anterior margin of zygomatic process of temporal bone 57 articular tuber 58 distalmost point on postglenoid process 59 posteriormost point of zygomatic process of temporal bone 60 foramen ovale (posterior inferior margin of pterygoid plate) 61 meeting point of zygomatic arch and alisphenoid on superior margin of pterygomaxillary fissure 62 meeting point of zygomatic arch, alisphenoid ,and frontal bone 63 bregma: junction of coronal and sagittal sutures 64 lambda: junction of sagittal and lamboid sutures 65 anterosuperior point of mandible between central incisors 66 anterosuperior point of mandible between lateral incisors 67 posteriormost point of lateral incisor alveolus 68 anteriormost point of canine alveolus 69 mesial P3: most mesial point on P3 alveolus, projected onto alveolar margin 70e73 contact points between adjacent premolars/molars, projected labially onto alveolar margin 74 posterior midpoint onto alveolar margin of M3 75e78 contact points between adjacent premolars/molars, projected lingually onto alveolar margin 79 superior tip of coronoid process A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 621

Table 2 (continued) L Definition 80e81 most lateral and most medial points on mandible condylar surfaces 82 anteriormost point on roughening for attachment of masseter on inferior margin of the angle of mandible 83 mandibular foramen 84 posteriormost point on superior area of insertion of medial pterygoid 85 region of insertion of genioglossus muscles (midline posteriormost point on upper ‘ridge behind incisors’) 86 region of insertion of geniohyoid muscles (midline posteriormost point on lower ‘ridge behind incisors’) bootstrap samples were built for each species (separate sexes), distance of a species and the sum of squared shape distances of and bootstrapped species mean shapes were computed. For in- all species measures the amount of shape disparity in the group stance, 100 bootstrap samples of n ¼ 37 were built using the that is accounted for by said species. This metric is called partial 37 females of C. ascanius and the mean was computed for disparity (Zelditch et al., 2004). For instance, among females, the each of them (hence, producing 100 bootstrap means for this (squared) shape distance to the grand mean of all species for A. species). Thus, for each species (with separate analyses for fe- nigroviridis is 0.00256, the C. diana distance is 0.00126, and males and males) 100 means are generated by bootstrapping the sum of (squared) distances of all species is 0.0633. Thus, the original sample and computing a new mean for each boot- A. nigroviridis explains about 4% [(100 0.00256)/0.0633] of strap sample. In general, the number of independent bootstrap the differences between the species mean shapes and their grand samples is given by (2n1)!/n!(n1) (Zelditch et al., 2004): it mean, whereas C. diana explains about 2%. Although a rigorous is 3 with n ¼ 2; 10 with n ¼ 3; 70 with n ¼ 4; 756 with n ¼ 5 use of the term ‘divergence’ implies a distance between a species and so on. The number of unique bootstrap means of the and its ancestor (which is unknown in the absence of fossils, and smallest samples will therefore be smaller than the number whose estimates based on present species traits and their putative of bootstraps (100). The sample of M. talapoin usefully illus- phylogeny can be strongly affected by errors: Martins, 1999), trates the effect of bootstrapping small samples. Just two spec- a more relaxed convention is adopted here. We use ‘divergence’ imens (tal1, tal2) are available in the female sample and to indicate how much a species contributes to total disparity rel- resampling with replacement can only produce three different ative to the grand mean of all species. Thus, in the present exam- bootstrap samples: tal1 þ tal1, tal2 þ tal2, and tal1 þ tal2 ple, the divergence in shape of A. nigroviridis relative to the grand (hence, three different means). These means might have mean might be said to be about twice that of C. diana. shapes identical (e.g., tal1 þ tal1) or very similar (tal1 þ tal2) The relative position in the shape space of female means to the real specimens. A PCA of observed and bootstrapped was compared to that of males. This was achieved by comput- mean shapes of each species was performed, and scatterplots ing a matrix correlation between the matrices of shape dis- of bootstrap means together with 95% confidence ellipses tances of the two sexes. Shape distances of mean shapes were drawn to show the relative position of the means together were also compared to the genetic distances. Genetic distances with the variation around their estimates. were reconstructed using the topology of the phylogenetic tree Cluster analyses were performed on the matrix of shape dis- of Tosi et al. (2005: 64, their Fig. 3). Tree branches are propor- tances of observed means and on bootstrapped species mean tional to the number of substitutions in the sequences of the shapes. Thus, the mean of the first bootstrap sample for A. nigro- w9.3 kb frangment of X-chromosomal DNA used by Tosi viridis was analysed together with that of the first bootstrap sam- et al. (2005) for inferring guenon phylogeny. Thus, measures ple of C. aethiops plus that of the first bootstrap sample of C. of branch lengths give a good approximation of the genetic ascanius and so on for all 23 species. This was done 100 times. distances in the tree and can be used to build a matrix of pair- Then, pairwise shape distances between the 23 species means of wise genetic distances (COPH module of NTSYSpc, additive the first bootstrap were computed and an UPGMA phenogram distances options). This matrix was used for comparisons calculated. This was done for all 100 bootstrap samples, produc- with shape distances including species in common between ing 100 phenograms. Eventually, a 50% majority rule consensus the two analyses and using Miopithecus talapoin and Cercoce- tree of the 100 phenograms (plus the observed) was built and bus agilis instead of M. ogouensis and C. atys in the genetic percentages of phenograms supporting the observed topology distance matrix. Miopithecus talapoin and M. ogouensis were shown. Bootstrap support of phenogram clusters provides were not considered separate species until 1997 (Kingdon, information on how consistently a cluster is found in pheno- 1997) and no information is available in Tosi et al. (2005) to grams when species means are computed after removing some assess whether the zoo specimen used in their genetic sample specimens and replacing them with others in the same sample. belongs to one or the other population. Miopithecus is strongly Thus, it indicates how inaccuracies of estimates of mean shapes divergent for skull shape but M. talapoin and M. ogouensis are (based on available samples) might influence the reconstruction relatively similar to each other. M. ogouensis is included in the of species phenetic relationships. shape dataset because of its larger sample. Cercocebus agilis The contribution of each species (separate sexes) to the gue- and C. atys are members of a monophyletic group of closely non and outgroup shape diversity (disparity) was examined. related species together with C. torquatus. Thus, their genetic Shape distances of every species mean to the grand mean of all differences should be negligible compared to those between species were computed. The ratio between the squared shape Cercocebus and the guenons. 622 A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637

Shape distances were also compared to ecological dis- 2) it was closer to the mean of species Y and within the 95th tances. Data on ecology were taken from the literature (Fedi- percentile of distances to the mean shape in species Y; gan and Fedigan, 1988; Gautier-Hion, 1988a,b; Isbell, 1998; 3) it was closer to the mean of species Y and outside the 95th Isbell et al., 1998a,b; Nakagawa, 1999; Chapman et al., percentile of distances of both species (X and Y). 2002; Curtin, 2002; Nunn, 2002; Nunn et al., 2004). Diet (in- sects, flowers/nectar, fruits/seeds, fibrous vegetation/leaves), Analyses were cross-validated using 50% hold-out samples habitat (tropical rain forest, deciduous forest, woodland, (Hair et al., 1998). Thus, 50% of the specimens (analysis sam- swamps, flooded, forest, gallery forest, scrub, scrub forest, sa- ple) were randomly selected and used for computing the dis- vanna), and locomotion (arboreal, terrestrial) were tabulated criminant functions (or the shape distances percentiles) and using categorical variables as in Jernvall and Wright (1998), these functions used for classifying the remaining 50% of in- with the importance of each dietary category, habitat, and lo- dividuals (hold-out sample). comotor mode ranked (up to a maximum of five) for each spe- The sample of C. aethiops was much larger than that of any cies. Euclidean distances based on ecological variables were other species. To test the effects of this large sample on anal- computed and a three-way Mantel test, which calculates the yses, DAs were repeated to include only a random subsample partial correlation between shape and ecology while holding of C. aethiops of equal size to the average species sample size the effect of phylogeny constant, was performed (Legendre (20 for females and 22 for males). and Legendre, 1998; Harmon et al., 2005; Kozak et al., 2005). This was done in order to factor out potentially con- Results founding effects of the phylogenetic hierarchy. In all cases, significance was assessed by comparing the correlation of Analysis of the pooled sample in the shape space and the actual matrices to those from 9,999 random permutations size-shape space (Harmon et al., 2005; Kozak et al., 2005; Rohlf, 2005). Species discrimination. Species discrimination was exam- The scatterplot (Fig. 2) indicates that those species that rep- ined using discriminant analyses (DA) of shape on samples resent opposite extremes of guenon size variation (smallest with separate sexes. A ‘standard’ DA was performed using species, M. talapoin and M. ogouensis, Fig. 2: line A1; largest the first 30 PCs of shape variance to generate Mahalanobis dis- species, E. patas, Fig. 2: line A2) can be discriminated along tances, which measure the differences between groups relative the first two PCs of shape (Fig. 2). The outgroup species to the within group variation. Mahalanobis distances were were separated from guenons along PC2, a result most clearly used to compute the number and percentage of specimens cor- seen in C. guereza (Fig. 2: line A3). In addition, Cercopithecus rectly classified according to species. hamlyni, an arboreal species of comparable size to most other In addition, a DA that used Procrustes shape distances to guenons but with a distinctive skull morphology, can also be assign individuals to species was performed. A similar appli- distinguished (Fig. 2: line A4). cation of shape distances for classification can be found in Examination of the first four axes of the size-shape space McNulty et al. (2006). This approach does not modify the spa- (Fig. 3) provides further insights into general patterns of gue- tial relationships of the specimens in the shape space, and thus non skull shape variation. To aid the interpretation of the series relates differences to their absolute magnitude rather than to of bivariate plots of Fig. 3, one can imagine ssPC1, ssPC2, and the observed variation (see Klingenberg and Monteiro, 2005, ssPC3 as if they were sides of a transparent box full of points for a discussion of properties of the shape space after the in which relative distances are proportional to size-shape dif- GPA and when GPA coordinates are used in general linear ferences. Examining variation along ssPC1 and ssPC2 is like models). Congruence of classifications in the transformed looking through the surface defined by the two longest sides shape space of Mahalanobis distances and in the original un- of the box. Examining ssPC2 and ssPC3 is analogous to look- transformed Procrustes shape space suggests robustness of re- ing through the top of the box (and so on for other ssPCs). sults and their independence of the analytical method. Thus, Thus, ssPC1 summarised the common allometric trajectory the following criteria were used for the ‘Procrustes DA’: that goes from Miopithecus to Erythrocebus (Fig. 3A: lines A specimen of species X was assigned to species X if: A1e2). Alternatively, ssPC2, ssPC3 and ssPC4 help visualise the most evident deviations from this trajectory. That is, 1) it was closer to the mean of X than to the mean of any ssPC2 separated guenons and outgroups (C. atys, Fig. 3B: other species; line B1; C. guereza, Fig. 3B: line B2) and, to some degree, 2) it was closer to the mean of a species Y (other than X) but C. hamlyni from most other guenons (Fig. 3B: line B1). And within 95th percentile of distances to the mean shape of ssPC3 and ssPC4 showed three main clusters: 1) outgroup spe- th species X and outside 95 percentile of distances to the cies (Fig. 3C: line C1), 2) C. aethiops (Fig. 3B, C: lines B3e mean shape of species Y. C2), and 3) all other species.

A specimen of species X was assigned to species Y if: Mean shapes relationships and divergence

1) it did not satisfy any of the conditions (above) for belong- Scatterplots of the first two PCs (species means with sepa- ing to X; rate sexes) summarise similarity relationships of guenons A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 623

Fig. 2. Scatterplots of PC1 and PC2 of shape variables of all species with pooled sexes. Percentages of explained variance in parentheses, abbreviations of species names in this and other figures are given in Table 1. Lines are drawn which emphasise main clusters (lines A1,2,3,4) described in the Results.

(Fig. 4). The PCA of female and male mean shapes indicated Excluding species with extremely small or large body masses similar phenetic relationships. This observation is consistent (Miopithecus and Erythrocebus), the guenon sample can be with the high correlation between matrices of female and subdivided into two size groups. Skulls larger than the median male mean shape distances (r ¼ 0.920, p ¼ 0.0001). Ninety- of species average sizes in both sexes were found in Allenopi- five percent confidence ellipses around means suggested large thecus nigroviridis, the terrestrial taxa C. aethiops, C. lhoesti, errors in estimates of species mean shapes for the smallest and C. preussi, and the arboreal C. diana, C. hamlyni, C. mitis, samples. For instance, large variation in bootstrap mean shapes C. neglectus, and C. nictitans. These species usually plot to the was found in C. erythrotis, for which fewer than ten specimens left of the dashed line shown in Fig. 4 (lines A4eB4). All other of each sex are available, while variation of means was very arboreal guenons had average skull sizes smaller than the me- small in C. aethiops and C. mitis, which represent the largest dian and are mostly found to the right of the dashed line. Con- samples in the analysis. sistent with the general trend of allometric shape variation, The colobine C. guereza was separated completely from all small and large species of Cercopithecus tend to have skulls other species (Fig. 4A, B: lines A1eB1). Both Miopithecus of shape intermediate between Miopithecus and Erythrocebus. species were also clearly isolated (Fig. 4A,B: lines A2eB2). Thus, the former are somewhat reminiscent of Miopithecus In addition, C. atys (the papionin outgroup), E. patas, and C. and the latter of Erythrocebus [compare, for instance, C. diana hamlyni had mean shapes in which the variation did not appre- and C. ascanius in Fig. 3 of the companion paper (Cardini and ciably overlap with that of other species (Fig. 4A,B: lines A3e Elton, 2008)]. B3). Miopithecus and E. patas lie at opposite extremes of the The discrimination of clusters of species mean shapes and range of guenon size variation. Miopithecus is characterised its correlation to size differences is effectively visualised in by a very short face with large orbits and a relatively large the phenograms of female and male mean shapes (Figs. 5 neurocranium (Figs. 5 and 6). This trend is reversed in E. patas and 6). Shape distances between means of any two species (Figs. 5 and 6). C. hamlyni also has a relatively long face. correlated to the corresponding absolute differences in skull However, compared to E. patas, its face is relatively flatter average size (female r ¼ 0.842, p ¼ 0.0001; male r ¼ 0.870, and its nasals longer. p ¼ 0.0001). The remaining species mean shapes cluster in the central The relative contribution of each species to the diversity in area of the plot and show partly overlapping ranges. Two species mean shapes is shown in Table 3 using percentages of main clusters can be identified (Fig. 4A,B: lines A4e5eB4e5), partial disparity. Females and males exhibited similar patterns. which largely corresponded to variation in skull size. The most distinctive skull shapes were found in the outgroup Fig. 3. Scatterplots of the first four principal components (ssPC) of the size-shape space with pooled sexes (percentages of explained variance in parentheses). A) Variation along the axis of the common allometry (ssPC1). B and C) Variation orthogonal to the axis of the common allometry. Note that axes in A, B, and C are not on the same scale. Lines are drawn which emphasise main clusters (A1,2;B1,2,3;C1,2) described in the Results. A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 625

Fig. 4. Scatterplots of PC1 and PC2 of species mean shapes with separate sexes (percentages of explained variance in parentheses; in both sexes, the correlation of PC1 with size is larger than 0.94, p < 0.00001, and smaller than 0.18 and not significant for all other PCs). For each species, variation around the mean is illustrated with 95% confidence ellipses of bootstrap mean shapes (not computed for M. talapoin females because n is too small). A, Female means. B, Male means. Lines are drawn which emphasise main clusters (lines A1,2,3,4,5;B1,2,3,4,5) described in the Results. and in the smallest and largest guenons (Erythrocebus and between species phenetic relationships and size also holds for Miopithecus). C. hamlyni and A. nigroviridis also had diver- the subsample of 17 species included in the comparison with gent shapes. All other species, arboreal and terrestrial, had genetic distances. Correlations were slightly smaller than similar magnitudes of shape differences to the grand mean. when all species were included but still remained high Cercopithecus nictitans, C. campbelli, and C. mona were (r 0.818, p ¼ 0.0001). Shape distances were compared to ge- among the least divergent, for both sexes. netic distances using a test for matrix correlation (Harmon Scatterplots of bootstrap mean shapes (Fig. 4), and unre- et al., 2005; Polly, 2005). Thus, the correlation between shape solved or poorly supported polytomies in phenograms (Figs. distances of all pairs of species mean shapes and correspond- 5 and 6), indicate that means of species with small samples ing genetic distances could be calculated (Fig. 7A2eB2). were strongly affected by sampling error. Thus, the smallest Corrections for possible deviations from linearity of the rela- samples (where n < 15) were excluded before comparing tionships between shape and genetic distances were not used shape similarity relationships with phylogeny. because nonlinear models (including logarithmic, power, and Shape distances were first compared with absolute differ- polynomial) did not produce large increases in the goodness 2 ences in skull size (Fig. 7A1eB1), to verify that the correlation of fit of the model (increase in r 0.060). Correlations 626 A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637

Fig. 5. Phenogram of the species female mean shapes with ‘bootstrap support’ of clusters (percentages of congruent branches of phenograms computed using mean shapes of 100 bootstrap samples of the original species sample). Species are subdivided in four groups based on average skull size and groups are shown using different background colours for species names (small: white; medium: light grey; large: dark grey; ‘extra’ large: black). A representative of each size group (in- dicated with S, M, L, XL) is shown by the phenogram with both pictures and surface rendering of the skull. The asterisk (*) indicates the outgroup species.

were 0.613 (p ¼ 0.0052) for females and 0.544 (p ¼ 0.0070) males (Fig. 7A3eB3). Finally, shape distances were compared for males. Another test was performed after regressing out to ecological differences with the effect of phylogeny held shape variance correlated to common evolutionary allometry constant (Fig. 7A4eB4). Correlations in the three-way Mantel of mean shapes. A matrix of Euclidean distances was com- tests were 0.182 for females and 0.343 for males (r ¼ 0.362 for puted on the residuals of the regression of shape onto size females and 0.116 for males, if nonallometric shape is used). and compared to the matrix of genetic distances. Correlations Tests for correlations between shape and ecology without con- between ‘nonallometric’ shape and genetic distances were trolling for phylogeny (i.e., simple correlations between matri- 0.695 (p ¼ 0.0008) for females and 0.746 (p ¼ 0.0003) for ces of shape and ecological distances) produced results very A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 627

Fig. 6. Phenogram of the species male mean shapes with ‘bootstrap support’ of clusters. See Fig. 5 for legend.

similar to three-way Mantel tests. None of the correlations sex tended to classify relatively poorly. Identification errors with ecological distances were highly significant (p > 0.01). were larger in species with larger shape variance (correlations between average percentages of misclassified specimens and Species discrimination shape variance were 0.487; p < 0.05) and smaller in those with strongly divergent shapes [correlations between average The standard DA and the discrimination based on shape percentages of misclassified specimens and partial disparities distances gave similar results (Table 4; Mahalanobis distances (Table 3) were 0.423; p < 0.05]. for comparison are given in Tables 5 and 6). Eighty to ninety C. diana, C. mona, and C. nictitans had classification errors percent of specimens were, on average, classified to the correct larger than 15% in both sexes. In C. ascanius, over 15% of fe- species. In species with the largest classification errors, each males and close to 15% of males were misclassified. The male 628 A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637

Table 3 Also, percentages of misclassified specimens in the DA that in- Partial disparities (percentages) with standard errors estimated using 100 boot- cluded all individuals versus those with a subsample of C. ae- straps (Zelditch et al., 2004). Asterisk (*) in this and other tables indicates thiops of equal size to the average species sample size differed, small samples (n < 15) on average, by less than two units (r 0.946, p < 0.001; differ- Species Female Male ence in means not significant in paired t-tests). Allenopithecus nigroviridis 4.0 0.5 4.3 0.4 Cercopithecus aethiops 1.8 0.1 1.5 0.1 Discussion Cercopithecus ascanius 1.6 0.2 1.9 0.2 Cercopithecus campbelli 1.2 0.1 0.7 0.1 Cercopithecus cephus 1.0 0.2 1.8 0.2 Size and shape relationships Cercopithecus diana 2.0 0.3 1.3 0.2 Cercopithecus erythrogaster* 2.6 0.6 2.7 0.5 Size appears highly influential in determining overall skull Cercopithecus erythrotis* 3.4 0.8 0.6 0.3 shape within the guenon clade. Alongside the strong, allomet- Cercopithecus hamlyni 4.8 0.4 5.9 0.6 Cercopithecus lhoesti 2.1 0.2 2.7 0.3 rically conserved sexual dimorphism evident in the sample Cercopithecus mitis 2.2 0.2 2.1 0.2 [discussed in the companion paper (Cardini and Elton, Cercopithecus mona 0.9 0.2 0.9 0.2 2008)], discrimination between large and small species oc- Cercopithecus neglectus 3.1 0.2 2.4 0.3 curred along the trajectory of common allometry summarised Cercopithecus nictitans 1.3 0.2 0.7 0.2 by the first axis of variation in size shape space (ssPC1). In or- Cercopithecus petaurista 1.4 0.2 1.5 0.2 Cercopithecus pogonias 1.7 0.2 1.6 0.1 dinations and phenograms, species mean shapes clustered Cercopithecus preussi* 3.0 0.8 2.5 0.6 largely according to size variation, although groups based on Cercopithecus sclateri* 2.4 0.5 3.2 0.5 phylogeny were also found (for instance, C. campbelli and Cercopithecus solatus* eeC. mona or C. ascanius and C. cephus). Clustering of species Erythrocebus patas 6.9 0.5 10.2 1.1 by size might be explained by either homoplasy or symplesio- Miopithecus ogouensis 16.9 0.8 16.4 0.8 Miopithecus talapoin* 10.1 0.7 14.9 0.9 morphy. Moreover, a propensity to have similar shapes in spe- Cercocebus atys 7.4 0.4 7.3 0.4 cies of similar size (‘similarity’) may be present in a clade Colobus guereza 18.2 0.7 13.0 0.6 where shape variation has a strong allometric component. Thus, although phenotypic means have changed during the ra- diation of guenons (as shown by high percentages of correctly C. campbelli, C. erythrotis, C. petaurista, and C. pogonias classified individuals in discriminant analyses, discussed fur- samples had classification errors in excess of 15%, and this ther below), a general pattern of changes in skull proportions also occurred in close to 15% of the females. In addition, correlated to size differentiation has been conserved. This is over 15% of C. preussi males were misclassified. All these also suggested by the almost parallel static (intraspecific) al- species, except C. preussi, belong to the arboreal guenon clade lometries observed in guenons and reported in the companion and have intermediate skull size. paper (Cardini and Elton, 2008), partly contrasting with stud- Results of the classifications of males were validated after ies on the sister clade of guenons, the papionins, in which excluding the five smallest samples (C. erythrogaster, C. eryth- intraspecific ontogenetic and interspecific adult scaling pat- rotis, C. preussi, C. sclateri, and M. talapoin) by repeating the terns were found to display a variable degree of divergence analysis using a 50% hold-out sample. The percentages of cor- (O’Higgins and Collard, 2001; Singleton, 2002). rectly identified specimens (hit ratios) in this sample were very The high correlation between the mean shape distance ma- similar to or higher than in the analysis that included all species. trices of males and females indicates that similarity relation- In the hold-out sample, hit ratios were smaller, and the total per- ships are alike in the sexes. Considered in the light of the centage of correctly classified specimens was 88.4% in the DA conserved allometric model of skull shape sexual dimorphism, (first 30 PCs, equal a priori probabilities) and 82.9% using the a simple interpretation for such high congruence rests on the PRD. Similarly, results of the female classifications were vali- fact that sex trajectories are almost parallel (Cardini and Elton, dated after excluding the three smallest samples (C. eryth- 2008). Thus, female and male mean shapes are different, but rogaster, C. preussi, M. talapoin) by repeating the analysis relative positions of species are similar because sexually di- with a 50% hold-out sample. As in the males, the hit ratios in morphic traits are produced by sliding species-specific shapes this sample were very similar to or higher than in the analysis along lines of shape variation that are nearly parallel. How- of all species, while in the hold-out sample hit ratios were ever, large variation of bootstrap mean shapes around observed smaller and the total percentage of correctly classified speci- mean shapes indicates that similarity relationships might be mens was 82.8% in the DA and 82.5% using the PRD. strongly affected by sampling error, and inferred patterns Including more than the first 30 PCs produced a negligible could thus be misleading. Interestingly, males seem to be increase in the hit ratios in all DA (all species or the subset more strongly affected by uncertainties in patterns of phenetic used for validation). In all cases, the matrix of Euclidean dis- relationships. A large number of clusters is weakly supported tances computed using only the first 30 PCs had a very high in the male phenogram. These are the clusters that are inconsis- correlation (r > 0.994) with the matrix of PRD computed in tently found in cluster analyses of bootstrapped mean shapes. the full shape space. Thus, relationships among specimens Males also have slightly lower percentages of correctly classi- are accurately summarised in the space of the first 30 PCs. fied specimens in discriminant analyses and a larger number .Crii .Etn/Junlo ua vlto 4(08 615 (2008) 54 Evolution Human of Journal / Elton S. Cardini, A. e 637

Fig. 7. Matrix correlations between shape distances (in units of Procrustes shape distances), absolute differences in size (mm), genetic distances (arbitrary unit of measure expressing relative lengths of tree branches from Tosi et al., 2005), and ecological distances (arbitrary units; see Material and methods). All correlations are highly significant (p < 0.01) except those with ecological distances. Scatterplots of shape distances between all pairs of species mean shapes and the corresponding differences in average size, genetic distances, or ecological distances. A) Females; B) Males. A1eB1: shape (including allometric shape) onto size; A2eB2: shape (including allometric shape) onto genetic distances; A3eB3: nonallometric shape onto genetic distances; and A4eB4: shape (including allometric shape) onto ecological distances after controlling for phylogeny. 629 630 A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637

Table 4 Miopithecus belongs to an ancient guenon lineage (Tosi Percentages of correctly classified specimens using discriminant analyses et al., 2005), but it is also the smallest living African monkey, based on discriminant functions computed using the first 30 PCs of shape commonly perceived as a ‘dwarf’ genus. Verheyen (1962) com- (which account for 77.1% and 78.2% of variance in females and males, respec- tively) or using Procrustes shape distances (PRD) mented on the ‘aberrant’ nature of Miopithecus cranial shape, explained by Shea (1992) as a product of rate hypomorphosis, Species Females Males a decrease in growth rate over a given time. The ecological con- PCs PRD PCs PRD ditions under which selection for small body size occurred in Allenopithecus nigroviridis 100.0 100.0 100.0 100.0 Miopithecus are still largely unexplored. Dietary niche parti- Cercopithecus aethiops 94.7 93.5 94.7 93.8 tioning might be one explanation, given the observation that Cercopithecus ascanius 73.0 73.0 87.2 84.6 Cercopithecus campbelli 87.5 90.6 80.6 80.6 over one third of the talapoin diet is insects (Gautier-Hion, Cercopithecus cephus 86.2 79.3 82.8 79.3 1978). However, as Shea (1992) points out, such supplementa- Cercopithecus diana 78.1 71.9 78.1 78.1 tion may not have been the prime mover in the evolution of Cercopithecus erythrogaster* 100.0 100.0 100.0 100.0 smaller body size. What proportion of its distinctive morphol- Cercopithecus erythrotis* 100.0 100.0 50.0 60.0 ogy results from its small (probably derived) size as opposed Cercopithecus hamlyni 100.0 100.0 93.3 100.0 Cercopithecus lhoesti 100.0 100.0 88.9 94.4 to its relatively long evolutionary history requires further Cercopithecus mitis 91.0 85.1 89.9 84.8 investigation. This is particularly true given a very recent phy- Cercopithecus mona 81.2 75.0 78.9 78.9 logenetic analysis on 11 species of guenons using short inter- Cercopithecus neglectus 100.0 100.0 100.0 96.3 spersed elements (Xing et al., 2007) that confirmed all the Cercopithecus nictitans 79.2 62.5 47.8 47.8 main clades reported by Tosi et al. (2005) but shed doubts on Cercopithecus petaurista 100.0 100.0 80.0 84.0 Cercopithecus pogonias 76.3 86.8 84.2 78.9 the evolutionary age of Miopithecus by suggesting it could be Cercopithecus preussi* 100.0 100.0 80.0 80.0 closely related to the arboreal guenons and thus may have Cercopithecus sclateri* 100.0 100.0 100.0 100.0 diverged more recently than originally thought. Erythrocebus patas 100.0 100.0 85.7 85.7 Patas monkeys have acquired distinct adaptations to terres- Miopithecus ogouensis 100.0 100.0 90.9 90.9 trial life and large home range areas (Gebo and Sargis, 1994; Miopithecus talapoin* 100.0 100.0 100.0 100.0 Cercocebus atys 100.0 100.0 100.0 100.0 Isbell et al., 1998b). Terrestrial primates tend to be larger than Colobus guereza 100.0 100.0 100.0 100.0 those that are arboreal, and increased body mass may confer Total % 90.6 88.9 88.3 87.3 advantages through protection from predation, increased ther- moregulatory efficiency, and buffering against food shortages. Home range area is often linked to habitat productivity, which of species that include misclassified individuals (16 compared to in many cases is a function of rainfall (Isbell et al., 1998b). nine in females). Although within species sample shape variance The large home range areas of patas monkeys are likely to is similar in the two sexes (female average variance ¼ 0.00299, be a response to relatively unproductive habitats (Isbell male average variance ¼ 0.00303, paired two-tailed t-test et al., 1998b). It has also been noted that E. patas is subject t22 ¼0.362, p ¼ 0.720), within species size variation is signif- to greater seasonality and a more unpredictable environment icantly larger in males (female average SD ¼ than is usually the case for guenons (Cords, 1987). Life history 3.2, male SD ¼ 5.5, paired two-tailed t-test t22 ¼4.194, strategies of E. patas appear to reflect this; it has a younger age p ¼ 0.0004). Larger variation in the size of males might be re- at first birth and shorter interbirth intervals than would be ex- lated to an extension of the growth period that contributes to sex- pected for a guenon of its body mass (Nakagawa et al., 2003). ual dimorphism and has as a by-product a higher frequency of If environmental selective pressures were intense enough to in- extreme phenotypes in species, like guenons, with a strong allo- fluence life history evolution they may well have also affected metric model of shape variation. These specimens are more body mass, which in turn appears to have contributed to the likely to be misclassified, as suggested by high percentages of divergent cranial morphology of E. patas. misclassified individuals of either sex (17.8% for males, The strong links between the extremes of shape and size di- 10.5% for females) who lay outside the 95% interval of within vergence suggest that allometry has played a central role in the species size variation. However, whether larger variation in evolution of guenon skull morphology. Indeed, in the species male size also contributes to increased uncertainties in clusters analysed, shape distances correlated most strongly with size of male mean shapes is less clear. variation, although shape was also significantly correlated Sampling error notwithstanding, the smallest and largest gue- with genetic distances. Correlations between shape and ge- non genera, Miopithecus and Erythrocebus, were consistently dis- netic distance became larger when size-independent shape var- tinguished from other guenon species, discriminated along the iation was analysed. This indicates that strongly divergent first two PCs of shape as well as being among the most divergent shapes that make similarity relationships inconsistent with in the partial disparity analysis. Miopithecus is at least as distant as phylogeny are partly a by-product of size divergence. Correla- the colobine outgroup when compared to the grand mean of all tions between size and shape were also much larger than those species for either sex. At the opposite extreme of size variation, between shape and ecology, reinforcing the importance of al- the skull of Erythrocebus patas is slightly less distinctive than lometry in determining skull shape in guenons. those of the Miopithecus species, but is still further from the grand Whether, or to what extent, guenon skull shape variation mean than is the skull of the papionin outgroup, C. atys. has been produced by simple extension or truncation of Table 5 Mahalanobis distances between centroids and Procrustes distances between species means are shown for females of all study species below and above the main diagonal, respectively. Abbreviation for species names are shown in Table 1 615 (2008) 54 Evolution Human of Journal / Elton S. Cardini, A. nig aet asc cam cep dia eryg* eryt* ham lhoe mit mon neg nic pet pog pre* scla* pat ogo tal* atys gue nig e 0.057 0.075 0.060 0.068 0.050 0.072 0.086 0.058 0.055 0.043 0.057 0.050 0.053 0.070 0.072 0.065 0.076 0.067 0.136 0.115 0.068 0.111 aet 7.36 e 0.058 0.049 0.050 0.043 0.058 0.060 0.065 0.042 0.049 0.043 0.050 0.044 0.050 0.056 0.052 0.060 0.064 0.119 0.095 0.075 0.100 asc 8.02 7.31 e 0.042 0.022 0.059 0.042 0.039 0.077 0.056 0.059 0.039 0.068 0.047 0.033 0.024 0.055 0.033 0.087 0.089 0.068 0.087 0.122 cam 7.28 7.36 5.17 e 0.037 0.049 0.049 0.057 0.057 0.050 0.051 0.024 0.051 0.044 0.039 0.040 0.055 0.052 0.075 0.101 0.082 0.070 0.116 cep 7.95 6.62 2.54 4.84 e 0.048 0.044 0.042 0.072 0.049 0.051 0.035 0.059 0.035 0.026 0.026 0.053 0.038 0.084 0.098 0.076 0.080 0.113 dia 7.07 5.97 5.92 5.69 4.89 e 0.062 0.067 0.061 0.042 0.028 0.046 0.037 0.025 0.048 0.059 0.055 0.061 0.066 0.131 0.108 0.076 0.096 eryg* 6.96 6.45 5.04 6.24 5.76 6.47 e 0.051 0.073 0.060 0.064 0.046 0.069 0.058 0.047 0.047 0.065 0.038 0.083 0.097 0.080 0.086 0.126 eryt* 8.76 6.82 4.13 5.51 4.52 6.31 5.40 e 0.092 0.064 0.070 0.053 0.079 0.058 0.044 0.045 0.064 0.042 0.097 0.098 0.074 0.100 0.118 ham 8.10 8.23 8.69 7.83 8.17 6.89 8.19 10.00 e 0.058 0.059 0.061 0.056 0.065 0.075 0.076 0.067 0.078 0.059 0.125 0.107 0.069 0.135 lhoe 7.53 5.82 6.62 7.09 5.88 5.32 7.17 7.09 6.47 e 0.043 0.048 0.051 0.043 0.054 0.057 0.042 0.058 0.068 0.125 0.103 0.071 0.110 mit 5.65 6.35 6.09 6.43 5.70 3.80 6.45 6.53 7.36 5.54 e 0.047 0.039 0.027 0.052 0.058 0.057 0.062 0.068 0.135 0.109 0.070 0.098 mon 6.92 6.34 4.81 3.31 4.73 5.21 5.62 5.39 8.30 6.97 5.84 e 0.052 0.039 0.039 0.033 0.052 0.045 0.072 0.102 0.081 0.073 0.115 neg 7.12 6.50 6.46 5.71 5.94 4.85 6.70 6.90 7.45 6.41 5.57 6.02 e 0.042 0.059 0.067 0.065 0.070 0.066 0.134 0.112 0.075 0.103 nic 6.81 5.88 4.97 5.44 3.83 2.87 6.39 5.55 7.80 5.39 3.18 4.79 4.84 e 0.038 0.047 0.055 0.052 0.074 0.124 0.099 0.074 0.098 pet 7.77 6.54 4.43 4.40 3.61 4.84 5.57 4.56 8.16 6.65 5.45 4.87 5.41 4.20 e 0.037 0.058 0.043 0.085 0.103 0.079 0.082 0.108 pog 7.98 7.45 3.39 4.98 3.71 6.46 6.25 5.29 9.26 7.49 6.66 3.55 6.97 5.32 5.08 e 0.059 0.041 0.087 0.091 0.070 0.086 0.122 pre* 6.96 6.50 6.47 6.74 6.05 5.81 7.33 7.43 6.74 4.46 5.86 6.46 7.21 5.96 7.12 7.24 e 0.063 0.076 0.112 0.090 0.085 0.123 scla* 8.47 7.17 3.75 6.53 4.79 6.22 3.75 4.34 8.89 6.68 6.45 5.74 7.04 5.65 5.33 5.47 7.50 e 0.087 0.099 0.077 0.091 0.123 pat 10.00 8.49 11.17 11.20 11.09 8.97 10.38 12.05 8.45 9.33 9.50 10.25 10.05 9.82 10.94 11.50 9.75 10.97 e 0.133 0.120 0.086 0.134 e ogo 13.40 13.48 10.09 10.56 10.50 13.28 11.37 11.21 13.58 14.22 14.13 10.75 13.20 12.96 11.28 9.80 12.47 11.58 15.60 e 0.065 0.141 0.190 637 tal* 11.05 10.59 7.48 8.13 8.03 10.90 9.01 8.26 10.88 11.48 11.28 8.43 10.78 10.46 8.65 7.48 9.85 9.04 13.75 6.02 e 0.125 0.160 atys 9.53 11.01 12.30 11.20 11.84 11.31 12.40 12.87 11.45 10.57 10.56 11.44 11.20 10.70 11.41 12.28 11.93 12.56 13.08 18.16 15.36 e 0.123 gue 14.70 13.06 14.63 14.03 14.32 13.03 14.88 13.33 16.28 14.34 13.18 14.41 13.50 13.41 12.95 15.22 15.30 14.77 16.79 20.41 17.43 14.66 e 631 632

Table 6 Mahalanobis distances between centroids and Procrustes distances between species means are shown for males of all study species below and above the main diagonal, respectively. Abbreviation for species names are shown in Table 1 615 (2008) 54 Evolution Human of Journal / Elton S. Cardini, A. nig aet asc cam cep dia eryg* eryt* ham lhoe mit mon neg nic pet pog pre* scla* pat ogo tal* atys gue nig e 0.061 0.085 0.062 0.084 0.058 0.089 0.068 0.067 0.051 0.049 0.062 0.050 0.063 0.080 0.078 0.059 0.089 0.076 0.149 0.144 0.074 0.097 aet 7.19 e 0.061 0.044 0.057 0.041 0.063 0.045 0.068 0.049 0.050 0.042 0.048 0.043 0.055 0.056 0.049 0.065 0.077 0.121 0.116 0.075 0.098 asc 8.69 6.81 e 0.043 0.023 0.058 0.036 0.035 0.091 0.072 0.068 0.047 0.072 0.044 0.031 0.028 0.069 0.039 0.116 0.086 0.083 0.096 0.116 cam 7.89 6.46 5.37 e 0.043 0.037 0.052 0.028 0.070 0.050 0.043 0.026 0.045 0.029 0.036 0.038 0.050 0.052 0.091 0.113 0.108 0.079 0.100 cep 8.66 6.51 2.76 5.12 e 0.057 0.033 0.032 0.092 0.072 0.069 0.043 0.070 0.041 0.030 0.029 0.068 0.040 0.116 0.091 0.087 0.091 0.112 dia 7.52 5.51 5.60 3.94 5.30 e 0.062 0.035 0.065 0.042 0.033 0.037 0.040 0.027 0.051 0.056 0.040 0.066 0.075 0.131 0.126 0.077 0.093 eryg* 9.16 7.21 4.81 6.23 4.40 6.07 e 0.041 0.094 0.078 0.073 0.051 0.075 0.050 0.038 0.043 0.072 0.045 0.117 0.093 0.089 0.099 0.119 eryt* 7.91 5.81 4.07 3.55 3.48 3.13 4.56 e 0.074 0.053 0.046 0.030 0.053 0.022 0.032 0.035 0.050 0.047 0.097 0.110 0.103 0.078 0.101 ham 8.41 8.26 9.55 7.41 9.59 7.06 9.68 7.77 e 0.054 0.060 0.080 0.059 0.072 0.090 0.092 0.065 0.102 0.071 0.149 0.142 0.068 0.133 lhoe 7.13 5.93 6.76 6.08 6.98 5.44 8.02 5.53 6.75 e 0.039 0.053 0.044 0.048 0.068 0.069 0.035 0.080 0.070 0.143 0.138 0.072 0.108 mit 6.12 5.38 6.20 5.51 6.46 4.39 6.76 4.71 7.78 5.86 e 0.044 0.036 0.037 0.061 0.065 0.042 0.079 0.069 0.143 0.136 0.070 0.097 mon 7.65 5.81 4.94 3.46 4.34 4.13 4.88 3.14 8.80 6.57 4.91 e 0.047 0.030 0.040 0.037 0.050 0.052 0.091 0.116 0.112 0.083 0.093 neg 6.77 6.33 6.57 4.72 6.32 5.22 7.11 5.59 7.48 6.76 5.15 5.04 e 0.045 0.065 0.067 0.045 0.078 0.068 0.141 0.137 0.075 0.097 nic 7.17 5.56 4.67 3.41 4.06 2.87 5.12 2.40 7.89 5.67 3.90 3.40 4.86 e 0.036 0.043 0.042 0.055 0.089 0.119 0.114 0.076 0.096 pet 8.48 6.68 5.12 3.79 4.80 5.23 5.32 3.98 8.62 7.13 5.93 4.65 5.74 3.87 e 0.032 0.064 0.042 0.111 0.098 0.093 0.092 0.102 pog 8.23 6.74 3.61 4.79 3.86 5.48 5.26 3.99 9.72 7.00 6.24 3.39 6.23 4.76 5.03 e 0.067 0.040 0.113 0.091 0.089 0.094 0.107 pre* 7.21 5.69 6.36 5.54 6.18 4.96 6.62 5.00 7.29 3.70 5.53 5.62 5.89 4.64 6.49 6.56 e 0.076 0.074 0.138 0.135 0.080 0.104 scla* 9.16 7.32 4.41 6.05 4.61 6.27 4.62 4.97 10.21 7.91 7.18 5.29 7.02 5.24 5.63 4.94 7.10 e 0.122 0.094 0.093 0.106 0.112 pat 9.52 7.82 10.78 9.74 11.03 8.36 10.65 9.62 8.56 8.72 8.00 9.40 8.82 9.40 10.19 10.74 9.23 11.32 e 0.178 0.175 0.095 0.126 e

ogo 13.28 11.83 8.60 10.30 9.43 11.69 9.77 10.58 13.38 12.75 12.81 10.34 11.99 11.30 10.31 8.58 11.94 10.06 15.21 e 0.044 0.155 0.175 637 tal* 12.04 10.88 8.43 9.16 9.08 10.65 9.71 9.53 11.54 11.66 11.82 9.76 11.12 10.35 9.09 8.44 11.35 10.13 14.11 4.71 e 0.146 0.170 atys 9.38 10.55 12.65 11.37 11.91 10.72 13.01 10.77 11.19 10.72 9.62 11.26 10.99 10.19 11.36 12.17 11.09 13.03 11.88 17.92 16.07 e 0.126 gue 13.05 12.79 14.30 13.20 13.79 12.70 14.86 12.96 15.58 14.84 12.75 13.15 13.07 12.63 12.34 13.73 14.70 14.51 15.94 18.90 17.24 14.15 e A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 633 a common allometric trajectory needs to be investigated using are weak, size and ecology are to some extent interrelated and ontogenetic series. The patterns of conserved allometries evi- it is difficult to partition their roles completely. They are also dent in guenons are strongly reminiscent of those found in associated with genetic distance, in that species that are less neotropical monkeys, although congruence with studies of pa- closely related might be expected to show greater differences pionins largely depends on whether static interspecific (Single- in size, ecology, and behaviour. The possibility of a combined ton, 2002) or ontogenetic (Collard and O’Higgins, 2001) contribution of ecology and evolutionary history is discussed allometries are considered. Comparisons of cranial trajectories below with respect to another, albeit moderately divergent, within papionins (Collard and O’Higgins, 2001; Singleton, guenon species, C. aethiops. 2002) and apes (Mitteroecker et al., 2004) demonstrate the It is apparent that genetic distance alone accounts for some presence of important size-related shape changes within pri- of the shape differences in guenon skulls. Not only are signifi- mates but do not support hypotheses of simple extension/trun- cant correlations of shape with genetic distances evident, but cation of a common allometric line in the taxa studied. In divergence of species mean shapes also increases with evolu- guenons, however, small divergence of shape trajectories (Car- tionary distance. In females, relative distances to the grand dini and Elton, 2008), correlations between magnitudes of size mean of all species were 18.2 for the colobine C. guereza, and shape sexual dimorphism, extreme divergence of shape in 7.4 for the papionin C. atys, 4.0 for the basal A. nigroviridis, species at extremes of size variation, and phenetic clusters mir- and below 3 for most other guenons. A similar observation of roring size groups suggest that, as in South American monkeys shape divergence being roughly proportional to evolutionary (Marroig and Cheverud, 2005), a large component of shape time can also be made for males. It is possible that the use of variation in guenon skulls has been produced following partial disparity (Zelditch et al., 2004), which measures dis- a path of least evolutionary resistance (Schluter, 2000) set tances between each mean shape and their grand mean, may in- by a strong covariation among allometric traits. In other troduce a bias if a large number of species are ‘clumped’ in words, size (probably selected for by dietary and other ecolog- a similar body mass range (in this case, small-to-medium), pro- ical factors, such as predation) has helped to determine mor- ducing a grand mean closer to the means of species in that phological change. This biasing of the direction of range. To avoid this, the average of all pairwise distances for evolutionary shape modification may have resulted in the con- each taxon can be used instead. When compared to partial dis- served morphology evident in many guenon taxa. Such biases, parity, the contribution of each species to total disparity was however, decay with time, and shape variation is progressively more ‘evenly’ spread across taxa, ranging from 2.6e2.5% to less constrained as the structure of trait covariance is modified 11.2e10.4% (females-males, respectively). Thus, the emphasis by selective pressures (Renaud et al., 2006). Thus, for in- on highly divergent species is indeed reduced. However, since stance, in partial disparity analysis the large shape divergence the metric has a correlation of 1.000 with partial disparity, its of A. nigroviridis, basal to all guenons (Tosi et al., 2005), has use does not change the general patterns observed in this study. more to do with phylogeny than with its size, which is about Analysis of the whole sample (including all guenon taxa average for a guenon. This could indicate that constraints on and both outgroups) in the shape space and size-shape space shape modification have lessened over time in Allenopithecus. indicated a good separation of the outgroup species. Phylog- On the other hand, the highly distinctive Miopithecus is both eny is likely to account for the discrimination of at least one a probable basal lineage and the smallest guenon. Its shape di- of the outgroup species, Colobus guereza. An expanded gonial vergence is likely to be the product of not only a long evolu- region of the mandible, short nasal bones, high vaulted cra- tionary history but also of strong selective pressures leading to nium, and larger interorbital distance are found in C. guereza a reduction in size and consequent allometric shape modifica- and most other colobines (Strasser and Delson, 1987). These tions of the skull. characters are considered ancestral compared to the derived condition, associated with lengthening of the face, seen in cer- The contributions of evolutionary distance and ecology copithecines (Verheyen, 1962; Strasser and Delson, 1987). Within the cercopithecines, papionins have a more elongated There are no compelling ecological and behavioural rea- face than guenons (Szalay and Delson, 1979). This trend of fa- sons to expect that guenons would be particularly conservative cial elongation is partly related to size and is especially evi- in their patterns of cranial and mandibular shape variation. dent in the largest African monkeys, the baboons (Papio) Members of the clade exploit a wide variety of ecological and mandrills (Mandrillus). Cercocebus atys, the papionin out- niches and exhibit considerable behavioural diversity. It is group in our analysis, is smaller than Papio and Mandrillus, thus surprising that small matrix correlations between shape and is more comparable in size to E. patas, but does not and ecology (and also, not shown, between size and ecology) have a particularly long face (which may explain why it is suggest no strong convergences in relation to diet, habitat, and less divergent than might be expected for a papionin). How- locomotion in the guenon clade as a whole. This is not to say ever, a relatively small neurocranium and deep mandible, that individual species fail to show skull features that have which are not associated with short nasals as in colobines, arisen in response to specific ecological and behavioural pres- seem to be among the characters that discriminate this species sures; ecology, for example, is likely have contributed directly from the others. or indirectly (via size) to skull shape in Erythrocebus and Mio- Within the guenon clade, C. hamlyni was clearly separated pithecus. Even if correlations between skull form and ecology on the first two PCs of the shape space, as well as to a certain 634 A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 degree on the second PC of the size-shape space (ssPC2). This 2005) does not simplistically equate with common ecological, distinctiveness was emphasised by its divergent shape as indi- biological, and behavioural responses to terrestriality. Re- cated by partial disparity and its isolation in the similarity search into the morphological variation of vervet skulls indi- analysis. Since C. hamlyni, the owl-faced monkey, is an aver- cates that spatial and environmental factors, particularly age-sized arboreal guenon, size alone is unlikely to account for rainfall, contribute to variation in both size and shape, al- the increased degree of morphological divergence in its skull. though a significant proportion of variance remains unex- C. hamlyni is unusual in that molecular data do not support its plained (Cardini et al., 2007). It is thus possible that vervet inclusion in either of the two major clades of arboreal guenons ecology has influenced skull shape, but further research into (Tosi et al., 2005). It might, therefore, be a representative of its ecomorphologydunfortunately beyond the scope of the the early radiation of the arboreal lineage. Such early separa- present studydis necessary to explore this hypothesis. tion from other arboreal species could help to explain its mor- Within the overall pattern of conserved allometric changes phological distinctiveness. Modern populations of C. hamlyni in guenon skull shape, it is clear that neither genetic distances are mostly restricted to a relatively small number of mountain (as a proxy for phylogeny) nor ecology provide simple expla- forests in eastern Congo, and the species is presently classified nations for the extreme divergence evident in some species, as near threatened or vulnerable, with one subspecies (C. h. ka- although their consideration does shed light on some morpho- huziensis) listed as endangered (Kingdon, 1997; IEA, 1998; logical differences, such as those between the basal Allenopi- Butynski and Members of the Primate Specialist Group, thecus and the rest of the tribe. Although complexities are 2000). Whether genetic bottlenecks might have contributed to introduced by their interrelationships with each other and fixation of autoapomorphic traits cannot be said until more is with size, ecology and evolutionary history remain important known about the genetics and evolutionary history of this spe- components in understanding at least some aspects of morpho- cies. Similarly, more information is required on its ecology and logical differentiation in guenons. The work reported here behaviour to assess whether its distinctive skull shape has highlights instances in which morphological divergence can- arisen through social or ecological pressures. not be attributed mostly to size. Crucially, these will help to Another species that appeared distinctive in the size-shape frame future research projects focusing on if, how, and why space (separated on ssPC3 and ssPC4 from the outgroup spe- ecological or phylogenetic differences contribute to skull mor- cies and the other guenons) was Cercopithecus aethiops.How- phology in specific taxa such as C. aethiops and C. hamlyni. ever, this species did not appear particularly divergent in any of the other analyses. C. aethiops is a relatively common spe- Species discrimination cies with a widespread distribution across Africa, so is well- represented in museum collections. Its sample size was larger The results of the phenetic analyses show that there is ap- than those of the other species, and it was therefore possible preciable shape differentiation in a number of guenon taxa, that it weighted the derivation of the main axes of the size- specifically Allenopithecus and Miopithecus, whose distinctive shape space. As with the discriminant analysis, a random morphologies have been noted in previous studies (Verheyen, subsample of C. aethiops specimens was used in place of 1962; Martin and MacLarnon, 1988; Shea, 1992), as well as E. the entire C. aethiops sample to ascertain whether this was patas, C. aethiops, and C. hamlyni. Identifying such extremes the case. The results (not shown) of this analysis showed of variation, however, does not help to determine whether that the separation of C. aethiops from other guenons was smaller but nonetheless informative differences can be found less defined than when the larger sample was used, and oc- in the skulls of other guenons, particularly those in the genus curred along the fourth ssPC rather than the first. Nonetheless, Cercopithecus. Good discrimination of guenon species based overall separation remained good, verifying its distinctiveness on skull shape was not an a priori expectation of this study. in the size-shape space. Other than in size, most species look very similar and diagnos- The C. aethiops face is relatively more elongated and the tic characters are virtually absent. Indeed, previous research cranial vault less enlarged than in many other guenons. Such (Verheyen, 1962) found only modest differentiation within Cer- divergence in skull shape might be attributable to a relatively copithecus, and in the present study, inspection of first axes of long history of genetic isolation, as indicated by branch shape (or size-shape) variation also showed relatively poor sep- lengths in Tosi et al.’s (2005) study. It is also possible that aration of most species. Nonetheless, percentages of correctly its morphological divergence from other species is somehow classified specimens according to species were high in the related to its distinct ecology. Vervet monkeys are highly eco- main analyses and the hold-out samples, regardless of which logically flexible, with eclectic diets and the ability, uncom- method was used for species discrimination. This indicates mon in other guenons, to inhabit marginal and disturbed that small quantitative differences that are large enough to dis- habitats (Fedigan and Fedigan, 1988). The vervet postcranium criminate species might well be present. It is highly likely that does not show the extreme adaptations to terrestriality evident a large number of morphometric descriptors, such as the 86 in the other terrestrial guenons, C. lhoesti and E. patas (Gebo landmarks measured on the cranium and mandible in this study, and Sargis, 1994). Also unlike the open-habitat adapted patas are necessary to pick up these slight but significant variations. monkey, vervets are medium-sized, with only moderate sexual All species with the largest errors in species identification dimorphism (Fedigan and Fedigan, 1988). It is apparent that belong to the arboreal guenon clade. This clade is approxi- the monophyly of the terrestrial guenons (Tosi et al., 2004, mately the same age as the terrestrial clade (4.6 versus 4.8 A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 635

Ma, according to the molecular clock of Tosi et al., 2005), but within the clade. One major question that emerges is why gue- has given rise to many more species, some probably through non skulls appear more evolutionarily conserved than those of rapid differentiation during the forest fragmentation of the their sister taxon, the papionins. An obvious difference Plio-Pleistocene (Hamilton, 1988). Cercopithecus ascanius in- between the two is the presence or absence of an elongated cludes several specimens of either sex which are misclassified muzzle. It is possible that whereas the largely terrestrial, as C. cephus and some females that are misclassified as C. er- open-habitat papionins responded to ecological and behaviou- ythrotis. These three species belong to the same terminal clade ral pressures through elaborating hard tissue features, leading of arboreal guenons in Tosi et al.’s (2005) molecular phylog- to divergent skull morphologies within the clade, the primarily eny. Thus, both morphological and genetic data suggest very arboreal guenons benefited more from altering facial coloura- close relationships among them. However, misclassification tion, soft tissue traits like whiskers, and vocalisations. Changes did not always fall along phylogenetic lines. Specimens of such as these may have been advantageous in relatively low- C. pogonias and C. mona were misclassified into several spe- light forest conditions, where hard tissue features might be cies in a pattern that had no clear relationship with phylogeny. less obvious than bright colouration or distinctive calls. None- Unexpectedly, C. campbelli and C. mona are not easily con- theless, the majority of guenon specimens included in our fused in discriminant analysis, despite strong similarities sample were correctly classified to species in discriminant (see phenograms). This suggests at least a moderate degree analyses. This suggests that most population boundaries indi- of morphological divergence between parapatric populations cated by vocalisations and soft tissue characters are also of these closely related species. reflected in hard tissue morphology, even though a large num- Misclassified specimens of C. diana are generally found in ber of accurate morphometric descriptors may be needed to C. nictitans, although the two species are not closely related. pick up these subtle differences. The significant correlation be- C. nictitans is the sister species of C. mitis, and although C. tween genetic distance and shape, although not as strong as the mitis is not extensively misclassifed, such specimens are found correlation between size and shape, indicates that phylogeny in C. nictitans. The molecular phylogeny of Tosi et al. (2005) influences skull shape within the guenons. The evolutionary suggests that C. mitis albogularis is actually more closely re- history of the clade has yet to be elucidated fully, hampered lated to C. nictitans than to other subspecies of C. mitis. This as it is by the paucity of the guenon fossil record. However, might explain why several specimens of C. mitis are classified further molecular analyses, particularly phylogeographic stud- as C. nictitans (up to five among females and up to three ies, could provide important insights into guenon speciation among males). However, among eight individuals classified and population genetics, which in turn could be mapped as C. mitis albogularis in museum databases, only one is mis- onto observed differences in hard tissue morphology. This classified as C. nictitans in the discriminant analysis. The lim- type of approach might help us to understand whether the dis- ited information on subspecies available in museums and the tinctive morphology of C. hamlyni, for example, is the product problematic taxonomy of the C. mitis-nictitans group (Napier, of stochastic events in its evolutionary history, or whether it is 1981) raise issues that need to be addressed through further re- due to ecological and social pressures. search on arboreal guenons. This is reinforced by the observa- Of all the factors included in our study that would potentially tion that misclassified specimens of C. nictitans are spread influence morphological differentiation, ecology was the least across four to six loosely-related species. significant. Given the inter- and intraspecific ecological varia- Among species in which females and males are misclassified tions in the tribe, this was unexpected. It is probable that ecol- most frequently, less than half of the wrongly-assigned speci- ogy does influence skull shape, and hints about this are given by mens seem to be found in the corresponding sister species. the divergence of not only E. patas and Miopithecus, that have This can be ascribed to short time since speciation. Overall, di- obviously distinct ecologies, but also by the modest differenti- versification of guenon skulls is modest, and is especially small ation of the eurytopic and geographically widespread C. ae- among arboreal species, but populations identified on the basis thiops. It is possible that the ‘broad brush’ categorisation of of pelage colour and geographic distribution do correspond to ecologies used in this study conceals important yet subtle dis- morphologically divergent lineages. In the broader context of tinctions between most taxa. Parallels can be drawn here with species recognition, morphological divergence must not be the discriminant analysis, in which a large number of detailed seen as a necessary property for inferring species boundaries, morphological descriptors were required to classify species. but as a line of evidence that needs to be compared with results Field research has provided a wealth of information about gue- from studies using other data and methods, that may either sup- non ecology as well as behaviour, but detailed studies of some port or contradict the hypothesis that a morphologically diver- of the more cranially distinctive taxa, like Miopithecus and C. gent population represents a ‘good species’ (de Queiroz, 2005). hamlyni, are yet to be undertaken, and more data on intraspe- cific variation are also required. Nonetheless, enough informa- Conclusions tion is currently available on well-studied species such as the vervet to address in more detail the links between small varia- Our analyses suggest that guenon skulls show an allometri- tions in ecology and skull shape. The future integration of de- cally conserved pattern of shape differentiation, but it is appar- tailed ecological, behavioural, and morphological data would ent that much remains to be discovered about the more subtle be highly beneficial to our understanding of the adaptations links between morphology, phylogeny, ecology, and behaviour and diversification of the guenons. 636 A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637

Acknowledgements Butynski, T.M., 2002. The guenons: an overview of diversity and taxonomy. In: Glenn, M.E., Cords, M. (Eds.), The Guenons: Diversity and Adaptation in Af- e A large number of people have contributed in different rican Monkeys. Kluwer Academic/Plenum Publishers, New York, pp. 3 13. Butynski, T., Members of the Primate Specialist Group, 2000. Cercopithecus ways to this study. We are deeply grateful to all of them, hamlyni ssp. kahuziensis. In: IUCN 2006 (Ed.), 2006 IUCN Red List of and apologise to those whom we might have forgotten to ac- Threatened Species. Available online at: www.iucnredlist.org. Downloaded knowledge. Special thanks are due to P. O’Higgins and on May 02, 2007. M. Bastir, both at the University of York; S. Cobb, University Cardini, A., Elton, S., 2008. Variation in guenon skulls (II): sexual dimor- e of Hull; C.E. Oxnard, University of Western Australia; S. Mar- phism. J. Hum. Evol. 54 (5), 638 647. Cardini, A., Elton, S., Skull functional and developmental modules: mosaic telli, University College London; E. Delson, City University of evolution and phylogenetic signal. Biol. J. Linn. Soc., in press. New York; S.R. Frost, University of Oregon; C.P. Groves, Cardini, A., Jansson, A.-U., Elton, S., 2007. Ecomorphology of vervet mon- Australian National University; C.J. Jolly and T.R. Disotell, keys: a geometric morphometric approach to the study of clinal variation. New York University, New York; F.J. Rohlf, State University J. Biogeogr. 34, 1663e1678. of New York; C.P. Klingenberg, University of Manchester, Chapman, C.A., Chapman, L.J., Cords, M., Mwangi Gathua, J., Gautier- Hion, A., Lambert, J.E., Rode, K., Tutin, C.E.G., White, L.J.T., 2002. Var- Manchester; P. Mitteroecker, F. L. Bookstein and K. Schaefer, iation in the diets of Cercopithecus species: differences within forests, University of Vienna; R.W. Thorington, Jr., L. Gordon, C.A. among forests, and across species. In: Glenn, M.E., Cords, M. (Eds.), Ludwig, J.G. Mead, C. Potter, D.F. Schmidt, A.L. Gardner, The Guenons: Diversity and Adaptation in African Monkeys. Kluwer C.A. Shaw, and M. Sitnik, all at the National Museum of Nat- Academic/Plenum Publishers, New York, pp. 325e350. ural History, Washington; N. Gilmore, Academy of Natural Cheverud, J.M., Richtsmeier, J.T., 1986. Finite element scaling applied to sex- ual dimorphism in rhesus macaque (Macaca mulatta) facial growth. Syst. Sciences, Philadelphia; J. Spence, T. Pacheco, and P.A. Bruna- Zool. 35, 381e399. uer, all at the American Museum of Natural History, New Collard, M., O’Higgins, P., 2001. Ontogeny and homoplasy in the papionin York; J.M. Chupasko and M.D. Omura, Museum of Compar- monkey face. Evol. Dev. 3, 322e331. ative Zoology of the Harvard University, Cambridge; M. Schu- Cords, M., 1987. Forest guenons and patas monkeys: male-male competition lenberg and J. Fooden, Field Museum, Chicago; R. Asher, in one-male groups. In: Smuts, B.B., Cheney, D.L., Seyfarth, R.M., Wrangham, R.W., Struhsaker, T. (Eds.), Primate Societies. University of I. Thomas, and R. Sastrawan, Museum fu¨r Naturkunde, Berlin; Chicago Press, Chicago, pp. 98e111. O. Ro¨her-Ertl, Staatliche Naturwissenschaftliche Sammlugen Curtin, S.H., 2002. Diet of the roloway monkey (Cercopithecus diana rolo- Bayerns, Munich; C. Rovati and S. Maretti, Museo di Storia way) in Bia National Park, Ghana. In: Glenn, M.E., Cords, M. (Eds.), Naturale di Pavia, Pavia; C.P.E. Zollikofer, M. Zefferer, The Guenons: Diversity and Adaptation in African Monkeys. Kluwer e M. Bragger, T. Weingrill, M. Gisi, and C. Zebig-Brunner, all Academic/Plenum Publishers, New York, pp. 351 371. ¨ de Queiroz, K., 2005. Different species problems and their resolution. BioEs- the University of Zurich; W. Van Neer and W. Wendelen, says 27, 1263e1269. Royal Museum for Central Africa, Tervuren; P. Jenkins, D. Dryden, I.L., Mardia, K.V., 1998. Statistical Shape Analysis. John Wiley and Hills and L. Tomsett British Museum of Natural History, Sons, New York. London; and J. Harrison and M. Harman, Powell-Cotton Mu- Elton, S., 2007. Environmental correlates of the cercopithecoid radiations. e seum, Birchington. Folia Primatol. 78, 344 364. Fedigan, L., Fedigan, L.M., 1988. Cercopithecus aethiops: a review of field We also very much appreciate the comments of the editor studies. In: Gautier-Hion, A., Bourliere, F., Gautier, J.-P., Kingdon, J. Susan Anto´n, Becky Ackermann, and several anonymous ref- (Eds.), A Primate Radiation: Evolutionary Biology of the African erees in improving this work, and thank them for their time Guenons. Cambridge University Press, Cambridge, pp. 394e411. and trouble. We are particularly grateful to one reviewer for Frost, S.R., Marcus, L.F., Bookstein, F.L., Reddy, D.P., Delson, E., 2003. Cra- their advice on and suggestions about how to recalculate the nial allometry, phylogeography, and systematics of large-bodied papionins (primates: cercopithecinae) inferred from geometric morphometric analy- partial disparity analysis so that the ‘clumped’ means did sis of landmark data. Anat. Rec. 275A, 1048e1072. not overly influence the analysis. Gautier-Hion, A., 1978. Food niches and coexistance in sympatric primates in This paper is dedicated to the memory of Walter N. Ver- Gabon. In: Chivers, D.J., Herbert, J. (Eds.), Recent Advances in Primatol- heyen in recognition of his great contribution to the under- ogy, vol. II. Academic Press, New York, pp. 270e286. standing of the evolution of form in African monkeys. We Gautier-Hion, A., 1988a. The diet and dietary habits of forest guenons. In: Gautier-Hion, A., Bourliere, F., Gautier, J.-P., Kingdon, J. (Eds.), A Pri- regret that the author of one of the most cited studies on cer- mate Radiation: Evolutionary Biology of the African Guenons. Cambridge copithecine monkey cranial variation, who was of extraordi- University Press, Cambridge, pp. 257e283. nary help to us with his advice and friendship in occasion of Gautier-Hion, A., 1988b. Polyspecific associations among forest guenons: eco- incidental, but unforgettable, meetings during early stages of logical, behavioural and evolutionary aspects. In: Gautier-Hion, A., this work, will not be able to see the outcome of a research Bourliere, F., Gautier, J.-P., Kingdon, J. (Eds.), A Primate Radiation: Evo- lutionary Biology of the African Guenons. Cambridge University Press, that he greatly helped to inspire. Cambridge, pp. 453e476. This study was funded by a grant from the Leverhulme Trust. Gebo, D.L., Sargis, E.J., 1994. Terrestrial adaptations in the postcranial skel- etons of guenons. Am. J. Phys. Anthropol. 93, 341e371. Grubb, P., Butynski, T.M., Oates, J.F., Bearder, S.K., Disotell, T.R., References Groves, C.P., Struhsaker, T.T., 2003. Assessment of the diversity of African primates. Int. J. Primatol. 24, 1301e1357. Adams, D.C., Slice, D.E., Rohlf, F.J., 2004. Geometric morphometrics: ten Hair, J.H., Anderson, R.E., Tatham, R.L., Black, W.C., 1998. Multivariate Data years of progress following the ‘revolution’. Ital. J. Zool. 71, 5e16. Analysis. Prentice Hall, Upper Saddle River, New Jersey. Bookstein, F.L., 1991. Morphometric Tools for Landmark Data. Cambridge Hamilton, A.C., 1988. Guenon evolution and forest history. In: Gautier- University Press, Cambridge, Massachusetts. Hion, A., Bourliere, F., Gautier, J.P., Kingdon, J. (Eds.), A Primate Radi- A. Cardini, S. Elton / Journal of Human Evolution 54 (2008) 615e637 637

ation: Evolutionary Biology of the African Guenons. Cambridge Univer- Mitteroecker, P., Gunz, P., Bookstein, F.L., 2005. Heterochrony and geometric sity Press, Cambridge, pp. 13e34. morphometrics: a comparison of cranial growth in Pan paniscus versus Harmon, L.J., Kolbe, J.J., Cheverud, J.M., Losos, J.B., 2005. Convergence and Pan troglodytes. Evol. Dev. 7, 244e258. the multidimensional niche. Evolution 59, 409e421. Nakagawa, N., 1999. Differential habitat utilization by patas monkeys (Eryth- IEA, 1998. African Mammals Databank e a Databank for the Conservation rocebus patas) and tantalus monkeys (Cercopithecus aethiops tantalus) and Management of the African Mammals, vols. 1e2. Report to the Direc- living sympatrically in northern Cameroon. Am. J. Primatol. 49, torate-General for Development (DGVIII/A/1) of the European Commis- 243e264. sion. Project No. B7e6200/94-15/VIII/ENV. Bruxelles. Nakagawa, N., Hideyuki Ohsawa, H., Muroyama, Y., 2003. Life-history pa- Isbell, L.A., 1998. Diet for a small primate: insectivory and gummivory in the (large) rameters of a wild group of West African patas monkeys (Erythrocebus patas monkey (Erythrocebus patas pyrrhonotus).Am.J.Primatol.45,381e398. patas patas). Primates 44, 281e290. Isbell, L.A., Pruetz, J.D., Lewis, M., Young, T.P., 1998b. Locomotor activity Napier, P.H., 1981. Catalog of Primates in the British Museum (Natural His- differences between sympatric patas monkeys (Erythrocebus patas) and tory) and Elsewhere in the British Isles, Part II: Family Cercopithecinae, vervet monkeys (Cercopithecus aethiops): implications for the evolution Subfamily Cercopithecinae. British Museum (Natural History), London. of long hindlimb length in Homo. Am. J. Phys. Anthropol. 105, 199e207. Nunn, C.L., 2002. A comparative study of leukocyte counts and disease risk in Isbell, L.A., Pruetz, J.D., Young,T.P., 1998a. Movements of vervets (Cercopithe- primates. Evolution 56, 177e190. cus aethiops) and patas monkeys (Erythrocebus patas) as estimators of food Nunn, C.L., Altizer, S.M., Sechrest, W., Jones, K.E., Barton, R.A., resource size, density and distribution. Behav. Ecol. Sociobiol. 42, 123e133. Gittleman, J.L., 2004. Parasite species richness and the evolutionary diver- Jernvall, J., Wright, P., 1998. Diversity components of impending primate sity of primates. Am. Nat. 162, 597e614. extinctions. Proc. Natl. Acad. Sci. 95, 11279e11283. O’Higgins, P., Collard, M., 2001. Sexual dimorphism and facial growth in Kendall, D.G., 1984. Shape manifolds, Procrustean metrics and complex pro- papionin monkeys. J. Zool., Lond. 257, 255e272. jective spaces. Bull. Lond. Math. Soc. 16, 81e121. O’Higgins, P., Jones, N., 1999. Morphologika. Tools for Shape Analysis. Kingdon, J.S., 1997. The Kingdon Field Guide to African Mammals. University College London, London. Academic Press, London. Polly, P.D., 2005. Development and phenotyopic correlations: the evolution of Klingenberg, C.P., 1996. Multivariate allometry. In: Marcus, L.F., Corti, M., tooth shape in Sorex araneus. Evol. Dev. 7, 29e41. Loy, A., Naylor, G.J.P., Slice, D.E. (Eds.), Advances in Morphometrics. Renaud, S., Auffray, J.-C., Michaux, J., 2006. Conserved phenotypic variation Plenum Press, New York, pp. 23e49. patterns, evolution along lines of least resistance, and departure due to Klingenberg, C.P., 1998. Heterochrony and allometry: the analysis of evolu- selection in fossil rodents. Evolution 60, 1701e1717. tionary change in ontogeny. Biol. Rev. 73, 79e123. Rohlf, F.J., 1998. On applications of geometric morphometrics to study of Klingenberg, C.P., Monteiro, L.R., 2005. Distances and directions in multidi- ontogeny and phylogeny. Syst. Biol. 47, 147e158. mensional shape spaces: implications for morphometric applications. Syst. Rohlf, F.J., 2003. TpsSmall. Department of Ecology and Evolution, State Uni- Biol. 54, 678e688. versity of New York, Stony Brook, New York. Available online at: http:// Kozak, K.H., Larson, A., Bonnett, R.M., Hramon, L., 2005. Phylogenetic anal- life.bio.sunysb.edu/morph/ ysis of ecomorphological divergence, community structure, and diversifi- Rohlf, F.J., 2005. NTSYSpc, Version 2.20f. Exeter Software, Setauket, New York. cation rates in dusky salamanders (Plethodontidae: Desmognathus). Rohlf, F.J., Marcus, L.F., 1993. A revolution in morphometrics. Trends Ecol. Evolution 59, 2000e2016. Evol. 8, 129e132. Leakey, M., 1988. Fossil evidence for the evolution of the guenons. In: Gaut- Rohlf, F.J., Slice, D.E., 1990. Extensions of the Procrustes method for the ier-Hion, A., Bourliere, F., Gautier, J.P., Kingdon, J. (Eds.), A Primate optimal superimposition of landmarks. Syst. Zool. 39, 40e59. Radiation: Evolutionary Biology of the African Guenons. Cambridge Uni- Schluter, D., 2000. The Ecology of Adaptive Radiation. Oxford University versity Press, Cambridge, pp. 7e12. Press, New York. Legendre, P., Legendre, L., 1998. Numerical Ecology, second ed. Elsevier, Shea, B.T., 1992. Ontogenetic scaling of skeletal proportions in the talapoin Amsterdam, pp. 552e560. monkey. J. Hum. Evol. 23, 283e307. Leigh, S.R., 2006. Cranial ontogeny of Papio baboons (Papio hamadryas). Singleton, A.M., 2002. Patterns of cranial shape variation in the Papionini Am. J. Phys. Anthropol. 130, 71e84. (Primates: Cercopithecinae). J. Hum. Evol. 42, 547e578. Leigh, S.R., Shah, N.F., Buchanan, L.S., 2003. Ontogeny and phylogeny in Slice, D.E., 1999. Morpheus (beta version). Department of Ecology and papionin primates. J. Hum. Evol. 45, 285e316. Evolution, State University of New York, Stony Brook, New York. Marcus, L.F., Hingst-Zaher, E., Zaher, H., 2000. Application of landmark morpho- SPSS for Windows, 2002. SPSS Inc., Version 11.5.0. SPSS Inc., Chicago, Il- metrics to skulls representing the orders of living mammals. Hystrix 11, 27e48. linois, USA. Marroig, G., Cheverud, J.M., 2001. A comparison of phenotypic variation and co- Strasser, E., Delson, E., 1987. Cladistic analysis of cercopithecid relationships. variation patterns and the role of phylogeny, ecology, and ontogeny during cra- J. Hum. Evol. 16, 81e99. nial evolution of New World monkeys. Evolution 55 (12), 2576e2600. Szalay, F.S., Delson, E., 1979. Evolutionary History of the Primates. Academic Marroig, G., Cheverud, J.M., 2005. Size as a line of least evolutionary resis- Press, New York. tance: diet and adaptive morphological radiation in New World monkeys. Tosi, A.J., Detwiler, K.M., Disotell, T.R., 2005. X-chromosomal window into Evolution 59 (5), 1128e1142. the evolutionary history of the guenons (Primates: Cercopithecini). Mol. Martin, R.D., MacLarnon, A.M., 1988. Quantitative comparisons of the Phylogenet. Evol. 36, 58e66. skull and teeth in guenons. In: Gaultier-Hion, A., Bourliere, F., Tosi, A.J., Melnick, D.J., Disotell, T.R., 2004. Sex chromosome phylogenetics Gautier, J.P., Kingdon, J. (Eds.), A Primate Radiation: Evolutionary Bi- indicate a single transition to terrestriality in the guenons (tribe Cercopi- ology of the African Guenons. Cambridge University Press, Cambridge, thecini). J. Hum. Evol. 46, 223e237. pp. 160e183. Verheyen, W.N., 1962. Contribution a la craniologie comparee des primates. Martins, E., 1999. Estimation of ancestral states of continuous characters: Les genres Colobus Illiger 1811 et Cercopithecus Linne 1758. Musee a computer simulation study. Syst. Biol. 48, 642e650. Royal de l’Afrique Centrale, Tervuren, Annales 8 (105), 1e255. McNulty, K.P., 2004. A geometric morphometric assessment of hominoid cra- Wood, B.A., Richmond, B., 2000. Human evolution: taxonomy and paleobiol- nia: conservative African apes and their liberal implications. Ann. Anat. ogy. J. Anat. 97, 19e60. 186, 429e433. Xing, J., Wang, H., Zhang, Y., Ray, D.A., Tosi, A.J., Disotell, T.R., McNulty, K.P., Frost, S.R., Strait, D.S., 2006. Examining affinities of the Batzer, M.A., 2007. A mobile element based evolutionary history of Taung child by developmental simulation. J. Hum. Evol. 51, 1e23. guenons (tribe Cercopithecini). BMC Biol. 5, 5. Mitteroecker, P., Gunz, P., Bernhard, M., Schaefer, K., Bookstein, F.L., 2004. Zelditch, M.L., Swiderski, D.L., Sheets, H.D., Fink, W.L., 2004. Geometric Comparison of cranial ontogenetic trajectories among great apes and Morphometrics for Biologists: a Primer. Elsevier Academic Press, San humans. J. Hum. Evol. 46, 679e697. Diego.