Biological Journal of the Linnean Society, 2010, ••, ••–••. With 6 figures

Size variation facilitates population divergence but does not explain it all: an example study from a widespread African monkey SARAH ELTON1*, JASON DUNN1 and ANDREA CARDINI1,2,3 1

Hull York Medical School, The University of Hull, Cottingham Road, Hull HU6 7RX, UK Dipartimento del Museo di Paleobiologia e dell’Orto Botanico, Universitá di Modena e Reggio Emilia, via Università 4, Modena 41100, Italy 3 Centre for Forensic Science, The University of Western Australia, 35 Stirling Highway, Crawley WA 6009, Australia 2

Received 14 December 2009; revised 25 May 2010; accepted for publication 25 May 2010

bij_1504

1..21

Subspecific variation is widespread in vertebrates. Within Africa, several mammals have extensive geographic distributions with attendant morphological, ecological, and behavioural variations, which are often used to demarcate subspecies. In the present study, we use a primate species, the vervet monkey, Cercopithecus aethiops, as a case study for intraspecific divergence in widespread mammals, assessed through hard tissue morphology. We examine intraspecific differences in size, shape, and non-allometric shape from a taxonomic perspective, and discuss the macroevolutionary implications of findings from microevolutionary analyses of geographic variation. A geometric morphometric approach was used, employing 86 three-dimensional landmarks of almost 300 provenanced crania. Many of the taxonomic differences in skull morphology between vervet populations appear to be related to geographic proximity, with subspecies at opposite extremes of a west-to-east axis showing greatest divergence, and populations from central and south Africa being somewhat intermediate. The classification rate from discriminant analyses was lower than that observed in other African primate radiations, including guenons as a whole and red colobus. Nonetheless, taxonomic differences in shape were significant and not simply related to either geography or size. Thus, although shifts in size may be an important first step in adaptation and diversification, with size responding more quickly than shape to environmental change, the six vervet taxa currently recognized (either as species or subspecies) are not simply allometrically scaled versions of one another and are probably best viewed as subspecies. Holding allometry constant when examining inter-population differences in shape may thus help to reveal the early stages of evolutionary divergence. The vervet case study presented here hence has relevance for future studies examining intraspecific differentiation in other large mammals, particularly through the methods used to identify small but biologically meaningful divergence, with attendant implications for conservation planning. © 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••.

ADDITIONAL KEYWORDS: African mammals – allometric shape – Cercopithecus – Chlorocebus – conservation – geography – guenons – Old World monkeys – taxonomy – vervets.

INTRODUCTION Subspecific variation is widespread in vertebrates. The majority of new taxonomic descriptions in well-described groups such as the birds are at the

*Corresponding author. E-mail: [email protected]

subspecific level (Winston, 1999). Examining subspecific variation can aid decisions about conservation priorities (Phillimore & Owens, 2006; Gippoliti & Amori, 2007; Cardini et al., 2009; Cardini & Elton, 2009a), because populations may show differences in morphology (Forsman, 1997), behaviour (Macedonia & Taylor, 1985; Ryan & Wilczynski, 1991), life history (Ballinger, 1979), and physiology (Tracey & Walsberg,

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

1

2

S. ELTON ET AL.

2000). In many widespread mammal groups, parapatric and allopatric populations exhibit morphological and behavioural variation but cannot easily be given a taxonomic rank because the magnitude of their differences is unknown and little, if any, genetic data are available. However, such ‘geospecies’ (Grubb, 2006) may represent incipient stages in a process of evolutionary radiation and may hide cryptic diversity in urgent need of protection. Assessing patterns of interpopulation and subspecific variation is crucial to understanding evolutionary divergence. By examining current subspecific ranges, the drivers of diversification may be revealed, providing information about biological responses to past and present environments and implicated evolutionary processes. Thus, to improve the understanding of speciation and the origin of phenotypic differences, as well as to measure the amount of disparity and diversity in nature, it is necessary to accurately quantify biological variation in these populations and investigate the processes by which it originated. Subspecific variation requires reduction in gene flux, as may occur in (but is not confined to) the presence of dispersal barriers (Lande, 1980). Even at a small scale, reduction in gene flow can cause marked intraspecific variation (e.g. in genetics of clutch size; Postma & van Noordwijk, 2005). Across extensive geographical ranges, homogenous gene pools cannot be sustained, which may result in clinal (James, 1970; Frost et al., 2003) or stepped clinal (Jolly, 1993) variation. In the absence of robust molecular data, a priori taxonomic groups (constructed on the basis of geography and soft tissue features) may provide indirect information about reproductive isolation and gene flow across the species range. Considering taxonomy as a contributing factor in analysis of morphological variation thus has the power to shed light on the differentiation of widespread mammals. In the present study, we use the vervet monkey (Cercopithecus aethiops ssp.) as a model for intraspecific divergence in mammals, assessed through hard tissue morphology. Taxonomy is used as a proxy for genetic differentiation because the few molecular studies focusing on vervets (Ruvolo, 1988; Shimada, Terao & Shotake, 2002) have insufficient geographic coverage for a study of this type. The vervet is widespread in sub-Saharan Africa with attendant interpopulation variation in behaviour, ecology, and soft tissue morphology (Cardini, Jansson & Elton, 2007; Fedigan & Fedigan, 1988; Groves, 2001; Grubb et al., 2003). A mostly longitudinal cline in skull morphology, largely related to size variation, has been identified (Cardini et al., 2007). However, using traditional methods of hard tissue analysis, including

qualitative comparisons, it is difficult to identify subspecies. This might explain the relative paucity of research on taxonomic signals in vervet craniodental and other skeletal material. Previous work on the whole guenon clade (Cardini & Elton, 2008a) (i.e. the group to which vervets belong) has shown that variations in the skull match taxonomic groups well, according to traditional definitions largely based on soft tissues and behaviour. Similar trends have been observed in the red colobus clade, Piliocolobus (Cardini & Elton, 2009a). This indicates that reproductive isolation translates into subtle yet highly discriminating differences in hard tissues, even in very closely-related species with short divergence times (Cardini & Elton, 2008a). However, as in a microevolutionary study of vervets (Cardini et al., 2007), a macroevolutionary investigation of guenon skulls indicated that the main pattern of interspecific relationships was again mostly dominated by size. It has not yet been determined whether the vervet, a species with several subspecies, also follows the pattern of obvious behavioural and external soft tissue differences translating into more subtle, yet discriminating, craniofacial variations. Nonetheless, k-means cluster analysis of geographic shape variation in a previous study suggested this possibility because arbitrary groups reminiscent of major taxonomic subdivisions of vervets were produced (Cardini et al., 2007). However, in that study, there was no separation of size related (allometric) and nonallometric components of shape variation. When examining morphological divergence within a group of closely-related animals, it is important to consider allometric influences. The work described in the present study is concerned only with adult form so that the type of allometry most relevant to this study is static allometry, which describes size-related shape differences in individuals of similar age within species (Klingenberg, 1996). There have been very few studies on the role of allometry in determining form in widespread African mammal clades, with much work to date undertaken on primates. In adult guenons, for example, the lateral transposition of largely conserved allometric trajectories is an important contributor to interspecific difference (Cardini & Elton, 2008a). Size changes have thus played an important role in morphological evolution across the guenons as a whole, a clade that has undergone an extensive adaptive radiation. Thorough analyses of population differences within a single African mammal species or superspecies are still lacking. Vervets are ideal candidates to investigate the first steps of the process of evolutionary divergence. Indeed, size and size-related shape changes over their wide geographic range have clearly

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

EVOLUTIONARY DIVERGENCE IN VERVET SKULLS played a role in determining form within the vervets: individuals from East Africa are smaller and have a more paedomorphic appearance (shorter face and a relatively larger and more rounded neurocranium) than those from West and central Africa. If differences between parapatric populations simply reflect size changes, the apparent complexity of divergence in shape, with its inherently multivariate nature, must be downplayed because it would actually be related to the effects of variation in a single trait (i.e. size). This becomes even more relevant when moving from a microevolutionary study, such as the present one, to a macroevolutionary perspective (Cardini & Elton, 2008a), where allometry can be so pervasive and dominating that it completely erases any phylogenetic signal (Roth, 1996; Gilbert & Rossie, 2007). Thus, it is important to take the effects of allometry into account when examining gradients of change (clines) in morphology and assessing their importance in producing patterns of taxonomic differences.

STUDY

AIMS

The present study augments the limited work undertaken on hard tissue differences among African mammal subspecies in general and vervet populations in particular. A sample of almost 300 skulls of adult vervets from all presently recognized subspecies, based on the taxonomic scheme of Grubb et al. (2003), was measured using three-dimensional anatomical landmarks, and geometric morphometrics was applied to quantify: (1) the magnitude, direction and significance of intraspecific differences in size and shape, as well as the accuracy of shape as a taxonomic predictor, and thus, indirectly, the degree of distinctiveness and separation of the putative subspecies, and (2) the pattern of similarity relationships among subspecies and its implications for the reconstruction of their evolutionary history. Findings from these

3

analyses were also used to examine the importance of allometry in producing taxonomic variation by rerunning analyses after controlling for size. Thus, the study also investigated the potential role of sizerelated shape changes in generating morphological disparity as a consequence of selective pressures on size. Patterns of morphological variation were interpreted in relation to geographic proximity.

MATERIAL AND METHODS DATA COLLECTION AND APPROACH Subspecies were defined sensu Grubb et al. (2003) who recognized six subspecies of vervet monkeys with a parapatric distribution in sub-Saharan Africa (Table 1). Specimens were assigned to subspecies based on geographic locality and subspecific ranges (Groves, 2001; Kingdon, 2003). Individuals from contact areas between Cercopithecus aethiops aethiops and Cercopithecus aethiops djamdjamensis and between Cercopithecus aethiops pygerythrus and Cercopithecus aethiops tantalus were considered hybrids by Kingdon (2003). Thus, they were classified into separate samples. Out of 296 adult specimens of Cercopithecus aethiops Linnaeus, 1758, 125 were females and 171 were males (Table 1). Information on museum collections from which specimens were obtained are provided in Cardini et al. (2007) and a detailed list of specimens used in the study is available from the authors upon request. For small samples, special care was taken in analysis and interpretation of results (sensu Cardini & Elton, 2008a). Three-dimensional coordinates of anatomical landmarks were directly collected by the same person on crania and mandibles using a 3D-digitizer (MicroScribe 3DX; Immersion Corporation). Landmarks were digitized only on the left side to avoid redundant information in symmetrical structures. The set (configuration) of 86 landmarks used for the analysis (Fig. 1, inset) is described in Cardini et al. (2007).

Table 1. Taxa used in the analysis and sample sizes

Species

Subspecies

Taxonomic authority

Number of females

Number of males

Cercopithecus aethiops

aethiops cynosuros djamdjamensis pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis pygerythrus ¥ tantalus Total

Linnaeus, 1758 Scopoli, 1786 Neumann, 1902 F. Cuvier, 1821 Linnaeus, 1766 Ogilby, 1841 – – –

11 20 4 46 13 16 5 10 125

7 18 3 77 30 24 5 7 171

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

4

S. ELTON ET AL.

Figure 1. Distribution of specimens (pooled sexes) used for the present study. Different symbols are used for different subspecies as appropriate. Diagrams of subspecific facial coloration modified from Osman Hill (1966). The landmark configuration on the vervet skull is shown in the inset.

Landmarks on crania and mandibles were digitized separately and aligned sensu Cardini et al. (2007). Measurement error and estimates of a small number of missing landmarks had negligible effects on the analysis (Cardini et al., 2007; Cardini & Elton, 2008c). To make the visualization of group differences using three-dimensional diagrams (see below) more effective, landmarks on the left side were reflected and tiny asymmetries (0.004% of size and 1.7% of shape sample variance) on midplane landmarks removed. A geometric morphometric approach was used (Rohlf & Marcus, 1993; Adams, Rohlf & Slice, 2004; Sanfilippo et al., 2009), for which an extensive introduction is provided elsewhere (Zelditch et al., 2004). Geometric morphometric analyses were performed using: Morpheus (Slice, 1999), NTSYS-pc 2.2N (Rohlf, 2007) and MORPHOLOGIKA (O’Higgins & Jones, 2006). Statistical analyses were performed using SPSS 15.0 (2006) and NTSYS-pc 2.2N (Rohlf, 2007). Statistical analysis for small samples was performed sensu Cardini & Elton (2008a).

SHAPE

VARIABLES

A principal component analysis (PCA) of shape variables was used, which identifies the axes of greatest variation in a sample, to reduce dimensionality in the analysis. This is because shape coordinates are redundant (seven degrees of freedom are lost in the generalized Procrustes analysis of three-dimensional

data) and often highly correlated. Thus, the number of variables was reduced by including only the first principal components of shape coordinates. The number of principal components to be analyzed was selected using scree plots (Hair et al., 1998). The first 25 PCs explained approximately 72% or more of total shape variance and the correlation of Euclidean distances based on 25 PCs and Procrustes shape distances in the full shape space was as large as or larger than 0.95.

STATISTICAL

ANALYSIS

Differences in size and shape Split-sex samples were used in all analyses as a result of the large degree of sexual dimorphism observed in C. aethiops, common to the guenons as a whole (Cardini et al., 2007; Cardini & Elton, 2008b). In this way, taking advantage of the relatively large sample sizes, correction factors for sexual dimorphism, which can make results harder to interpret, were avoided. This also allowed observed patterns to be ‘double checked’ by comparing results from females and males. Permutation tests were used to assess pairwise the significance of differences in size and shape between samples. For size, differences were tested using a t-test, where the observed t statistic was compared with the t-distribution obtained by repeating the test a large number of times after randomly permuting group membership (Manly, 1997). The significance

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

EVOLUTIONARY DIVERGENCE IN VERVET SKULLS level was given by the frequency with which randomized t-tests produced t-statistics as large as or larger than the observed t. For shape, differences in sample means were tested with a nonparametric analysis of variance where the sum-of-squares explained by group membership in the data was compared to that for random permutations of group membership (Slice, 1999; Fontaneto, Melone & Cardini, 2004). The significance level of a difference between two samples was given by the frequency with which a random permutation of the group affiliation of specimens explained as much as or more variation between pseudosamples than observed in the original samples. The advantage of using permutation tests instead of standard parametric statistics is that they are less strongly affected by the large number of shape variables relative to sample size. Nevertheless, differences were also tested using standard parametric tests (analysis of variance and post hoc tests for size; discriminant analysis and tests of pairwise differences using Mahalanobis distances for shape) and the results were found to be mostly congruent with but slightly more liberal than those from permutation tests. Discriminant analysis of shape was also used to compute percentages of specimens correctly classified (‘hit-ratios’) according to subspecies. Results from discriminant analyses were cross-validated using jacknife (‘leave-one-out’ option) in SPSS (2006). Mean shape similarity relationships Size variation was summarized using box-plots. Similarity relationships among subspecies were summarized using a PCA on the matrix of mean shape variables. A mean shape is computed by taking the sample average of shape coordinates from either the full set of shape variables or, in analyses done after reducing dimensionality, the reconstruction of shapes based on a subset of PCs. Scatterplots of subspecies mean shapes, together with their bootstrap confidence ellipses (Cardini & Elton, 2008a), were used to show the relative positions of means and variation around their estimates along major axes of shape variance. Similarity relationships of subspecies were also summarized by superimposing a minimum spanning tree (Rohlf, 1970; Frost et al., 2003) from the matrix of mean shape Procrustes distances on a plot of subspecies geographic centroids (i.e. average longitude and latitude within each subspecific sample). Clusters with a high (> 50%) bootstrap percentage of node repeatability (Cardini & Elton, 2008a, c) are emphasized using grey coloured backgrounds. High repeatability refers to similarity relationships that are less strongly affected by sampling error (Cole, Lele & Richtsmeier, 2002; Caumul & Polly, 2005; Cardini & Elton, 2008a, c).

CONTROLLING

5

FOR ALLOMETRY

Group differences were also tested while holding the effect of allometry constant. A classical multivariate analysis of covariance (MANCOVA) was performed using shape (first 25 PCs) as the dependent variable, taxon as the grouping variable and the natural logarithm of centroid size (the measure of dispersion of landmarks around the centroid of the landmark configuration) as the covariate. Differences in slopes of static allometric trajectories (best fit regression lines of shape onto size within each group) were compared within each sex by looking at the significance of the interaction between groups (vervet subspecies) and the covariate. If slopes were not significant, the MANCOVA was repeated after removing the interaction term. Grouping variable significance would then indicate differences in intercepts and thus parallel allometric trajectories. This implies that groups are significantly different even when the effect of allometry is held constant. By contrast, nonsignificant intercepts would lie on the same allometric trajectories and possible group differences would be a result of allometric scaling. To assess the effect of small samples on the analyses, all tests were repeated after excluding smallest samples (N ⱕ 5). Because they led to similar results, the output of these tests is not shown. This model was then used to reconstruct allometry controlled (AC) shapes, with group differences tested in the same way as for full shapes, and similarity relationships estimated based on AC means. Two slightly different methods were employed to reconstruct AC shapes (indicated by AC1 and AC2); one an extension of the tests for slopes and intercepts for geometric morphometric shape (Zelditch et al., 2004) and a slightly modified version that regressed population mean shapes onto corresponding mean sizes to obtain residual non-allometric shape variation (Cardini & Elton, 2008a, b). The Zelditch et al. (2004) method aims to reconstruct shapes of size S according to group-specific allometric trajectories and the variation left unexplained by the same trajectories. The size S used to estimate predictions is arbitrary, and provided slopes are not significant the choice does not make any difference to the tests but has an effect on the visualization. Thus, the grand mean of size was used as S, pooling all samples, as this is the size that on average is closer to all specimens. PC scores predicted for each subspecies at size S (PCS) by the MANCOVA model were calculated. PCS scores were then multiplied by the corresponding eigenvectors and added back to the grand mean shape to reconstruct shape coordinates on subspecies’ allometric trajectories at size S, thus producing one allometry controlled shape

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

P-values, estimated using 10 000 random permutations, are shown below the main diagonal and percentages of variance explained by group membership. P values, found to be significant after a sequential Bonferroni correction for multiple comparisons using Holm’s method (Howell, 2002) are shown in italics, whereas asterisks (*) denote the smallest samples (N ⱕ 5). †S is the percentage of significant pairwise comparisons of a subspecies to all others. ‡E is the average percentage of variance explained in pairwise comparisons of a subspecies to all others.

25.2% 0.3% 19.4% 13.4% 4.4% 3.8% 45.9% – 29.0% 44.3% 8.1% 4.3% 55.5% 56.8% – 0.0071 43.4% 2.1% 30.4% 28.0% 0.1% – 0.0002 0.3435 42.2% 2.7% 30.1% 26.2% – 0.8669 0.0005 0.3305 0.5% 22.5% 0.1% – 0.0001 0.0001 0.1430 0.0046 2.0% 19.8% – 0.8769 0.0238 0.0108 0.4680 0.1121 28.2% – 0.0316 0.0001 0.3527 0.3978 0.0002 0.7825 – 0.0022 0.6184 0.6052 0.0010 0.0003 0.0316 0.0234 24.4 17.1 15.7 13.6 23.0 23.5 34.8 16.1

(7)* (6) (5) (4) (3)* (2) (1) E‡ (%)

42.9 42.9 0.0 42.9 42.9 42.9 42.9 0.0

The results of permutation tests for differences between subspecies are shown in Tables 2, 3 (size) and Tables 4, 5 (shape), together with the percentage of variance explained by group membership. Size variation within subspecies is summarized in Figure 2. Females and males showed strong congruence in the pattern of size variation. On average, less than 50% of pairwise comparisons were significant after a sequential Bonferroni correction and approximately 20% of variance was explained by group membership. The lack of significance in pairwise comparisons was at least in part related to low

aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus

AND DISCRIMINATION

(1) (2) (3) (4) (5) (6) (7) (8)

IN SIZE AND SHAPE: SIGNIFICANCE

S† (%)

RESULTS DIFFERENCES

Subspecies

for each subspecies. Finally, residuals from the MANCOVA were added back to these predicted shapes of each sample to create model population samples for which the effect of allometry was held constant. This model cannot be easily applied if slopes are different because predictions vary in different ways depending on the skull size of different samples (significant interaction). Thus, a different S would produce different outcomes and the criterion for choosing a given S would have to be explicitly justified and biologically meaningful. By contrast, if allometric trajectories are parallel (significant intercept), differences in predicted shapes for a given size will be identical, independent of the skull size S chosen to predict the score of the independent variable. Shapes computed this way (AC1) were used for further testing in pairwise permutation tests, discriminant analysis, and so on. The second method (AC2) considered the effect of evolutionary allometry. In a narrow sense, sensu Klingenberg (1996), evolutionary allometry refers to covariation along the branches of a phylogenetic tree. However, if we assume that subspecies reflect reductions in gene fluxes and may represent the initial steps of the process of evolutionary divergence, the same rationale as for ‘narrow sense evolutionary allometry’ can be applied. For this type of allometric variation, the effect of scaling is often removed by regressing population mean shapes onto corresponding mean sizes to obtain residual non-allometric shape variation (Cardini & Elton, 2008a, b). This approach was applied to vervet subspecies. Because we aimed to test population differences, these residuals were added to the grand mean of all samples to derive non-allometric mean shapes and residuals of the MANCOVA added as before. As for AC1, shapes derived in this way (AC2) were subjected to further tests including pairwise permutation tests and discriminant analysis.

(8)

S. ELTON ET AL.

Table 2. Pairwise tests of mean size differences between females of study taxa

6

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

P-values, estimated using 10 000 random permutations, are shown below the main diagonal and percentages of variance explained by group membership. P values, found to be significant after a sequential Bonferroni correction for multiple comparisons using Holm’s method (Howell, 2002) are shown in italics, whereas asterisks (*) denote the smallest samples (N ⱕ 5). †S is the percentage of significant pairwise comparisons of a subspecies to all others. ‡E is the average percentage of variance explained in pairwise comparisons of a subspecies to all others.

52.2% 0.5% 26.4% 13.2% 1.6% 3.8% 61.8% – 4.9% 59.3% 12.8% 1.6% 41.9% 29.7% – 0.0038 25.5% 3.7% 7.4% 14.8% 1.3% – 0.0021 0.2881 37.6% 0.9% 13.2% 23.6% – 0.4155 0.0001 0.4653 0.6% 22.2% 0.0% – 0.0001 0.0004 0.2639 0.0007 3.7% 23.2% – 0.8793 0.0355 0.1672 0.5000 0.1470 52.4% – 0.0235 0.0001 0.5305 0.2273 0.0001 0.7451 – 0.0003 0.7059 0.5091 0.0001 0.0034 0.5291 0.0069 28.6 42.9 0.0 57.1 42.9 28.6 42.9 14.3 aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus (1) (2) (3) (4) (5) (6) (7) (8)

25.3 23.1 12.4 10.8 17.1 12.3 30.3 22.8

(2) (1) E‡ (%) S† (%) Subspecies

Table 3. Pairwise tests of mean size differences between males of study taxa

(3)*

(4)

(5)

(6)

(7)*

(8)

EVOLUTIONARY DIVERGENCE IN VERVET SKULLS

7

statistical power in comparisons of small samples (e.g. C. a. djamdjamensis). In both sexes, subspecies were significantly different in a standard parametric analysis of variance (females, F7,117 = 8.738, P = 1.3 ¥ 10-8; males, F7,163 = 10.868, P = 3.2 ¥ 10-11). Results of post hoc tests (not shown) were almost identical to those of permutation tests. Subspecies from west and central Africa (Cercopithecus aethiops sabaeus, C. a. tantalus and Cercopithecus aethiops cynosuros) tended to be larger than those from east and south-east Africa. Putative hybrids between small-sized subspecies (C. a. aethiops ¥ C. a. djamdjamensis) were also small. Those between small and large ones (C. a. pygerythrus ¥ C. a. tantalus), in contrast, were large. Results of pairwise tests for differences in shape were also largely congruent in females and males. As for size, comparisons of smallest samples were often nonsignificant. However, for shape, there was more variability in percentages of significant tests. These, after a sequential Bonferroni correction for multiple comparisons, ranged within each subspecies from less than 15% (C. a. djamdjamensis, some of the putative hybrids, and males of C. aethiops after controlling for allometry using the AC2 method) to 85% or more (C. aethiops sabaeus). Percentages of variance explained by group differences, in contrast, were fairly similar across subspecies and, on average, were mostly in the range 7–8%. Static allometries accounted for little less than 20% of variance (females: 18.4%; males: 19.2%) in the PCs used for AC tests. Controlling for allometry (AC1, AC2) did not make a remarkable difference to the results, with most of the highly significant comparisons (which included approximately 45% and 30% of tests in females and males, respectively) being the same as when full shapes were used. Parametric tests were also performed for shape. Discriminant analyses were highly significant (without controlling for allometry: females, Wilks’ l = 0.0128, F175,640 = 3.312, P = 4.4 ¥ 10-28; males, Wilks’ l = 0.0365, F175,951 = 3.433, P = 7.2 ¥ 10-34), as well as most of the pairwise tests based on Mahalanobis distances (not shown). These were clearly more liberal than permutation tests with, overall, only 14% and 39% of tests nonsignificant in females and males, respectively. Tests of significance of discriminant analyses after controlling for allometry (AC1, AC2) (not shown) were very similar to those where allometry was not held constant. Table 6 show percentages of correctly classified specimens according to subspecies in discriminant analyses of shape. On average, at least 75% of individuals were correctly classified in a priori subspecific groups and females did slightly better than males. However, when results were cross-validated

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

71.4 42.9 14.3 42.9 100.0 57.1 14.3 28.6

aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus

aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus

aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus

tot. (1) (2) (3) (4) (5) (6) (7) (8)

AC1 (1) (2) (3) (4) (5) (6) (7) (8)

AC2 (1) (2) (3) (4) (5) (6) (7) (8) 9.4 7.4 10.5 4.6 9.4 7.2 8.5 8.9

9.4 7.1 10.2 3.8 10.9 6.6 7.9 8.4

8.6 6.3 9.7 3.7 10.7 6.5 8.0 7.7

E‡ (%)

– 0.0002 0.0726 0.0001 0.0004 0.0203 0.0111 0.0012

– 0.0001 0.2042 0.0001 0.0001 0.0228 0.0811 0.0009

C 0.0002 0.2965 0.0008 0.0002 0.0010 0.2800 0.0026

(1)

9.0% – 0.0237 0.0002 0.0001 0.0003 0.3436 0.0006

9.4% – 0.0088 0.0766 0.0003 0.0010 0.1563 0.0037

8.6% – 0.0048 0.0039 0.0004 0.0023 0.1423 0.0100

(2)

10.6% 7.8% – 0.0040 0.0096 0.3178 0.0156 0.1635

8.9% 8.5% – 0.0096 0.0004 0.2013 0.1861 0.1174

7.8% 8.2% – 0.0244 0.0018 0.0420 0.3111 0.0836

(3)*

5.7% 4.2% 4.4% – 0.0002 0.0001 0.172 0.0011

5.1% 2.4% 4.1% – 0.0001 0.0022 0.1498 0.0226

3.7% 2.9% 3.2% – 0.0001 0.0001 0.2287 0.0079

(4)

10.9% 9.0% 12.2% 4.9% – 0.0001 0.4997 0.0001

14.7% 8.5% 16.8% 5.6% – 0.0008 0.0633 0.0005

14.7% 6.9% 16.3% 6.5% – 0.0011 0.0010 0.0011

(5)

7.0% 8.3% 5.9% 6.1% 10.8% – 0.5617 0.0178

7.0% 7.3% 6.6% 3.8% 10.0% – 0.4082 0.0575

8.1% 5.9% 8.0% 4.3% 7.9% – 0.2375 0.0820

(6)

11.9% 4.6% 22.6% 2.6% 5.6% 4.5% – 0.3127

9.7% 5.5% 15.8% 2.8% 9.0% 5.2% – 0.4081

7.4% 5.2% 13.7% 2.3% 13.0% 5.8% – 0.1757

(7)*

10.8% 8.9% 10.2% 4.3% 12.4% 7.7% 7.9% –

11.0% 7.8% 10.8% 3.2% 11.9% 6.5% 7.4% –

10.2% 6.1% 11.0% 3.2% 9.4% 5.5% 8.6% –

(8)

P-values, estimated using 10 000 random permutations, are shown below the main diagonal and percentages of variance explained by group membership. P-values, found to be significant after a sequential Bonferroni correction for multiple comparisons using Holm’s method (Howell, 2002) are shown in italics, whereas asterisks (*) denote the smallest samples (N ⱕ 5). †S is the percentage of significant pairwise comparisons of a subspecies to all others. ‡E is the average percentage of variance explained in pairwise comparisons of a subspecies to all others. Significant results in all three sets of analyses are emphasized using a grey background.

57.1 71.4 0.0 71.4 71.4 42.9 0.0 57.1

57.1 42.9 14.3 42.9 85.7 42.9 0.0 28.6

S† (%)

Subspecies

Table 4. Pairwise tests of mean shape differences between females of study taxa using total and allometrically controlled (AC1, AC2) shape variance

8 S. ELTON ET AL.

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

28.6 71.4 0.0 28.6 85.7 28.6 28.6 14.3

aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus

aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus

aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus

tot. (1) (2) (3) (4) (5) (6) (7) (8)

AC1 (1) (2) (3) (4) (5) (6) (7) (8)

AC2 (1) (2) (3) (4) (5) (6) (7) (8)

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

8.2 6.9 10.0 3.2 7.9 5.6 7.7 8.1

9.3 6.8 8.6 2.5 8.3 5.1 6.1 7.8

8.5 6.8 8.3 2.7 7.8 4.4 9.2 8.9

E‡ (%)

– 0.1397 0.0854 0.0053 0.0008 0.0904 0.0959 0.1561

– 0.0021 0.3682 0.0107 0.0001 0.0507 0.1612 0.0104

– 0.0009 0.4611 0.0318 0.0001 0.1617 0.5146 0.0056

(1)

5.7% – 0.0743 0.0001 0.0001 0.0001 0.4056 0.018

9.5 – 0.045 0.0075 0.0001 0.0001 0.6578 0.1020

9.2% – 0.0597 0.0001 0.0001 0.0007 0.0020 0.1217

(2)

15.9% 7.6% – 0.0102 0.0054 0.1162 0.0728 0.3148

11.9 8.4 – 0.1085 0.0006 0.2273 0.6375 0.2546

11.2% 7.1% – 0.1937 0.0035 0.5387 0.2507 0.2524

(3)*

2.4% 3.1% 2.3% – 0.0001 0.0001 0.3503 0.0124

2.3% 2.2% 1.8% – 0.0001 0.0018 0.7123 0.3005

1.8% 3.1% 1.5% – 0.0001 0.0086 0.3322 0.0070

(4)

6.8% 11.9% 6.7% 6.4% – 0.0001 0.0506 0.0001

10.9% 11.1% 7.8% 6.7% – 0.0001 0.1022 0.0001

10.3% 7.8% 6.5% 7.2% – 0.0001 0.0002 0.0002

(5)

4.7% 7.1% 5.3% 4.5% 8.7% – 0.7094 0.0121

5.2% 6.5% 4.6% 2.3% 8.2% – 0.4254 0.068

4.1% 5.6% 3.6% 1.8% 6.8% – 0.4741 0.0358

(6)

12.4% 4.7% 20.1% 1.2% 4.7% 2.9% – 0.5985

11.5% 3.8% 12.6% 1.0% 4.2% 3.6% – 0.8506

8.9% 9.6% 15.8% 1.3% 9.5% 3.5% – 0.0093

(7)*

9.7% 8.0% 12.4% 2.2% 10.3% 6.3% 8.0% –

13.8% 6.1% 13.3% 1.4% 9.0% 5.1% 6.1% –

14.4% 5.4% 12.4% 2.2% 6.6% 5.1% 16.0% –

(8)

P-values, estimated using 10 000 random permutations, are shown below the main diagonal and percentages of variance explained by group membership. P-values, found to be significant after a sequential Bonferroni correction for multiple comparisons using Holm’s method (Howell, 2002) are shown in italics, whereas asterisks (*) denote the smallest samples (N ⱕ 5). †S is the percentage of significant pairwise comparisons of a subspecies to all others. ‡E is the average percentage of variance explained in pairwise comparisons of a subspecies to all others. Significant results in all three sets of analyses are emphasized using a grey background.

14.3 42.9 0.0 42.9 71.4 42.9 0.0 14.3

28.6 42.9 14.3 28.6 85.7 42.9 0.0 14.3

S† (%)

Subspecies

Table 5. Pairwise tests of mean shape differences between males of study taxa using total and allometrically controlled (AC1, AC2) shape variance

EVOLUTIONARY DIVERGENCE IN VERVET SKULLS

9

10

S. ELTON ET AL.

Figure 2. Box-plots of female (A) and male (B) centroid size of subspecies samples. Skull centroid size, which is highly correlated to cranial length (Cardini & Elton, 2008b), is used as a proxy for body size, in line with other similar work on primates (Fooden & Albrecht, 1993).

percentages of correctly classified specimens dropped to slightly more than 50% in both sexes, which is nevertheless significantly better than chance (i.e. 12.5%, assuming equal probability of being classified by chance in the right group or, on average, 14–16% with 20–22% 95th upper percentile if computed using simulations taking uneven sample size into account (White & Ruttenberg, 2007). Again, controlling for allometry (AC1, AC2) did not appear to make any appreciable difference, with similar ranges of correctly classified specimens, smallest hit-ratios in smallest samples and a sharp drop in classification accuracy when the results were cross-validated.

MEAN

SHAPE SIMILARITY RELATIONSHIPS

Similarity relationships based on skulls are shown in Figures 3 and 4. Ninety-five percent confidence intervals around means mostly overlapped on PC1–PC2, with the main exception being C. a. djamdjamensis in both females and males. A better separation, however, was achieved on higher axes in males but not in females. In the latter, only the putative hybrid C. a. pygerythrus ¥ C. a. tantalus stood apart from all other groups. In the former, confidence intervals showed little overlap but subspecies did not form clear clusters.

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

EVOLUTIONARY DIVERGENCE IN VERVET SKULLS

11

Table 6. Percentages of correctly classified specimens in discriminant analyses. Females

Males

Shape

Species

Analysis

Jacknife

Analysis

Jacknife

Total

aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus Total

100.0 80.0 100.0 78.3 100.0 87.5 100.0 90.0 86.4

72.7 40.0 50.0 60.9 69.2 43.8 40.0 40.0 54.4

100.0 83.3 100.0 63.6 100.0 58.3 100.0 85.7 75.4

71.4 66.7 0.0 51.9 93.3 25.0 0.0 14.3 53.8

AC1

aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus Total

100.0 80.0 100.0 71.7 100.0 75.0 100.0 100.0 83.2

81.8 40.0 50.0 45.7 76.9 37.5 20.0 60.0 50.4

100.0 77.8 100.0 54.5 96.7 62.5 80.0 100.0 70.8

71.4 66.7 33.3 39.0 90.0 33.3 0.0 0.0 48.5

AC2

aethiops cynosuros djamdjamensis* pygerythrus sabaeus tantalus aethiops ¥ djamdjamensis* pygerythrus ¥ tantalus Total

100.0 90.0 100.0 91.3 92.3 81.3 100.0 100.0 92.0

63.6 75.0 25.0 82.6 53.8 31.3 20.0 50.0 63.6

85.7 83.3 100.0 83.1 96.7 70.8 60.0 85.7 83.6

57.1 61.1 33.3 74.0 90.0 54.2 0.0 14.3 63.7

Discriminant functions were computed using the first 25 principal components (PCs) of shape [using total and allometric controlled (AC1, AC2) shape variance] and cross-validated using jacknife. Percentages of correctly classified specimens larger than 50–75% (i.e. four- to six-fold higher than expected by chance) are emphasized using light and dark grey backgrounds, respectively

To relate differences among subspecies to geographic distribution, geographic centroids of samples were computed and similarity relationships summarized using a minimum spanning tree based on shape distances (Figs 5, 6). In the same figures, diagrams of mean shapes are shown to visualize main trends of geographic variation. Similarities were found between neighbouring populations lying mostly on a west-to-east axis. Thus, in males C. a. sabaeus was relatively similar to C. a. tantalus, which in turn showed similarities with C. a. pygerythrus. In both sexes, C. a. pygerythrus showed the largest number of connections not only with subspecies in the Horn of Africa, but also with C. a. cynosuros in the south. The most evident difference between females and males was the similarity of female C. a. sabaeus to C. a. cynosuros instead of the geographically closer C. a. tantalus.

Very few within-sample similarity relationships suggested by the minimum spanning tree were supported by a high proportion of node repeatability. This means that, after bootstrapping, few clusters were consistently found that were in common to all pseudosamples. These were: (1) a group including all subspecies to the exclusion of C. a. djamdjamensis; (2) a cluster including C. a. cynosuros, C. a. pygerythrus, and C. a. tantalus in males but only the first two in females; and (3) in males only, a cluster including C. a. pygerythrus and C. a. tantalus. Interestingly, none of these clusters was homogeneous for skull size and actually included a mix of small (C. a. pygerythrus) and large (C. a. cynosuros and C. a. tantalus) subspecies. Controlling for allometry using AC1 produced the same supported clusters as using all shape information. AC2 also produced almost identical supported clusters as in the other analyses, with the exception

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

12

S. ELTON ET AL.

female mean shapes 0.03 0.02

PC2 (10.7%)

0.01 0.00

*

-0.01 -0.02 -0.03 -0.04 -0.04 -0.03 -0.02

-0.01

0.00

0.01

0.02

0.03

0.04

0.05

0.06

PC1 (28.6%)

(A) 0.02

PC4 (7.8%)

0.01

* 0.00

-0.01

-0.02 -0.03

(B)

-0.02

-0.01

0.00

0.01

0.02

PC3 (9.4%)

Figure 3. Female subspecies mean shapes. Scatterplots of the first principal components (PCs) of shape variables (percentages of explained variance in parentheses). For each subspecies, variation around the mean is illustrated with 95% bootstrap confidence ellipses. (A) PC1 and PC2; (B) PC3 and PC4. Different symbols are used for different subspecies as appropriate.

being C. a. tantalus, which was still part of the largest subspecies group (all subspecies but C. a. djamdjamensis) but no longer clustered either with C. a. pygerythrus or next to C. a. cynosuros (results not shown). Shape variation on a west-to-east axis can be summarized by comparing C. a. sabaeus with C. a. pygerythrus. Western populations tend to have long muzzles with prominent mandibles. Eastern populations tend to have short faces and relatively expanded

neurocrania. For size, C. a. cynosuros was as large as C. a. sabaeus and, for shape, somewhat similar in the pattern of facial expansion and neurocranial reduction. However, C. a. cynosuros had several distinctive traits. For example, the mandibular corpus of C. a. cynosuros was not as deep as in C. a. sabaeus and the gonial angle was less obtuse. Also, in C. a. cynosuros, nasals formed a relatively pronounced angle, with the frontal bones and the tip of the snout projected

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

EVOLUTIONARY DIVERGENCE IN VERVET SKULLS

13

male mean shapes 0.03 0.02

PC2 (16.2%)

0.01

*

0.00 -0.01 -0.02 -0.03 -0.04

-0.05 -0.04 -0.03 -0.02 -0.01 0.00

0.01

0.02

0.03

0.04

PC1 (26.6%)

(A) 0.03

0.02

PC4 (8.1%)

0.01

* 0.00

-0.01

-0.02

-0.03 -0.02

(B)

-0.01

0.00

0.01

0.02

PC3 (9.5%)

Figure 4. Male subspecies mean shapes. Scatterplots of the first principal components (PCs) of shape variables (percentages of explained variance in parentheses). For each subspecies variation around the mean is illustrated with 95% bootstrap confidence ellipses. (a) PC1 and PC2; (b) PC3 and PC4.

forward. Differences relative to other subspecies were similar in both sexes.

DISCUSSION SIZE, SHAPE, AND VERVET TAXONOMY The present study aimed to assess the occurrence and degree of difference in skull form between vervet populations, as a model for subspecific variation in

African mammals with wide distributions, and to test whether size was the dominating factor in the observed patterns. Vervet skull morphology undoubtedly varied across Africa. Some of the most striking differences, in both size and shape, were seen at the extremes of their longitudinal distribution. Cercopithecus a. sabaeus, C. a. tantalus, and C. a. cynosuros, the three western-most subspecies, were large compared to subspecies from the east. The distinctive cranial

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

14

S. ELTON ET AL.

female mean shapes (x5-16) well-supported clusters (node repeatibility ≥ 50%) all subspecies except djamdjamensis cynosuros + pygerythrus

latitude

12.75

sabaeus (W)

aethiops djamdjam.* aeth.×djam.*

tantalus pyg.×tant. -1.47

pygerythrus (E) -15.70 -14.59

12.78

cynosuros (SW)

40.15

longitude

Figure 5. Plot of female geographic centroids for subspecies samples. A minimum spanning tree computed on the matrix of mean shape Procrustes distances is superimposed to summarize similarity relationships. Clusters with bootstrap percentage of node repeatability larger than 50% were emphasized using grey coloured backgrounds. Shape differences are visualized using transformations from the grand mean to the mean of three subspecies chosen as representatives of main trends in geographic variation: Cercopithecus aethiops sabaeus for west Tropical Africa (W), Cercopithecus aethiops cynosuros for south-west Africa (SW), and Cercopithecus aethiops pygerythrus for east Africa (E). Transformations are magnified and shown together with thin plate spline deformation grids drawn on three planes parallel to either the midsagittal plane (dorsal view) or the occusal plane (side view). Symbols for subspecies were drawn using circles proportional to the average skull size of a subspecies. Allometry control (AC) produced the same supported clusters in both AC1 and AC2 as using all shape information (results not shown).

shapes of C. a. sabaeus (in the west) and, to a lesser extent, C. a. aethiops (in the east) were evident not only through tests of mean shape differences, but also by discriminant analyses in which percentages of correctly classified individuals did not drop dramatically when cross-validated. This suggests a degree of congruence between taxonomic and geographic distances. The pronounced pattern of taxonomic variation from west to east is in agreement with previous observations (Cardini et al., 2007) that the main axis of clinal variation in vervets is longitudinal. Primary productivity, which is higher in west and central tropical Africa, may contribute to the observed differences, with the abundance of resources possibly facili-

tating larger body size (Cardini et al., 2007, 2010). A similar relationship has been noted in African hartebeest, another widespread mammal with several allopatric subspecies (Capellini & Gosling, 2007). Pronounced changes in size may then influence shape change through allometry. The cranial shape differences in small forms such as C. a. aethiops and C. a. pygerythrus versus the larger C. a. sabaeus, as noted elsewhere (Cardini et al., 2007), were mainly related to a trend of neurocranial expansion and facial reduction. This trend towards a more paedomorphic shape in smaller vervet subspecies corresponds to other comparisons of larger and smaller representatives of closely related primates (Cardini & Elton, 2008a, 2009b) and mammals

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

EVOLUTIONARY DIVERGENCE IN VERVET SKULLS

15

male mean shapes (x5-16) well-supported clusters (node repeatibility ≥ 50%) all subspecies except djamdjamensis cynosuros + pygerythrus + tantalus pygerythrus + tantalus

latitude

sabaeus (W)

aethiops

13.46

djamdjam.* aeth.×djam.*

tantalus

pyg.×tant. -1.77

pygerythrus (E) -17.00 -13.48

13.01

cynosuros (SW)

39.49

longitude

Figure 6. Plot of male geographic centroids for subspecies samples. A minimum spanning tree computed on the matrix of mean shapes Procrustes distances is superimposed to summarize similarity relationships. Clusters with bootstrap percentage of node repeatability larger than 50% were emphasized using grey coloured backgrounds. Shape differences are visualized using transformations from the grand mean to the mean of three subspecies chosen as representatives of main trends in geographic variation: C. a. sabaeus for west Tropical Africa (W), C. a. cynosuros for south-west Africa (SW) and C. a. pygerythrus for east Africa (E). Transformations are magnified and shown together with thin plate spline deformation grids drawn on three planes parallel to either the midsagittal plane (dorsal view) or the occusal plane (side view). Symbols for subspecies were drawn using circles proportional to the average skull size of a subspecies magnified by an arbitrary non-linear factor in order to emphasize differences and facilitate comparisons. Controlling allometry produced the same supported clusters in both AC1 and AC2 as using all shapes information (results not shown).

(A. Cardini, unpubl. observ.), and might occur because of hypomorphosis (sensu Shea, 1992). Research on Neotropical primates suggests that size is a ‘line of least evolutionary resistance’ (Marroig & Cheverud, 2005) and a rapid way to shift niches and respond to environmental change. In the guenon clade as a whole, the large size difference between the smallest species (the talapoin monkey) and the biggest (the patas monkey) is definitely adaptive; size reduction in the talapoin allows it to exploit an insectivorous niche whereas the relatively large size of the patas monkey is advantageous in the very open environments it inhabits (Cardini & Elton, 2008a). In combination with this evolutionary trend, phenotypic plasticity, including changes to size in an individual’s

lifetime, may influence fitness and help move between new peaks in the adaptive landscape (Borenstein, Meilijson & Ruppin, 2006). Work on Drosophila suggests that size may respond more readily than shape to variation in resources (Breuker, Patterson & Klingenberg, 2006), and complex developmental regulation may make shape more resilient in rapidly changing environments (Debat, Debelle & Dworkin, 2009). A key question is whether vervet shape differences at geographic extremes, which follow size differences, are exclusively allometric. When controlling for allometry, and hence indicating shape variance with the effects of size held constant, test outputs were largely similar to those when all shape information was used.

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

16

S. ELTON ET AL.

For example, C. a. aethiops and C. a. pygerythrus (males and females) still differed from C. a. sabaeus in pairwise tests. This indicates that subspecific differences are not simply a result of variations in size or size-related shape, probably reflecting variations with a profound genetic basis and longer evolutionary history than might be the case only if changes had occurred rapidly along gradients of environmental change. The impressive congruence in these and other pairwise comparisons regardless of controlling for allometry reinforces the view that, between vervet subspecies, size alone did not drive all morphological similarities and differences. Further support for this comes from the observation that, although C. a. cynosuros shared some aspects of facial expansion and neurocranial reduction with another large, western species, C. a. sabaeus, the males and females of the two subspecies were significantly different in all pairwise comparisons. By contrast, C. a. pygerythrus, despite having a relatively small size, did not have a cranial shape that differed consistently from its considerably larger neighbouring subspecies: in pairwise comparisons, there were no significant shape differences between C. a. pygerythrus and C. a. cynosuros females, or between C. a. pygerythrus and C. a. tantalus males.

SIMILARITY

AND GEOGRAPHIC PROXIMITY

Another aim of the present study was to assess how the pattern of similarity relationships between subspecies is related to geographic proximity and clines and whether it was mostly determined by size. The answer is certainly negative because both AC methods produced support for clusters unappreciably different from those using all shape information. The distinctiveness of C. a. aethiops (in the east) and C. a. sabaeus (in the west) suggests, not unexpectedly, that geographic distance contributes to morphological divergence. However, there was no straightforward relationship between simple geographic proximity and morphological similarity. Taken as a whole, skull shape divergence among the neighbouring C. a. pygerythrus, C. a. tantalus, and C. a. cynosuros appeared to be fairly subtle. The relative lack of differentiation between these three subspecies resonated with the results of the discriminant analyses, which showed that, although percentages of correctly-classified specimens were fairly high, at least in females, groupings were far from secure under cross-validation. Thus, morphological variation across the broad geographic range of vervets might be small and gradual (i.e. clinal), and partly driven by changes in the environment. This is also suggested by the fact that a large proportion of size variation

is explained by spatially structured environmental variation (Cardini et al., 2007). Cercopithecus a. pygerythrus and C. a. tantalus comprised a strong male cluster with C. a. cynosuros, although there were still significant pairwise mean shape differences (evident in males and females) between C. a. tantalus and C. a. cynosuros. Mean shapes visualized on PC3 and PC4, especially in females, also hinted at this differentiation. The ranges of C. a. tantalus (found in a band across east to west central Africa) and C. a. cynosuros are not contiguous as a result of the presence of the Congo Basin rainforest, one of the few sub-Saharan African regions not exploited by vervets (Fedigan & Fedigan, 1988). With a semi-terrestrial, wide ranging, behaviourally, and ecologically flexible primate, especially one as ready to exploit numerous habitats as the vervet, it is difficult to accurately pinpoint how geographic barriers, topographic changes and forest fragmentation influenced dispersal and population differentiation. The complexity of this is exemplified by the observation that even small changes to the ecology of an area may alter vervet exploitation of it. For example, vervets are considered to have moved into the arid Karoo of southern Africa between 1900 and 1970 because of the introduction of non-native trees, food crops and increased water availability (Milton, Zimmermann & Hoffmann, 1999). Vervets in the past would probably have taken similar opportunities for expansion or range shift during periods of rapid environmental change. Given enough genetic variation and time, vervet ecological plasticity may have facilitated adaptation because movement (precipitated for example by climate change) into sub-optimal environments and subsequent exploitation may lead to groups becoming better adapted to those environments. Considering such flexibility reinforces the difficulties of predicting their dispersal patterns and population barriers. Nonetheless, some assumptions can be made with a reasonable degree of certainty. Since the modern vervet distribution skirts round the Congo Basin (Fedigan & Fedigan, 1988), it is very likely that this geographic barrier prevents direct gene flow between C. a. tantalus and C. a. cynosuros, resulting in cranial shape differences important enough to be significant even after controlling for allometry. Most of the largest population samples showed significant shape differences, unrelated to size (which varies more readily with the environment), so it is likely that clines are only part of the reason why vervets show such a large amount of variation across their range. Thus, vervet subspecies are unlikely to simply represent somewhat arbitrary subdivisions of a continuous range of variation, and their differences may not only reflect size differences in response

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

EVOLUTIONARY DIVERGENCE IN VERVET SKULLS to gradients of environmental change but also be a consequence of reduced genetic fluxes during the evolutionary history of this species. If population differences have a genetic basis, geographic proximity also has implications for hybridization. Along the northern and western shores of Lake Victoria there is a well documented hybrid zone between C. a. pygerythrus and C. a. tantalus (Kingdon 2003). The specimens classed as pygerythrus ¥ tantalus hybrids in this study did not differ significantly in shape from either of the two subspecies in the majority of analyses. Hybridization is not unusual in primate taxa, and a number of studies (Smith & Scott, 1989; Cheverud, Jacobs & Moor, 1993; Kohn, Langton & Cheverud, 2001; Ackermann, Rogers & Cheverud, 2006) have detected a degree of heterosis, manifested by large body size, in hybrid primates. In the present study, males and females from the contact zone between C. a. tantalus and C. a. pygerythrus tended to be large, possibly indicating heterosis or, alternatively, asymmetric mating. By contrast, individuals from the easterly C. a. aethiops and C.a.djamdjamensis contact zone were small, even relative to their parent populations. This result is very difficult to interpret because the cranial sizes of the C. a. djamdjamensis specimens included in the present study were highly variable, probably as a result of the relatively small sample, which has the potential to profoundly alter results in some geometric morphometric analyses (Cardini & Elton, 2007). Additionally, reports of hybridization between C. a. aethiops and C. a. djamdjamensis are yet to be confirmed (Butynski et al., 2008). This notwithstanding, the smaller observed sizes could indicate a degree of dysgenesis in hybridizing groups.

IMPLICATIONS

OF CRANIAL MORPHOLOGY FOR

TAXONOMY AND CONSERVATION BIOLOGY

Hard tissue morphology is unlikely to be the most influential factor in classifying extant primates, especially in taxa that have diverged relatively recently. Nonetheless, the data reported in the present study contribute to the debate over modern vervet taxonomy. Groves (2001) recognized the six taxa used in the present study as species, primarily on the basis of external soft tissue features. However, other studies treat them as a single species with numerous variants (Grubb et al., 2003), not least because vervet populations are difficult to distinguish vocally (Struhsaker, 1970). In their study on marmoset crania, Marroig, Cropp & Cheverud (2004) suggested that discrimination lower than 75% between taxa is indicative of subspecific status. In red colobus taxa, the high hitratios in discriminant analyses (100% in many

17

groups) fit well with recent genetic data (Ting, 2008) on their taxonomic affinities, supporting their classification as a cluster of mostly good species (Cardini & Elton, 2009a, b). Hit-ratios from discriminant analyses of vervets were approximately 75–85% but only slightly larger than 50% when cross-validated. This is only marginally better than the average 30–60% cross-validated hit-ratios from a study of cranial variation of sub-Saharan populations of humans (Franklin, Cardini & Oxnard, 2010) and can be contrasted with much higher classification and cross validation rates within the guenons as a whole (Cardini & Elton, 2008a). Thus, on the basis of cranial morphology at least, the six most commonly recognized vervet taxa are better classed as subspecies rather than species. Nevertheless, vervet taxonomic groups are more than local variants of a single skull ‘plan’ that varies according to size and geography. This is because controlling for allometry not only makes little difference in tests, but also the gradient in clinal shape appears to become steeper close to the contact areas of subspecies, as indicated previously (Cardini et al., 2007). Albeit using very small samples, the small number of genetic studies of vervets show some evidence for divergence, although the magnitude was similar to that found within a single subspecies (Shimada et al., 2002). On the basis of other genetic studies of primates (Guillén, Barrett & Takenaka, 2005), these data might suggest that vervet taxa diverged from one another relatively recently or that populations have experienced a complex history of separation and amalgamation. Vervets are not the only African mammal likely to have experienced this. For example, based on mitochondrial DNA, a similar complex history is suggested for the kob antelope, with differentiation in refugia followed by secondary contact and hybridization creating challenges for interpretation of subspecific divergence (Lorenzen et al., 2007). Unlike the vervet, however, it has been suggested that variation among kob populations is not sufficient to sustain some current subspecific distinctions (Lorenzen et al., 2007). Lack of clarity over taxonomy in many animal groups, in Africa and elsewhere, has the potential to confound accurate assessments of diversity for conservation purposes. In the case of the widespread, eurytopic and abundant vervet, which appears to survive well in anthropogenically modified habitats, it is tempting to dismiss the need for future conservation efforts and therefore the importance of accurately assessing inter-population variation. However, the fact that C. aethiops as a whole is currently defined in the Red List as ‘lower risk/least concern’ should be no bar to

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

18

S. ELTON ET AL.

research into its population history and groupings. In analysis considering human population densities and geographic ranges, it is ranked as being an ‘intermediate’ conservation priority (Harcourt & Parks, 2003). It has also been noted (Harcourt, 2006) that the future survival of primate taxa that are common today may rely on conservation of several populations across the range. Catch-all taxonomic classifications may conceal significant variation in conservation priority across subSaharan Africa (sensu Harcourt, 2006). The Bale monkey (Cercopithecus a. djamdjamensis) exemplifies how important it is to consider inter-population variation when assessing the conservation status of taxa with wide geographic distributions. Found over a restricted highland area in Ethiopia, the Bale monkey is the only vervet subspecies included on the Red List as ‘vulnerable’, and the need for further research on the subspecies has been highlighted (Butynski et al., 2008). The Bale monkey has been noted as a particularly morphologically distinctive member of the vervet group (Groves, 2001; Kingdon, 2003; Butynski et al., 2008) and, although this must be verified on a large sample (Cardini & Elton, 2007), the results of the present study may be consistent with this assertion. In males and females, its mean shape formed a distinct group on PC1 and PC2, which is also reflected in the cluster that grouped all species except C. a. djamdjamensis together. The bamboo forest habitat of Cercopithecus a. djamdjamensis in the Bale Mountains (an area of high endemism and vegetation diversity) is highly unusual compared to those of other vervets (Butynski et al., 2008; Mekonnen et al., 2010). Fossil pollen records from the late Pleistocene and early Holocene indicate a complex history of vegetative change in the region, which alternated between glacial conditions, Afroalpine vegetation and woody vegetation, with the flora responding to shifts in moisture availability (Umera et al., 2007). The highly changeable nature of the environment in this region has been hypothesized to be the cause of the high levels of endemism (Umera et al., 2007). It is possible that this also influenced morphological divergence in the Bale monkey, as well as chromosomal divergence in several species of endemic rodents (Corti et al., 1995). The divergence date of the Bale monkey is unknown but environmental shifts of the type indicated by pollen records (Umera et al., 2007), caused by Pleistocene climate change, may have periodically or more permanently separated it from other vervet populations, accelerating its morphological evolution. Extending this model to the vervets as a whole, and assuming that Pleistocene forms did not exploit primary rainforest, any expansion of forest in inter-

glacial periods may have isolated populations that were contiguous during glacials. Similarly, assuming that the behaviour of Pleistocene vervets was similar to modern conspecifics, extremely dry/desert areas would have been outside their normal range. Extremely xeric environments, as might emerge, for example, during glacial periods in areas previously covered by savanna, would thus also act as a barrier to dispersal. Such environmental shifts may have contributed to population divergence in vervets. Detailed environmental modelling at regional and local scales combined with morphological and phylogeographic studies will help to reveal the complex and so far poorly understood evolutionary and population history of the vervets across Africa. This in turn will augment our knowledge of its morphological differentiation along with classification and taxonomy.

CONCLUSIONS The present study illustrates the potential of examining variation within widespread mammal species to identify biologically meaningful differences and their origin. Although, in our vervet sample, there was a degree of congruence between geographic and taxonomic proximity, it is too simplistic to assume a straightforward trend whereby geography and environmental differences such as rainfall cause shifts in body size that then alter shape. Size differences may well help to start and facilitate adaptive divergence. These differences can be genetic, plastic or a mix of the two. Indeed, even if size variation were largely plastic, plasticity has been shown to potentially facilitate adaptive evolution (Scovillea & Pfrendera, 2010). However, despite the mounting evidence for a link between size variation and rainfall in African mammals, as seen in ungulates (Capellini & Gosling, 2007) as well as in several primate species (Cardini et al., 2007; Cardini & Elton, 2009a, b), taxonomic differences in vervets were still highly significant when the effect of size on shape was held constant. Methods such as Geometric Morphometrics (GMM) that capture and visualize shape variation by efficiently separating size and shape tend to focus on testing differences in shape. However, the present study shows that both size and shape, as well as their interaction, need to be considered for a proper understanding of quantitative variation in morphology, and that this can readily be accomplished using available methods. The effect of size on shape needs to be taken into account, especially because strong selection on size and the concomitant changes in shape as a result of allometry can reduce or even erase the phylogenetic signal in morphological data (Gilbert & Rossie, 2007; Cardini & Elton, 2008a). Further investigation of this

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

EVOLUTIONARY DIVERGENCE IN VERVET SKULLS in other widespread mammals may help to reveal the relative importance of size-driven shape change, interactions between size and shape, and subspecific differentiation.

ACKNOWLEDGEMENTS We are grateful to all the collection managers and curators who allowed us to study their collections and also provided help. We also thank very much two anonymous referees whose comments greatly helped to improve the paper. This study was funded by grants from the Leverhulme Trust and the Royal Society.

REFERENCES Ackermann RR, Rogers J, Cheverud JM. 2006. Identifying the morphological signatures of hybridization in primate and human evolution. Journal of Human Evolution 51: 632–645. Adams DC, Rohlf FJ, Slice DE. 2004. Geometric morphometrics: ten years of progress following the ‘revolution’. Italian Journal of Zoology 71: 5–16. Ballinger RE. 1979. Intraspecific variation in demography and life history of the Lizard, Sceloporus jarrovi, along an altitudinal gradient in Southeastern Arizona. Ecology 60: 901–909. Borenstein E, Meilijson I, Ruppin E. 2006. The effect of phenotypic plasticity on evolution in multipeaked fitness landscapes. Journal of Evolutionary Biology 19: 1555–1570. Breuker CJ, Patterson JS, Klingenberg CP. 2006. A single basis for developmental buffering of Drosophila wing shape. PloS ONE 1: e7. Butynski TM, Gippoliti S, Kingdon J, De Jong Y. 2008. Chlorocebus djamdjamensis. In: IUCN 2009. IUCN Red List of Threatened Species, Version 2009.1. Available at: http:// www.iucnredlist.org. Downloaded on 3 October 2009. Capellini I, Gosling LM. 2007. Habitat primary production and the evolution of body size within the hartebeest clade. Biological Journal of the Linnean Society 92: 431– 440. Cardini A, Elton S. 2007. Sample size and sampling error in geometric morphometric studies of size and shape. Zoomorphology 126: 121–134. Cardini A, Elton S. 2008a. Variation in guenon skulls (I): species divergence, ecological and genetic differences. Journal of Human Evolution 54: 615–637. Cardini A, Elton S. 2008b. Variation in guenon skulls (II): sexual dimorphism. Journal of Human Evolution 54: 638– 647. Cardini A, Elton S. 2008c. Does the skull carry a phylogenetic signal? Evolution and modularity in the guenons. Biological Journal of the Linnean Society 93: 813–834. Cardini A, Elton S. 2009a. The radiation of red colobus monkeys (Primates, Colobinae): morphological evolution in

19

a clade of endangered African primates. Zoological Journal of the Linnean Society 157: 197–224. Cardini A, Elton S. 2009b. Geographic and taxonomic influences on cranial variation in red colobus monkeys (Primates, Colobinae): introducing a new approach to ‘morph’ monkeys. Global Ecology and Biogeography 18: 248–263. Cardini A, Jansson A-U, Elton S. 2007. A geometric morphometric approach to the study of ecogeographical and clinal variation in vervet monkeys. Journal of Biogeography 34: 1663–1678. Cardini A, Diniz Filho JAF, Polly PD, Elton S. 2010. Biogeographic analysis using geometric morphometrics: clines in skull size and shape in a widespread African arboreal monkey. In: Elewa AMT, ed. Morphometrics for nonmorphometricians. Heidelberg: Springer-Verlag. Cardini A, Nagorsen D, O’Higgins P, Polly PD, Thorington Jr RW, Tongiorgi P. 2009. Detecting biological uniqueness using geometric morphometrics: an example case from the Vancouver Island marmot. Ecology, Ethology and Evolution 21: 209–223. Caumul R, Polly PD. 2005. Phylogenetic and environmental components of morphological variation: skull, mandible and molar shape in marmots (Marmota, Rodentia). Evolution 59: 2460–2472. Cheverud JM, Jacobs SC, Moor AJ. 1993. Genetic differences among subspecies of the saddle-back tamarin (Saguinus fuscicollis). American Journal of Primatology 31: 23–39. Cole TMI, Lele S, Richtsmeier JT. 2002. A parametric bootstrap approach to the detection of phylogenetic signals in landmark data. New York, NY: Taylor and Francis. Corti M, Civitelli M, Bekele A, Castiglia R, Capanna E, Capanna E. 1995. The chromosomes of three endemic rodents of the Bale mountains, South Ethiopia. Rendiconti Lincei 6: 157–164. Debat V, Debelle A, Dworkin I. 2009. Plasticity, canalization and developmental stability of the Drosophila wing: joint effects of mutations and of developmental temperature. Evolution 63: 2864–2876. Fedigan L, Fedigan LM. 1988. Cercopithecus aethiops: a review of field studies. In: Gautier-Hion A, Bourliere F, Gautier J-P, Kingdon J, eds. A primate radiation. Evolutionary biology of the African guenons. Cambridge: Cambridge University Press. Fontaneto D, Melone G, Cardini A. 2004. Geometric morphometrics study of the jaws in microscopic aquatic pseudocoelomates: shape diversity in the trophy of different species of Rotaria (Rotifera, Bdelloidea). Italian Journal of Zoology 71: 63–72. Fooden J, Albrecht GH. 1993. Latitudinal and insular variation of skull size in crab-eating macaques (Primates, Cercopithecidae: Macaca fascicularis). American Journal of Physical Anthropology 92: 521–538. Forsman A. 1997. Rejection of non-adaptive hypotheses for intraspecific variation in trophic morphology in gape-limited predators. Biological Journal of the Linnean Society 62: 209–223. Franklin D, Cardini A, Oxnard CE. 2010. A geometric

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

20

S. ELTON ET AL.

morphometric approach to the quantification of population variation in sub-Saharan African crania. American Journal of Human Biology 22: 23–35. Frost SR, Marcus LF, Bookstein FL, Reddy DP, Delson E. 2003. Cranial allometry, phylogeography, and systematics of large-bodied papionins (primates: Cercopithecinae) inferred from geometric morphometric analysis of landmark data. The Anatomical Record Part A: Discoveries in Molecular, Cellular, and Evolutionary Biology 275A: 1048–1072. Gilbert CC, Rossie JB. 2007. Congruence of molecules and morphology using a narrow allometric approach. PNAS 17: 2911910–2911914. Gippoliti S, Amori G. 2007. The problem of subspecies and biased taxonomy in conservation lists: the case of mammals. Folia Zoologica 56: 113–117. Groves CP. 2001. Primate taxonomy. Washington, DC: Smithsonian Institute Press. Grubb P. 2006. Geospecies and superspecies in the African primate fauna. Primate Conservation 20: 75–78. Grubb P, Butynski TM, Oates JF, Bearder SK, Disotell TR, Groves CP, Struhsaker TT. 2003. Assessment of the diversity of african primates. International Journal of Primatology 24: 1301–1357. Guillén AKZ, Barrett GM, Takenaka O. 2005. Genetic diversity among African great apes based on mitochondrial DNA sequences. Biodiversity and Conservation 14: 2221– 2233. Hair JH, Anderson RE, Tatham RL, Black WC. 1998. Multivariate data analysis. Englewood Cliffs, NJ: Prentice-Hall. Harcourt AH. 2006. Rarity in the tropics: biogeography and macroecology of the primates. Journal of Biogeography 33: 2077–2087. Harcourt AH, Parks SA. 2003. Threatened primates experience high human densities: adding an index of threat to the IUCN Red List criteria. Biological Conservation 109: 137–149. Howell DC. 2002. Statistical methods for psychology. Fifth Edition. Wadsworth: Pacific Grove, CA. James FC. 1970. Geographic size variation in birds and its relationship to climate. Ecology 51: 365–390. Jolly C. 1993. Species, subspecies and baboon systematics. In: Kimbel W, Martin L, eds. Species, species concepts and primate evolution. New York: Wiley, 67–107. Kingdon J. 2003. The Kingdon field guide to African mammals. London: Christopher Helm. Klingenberg CP. 1996. Individual variation of ontogenies: a longitudinal study of growth and timing. Evolution 50: 2412–2428. Kohn LAP, Langton LB, Cheverud JM. 2001. Subspecific genetic differences in the saddle-back tamarin (Saguinus fuscicollis) postcranial skeleton. American Journal of Primatology 54: 41–56. Lande R. 1980. Genetic variation and phenotypic evolution during allopatric speciation. American Naturalist 116: 463– 479. Lorenzen ED, Neergaard RD, Arctander P, Siegismund

HR. 2007. Phylogeography, hybridization and Pleistocene refugia of the kob antelope (Kobus kob). Molecular Ecology 16: 3241–3252. Macedonia JM, Taylor LL. 1985. Subspecific divergence in a loud call of the ruffed lemur (Varecia variegata). American Journal of Primatology 9: 295–304. Manly BFJ. 1997. Randomization, bootstrap, and Monte Carlo methods in biology. London: Chapman & Hall. Marroig G, Cheverud JM. 2005. Size as a line of least evolutionary resistance: diet and adaptive morphological radiation in New World monkeys. Evolution 59: 1128– 1142. Marroig G, Cropp S, Cheverud JM. 2004. Systematics and evolution of the jacchus group of marmosets (Platyrrhini). American Journal of Physical Anthropology 123: 11–22. Mekonnen A, Bekele A, Fashing PJ, Hemson G, Atickem A. 2010. Diet, activity patterns, and ranging ecology of the bale monkey (Chlorocebus djamdjamensis) in Odobullu Forest, Ethiopia. International Journal of Primatology 31: 339–362. Milton SJ, Zimmermann HG, Hoffmann JH. 1999. Alien plant invaders of the Karoo: attributes, impacts and control. In: Dean WRJ, Milton SJ, eds. The Karoo: ecological patterns and processes. Cambridge: Cambridge University Press, 274–287. O’Higgins P, Jones N. 2006. Tools for statistical shape analysis. Hull/York: Hull York Medical School. Osman Hill WC. 1966. Primates: comparative anatomy and taxonomy Catarrhini, Cercopithecoidea, Cercopithecinae. Edinburgh: Edinburgh University Press. Phillimore AB, Owens IPF. 2006. Are subspecies useful in evolutionary and conservation biology? Proceedings of the Royal Society of London Series B, Biological Sciences 273: 1049–1053. Postma E, van Noordwijk AJ. 2005. Gene flow maintains a large genetic difference in clutch size at a small spatial scale. Nature 433: 65–68. Rohlf FJ. 1970. Adaptive hierarchical clustering schemes. Systematic Zoology 19: 58–82. Rohlf FJ. 2007. NTSysPC. 2.20N. Ed. Setauket, NY: Exeter Software. Rohlf FJ, Marcus LF. 1993. A revolution in morphometrics. Trends in Ecology & Evolution 8: 129–132. Roth VL. 1996. Cranial integration in the Sciuridae. American Zoologist 36: 14–23. Ruvolo M. 1988. Genetic evolution in the African guenons. New York, NY: Cambridge University Press. Ryan MJ, Wilczynski W. 1991. Evolution of Intraspecific Variation in the Advertisement Call of a Cricket Frog (Acris crepitans, Hylidae). Biological Journal of the Linnean Society 44: 249–271. Sanfilippo PG, Cardini A, Hewitt AW, Crowston JG, Mackey DA. 2009. Optic disc morphology – rethinking shape. Progress in Retinal and Eye Research 28: 227– 248. Scovillea AG, Pfrendera ME. 2010. Phenotypic plasticity facilitates recurrent rapid adaptation to introduced preda-

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••

EVOLUTIONARY DIVERGENCE IN VERVET SKULLS tors. Proceedings of the National Academy of Sciences of the United States of America 107: 4260–4263. Shea BT. 1992. Developmental perspective on size change and allometry in evolution. Evolutionary Anthropology 1: 125–134. Shimada M, Terao K, Shotake T. 2002. Mitochondrial sequences diversity within a subspecies of savanna monkeys (Cercopithecus aethiops) is similar to that between subspecies. Journal of Heredity 93: 9–18. Slice DE. 1999. Morpheus, beta edn. New York, NY. Smith DG, Scott LM. 1989. Heterosis associated with regional crossbreeding between captive groups of rhesus macaques. American Journal of Primatology 19: 255–260. SPSS. 2006. SPSS for Windows, Version 15.0. Chicago, IL: SPSS Inc. Struhsaker T. 1970. Phylogenetic implications of some vocalisations of Cercopithecus monkeys. London: Academic Press. Ting N. 2008. Mitochondrial relationships and divergence dates of the African colobines: evidence of Miocene origins

21

for the living colobus monkeys. Journal of Human Evolution 55: 312–325. Tracey RL, Walsberg GE. 2000. Prevalence of cutaneous evaporation in Merriam’s kangaroo rat and its adaptive variation at the subspecific level. Journal of Experimental Biology 203: 773–781. Umera M, Lamb HF, Bonnefille R, Lezine A-M, Tierceline J-J, Gibert E, Cazet J-P, Watrin J. 2007. Late Pleistocene and Holocene vegetation history of the Bale Mountains, Ethiopia. Quaternary Science Reviews 26: 2229– 2246. White JW, Ruttenberg BI. 2007. Discriminant function analysis in marine ecology: some oversights and their solutions. Marine Ecology Progress Series 329: 301– 305. Winston J. 1999. Describing species. New York: Colombia University Press. Zelditch ML, Swiderski DL, Sheets HD, Fink WL. 2004. Geometric morphometrics for biologists: a primer. San Diego, CA: Elsevier Academic Press.

© 2010 The Linnean Society of London, Biological Journal of the Linnean Society, 2010, ••, ••–••