Global Ecology and Biogeography, (Global Ecol. Biogeogr.) (2009) 18, 248–263 Blackwell Publishing Ltd

RESEARCH PAPER

Geographical and taxonomic influences on cranial variation in red colobus monkeys (Primates, Colobinae): introducing a new approach to ‘morph’ monkeys Andrea Cardini1,2* and Sarah Elton2

1

Museo di Paleobiologia e dell’Orto Botanico, Universitá di Modena e Reggio Emilia, via Università 4, 41100, Modena, Italy, 2Hull York Medical School, The University of Hull, Cottingham Road, Hull HU6 7RX, UK

ABSTRACT

Aim To provide accurate but parsimonious quantitative descriptions of clines in cranial form of red colobus, to partition morphological variance into geographical, taxonomic and structured taxonomic components, and to visually summarize clines in multivariate shape data using a method which produces results directly comparable to both univariate studies of geographical variation and standard geometric morphometric visualization of shape differences along vectors. Location Equatorial Africa. Methods Sixty-four three-dimensional cranial landmarks were measured on 276 adult red colobus monkeys sampled over their entire distribution. Geometric morphometric methods were applied, and size and shape variables regressed onto geographical coordinates using linear and curvilinear models. Model selection was done using the second-order Akaike information criterion. Components of variation related to geography, taxon or their combined effect were partitioned using partial regresssion. Multivariate trends in clinal shape were summarized using principal components of predictions from regressions, plotting vector scores on maps as for univariate size, and visualizing differences along main axes of clinal shape variation using surface rendering. Results Significant clinal variation was found in size and shape. Clines were similar in females and males. Trend surface analysis tended to be more accurate and parsimonious than alternative models in predicting morphology based on geography. Cranial form was relatively paedomorphic in East Africa and peramorphic in central Africa. Most taxonomic variation was geographically structured. However, taxonomic differences alone accounted for a larger proportion of total explained variance in shape (up to 40%) than in size (≤ 20%). Main conclusions A strong cline explained most of the observed size variation and a significant part of the shape differences of red colobus crania. The pattern of geographical variation was largely similar to that previously reported in vervets, despite different habitat preferences (arboreal versus terrestrial) and a long period since divergence (c. 14–15 Myr). This suggests that some aspects of morphological divergence in both groups may have been influenced by similar environmental, geographical and historical factors. Cranial size is likely to be evolutionarily more labile and thus better reflects the influence of recent environmental changes. Cranial shape could be more resilient to change and thus better reflects phylogenetically informative differences.

*Correspondence: Andrea Cardini. E-mail: [email protected]

248

Keywords Akaike information criterion, Central Africa, clinal variation, cranial shape, cranial size, curvilinear models, geometric morphometrics, Piliocolobus, partial regression, surface rendering.

DOI: 10.1111/j.1466-8238.2008.00432.x © 2009 The Authors Journal compilation © 2009 Blackwell Publishing Ltd www.blackwellpublishing.com/geb

‘Morphing’ red colobus cranial variation INTRODUCTION Understanding and appreciating clinal variation in size and shape is a key aspect of biogeographical research. Recent studies have demonstrated the utility of advanced morphometric techniques in the study of clinal variation (Fadda & Corti, 2001; dos Reis et al., 2002; Frost et al., 2003; Monteiro et al., 2003; Santos et al., 2004; Cardini et al., 2007). However, although size differences are relatively easy to illustrate, it is much more challenging to accurately display spatial variation in shape. Visualizing differences is considered to be one of the strengths of geometric morphometrics, so developing efficient methods for illustrating multivariate predictions of clinal shape is crucial to the development of morphometric research within biogeography. A fundamental aim of this paper is therefore to report a new method that uses a principal components analysis of clinal shapes to produce a visual summary that is directly comparable with results from classical biogeographical studies of single variables (e.g. size). By building on previous work (Cardini et al., 2007), this method takes full advantage of powerful visualization tools like surface rendering to illustrate three-dimensional shape variation along vectors. We apply this new method to a geometric morphometric study of clinal variation in a group of African primates, the red colobus monkeys, and examine the effects of geography alongside taxonomy to investigate the factors that might influence their cranial form. Red colobus [Procolobus (Piliocolobus) de Rochebrune 1887 – Primates, Cercopithecidae] are medium-sized colobine monkeys that are patchily distributed within the moist lowland forests of west, central and east equatorial and tropical Africa (Davies & Oates, 1994). They are currently considered to show more biological variability than is found in a single species and the prevailing trend is to treat them as a superspecies divided into assemblages (Grubb et al., 2003; Grubb, 2006). Grubb et al. (2003) recognized four main Piliocolobus groups, which largely reflect the allopatric or parapatric distribution of their members (Fig. 1a): (1) Piliocolobus badius, consisting of three populations living in West Africa from Senegal to Ghana; (2) Piliocolobus pennantii, including four populations from western equatorial Africa; (3) several populations found to the east of the range of P. pennatii, which together form the ‘central equatorial African assemblage’ (of these, Piliocolobus sp. ellioti, Piliocolobus sp. foai, Piliocolobus sp. oustaleti, Piliocolobus sp. tephrosceles and Piliocolobus sp. tholloni were included in the study reported here); (4) an eastern assemblage comprising three small and isolated populations in Kenya and Tanzania (Piliocolobus gordonorum, Piliocolobus kirkii and Piliocolobus rufomitratus). Geographical factors probably played a major role in producing the morphological variation and taxonomic complexity evident in red colobus today. Within West and Central Africa, red colobus populations tend to live in reasonably close proximity and are separated to a variable extent by rivers or mountains. These often contiguous populations presumably diverged because of reductions in gene flux during Pleistocene glacials (Verheyen, 1957, 1962; Rodgers et al., 1982; Colyn, 1991; Gautier-Hion et al., 1999). Mayr & O’Hara (1986) and Hamilton

(1988) argued that aridity during glacial maxima caused tropical forest contractions and confined forest fauna to refugia. These potentially included highland areas (Mayr & O’Hara, 1986) like the Fouta Djallon highlands (Guinea, West Africa), the Adamawa Plateau (Cameroon, western equatorial Africa) and the mountains between the Rift Valley and the Lualaba River (central and eastern equatorial Africa). From there, according to this hypothetical scenario, lowlands were recolonized during interglacials, when forests expanded. Indeed, ranges of assemblages of red colobus (Grubb et al., 2003) are more or less centred around those mountain refugia. Forest expansion interrupted the process of divergence and created contact areas like those found today in some regions of Central Africa (see map in Gautier-Hion et al., 1999, p. 81). Patches of lowland forests along major rivers could also have allowed the survival of small populations of forest animals (Colyn, 1991; Colyn et al., 1991). River barriers were and still are crucial to the evolution of red colobus and other Central African primates (Colyn, 1991). The existence of several differentiated taxa within interfluvial blocks of central equatorial Africa suggest that during the last dry climatic period, populations may have survived in isolated patches of lowland forest within the Congo– Lualaba river basin (Colyn, 1991). Thus, in this region, primate diversity might not only be due simply to emigration during times of forest expansion from a major mountain refugium in the Rift Valley, but also to the survival of islands of lowland forests during the arid interglacials. Phylogenetic reconstructions based on vocalizations (Struhsaker, 1981) give tentative support to the idea that refugia, particularly mountain regions, acted as major originating centres for modern red colobus populations. Interestingly, the sole published molecular phylogenetic analysis (Ting, 2008) is fairly congruent with this earlier phylogeny, supporting an initial Late Pliocene split of the red colobus clade into three main lineages, which approximately correspond to the western, western equatorial and central equatorial/eastern assemblages, followed by Pleistocene radiations within the three clades. Quantifying clinal variation in cranial form is an important element in understanding Piliocolobus population history and biogeography. Although studies of cranial variability within red colobus were conducted in the past (Verheyen, 1957; Colyn, 1991), their main purpose was to assess the validity of taxa proposed on the basis of pelage colour rather than examining broader spatial patterns. As briefly reviewed in Cardini et al. (2007), several primate species, including Brazilian tufted-eared marmosets, Malagasy sifakas, Kenyan vervets, pig-tailed and crab-eating macaques in Southeast Asia, Japanese macaques, African baboons and vervets are featured in the vast literature on clinal and ecogeographical size variation. However, only a few of these studies investigated African monkeys and none of them concerned strictly arboreal species. Therefore, investigation of clinal variation will not only shed light on red colobus morphological variation but will also help to determine whether any general patterns exist within the African primates studied to date. Although temperature is highlighted as being a contributor to size in higher-latitude Asian monkeys (e.g. Albrecht, 1982; Fooden & Albrecht, 1993; Rae et al., 2003), within African

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

249

A. Cardini and S. Elton

Figure 1 (a) Distribution of red colobus taxa (modified from Colyn, 1991). (I) Piliocolobus badius (western Tropical Africa): 1, Piliocolobus badius temminckii; 2, Piliocolobus badius badius; 3, Piliocolobus badius waldroni; (II) Piliocolobus pennatii (western equatorial Africa): 4, Piliocolobus pennatii epieni; 5, Piliocolobus pennatii pennantii; 6, Piliocolobus pennatii preussi; 7, Piliocolobus pennatii bouveri; (III) central African assemblage: 8, Piliocolobus sp. tholloni; 9, Piliocolobus sp. oustaleti; 10, Piliocolobus sp. parmentieri; 11, Piliocolobus sp. lulindicus-foai; 12, Piliocolobus sp. langi-ellioti; 13, Piliocolobus sp. tephrosceles; (IV) eastern African species: 14, Piliocolobus gordonorum; 15, Piliocolobus rufomitratus; 16, Piliocolobus kirkii. Grey areas are putative Pleistocene mountain refugia from Mayr & O’Hara (1986). (b) Geographical distribution of Piliocobus female and male specimens with different symbols according to taxonomy. Hammer–Aitoff equal area map projection is used in this and the following figures.

monkeys, temperature appears to influence size and/or shape less than other environmental factors, including increased access to food (Turner et al., 1997) and rainfall as a proxy of habitat productivity (see Cardini et al., 2007, and references therein for a short review on this subject). Elton (2008) argued that the

250

differing relationships between primate morphology and environment in Africa and Asia are logical, given that precipitation is a more important component of climatic variation than temperature at low latitudes (deMenocal & Bloemendal, 1995).

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

‘Morphing’ red colobus cranial variation The extensive geographical distribution of the African red colobus from southern Senegal in the west to Kenya and Tanzania in the east makes this taxon an ideal candidate for the investigation of clinal variation in a tropical forest monkey. Thus, three-dimensional coordinates of anatomical landmarks were measured on crania of a large sample of red colobus including most populations, all endangered East African taxa and at least one representative for each of the main assemblages. Geometric morphometrics methods were then applied alongside a new method of visualization to investigate: whether any significant cline in size (1) or shape (2) occurred, using linear and curvilinear regression models and applying information criteria for model selection; (3) whether components related to taxon, geography or their combined effect could be partitioned using partial regression to better understand the relationship between geographical and taxonomic differences.

Table 1 Taxa used in the analysis and sample sizes. n Genus/subgenus

Species

Subspecies

Females

Males

Procolobus Piliocolobus

badius

badius temminckii waldroni – – bouveri epieni preussi – ellioti foai oustaleti tephrosceles tholloni

36 10 15 4 33 1 – 31 5 16 3 9 7 4

23 4 6 – 10 – 1 10 1 18 6 5 16 2

gordonorum kirkii pennantii

rufomitratus sp.*

MATERIAL AND METHODS Sample The sample, described in Nowak et al. (2008), comprised 276 adult specimens (of which 174 were female and 102 male) derived from museum and field collections in Zanzibar. Following the classification scheme of Grubb et al. (2003), taxa were identified on the basis of geographical distribution (taken from the distribution maps in Colyn, 1991) and the published taxonomy of particular specimens (Colyn, 1991). These assignments were generally congruent with information from museum catalogues. Any specimens that could not be taxonomically assigned with confidence were excluded from the sample, and the utmost care was taken to correctly identify species and define a priori groups, a necessary step given the unstable and unclear taxonomy of red colobus monkeys. As detailed in Table 1 and shown in Fig. 1(b), most groups of Piliocolobus monkeys were included in the sample. These groups were then analysed as if all populations within Piliocolobus were of equal taxonomic rank (i.e. without discriminating between putative species versus subspecies – see below). Data collection and geometric morphometrics One of us (A.C.) collected three-dimensional coordinates of anatomical landmarks on crania and mandibles using a 3D digitizer (MicroScribe 3DX, Immersion Corporation). Landmarks were digitized only on the left side to avoid redundancy of information in symmetric structures and increase the number of measured specimens. The set (configuration) of 64 landmarks used for the analysis is shown in Fig. 2; landmarks are described in Nowak et al. (2008). These correspond to a subset of cranial landmarks used in a series of studies on skull variation in Old World monkeys (Cardini et al., 2007; Cardini & Elton, 2008a,b,c) and already employed to analyse variation in red colobus (Nowak et al., 2008). As discussed in Nowak et al. (2008), a very small percentage of specimens with one to four missing landmarks (substituted

*The following populations were considered of uncertain taxonomic status by Grubb et al. (2003).

with within-sex species means), alongside negligible measurement error, did not introduce any appreciable error in either size or shape. Geometric morphometric analyses were performed using Morpheus (Slice, 1999), TPSSmall 1.20 (Rohlf, 2006a), NTSYSpc 2.2L (Rohlf, 2006b) and Morphologika (O’Higgins & Jones, 2006). Variations in the form of the landmark configurations were examined using Procrustes-based geometric morphometrics, rather than angle- or distance-based approaches, because this method has desirable statistical properties as detailed in Franklin et al. (2008) and provides an efficient separation of size and shape components of form differences (Adams et al., 2004). Further, by applying appropriate mathematical functions (e.g. the thin plate spline) to warping of images or grids, they allow localization of shape differences between pairs of landmark configurations. An extensive introduction to the theory of geometric morphometrics and its applications in biology is provided in Cardini et al. (2007) and Cardini & Elton (2008a,c). Shape variables Due to the large degree of sexual dimorphism evident in colobines (Cardini & Elton, in press), all analyses were performed using split-sex samples. Using separate sexes may result in a loss of statistical power within smaller samples but it has at least two advantages. First, it avoids the use of corrections for sexual dimorphism, which often makes results less easy to interpret. Second, it allows the verification of observed patterns through the comparison of results from females and males. Whenever congruencies are found, findings are corroborated and confidence is subsequently increased. A principal components analysis (PCA) of shape variables was used, which identifies the axes of greatest variation in a sample,

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

251

A. Cardini and S. Elton Statistical analyses: geographical variation in size and shape

Figure 2 Landmark configuration (modified from Cardini et al., 2007).

to reduce dimensionality in the analysis of shape coordinates. Thus, the number of variables was reduced by including only the first several principal components of the 192 shape coordinates from the Procrustes analysis of the raw data. The number of principal components to be analysed was selected by measuring the correlation between the matrix of Procrustes shape distances in the full shape space and pairwise Euclidean distances in the reduced shape space (5, 10, 15 principal components, and so on). Plots of correlation coefficients onto the number of components can be used in a similar way to scree plots to select how many variables summarize most shape variation (Fadda & Corti, 2000; Cardini et al., 2007). Depending on the data set used, 30–35 principal components provided an accurate but reasonably parsimonious summary of total shape variation (> 80% of total variance; correlation with distances in the full shape space ≥ 0.995). Issues with non-homogeneous samples due to sexual dimorphism or an exceedingly large number of shape variables in statistical analyses are relatively easy to address (as shown above). Taking into account the non-independence of samples due to phylogeny in statistical tests, in contrast, is less straightforward. Rohlf (2006c) recently reviewed phylogenetic comparative methods and suggested that some problems could be addressed by using more complex models. None of these, however, could have been fruitfully applied in our analysis, because a phylogeny including all study taxa is not yet available. The very recent publication of Ting’s (2008) molecular analysis of phylogenetic relationships among African colobines represents a big step forward in this respect, but does not include about a third of the populations analysed in our study. Thus, at present, benefits of a better taxonomic coverage and avoidance of unnecessary loss of information (with concomitant big gaps in the spatial distribution of our data) outweigh the potential advantage of applying phylogenetic comparative methods to a small set of taxa.

252

Trend surface analysis (TSA; Legendre & Legendre, 1998; Ruggiero & Kitzberger, 2004; Botes et al., 2006; Cardini et al., 2007) was used for fitting geographical coordinates to variation in skull size and shape, taking into account nonlinearities. Thus, size or shape variables were regressed onto a third-order polynomial of latitude and longitude and non-significant terms were removed one by one until all terms of the multiple regression were significant. Results of TSA were compared with those of both simple linear models and the full third-order polynomial expansion of geographical coordinates to decide whether TSA was both parsimonious and informative compared with alternative models. Thus, size or shape variables were regressed onto geographical coordinates, and the second-order Akaike information criterion (AICc; Burnham & Anderson, 1998; Mazerolle, 2004) was used to compare the goodness of fit of the models. AICc is a measure based on information theory and derived from the concept of entropy in physics. Briefly, it measures the lack of fit of the data (sum of squared residuals in a regression) to a given model, where the model is penalized in proportion to the number of parameters it employs. Thus, the best model compared with available alternatives is the one with the lowest AICc (AICcmin). For multivariate data, Burnham & Anderson (1998) suggested that AICc = AIC + 2k(k + v)/(N × p – k – v), where AIC = N × log(residual sum of squares/N) + 2k, k is the number of parameters in the model, N is sample size and 1 ≤ v ≤ p(p + 1)/ 2, with p the number of dependent variables. In this study, AICc values were always calculated using v = 1, as in the univariate case. This was done for simplicity as AICc values (not shown) obtained with v = p(p + 1)/2 were virtually identical to those with v = 1. Finally, the relative level of support of different models was evaluated by ΔAICc = AICc– AICcmin and Akaike weights (Burnham & Anderson, 1998). Akaike weights provide another measure of the strength of evidence (likelihood) for each model and approximately represent the ratio of ΔAICc values for each model relative to the whole set of candidate models. Burnham & Anderson (1998) suggest that models with ΔAICc values of 0–2 provide similar support, whereas ΔAICc > 2 indicate substantially less support than the best model. Specimens were plotted according to geographical coordinates on a map of Africa using Arcview GIS 3.2 (ESRI, Redlands, CA). Clinal variation predicted by the selected model was illustrated with grey-scale symbols on the map. Size, which is univariate, can be easily described by a single variable. Thus, grey symbols of a tone proportional to the size of the skull predicted by geography were used. The visualization of nonlinear clines is less straightforward for shape, which is multivariate. Thus, we developed a modified version of the method described by Cardini et al. (2007) to more effectively summarize the main trends of clinal variation and allow their visualization using surface rendering of shapes described by three-dimensional anatomical landmarks. As for size, clinal variation in shape predicted by the selected model was first computed and scores of predictions were saved. The variables

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

‘Morphing’ red colobus cranial variation Table 2 Clinal variation in cranial size: comparison of different models using percentages of size variance explained (R2, squared correlation coefficient; adjusted R2, squared correlation coefficient adjusted for the number of predictors) by geography, significance of tests, second order Akaike information criterion (AICc), delta AICc (ΔAICc = AICc – AICcmin) and Akaike weights (Wi). 2

Sex

Predictors*

R2 (%)

R adj. (%)

d.f.1

d.f.2

F

P

AICc

ΔAICc

Wi

Females

x 2, x 3, y 2, xy, x 2y Full polynomial x, y x, x 2, y 2, xy, x 2y, xy 2 Full polynomial x, y

60.4 61.8 12.1 71.2 71.2 11.0

59.2 59.7 11.1 69.3 68.4 9.2

5 9 2 6 9 2

168 164 171 95 92 99

51.293 29.463 11.751 39.049 25.251 6.088

4.3 × 10–32 4.9 × 10–30 0.00002 1.4 × 10–23 2.9 × 10–21 0.0032

340.9 347.2 394.4 200.5 207.9 240.8

0.0 6.3 53.5 0.0 7.4 40.3

0.959 0.041 0.000 0.975 0.025 0.000

Males

*In this and the next table, significant terms in the trend surface are shown using x for longitude and y for latitude.

describing the predicted cline in shape were then subjected to a PCA (‘geo-shape PCA’ or gsPCA) in order to summarize most of the variation predicted by the model with a few variables. Eventually, variation along gsPCA axes (gsPC1 and gsPC2), which together accounted for most of the clinal variation in shape, was illustrated using both (a) grey-scale colour symbols of a tone proportional to the score of gsPC1 (or gsPC2) for individuals plotted on a map of Africa and (b) surface renderings of shapes corresponding to individuals at opposite extremes of gsPC1 (or gsPC2). To aid interpretation of (b) relative to (a), shapes were shown using the same grey tone as for symbols on the map. Thus, for instance, if lowest scores on gsPC1 were shown using light grey symbols on the map, light grey was also used to for surface rendering of the shape predicted for the negative extreme of gsPC1. This allowed: (1) mapping of clinal shape in a fashion similar to clinal size, and (2) visualization of geographical shape variation as is commonly done in geometric morphometrics by using predictions for shapes along a vector (e.g. for PCA axes Milne & O’Higgins, 2002; Franklin et al., 2007a; Wroe & Milne, 2007; Cardini & Elton, 2008a,c; and for ontogenetic vectors O’Higgins & Jones, 1998; Collard & O’Higgins, 2001; Cardini & Thorington, 2006; Cobb & O’Higgins, 2007; Franklin et al., 2007b). Statistical analyses: taxonomic and geographical components of size and shape variation Partial linear regression was used to assess the effects of spatial structuring of variables and estimate the amount of skull size or shape variation attributable to one set of factors (taxon) once the effects of the other factors (geography) were taken into account. This method is commonly used in biogeographical studies to partition the effects of geography and environment (Legendre & Legendre, 1998; Ruggiero & Kitzberger, 2004; Botes et al., 2006; Cardini et al., 2007). Thus, geographical predictors selected in steps 1–2 (spatial component) and grouping variables for populations were combined. Morphological variation was partitioned into four components: (1) non-taxonomic spatial (proportion of variance exclusively explained by geography); (2) spatially structured taxonomic (proportion of variance explained by both geography and taxon); (3) non-spatial taxonomic (proportion of variance explained exclusively by

taxon); (4) unexplained variation (proportion of variance explained by the effect of other factors). RESULTS Clinal variation in size Significant clinal variation in cranial size was found both in females and males (Table 2). The inclusion of curvilinear terms increased the proportion of variance in size explained by geography from about 10% to about 60–70%. TSA had the lowest AICc, and ΔAICc was close to seven for polynomial models and larger than 40 for the simple linear regressions on longitude and latitude. Thus, these models were either weakly supported (range 3–7) or very unlikely (> 10) according to the guidelines suggested by Burnham & Anderson (1998). Akaike weights of TSA models were larger than 0.95. This indicates that, given the data, they had a more than 95% chance of being the best models among those considered in the analysis. Thus, only results from TSA were considered. Patterns of clinal variation in size predicted by TSA are shown in Fig. 3. Visual inspection of plots (Fig. 3a,b) and also the high correlation (r = 0.753, P = 0.002) between female and male predictions of clinal size for geographical coordinates corresponding to population centroids (i.e. averages of geographical coordinates within each population) suggested a very similar trend in both sexes. Cranial size was intermediate in West Africa, larger in central equatorial Africa and smaller in East Africa. However, in males there was a progressive increase in size from west to central equatorial Africa, whereas in females, size increased from Senegal to Cameroon, decreased slightly in western Congo and finally increased again in eastern Congo and Uganda. The most conspicuous reduction in size in both sexes occurred in East Africa, and this finding was supported even after excluding the two smallest eastern populations (P. kirkii and P. rufomitratus) from the analysis. Clinal variation in shape Clinal variation in shape was highly significant in both sexes (Table 3). As for size, the inclusion of curvilinear terms led to an

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

253

A. Cardini and S. Elton

Figure 3 Patterns of clinal variation in female (a) and male (b) size of Piliocobus: predictions of trend surface analysis (TSA). Grey scale of symbols is according to increasing size. Table 3 Clinal variation in skull shape: comparison of different models* using percentages of shape variance explained (% ex.), significance in multivariate tests, second-order Akaike information criterion, delta AICc and Akaike weights. See Table 2 for abbreviations. Sex

Predictors

% ex.

Wilks λ

F

d.f.1

d.f.2

P

AICc

ΔAICc

Wi

Females

x, x 2, x 3, y, y 2, xy, x 2y, xy 2 Full polynomial x, y x, x 2, y 2, xy, x 2y, xy 2 Full polynomial x, y

26.7 27.2 10.2 25.1 32.0 8.8

2.3 × 10–4 1.8 × 10–4 6.5 × 10–2 3.4 × 10–4 7.6 × 10–5 5.9 × 10–2

7.120 6.188 11.426 6.340 4.185 7.306

280 315 70 180 270 60

1040.9 1157.8 274.0 397.7 566.0 140.0

8.7 × 10−122 1.3 × 10−116 2.2 × 10−50 4.8 × 10−53 8.6 × 10−47 2.5 × 10−22

–444.2 –442.3 –441.9 –250.3 –247.1 –250.8

0.0 1.8 2.3 0.5 3.7 0.0

0.581 0.232 0.187 0.402 0.082 0.516

Males

*Multivariate multiple regressions were performed using the first 35 principal components of shape variables for females and the first 30 for males. The same number of principal components was used in the partial regression (Table 4).

evident increase in the amount of shape variance explained by geography (from 10% or less to more than 25%). However, AICc was fairly similar across models. TSA was the best model in females and ΔAICc was about 2 for linear and full polynomial

254

models. In males, however, the linear model had the lowest AICc but was about as good as TSA (ΔAICc < 2) whereas the full polynomial model was substantially less supported than the other two (ΔAICc > 2). Thus for shape, information criteria were not as

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

‘Morphing’ red colobus cranial variation conclusive as for size in model selection. However, TSA was almost three times more likely than alternative models in females and about as likely as the linear regression in males. Thus, as TSA produced a much better fit to the data and was either more likely or as likely as alternative models, this model was adopted in all analyses. A visual summary of the pattern of clinal shape variation according to predictions of the trend surface analysis is shown in Figs 4 and 5 using the first two gsPCs which together explained about 65% of clinal shape variation. As for size, patterns of female and male ‘geographical shape’ variation were largely congruent. This was suggested by the visual inspection of plots (Figs 4 & 5) and also by the high correlation (r = 0.827, P = 0.0001) between matrices of shape distance based on female and male predictions of clinal shape for geographical coordinates corresponding to the population centroids from our sample. In both sexes clinal variation summarized by gsPC1 was largely congruent with the cline in size. This was suggested again by evident similarities in patterns (Figs 3 and 4), and by a very high correlation between gsPC1 and predictions of TSA for size (rfemales = 0.932, P < 0.00001; rmales = 0.981, P < 0.00001). Consistently, populations with the smallest and largest size tended to score lowest and highest on gsPC1. This means that in both sexes, the small P. kirkii was on one extreme of gsPC1 and either the large P. p. preussi (females) or specimens from eastern Congo and Uganda (males) were on the opposite extreme. Clinal variation was visualized by morphing shapes along gsPC1. Thus, East African populations tended to have a short orthognathous face with large orbits and a relatively small palate. The neurocranium, in contrast, was long and wide. Specimens from central equatorial Africa were prognathous with a deep upper jaw and fairly short nasals, and had a small neurocranium with a pronounced narrowing of the anterior temporal fossa. Individuals from West Africa were somewhat intermediate between east and central equatorial populations. Clinal variation summarized by gsPC2 also followed a direction common to both sexes. This time the vector mostly picked up differences between East and West African populations, and mainly concerned facial depth (more pronounced in the west) and nasal length (longer in the east). In the west this trend was most pronounced in P. badius badius but was also evident in P. badius temminckii. Pilocolobus badius waldroni, in contrast, scored next to zero on gsPC2 and was thus intermediate between the other two western taxa and the eastern populations. Taxonomic and geographical components of size and shape variation Results of partial regressions used to partition components of size and shape variation are shown in Table 4. With the exception of the non-spatial taxonomic component of size, patterns were very similar in females and males. Spatially structured taxonomic variation accounted for most of the variation in size (c. 60–70%). Less than 30% of total variance remained unexplained. Non-taxonomic spatial variation, which only relates to geography, was very small

Table 4 Partial regressions of size and shape onto geography and taxon: percentages of variance explained by different components. Size

Shape

Components

Females

Males

Females

Males

(1) Geography only (2) Common (3) Taxon only (4) Unexplained

2.2 58.2 12.7 26.9

2.0 69.2 2.8 26.0

4.2 22.6 9.0 64.2

5.6 19.6 14.5 60.4

(c. 2%). In contrast, non-spatial taxonomic variation in size was fairly large in females (c. 13%) but small in males (c. 3%). Indeed, in the full model (spatial and taxonomic predictors all included), the effect of taxon was significant in females but not in males (females, F12,156 = 6.117, P = 9.7 × 10–9; males, F11,84 = 0.831, P = 0.609). For shape, results for females and males were almost identical. Most of the variance remained unexplained (c. 60%). As for size, spatially structured taxonomic variation accounted for more variation (c. 20%) than either non-taxonomic spatial (c. 5%) or non-spatial taxonomic (c. 9–15%) variation. However, for shape the non-spatial taxonomic component accounted for up to half of the total amount of explained variance, whereas for size it was only up to one-sixth. In the full model, the effect of taxon on shape was highly significant in both sexes (females, Wilks λ = 0.00476, F420,1394.5 = 1.992, P = 9.9 × 10–21; males, Wilks λ = 0.000302, F330,602.1 = 2.167, P = 1.2 × 10–16). DISCUSSION ‘Geo-shape PCA’ and the visualization of clinal shape The increasing use of complex, multivariate morphological data in biogeographical studies (e.g. Frost et al., 2003; Cardini et al., 2007) has led to the need for efficient methods for illustrating multivariate predictions of clinal shape. We suggest that the ‘geo-shape PCA’ proposed in this paper is an effective way to summarize and visualize clinal shape variation in geometric morphometric studies. Geometric morphometrics is extensively used and has a number of advantages over traditional morphometrics. It is powerful in statistical tests and efficient in separating size and shape components of form, and also allows the use of various kinds of diagrams for an effective visualization of shape variation in terms of its geometry (Adams et al., 2004). Many classical applications of statistical analyses in biology were developed in order to predict a single variable using one or more explanatory predictors. Often, these methods can be easily extended to multivariate shape data. However, the last step of the analysis, where shape differences are summarized and findings visualized back in the space of the landmark configuration (Bookstein, 2000), is not necessarily straightforward, as exemplified by studies of nonlinear clinal variation. Predictions from clines need to be plotted on a geographical map to show where

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

255

A. Cardini and S. Elton

Figure 4 (a, b) Patterns of clinal variation in female (a) and male (b) shape of Piliocobus: predictions of TSA summarized by ‘geo-shape principal components’ (gsPCs). The first major axis (gsPC1) is shown which summarizes 42.9% (females)/41.5% (males) of clinal variation in shape predicted by the best fit trend surface. Grey scale of symbols on the map is according to increasing gsPC score. Shape changes at extremes of the axis (maximum above and minimum below) are shown with surface renderings using three views (lateral, frontal and ventral). Consistent with the grey scale of symbols on the map, dark and light grey tones are used for shapes corresponding, respectively, to the positive (largest gsPC score – females and males, respectively, 0.0258, 0.0357) and negative (smallest gsPC score – females and males, respectively, −0.0271, −0.0373) extremes of the axis.

256

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

‘Morphing’ red colobus cranial variation

Figure 5 (a, b) Patterns of clinal variation in female (a) and male (b) shape of Piliocobus: predictions of TSA. The second major axis (gsPC2) is shown which summarizes 23.7% (females)/23.5% (males) of clinal variation in shape predicted by the best fit trend surface. Grey scale of symbols on the map is according to increasing gsPC score. Shape changes at extremes of the axis (maximum above and minimum below) are shown with surface renderings using three views (lateral, frontal and ventral). Consistent with the grey scale of symbols on the map, dark and light grey tones are used for shapes corresponding respectively to the positive (largest gsPC score – females and males, respectively, 0.0161, 0.0238) and negative (smallest gsPC score – females and males, respectively, –0.0232, –0.0220) extremes of the axis.

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

257

A. Cardini and S. Elton and how the dependent variable changes as a function of geographical coordinates. This is very easy when there is only one dependent variable, like size, which can be plotted using different symbols/colours for different scores of predictions. In contrast, because each shape is described by many variables, shapes predicted by locality cannot be illustrated simply by varying symbols on a map, even if there is only one shape for each locality. One way to visualize clinal shape would be to directly plot shapes (instead of symbols) on the map. This would be analogous to plotting scores (i.e. numbers instead of grey-scale symbols) for size predictions. However, this would lead to figures overcrowded with shape diagrams or numbers which would be difficult or even impossible to interpret. An interactive visualization may provide a much better alternative, where one can simply drag a pointer on a digital map and display clinal shape using diagrams that change as a function of geographical coordinates. This is similar to visualization tools already available in standard statistical software like Morphologika (O’Higgins & Jones, 2006) or TPSRelw (Rohlf, 2007) to show shapes in any point of the two-dimensional space implied by any pair of PCs from a PCA of the original superimposed shape coordinates. However, for publications, this method requires an electronic format with either specific software or digital animations, and is not applicable to standard articles printed on paper or made available as pdf files. To circumvent this problem, Cardini et al. (2007) developed a method which employs a k-means cluster analysis of predicted shapes and surface rendering diagrams for cluster means to respectively summarize and visualize the main trends of clinal shape variation. Thus, they used a k-means cluster analysis on variables which described the predicted cline in shape in order to discriminate homogeneous groups of shapes predicted by geography. Groups obtained in the k-means cluster analysis were then plotted on a map using different symbols, and variation within each group of ‘geographical shapes’ was summarized by using its mean, which was then visualized using surface rendering. This method can be seen simply as a shortcut to produce a summary of a continuous pattern of multivariate clinal shape variation on paper. However, the effectiveness of this visualization will partly depend on the arbitrary choice of an appropriate number of a priori groups for the k-means cluster analysis, which cannot be too large (accurate but not easy to interpret) or too small (inaccurate but easier to interpret). Also, the outcome will not be directly comparable to either the output of the analysis of clinal variation in size or to any standard visualization of shape differences using predictions of shapes for extreme points along main axes of variation. This way of displaying shape differences along vectors is standard in geometric morphometrics and customary in studies like those involving ordinations or regressions (see Methods). The ‘geo-shape PCA’ method proposed in this study overcomes some of these difficulties by taking clinal shape predictions, doing a second PCA on these scores and using resulting PCs (gsPCs) for both plots on maps and surface rendering along vectors. Thus, the main component of clinal shape can be shown on a map using grey-scale symbols of a tone proportional to gsPC1 scores, as was done for clinal size and as is

258

commonly done in standard software for the analysis of spatial data (e.g. sam, Rangel et al., 2006). Further components (gsPC2, gsPC3 etc.) are similarly illustrated in separate plots. Correlations between gsPCs and clinal size can be used to quickly summarize the degree of congruence of size and shape clines. Finally, shapes which correspond to opposite extremes of clinal variation summarized by gsPC1 (gsPC2, gsPC3 and so on) are computed and plotted next to the map with, for instance, dark grey symbols corresponding to the positive extreme and light grey to the negative one. This last operation strictly mirrors the ‘standard’ way of visualizing shape variation along a vector (as found in PCs, canonical variates, partial least squares and multivariate regression vectors) and achieves the last goal of geometric morphometric analyses: ‘Shape differences are summarized by multivariate analysis of shape coordinates, and findings are visualized by quantitative diagrams back in the plane or space where the landmark data originated’ (Bookstein, 2000). In conclusion, gsPCA provides a simple way for summarizing and visualizing clinal shape variation, produces results directly comparable to those of classical biogeographical studies of single variables (e.g. size) but takes full advantage of powerful visualization tools like surface rendering by using the same underlying techniques commonly employed for visualization of shape differences in geometric morphometric analyses (i.e. diagrams of shapes predicted for opposite extremes of a vector). Clinal size and shape, and geographical and taxonomic components of variation Significant clinal variation was found in red colobus crania. The cline in size was particularly pronounced and explained most of the variance in the sample. The cline in shape was also significant but accounted for a smaller proportion of shape variance. For both size and shape, the major trend was longitudinal rather than latitudinal, although latitude did contribute to some of the variance explained by both the linear and curvilinear models. The dominance of longitude in clines may be partly explained by the fact that the major axis of the distribution range is also longitudinal (about three times longer than latitude). However, Cardini et al. (2007) also reported a prevalence of longitude in determining a morphological cline in vervet skulls. Vervet populations are found over most of sub-Saharan Africa, in contrast to red colobus which are confined to tropical forests in the equatorial belt. Thus, in vervets, the longitudinal extension of the range is about the same as the latitudinal one and a larger contribution of longitude to the vervet cline cannot simply be explained by the length of the main geographical distribution axis. The fact that, despite ecological differences, most geographical variation occurred both in vervets and red colobus on a west to east axis is especially interesting if one considers that, generally, the largest differences in temperature are latitudinal. Bergmann’s rule predicts that size increases with latitude (as a surrogate for temperature) but, if cranial size is taken as a proxy for body size, neither of these primate taxa appear to follow this rule. In vervets there was only a weak trend towards size increase in populations

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

‘Morphing’ red colobus cranial variation further from the equator, whilst in red colobus the largest individuals were found around the equator with populations north and south of it generally being smaller. Closer inspection of the longitudinal cline in red colobus size reveals further similarities with vervets. In both groups, the largest individuals are mostly from central equatorial Africa in a range that goes from the west coast (c. 10° E) to the Rift Valley (c. 30° E). East of the Rift Valley, red colobus size becomes much smaller, again mirroring the vervet trend. This pattern was evident even when the insular (and small) P. kirkii and the small mainland P. ruformitratus were excluded from the analysis. Cardini et al. (2007) found rainfall together with seasonality to be consistently important predictors of size in vervet skulls. Rainfall, as a proxy for habitat productivity, might thus be central to intraspecific size variations in other primate groups, even those that are phylogenetically well separated or have significant ecological differences. To put it simply, animals may grow larger where productivity is higher, and this association is expected for both terrestrial and arboreal taxa. Since, the largest representatives of Piliocolobus inhabit central African regions with highest average annual rainfall (Fig. 6b), the effects of productivity on body size could help to explain some of the similarities in the red colobus and vervet clines. Nonetheless, the general patterns observed in red colobus and vervets are not identical, and the links with rainfall and hence productivity are not equally strong across the red colobus range. In high-rainfall west tropical Africa, from Ghana to Senegal, red colobus are medium-small compared with other Piliocolobus, whereas vervets are comparatively large. It is possible that environmental or ecological variables other than rainfall and temperature (or indeed interaction effects between variables) influence red colobus size in West Africa. Consideration of additional variables in future work may help to fully understand similarities and incongruences between clines in cranial size of red colobus and vervets. It is also possible (indeed probable) that the patterns we observe today not only reflect adaptations to the environment in the present or in the recent past but also events that took place earlier in the evolutionary history of Piliocolobus. The intense climatic fluctuations of the Pleistocene and attendant changes to African habitats including tropical forest may have been an important factor in primate speciation and differentiation (Hamilton, 1988). Forest reduction in Africa during glaciations presumably confined tropical forest fauna to refugia (Mayr & O’Hara, 1986). Thus, the taxa were probably split into isolated populations in a fragmented habitat made up of forest patches with high population densities. There, as often happens for large mammals on islands (Lomolino, 2005), a reduction in size might have been advantageous for reducing competition for food. A reduction in size in small peripheral populations is likely to have occurred at least a couple of times in the evolutionary history of red colobus (Rodgers et al., 1982; Colyn, 1991), since red colobus taxa with the smallest size are found in small and relict populations, like those of the Tana River and Zanzibar Archipelago. If a relatively small size in western red colobus is related to a history of geographical isolation within Pleistocene forest ‘islands’, a

similar size reduction would not be expected for vervets, as their terrestriality would have enabled them to move between forest patches, and thus allow them to avoid strong competition within restricted areas. Within red colobus crania there were clear geographical differences. However, compared with size, much less cranial shape was explained by geography. This is perhaps unsurprising as size tends to be highly plastic and adaptive; this, again, is consistent with findings from vervets (Cardini et al., 2007). Nevertheless, size and shape are seldom uncorrelated and clinal variation in shape corresponded well to expectations based on size in the presence of allometry. Thus, small red colobus from West and, especially, East Africa had short muzzles, large orbits and expanded cranial vaults relative to populations from central equatorial Africa. This size-related aspect of shape variation follows a common trend in primates and other mammals, where it is often observed that within a group of closely related populations, those smaller in size tend to be somewhat paedomorphic. This has been demonstrated in papionins (Collard & O’Higgins, 2001; Singleton, 2002), guenons (Cardini & Elton, 2008a), ungulates (Emerson & Bramble, 1993) and sciurid rodents (Cardini & O’Higgins, 2004; Cardini et al., 2005). In the smallest guenon, Miopithecus, rate hypomorphosis – a decrease in growth rate over a given time – could account for the observed paedomorphy (Shea, 1992). However, evolutionary explanations for paedomorphy are far from clear. Functional constraints and allometric scaling might be involved if, for instance, a relatively smaller brain in larger animals allows better accommodation of their proportionally larger masticatory muscles (Emerson & Bramble, 1993). It might also be related to some fundamental and highly conserved process of skull ontogeny, since ‘strong parallels between intraspecific ontogenetic allometry and interspecific scaling relationships have been documented in a number of tetrapod groups ... as ... diverse as cichlid fishes, plethodontid salamanders, iguanid lizards, anteaters, dogs and great apes’ (Emerson & Bramble, 1993, p. 399). When variation in red colobus cranial form was partitioned, geographically structured taxonomic variation was the major component influencing both size and shape. This observation indicates that population differences are largely consistent with those expected based on geographical barriers like rivers and mountains, and suggests that in the past these same barriers may have had a crucial role in reducing gene fluxes (Rodgers et al., 1982; Colyn, 1991; Colyn et al., 1991). The importance of taxonomy as an explanatory factor in shape is consistent with Ting’s (2008) hypothesis that the evolutionary history of red colobus may actually be longer than traditionally thought, with the clade originating in the late Miocene and radiating in the Plio-Pleistocene. Our confidence in the patterns observed in this study is further strengthened by the congruence between proportions of variance explained by different components in males and females. The only remarkable difference (the taxonomic component of size, large in females and small in males) might have occurred simply because of sampling error, likely to be greater in the smaller male sample. Interestingly, taxonomic differences alone accounted for a much larger proportion of total explained variance in shape

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

259

A. Cardini and S. Elton

Figure 6 Average annual temperature (a) and average annual rainfall (b) from HADCRUT2V Data. Images provided by the NOAA-ESRL Physical Sciences Division, Boulder Colorado from their website at http://www.cdc.noaa.gov/: (a), http://www.cdc.noaa.gov/Data/ descriptions/HADCRUTEM2.html; (b) http://www.cdc.noaa.gov/cdc/ data.cmap.html.

(one-quarter to almost a half) than in size (less than one-sixth). This may be due to the evolutionary lability of size, in that size tends to reflect adaptations and differences from the recent past and is therefore prone to convergence. Shape, in contrast, thanks to its inherently multivariate complexity, is less likely to be easily and frequently modified and is thus more likely to retain information useful for taxonomic identification and phylogenetic inference. The effectiveness of shape in discriminating taxonomic differences in red colobus has been demonstrated

260

elsewhere by cross-validated discriminant analyses of cranial shape, in which a high percentage of specimens were correctly classified to population (> 84%) compared with cranial size (< 43%; Cardini & Elton, in press). Thus, size may be better in detecting the effects of recent environmental changes on form, whereas shape may better reflect the differences created by older evolutionary events. If so, it is likely that the study of shape differences among red colobus can be much more informative about the evolutionary history of these endangered populations than

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

‘Morphing’ red colobus cranial variation Struhsaker (1981) believed when he issued his warning that morphological comparisons of skulls in red colobus might be hampered by the remarkable variation of form within populations. ACKNOWLEDGEMENTS This paper is dedicated to the memory of Marco Corti (1950– 2007) in recognition of his great contribution to the development and application of geometric morphometrics to the study of systematics and the mechanisms of speciation in mammals. He was a pioneer, clearing the path and serving as an inspiration for many mammalogists. His advice and example was invaluable for us and for many young zoologists who were struggling to learn geometric morphometrics and multivariate statistics. We are deeply grateful to all museum curators and collection managers who allowed and helped us to study their collections. Among them, special thanks to Hans-Walter Mittmann (Staatliches Museum für Naturkunde, Karlsruhe) for sending us specimens on loan during our visit at the Museum für Naturkunde in Berlin. Wim Wendelen (Royal Museum for Central Africa, Tervuren) and Olav Olav Röhrer-Ertl (formerly at Staatliche Naturwissenschaftliche Sammlugen Bayerns, Munich) provided invaluable help with specimen identification and advice on Procolobus collections, and Cristina Murari (University of Modena and Reggio Emilia) provided crucial support for running computer analyses on Linux workstations. We thank Kate Nowak for her assistance and collaboration in the earlier part of this study. Claudio Gentilini, Roberta Cantaroni, Costantino Crescimanno, Andrea Ghidoni and Maria Teresa Martinelli (all of them at the University of Modena and Reggio Emilia) were also of great help solving computer and network problems. Craig Ludwig (National Museum of Natural History, Washington), Emiliano Brunner and Paolo Colangelo (University of Rome), Damiano Preatoni and Adriano Martinoli (University of Insubria), and Andrew Marshall (University of York) were all of great help during various stages of this study. We are also very grateful to the Zanzibar authorities in the Department of Commercial Crops, Fruits and Forestry (DCCFF). We would like to thank Nelson Ting (City University of New York) and his co-authors in the study of the molecular systematics of Piliocolobus very much for sharing their preliminary results with us. Finally, we are very grateful to the editor, Kate Jones, David Polly and two other referees whose comments and suggestions improved this paper. This study was funded by a grant from the Leverhulme Trust and the Ruggles-Gates Fund for Biological Anthropology. REFERENCES Adams, D.C., Slice, D.E. & Rohlf, F.J. (2004) Geometric morphometrics: ten years of progress following the ‘revolution’. Italian Journal of Zoology, 71, 5–16. Albrecht, G.H. (1982) The relationship of size, latitude and habitat in the South American primate Callithrix jacchus. American Journal of Physical Anthropology, 57, 166. Bookstein, F.L. (2000) Morphometrics. Encyclopedia of life sciences. John Wiley, Chichester. http://www.els.net/

Botes, A.M.A., McGeoch, H., Robertson, G., van Niekerk, A., Davids, H.P. & Chown, S.L. (2006) Ants, altitude and change in the northern Cape Floristic Region. Journal of Biogeography, 33, 71–90. Burnham, K.P. & Anderson, D.R. (1998) Model selection and inference: a practical information-theoretic approach. Springer, New York. 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 and in the guenons. Biological Journal of the Linnean Society, 93, 813–834. Cardini, A. & Elton, S. (in press) The radiation of red colobus monkeys (Primates, Colobinae): morphological evolution in a clade of endangered African primates. Zoological Journal of the Linnean Society. Cardini, A. & O’Higgins, P. (2004) Patterns of morphological evolution in Marmota (Rodentia, Sciuridae): geometric morphometrics of the cranium in the context of marmot phylogeny, ecology and conservation. Biological Journal of the Linnean Society, 82, 385–407. Cardini, A. & Thorington, R.W. Jr (2006) Post-natal ontogeny of the marmot (Rodentia, Sciuridae) cranium: allometric trajectories and species divergence. Journal of Mammalogy, 87, 201– 216. Cardini, A, Hoffmann, R.S. & Thorington, Jr R.W. (2005) Morphological evolution in marmots (Rodentia, Sciuridae): size and shape of the dorsal and lateral surfaces of the cranium. Journal of Zoological Systematics and Evolutionary Research, 43, 258–268. Cardini, A., Jansson, A-U. & Elton, S. (2007) Ecomorphology of vervet monkeys: a geometric morphometric approach to the study of clinal variation. Journal of Biogeography, 34, 1663–1678. Cobb, S.N. & O’Higgins, P. (2007) The ontogeny of sexual dimorphism in the facial skeleton of the African apes. Journal of Human Evolution, 53, 176–190. Collard, M & O’Higgins, P (2001) Ontogeny and homoplasy in the papionin monkey face. Evolution and Development, 3, 322– 331. Colyn, M.M. (1991) L’importance zoogeographique du Bassin du Fleuve Zaire pour la spéciation: le cas des primates simiens. Annales de le Musée Royal de l’Afrique Centrale, Tervuren, Sciences Zoologiques, 264, 1–250. Colyn, M., Gautier–Hion, A. & Verheyen, W. (1991) A re-appraisal of palaeoenvironmental history in Central Africa: evidence for a major fluvial refugium in the Zaire Basin. Journal of Biogeography 18, 403–407. Davies, A.G. & Oates, J.F. (eds) (1994) Colobine monkeys: their ecology, behaviour and evolution. Cambridge University Press, Cambridge. Elton, S. (2008) The environmental context of human evolutionary history in Eurasia and Africa. Journal of Anatomy, 212, 377– 393.

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

261

A. Cardini and S. Elton Emerson, S.B. & Bramble, D.M. (1993) Scaling, allometry, and skull design. The skull, Volume 3 (ed. by J. Hanken and B.K. Hall), pp. 384–421. University of Chicago Press, Chicago. Fadda, C. & Corti, M. (2000) Three dimensional geometric morphometric study of the Ethiopian Myomys–Stenocephalemys complex (murinae, rodentia). Hystrix, 10, 131–143. Fadda, C. & Corti, M. (2001) Three-dimensional geometric morphometrics of Arvicanthis: implications for systematics and taxonomy. Journal of Zoological Systematics and Evolutionary Research, 39, 235–245. Fooden, J. & Albrecht, G.H. (1993) Latitudinal and insular variation of skull size in crab-eating macaques (Primates, Cercopithecidae: Macaca fascicularis). American Journal of Physical Anthropology, 92, 521–538. Franklin D., Cardini A., O’Higgins P., Oxnard, C.E. & Dadour, I. (2007a) Mandibular morphology as an indicator of human subadult age: geometric morphometric approaches. Forensic Science, Medicine and Pathology, 52, 1015–1019. Franklin, D., Freedman, L., Milne, N. & Oxnard, C.E. (2007b) Geometric morphometric study of population variation in indigenous southern African crania. American Journal of Human Biology, 19, 20–33. Franklin, D., Cardini, A., O’Higgins, P., Oxnard, C.E. & Dadour, I. (2008) Mandibular morphology as an indicator of human subadult age: geometric morphometric approaches. Forensic Science, Medicine and Pathology, 4, 91–99. Frost, S.R., Marcus, L.F., Bookstein, F.L., Reddy, D.P. & 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, 275A, 1048–1072. Gautier-Hion, A., Colyn, M. & Gautier, J.-P. (1999) Histoire Naturelle des Primates d’Afrique Centrale. Ecofac, Gabon. Available as pdf file at: http://www.ecofac.org/Biblio/Download/Guides/ PrimatesGuide.pdf Grubb, P. (2006) Geospecies and superspecies in the African primate fauna. Primate Conservation, 20, 75–78. Grubb, P., Butynski, T.M., Oates, J.F., Bearder, S.K., Disotell, T.R., Groves, C.P. & Struhsaker, T.T. (2003) Assessment of the diversity of African primates. International Journal of Primatology, 24, 1301–1357. Hamilton, A.C. (1988) Guenon evolution and forest history. A primate radiation. Evolutionary biology of the African guenons (ed. by A. Gautier-Hion, F. Bourlierey, J-P. Gautier and J. Kingdon), pp. 13–34. Cambridge University Press, Cambridge. Legendre, P. & Legendre, L. (1998) Numerical ecology, 2nd edn. Elsevier, Amsterdam. Lomolino, M.V. (2005) Body size evolution in insular vertebrates: Generality of the island rule. Journal of Biogeography, 32, 1683–1699. Mayr, E. & O’Hara, R.J. (1986) The biogeographical evidence supporting the Pleistocene forest refugium hypothesis. Evolution, 40, 55–67. Mazerolle, M.J. (2004) Making sense out of Akaike’s information criterion (AIC): its use and interpretation in model selection and inference from ecological data. Thesis, Appendix 1,

262

Université Laval, Québec (http://www.theses.ulaval.ca/2004/ 21842/apa.html). deMenocal, P.B. & Bloemendal, J. (1995) Plio-Pleistocene climatic variability in subtropical Africa and the palaeoenvironment of hominid evolution: a combined data-model approach. Paleoclimate and evolution, with emphasis on human origins (ed. by E.S. Vrba, G.H. Denton, T.C. Partridge and L.H. Burckle), pp. 262–288. Yale University Press, New Haven, CT. Milne, N, & O’Higgins, P. (2002) Inter-specific variation in Macropus crania: form, function and phylogeny. Journal of Zoology, London, 256, 523–535. Monteiro, L.R., Duarte, L.C. & dos Reis, S.F. (2003) Environmental correlates of geographical variation in skull and mandible shape of the punaré rat Thrichomys apereoides (Rodentia: Echimyidae). Journal of Zoology, London, 261, 47–57. Nowak, K., Cardini, A. & Elton S. (2008) Evolutionary acceleration in an endangered African primate: speciation and divergence in the Zanzibar Red Colobus (Primates, Colobinae). International Journal of Primatology, DOI: 10.1007/s10764-008-9306-1. O’Higgins, P. & Jones, N. (1998) Facial growth in Cercocebus torquatus: an application of three dimensional geometric morphometric techniques to the study of morphological variation. Journal of Anatomy, 193, 251–272. O’Higgins, P. & Jones, N. (2006) Morphologika, tools for statistical shape analysis. Hull York Medical School (http://hyms.fme. googlepages.com/resources). Rae, T.T., Hill, R.A., Hamada, Y. & Koppe, T. (2003) Clinal variation of maxillary sinus volume in Japanese macaques (Macaca fuscata). American Journal of Primatology, 59, 153–158. Rangel, T.F.L.V.B., Diniz-Filho, J.A.F. & Bini, L.M. (2006) Towards an integrated computational tool for spatial analysis in macroecology and biogeography. Global Ecology and Biogeography, 15, 321–327. dos Reis, S.F., Duarte, L.C., Monteiro, L.R. & Von Zuben, F.J. (2002) Geographic variation in cranial morphology in Thrichomys apereoides (Rodentia: Echimyidae): II. Geographic units, morphological discontinuities, and sampling gaps. Journal of Mammalogy, 83, 345–353. Rodgers, W.A., Owen, C.F. & Homewood, K.M. (1982) Biogeography of East African forest mammals. Journal of Biogeography, 9, 41–54. Rohlf, F.J. (2006a) TpsSmall. Department of Ecology and Evolution, State University of New York, Stony Brook, NY (http:// life.bio.sunysb.edu/morph/). Rohlf, F.J. (2006b) NTSYSpc, version 2.20L. Exeter Software, Setauket, NY. Rohlf, F.J. (2006c) A comment on phylogenetic correction. Evolution, 60, 1509–1515. Rohlf, F.J. (2007) TpsRelw. Department of Ecology and Evolution, State University of New York, Stony Brook, NY (http:// life.bio.sunysb.edu/morph/). Ruggiero, A. & Kitzberger, T. (2004) Environmental correlates of mammal species richness in South America: effects of spatial structure, taxonomy and geographic range. Ecography, 27, 401–416.

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

‘Morphing’ red colobus cranial variation Santos, M., Iriarte, P.F., Céspedes, W., Balanyà, J., Fontdevila, A. & Serra, L. (2004) Swift laboratory thermal evolution of wing shape (but not size) in Drosophila subobscura and its relationship with chromosomal inversion polymorphism. Journal of Evolutionary Biology, 17, 841–855. Shea, B.T. (1992) Ontogenetic scaling of skeletal proportions in the talapoin monkey. Journal of Human Evolution, 23, 283– 307. Singleton, A.M. (2002) Patterns of cranial shape variation in the Papionini (Primates: Cercopithecinae). Journal of Human Evolution, 42, 547–578. Slice, D.E. (1999) Morpheus, beta version. Department of Ecology and Evolution, State University of New York, Stony Brook, NY. Struhsaker, T.T. (1981) Vocalizations, phylogeny and palaeogeography of red colobus monkeys (Colobus badius). African Journal of Ecology, 19, 265–283. Ting, N. (2008) Mitochondrial relationships and divergence dates of the African colobines: evidence of Miocene origins for the living colobus monkeys. Journal of Human Evolution, 55, 312–325. Turner, T.R., Anapol, F. & Jolly, C.J. (1997) Growth, development, and sexual dimorphism in vervet monkeys (Cercopithecus aethiops) at four sites in Kenya. American Journal of Physical Anthropology, 103, 19–35. Verheyen, W.N. (1957) Bijdrage tot de Craniometrie van Colobus badius (Kerr, 1792). Annales du Musée Royal du Congo Belge, Tervuren, 8(62), 1–105.

Verheyen, W.N. (1962) Contribution a la craniologie comparee des primates. Les genres Colobus Illiger 1811 et Cercopithecus Linne 1758. Annales Du Musée Royal Du Congo Belge, Tervuren, 8(105), 1–255. Wroe, S. & Milne, N. (2007) Convergence and remarkable constraint in the evolution of mammalian carnivore skull shape. Evolution, 61, 1251–1260.

BIOSKETCHES Andrea Cardini is a lecturer in animal biology at the University of Modena and Reggio Emilia and honorary fellow of the Hull York Medical School, UK. His main interest is the study of the phylogenetic signal in animal morphology and his work focuses on the ontogenetic, ecological, demographic and biogeographical factors which may have an influence on form and contribute to set the tempo and mode of morphological evolution. Sarah Elton is senior lecturer in anatomy at the Hull York Medical School, UK. Her work focuses on ecological and morphological variation in modern and fossil monkeys.

Editor: Kate E. Jones

© 2009 The Authors Global Ecology and Biogeography, 18, 248–263, Journal compilation © 2009 Blackwell Publishing Ltd

263