Michelle SingletonDepartment of Anatomy,Midwestern University, 55531st Street, Downers Grove,Illinois 60515, U.S.A.E-mail:msingl@midwestern.edu

Received 10 January 2001Revision received26 October 2001 andaccepted 12 December 2001

Keywords: Papionini,mangabey diphyly, cranialhomoplasy, allometry,geometric morphometrics.

Patterns of cranial shape variation in thePapionini (Primates: Cercopithecinae)

Traditional classifications of the Old World monkey tribe Papionini(Primates: Cercopithecinae) recognized the mangabey generaCercocebus and Lophocebus as sister taxa. However, molecular studieshave consistently found the mangabeys to be diphyletic, withCercocebus and Mandrillus forming a clade to the exclusion of all otherpapionins. Recent studies have identified cranial and postcranialfeatures which distinguish the Cercocebus–Mandrillus clade, howeverthe detailed similarities in cranial shape between the mangabeygenera are more difficult to reconcile with the molecular evidence.Given the large size differential between members of the papioninmolecular clades, it has frequently been suggested that allometriceffects account for homoplasy in papionin cranial form. A combina-tion of geometric morphometric, bivariate, and multivariate methodswas used to evaluate the hypothesis that allometric scaling contributesto craniofacial similarities between like-sized papionin taxa. Patternsof allometric and size-independent cranial shape variation weresubsequently described and related to known papionin phylogeneticrelationships and patterns of development.

Results confirm that allometric scaling of craniofacial shapecharacterized by positive facial allometry and negative neurocranialallometry is present across adult papionins. Pairwise comparisons ofregression lines among genera revealed considerable homogeneity ofscaling within the Papionini, however statistically significant differ-ences in regression lines also were noted. In particular, Cercocebus andLophocebus exhibit a shared slope and significant vertical displacementof their allometric lines relative to other papionins. These findingsgive no support to narrowly construed hypotheses of uniquely sharedpatterns of allometric scaling, either between sister taxa or across allpapionins. However, more general allometric trends do appear toaccount for a substantial proportion of papionin cranial shape vari-ation, most notably in those features which have influenced tra-ditional morphological phylogenies. Examination of size-uncorrelatedshape variation gives no clear support to molecular phylogenies,but underscores the absence of morphometric similarities betweenthe mangabey genera when size effects are controlled. Patterns ofallometric and size-uncorrelated shape variation indicate conserva-tism of cranial form in non-Theropithecus papionins, and suggest thatPapio represents the primitive morphometric pattern for the Africanpapionins. Lophocebus exhibits a divergent morphometric pattern,clearly distinguishable from other papionins, most notably Cercocebus.These results clarify patterns of cranial shape variation among theextant Papionini and lay the groundwork for studies of related fossiltaxa.

� 2002 Elsevier Science Ltd

Journal of Human Evolution (2002) 42, 547–578doi:10.1006/jhev.2001.0539Available online at http://www.idealibrary.com on

0047–2484/02/050547+32$35.00/0 � 2002 Elsevier Science Ltd

548 .

Introduction

The Old World monkey tribe Papionini(Primates: Cercopithecinae) comprises amonophyletic group of six extant genera.While Macaca is widely acknowledged asrepresenting the sister taxon to the Africanpapionins (Strasser & Delson, 1987;Morales & Melnick, 1998), relationshipsamong the latter taxa have been a sourceof controversy. Traditional classificationsrecognized the mangabey genera Cercocebusand Lophocebus as sister taxa and frequentlyaccorded them congeneric status on thebasis of shared morphological traits(Thorington & Groves, 1970; Szalay &Delson, 1979). However, molecular studieshave consistently found the mangabeys to bediphyletic, with Cercocebus and Mandrillusforming a clade to the exclusion of all otherpapionins (Harris, 2000). Recent studieshave identified a number of features whichdistinguish the Cercocebus–Mandrillus clade(Fleagle & McGraw, 1999; Groves, 2000);however the detailed similarities in cranialshape between the mangabey genera haveyet to be reconciled with molecular phylog-enies. Because of the large size differentialbetween members of the papionin molecularclades, it frequently has been suggestedthat allometric effects might account forhomoplasies in papionin cranial shape (Shah& Leigh, 1995; Lockwood & Fleagle, 1999;Harris, 2000; Ravosa & Profant, 2000).

In an effort to circumvent certain limi-tations of traditional allometric studies, ageometric morphometric study of papionincraniofacial morphology was undertaken. Acombination of geometric morphometric,bivariate, and multivariate techniques wasapplied in order to identify multivariate scal-ing patterns not discernable by conventionalbivariate regression analysis. The objectivesof this study were twofold. The first was toevaluate the supposition that homoplasy inpapionin craniofacial shape is attributableto shared patterns of allometric scaling,

either within individual clades or across allpapionins. The second was to describepatterns of allometric and residual (size-uncorrelated) cranial shape variation andinterpret these patterns in the light ofprior studies of papionin ontogeny andphylogeny.

Background

The Old World monkey tribe PapioniniBurnett, 1828 (Primates, Cercopithecinae)encompasses six extant genera: Macaca, themacaques; Cercocebus and Lophocebus, themangabeys; Mandrillus, including mandrillsand drills; Papio, the savannah baboons;and Theropithecus, the gelada baboon. Thepapionins have long been recognized as amonophyletic group on the basis of sharedtraits including relatively flaring molars,broad nasal apertures, relatively long faces,a tendency toward terrestriality, and adiploid chromosome number of 42 (Kuhn,1967; Hill, 1974; Delson, 1975a,b; Szalay& Delson, 1979; Dutrilleaux et al., 1982;Strasser & Delson, 1987). Macaca, whichlacks well-developed maxillary and man-dibular facial fossae and is thought to retaina number of primitive cercopithecine mor-phological features, is considered the sistertaxon to the African papionins (Szalay &Delson, 1979; Strasser & Delson, 1987;Morales & Melnick, 1998). This relation-ship is confirmed by an array of molecularevidence including immunological dis-tances, protein polymorphisms, DNA–DNAhybridization, amino acid sequences, andnuclear and mitochondrial genetic studies(Disotell, 1994; van der Kuyl et al., 1995;Harris & Disotell, 1998; Harris, 2000).

The resolution of phylogenetic relation-ships among the African papionins has beenmore problematic. The mangabeys wereoften classified in a single genus, Cercocebus,on the basis of shared morphologicalfeatures including moderate body size,

549

(a)

CercocebusLophocebusPapioMandrillusTheropithecusMacaca (b)

PapioMandrillusTheropithecusCercocebusLophocebusMacaca

(c)

PapioTheropithecusLophocebusCercocebusMandrillusMacaca (d)

LophocebusPapioTheropithecusCercocebusMandrillusMacaca

Figure 1. Cladograms depicting alternate hypotheses of phylogenetic relationships among the Papionini.Traditional phylogenies exemplified by (a) Strasser & Delson (1987) and (b) Delson & Dean (1993)recognized the mangabey genera Cercocebus and Lophocebus as sister taxa. Molecular phylogenies,including (c) Disotell (1994) and (d) Harris & Disotell (1998), reconstruct mangabeys as diphyletic, withCercocebus most closely related to Mandrillus.

relatively short muzzles, excavated sub-orbital fossae, and retention of a long tail(Thorington & Groves, 1970; Hill, 1974;Szalay & Delson, 1979). However, consist-ent morphological and behavioral differ-ences were noted between two recognizedspecies groups—a semi-terrestrial torquatus–galeritus group and a highly arborealalbigena–atterimus group (Schwarz, 1928;Jones & Sabater Pi, 1968; Hill, 1974)—which were sometimes accorded subgenericstatus (Szalay & Delson, 1979). Groves(1978) resurrected the genus LophocebusPalmer, 1903, to receive the albigena–atterimus species group, which he diagnosedon the basis of skeletal, behavioral, andreproductive traits. Still, Cercocebus andLophocebus were generally considered closelyrelated (Kuhn, 1967; Szalay & Delson,1979; Strasser & Delson, 1987), and manyphylogenetic hypotheses reconstructed themangabeys as sister taxa to the exclusion ofthe remaining African papionins [Figure1(a), (b)].

By contrast, molecular studies have con-sistently indicated that mangabeys arediphyletic [Figure 1(c), (d)]. Beginning with

the earliest protein electrophoresis andimmunological studies (Barnicott & Hewett-Emmett, 1972; Cronin & Sarich, 1976;Hewett-Emmett et al., 1976), analysesof chromosome structure, amino acidsequences, mitochondrial and nuclear DNAhave all linked Cercocebus and Mandrillusto the exclusion of Lophocebus, Papio andTheropithecus (Dutrilleaux et al., 1979, 1982;Disotell et al., 1992; Disotell, 1994; vander Kuyl et al., 1995; Harris & Disotell,1998; Page et al., 1999; Harris, 2000).Cladistic relationships among the lattertaxa are somewhat less clear. Trees basedon immunological distances and certainDNA sequences (CO II and �-globin)support a Papio–Theropithecus sister relation-ship (Cronin & Sarich, 1976; Disotell,1994; Page et al., 1999). Three of the fournuclear sequences (��-� globin intergenicregion; � 1,3 GT; IRBP) analyzed by Harris& Disotell (1998) support a Papio–Lophocebus sister relationship. In a recentsynthesis of published molecular data,Harris (2000) found that both molecularconsensus and total molecular evidencetrees supported the Lophocebus–Papio clade,

550 .

although a Lophocebus–Papio–Theropithecustrichotomy could not be rejected statisti-cally. This point of uncertainty notwith-standing, mangabey diphyly appears wellsupported.

The obvious discrepancy between mol-ecular and morphological classifications ofthe papionins has led to renewed interest inthe behavior, ecology, reproductive biology,and especially the comparative morphologyof this group. Nakatsukasa (1994, 1996)identified a number of postcranial featuresdistinguishing Cercocebus from Lophocebus,and Fleagle & McGraw (1999) establishedthat Cercocebus more closely resemblesMandrillus in these same features. They alsonoted the presence of enlarged fourthpremolars in this clade in comparison withPapio and Lophocebus. Based on outgroupcomparisons with Macaca nemestrina, theyjudged the postcranial traits primitive andthe dental dimensions derived, and theyattributed this suite of features to a sharedecological adaptation, namely terrestrialforaging and hard object feeding in a tropi-cal forest milieu (Fleagle & McGraw, 1999).More recently, additional qualitative cranialfeatures have been put forward as potentialsynapomorphies of a Cercocebus–Mandrillusclade (McGraw & Fleagle, 2000; Groves,2000).

The marked similarities in cranial shapeand proportion which distinguish Cerco-cebus and Lophocebus from other Africanpapionins have proven less tractable toanalysis, giving rise to evolutionary scenariosinvoking various combinations of primitiveretention and parallel evolution within thetwo African molecular clades. Based on out-group comparisons, it has been suggestedthat the moderate facial length seen inmangabeys—like moderate body size andthe presence of a long tail—is primitive forall papionins (Disotell, 1994; Harris &Disotell, 1998; Harris, 2000). This wouldimply that Papio and Mandrillus experiencedparallel increases in facial length. However,

Groves (1978) and Kingdon (1997) haveproposed that long faces are primitive forAfrican papionins. In this scenario, theCercocebus and Lophocebus lineages wouldhave experienced parallel reductions infacial length, with the excavated suborbitalfossae found in these taxa perhaps resultingfrom this secondary facial shortening (Harris& Disotell, 1998; Harris, 2000).

Allometric scaling of cranial proportionsis well-documented within the Papionini.Freedman (1962) showed that adult faciallength scales positively relative to calvarialength both between males and females andacross (sub)species of Papio. He subse-quently related differences in cranial propor-tions among Papio varieties to clinal trendsin body size (Freedman, 1963). Ontogeneticstudies of macaques have consistentlyshown that males and females follow acommon trajectory characterized by positivefacial allometry, negative allometry ofneurocranial dimensions, and dorsal ro-tation of the maxilla and palate, withsexual dimorphism between adult cranialform resulting from male hypermorphosis(MacNamara et al., 1976; Bookstein, 1985;Cochard, 1985; Cheverud & Richtsmeier,1986). Comparative developmental studieshave found Papio and Macaca to have simi-lar craniofacial growth rates and a commonontogenetic trajectory characterized byprogressive increase in cranial base angle,relatively rapid forward growth of themuzzle, proportional increase in lowerfacial height, and decelerating neuro-cranial growth (Swindler & Sirianni, 1973;Swindler et al., 1973). Thus, many observeddifferences in adult cranial morphologies ofmacaques and baboons are construed asallometric consequences of Papio’s largerabsolute body size.

It has frequently been suggested thatallometric effects might also account forhomoplasies in mangabey cranial shape(Shah & Leigh, 1995; Lockwood & Fleagle,1999; Harris, 2000; Ravosa & Profant,

551

2000). Comparing adult male body weights,Mandrillus sphinx averages 34 kg; M.leucophaeus, 20 kg; and Papio hamadryassubspecies range from a mean weight of16 kg (P. h. kindae) to 31 kg (P. h. anubis)(Delson et al., 2000). By comparison, malesof the larger Cercocebus varieties average only12 kg (C. torquatus torquatus); and Lophoce-bus males average no more than 9 kg (L.albigena zenkeri) (Delson et al., 2000). Giventhis marked size difference, it is highlyplausible that craniofacial similaritiesbetween like-sized members of disparatepapionin clades can be explained in terms ofinterspecific allometric scaling. In particular,it has been suggested that ontogeneticscaling—whereby adults of papionin sistertaxa occupy different points along a sharedgrowth trajectory (Shea, 1985)—may be amajor factor in mangabey cranial homoplasy(Shah & Leigh, 1995). Shah & Leigh’s(1995) initial efforts to test the hypothesis ofontogenetic scaling within the Cercocebus–Mandrillus clade were inconclusive. Theyfound that Mandrillus and Cercocebus shareda common ontogenetic trajectory for neuro-cranial dimensions, but facial growth inMandrillus more closely resembled that ofPapio. The authors concluded that scalingpatterns among these taxa were complex,and that global ontogenetic scaling couldnot adequately account for differences incranial shape between Cercocebus andMandrillus.

Minor variations in scaling patterns ofindividual linear dimensions within andbetween cranial regions can unnecessarilycomplicate interspecific allometric compari-sons. By design, linear measures reducecomplex spatial relationships to unidimen-sional values lacking geometric context. Theresulting atomization of form can generateapparently contradictory scaling patterns inwhat are, by definition, developmentallyintegrated morphologies. By contrast,landmark-based geometric morphometricanalysis preserves spatial relationships,

permitting the simultaneous analysis ofdimension and relative position, i.e., sizeand shape (Rohlf & Marcus, 1993). Byrelating geometrically derived shape vari-ables to cranial size, it might be possibleto identify global scaling patterns not dis-cernable by traditional allometric analyses.Geometric analysis has the additionaladvantage of permitting direct visualizationof shape trends—both allometric andnonallometric—in the original specimenspace, thus facilitating description of results.

The present study applies geometric tech-niques to the description of allometric andsize-uncorrelated cranial shape variation inadult papionins. Following Cheverud’s(1982) demonstration that patterns of staticadult allometry cannot be assumed to reflectontogenetic processes, ontogenetic studieshave come to be viewed as the allometricgold standard; however, it remains necessaryto document patterns of adult interspecificallometry. Our understanding of the phylo-genetic affinities of fossil papionins such asParapapio and Paradolichopithecus (Szalay &Delson, 1979; Fleagle, 1999) could be sig-nificantly improved by interpretation withinthe context of papionin allometric shapevariation. But, given the rarity of adequatefossil primate ontogenetic sequences, theseforms can only be evaluated on the com-parative basis of adult interspecific scalingpatterns. To this end, this paper documentsadult patterns of cranial shape variation inpapionin primates, interprets these patternsin light of prior studies of papionin ontogenyand phylogeny, and considers their impli-cations for the development and evolution ofpapionin cranial form.

Materials and methods

Data collectionThe sample comprised 238 adult indi-viduals from all six extant papionin genera(Table 1). The sample was largely limitedto wild-shot adult individuals of known

552 .

provenience. However, given the scarcity ofTheropithecus gelada in museum collections,four zoo specimens (two male, two female)lacking obvious pathology and judged topresent normal, ‘‘wild-type’’ morphologywere included in the sample.

Data consisted of three-dimensionalcranial landmarks recorded using a Micro-scribe 3-DX digitizer and InScribe-32 soft-ware (Immersion Corp., San Jose, CA).This study was conducted in collaborationwith the Morphometrics Research Group ofthe New York Consortium in EvolutionaryPrimatology, and cranial landmark datawere collected by several observers under acommon protocol developed by Frost et al.(in prep.). Each skull was first mounted toprovide access to its dorsal aspect and osteo-metric landmarks recorded followed by fourarbitrarily chosen, noncoplanar, noncol-linear registration landmarks. The skull wasthen remounted to gain access to its ventralaspect and the remaining landmarks wererecorded along with the same four regis-tration landmarks. The registration land-marks were used to align the dorsal andventral aspects of the skull within a commoncoordinate system, yielding a complete con-figuration. Realignment was performed on a

Study sample

Taxonn

Collection*Female Male

Cercocebus galeritus agilis 10 7 AMNHCercocebus torquatus torquatus 6 13 AMNH PCMLophocebus albigena johnstoni 12 21 AMNHMacaca fascicularis 17 26 AMNH UCMVZMandrillus leucophaeus 8 19 BMNH FMNH FSMMandrillus sphinx 8 13 AMNH BMNH FSM PCMPapio hamadryas anubis 21 41 AMNH FMNH NMNH UCMVZTheropithecus gelada 4 11 AMNH FMNH LHES NMNH

*Collection abbreviations: AMNH=American Museum of Natural History;BMNH=British Museum (Natural History); FMNH=Field Museum of NaturalHistory; FSM=Senckenberg Natural History Museum—Frankfurt; LHES=Laboratory for Human Evolutionary Studies—UC Berkeley; PCM=Powell–CottonMuseum; NMNH=National Museum of Natural History; UCMVZ=University ofCalifornia Museum of Vertebrate Zoology.

Table 1

Silicon Graphics O2 workstation usingdedicated software combining UNIX andMatlab 4.2c routines (Frost et al., in prep.).This dorsal–ventral landmark registration(DVLR) procedure allows all regions ofthe skull to be digitized without the useof cumbersome mounting equipment.

Following Frost et al. (in prep.), elevenmidsagittal and 17 bilateral landmarks wererecorded, yielding a maximum of 45 land-marks for each specimen (Figure 2, Table2). At present, many geometric morpho-metric programs are unable to accommo-date missing data. Where specimens aremissing landmarks, whether due to ante-mortem pathology or postmortem damage,it is necessary either to exclude the specimenfrom analysis or to reconstruct the missingdata. For specimens with localized damage,missing values for paired landmarks can beestimated by ‘‘reflecting’’ the correspondinglandmarks from the opposite side to producecomplete configurations. Specimens withmissing data were read into GRF-ND (Slice,1999) and Bookstein shape coordinates(Bookstein, 1991) were computed relativeto an inion–prosthion baseline using bregmaas the third landmark and invoking the‘‘No Scale’’ option to preserve the original

553

Figure 2. Osteological landmarks used in this study mapped on to a representative papionin skull (femaleMacaca fascicularis). Landmarks are defined following Frost et al. (in prep.); abbreviations follow Table 2.Line drawings of female Macaca fascicularis skull courtesy of J. Michael Plavcan (� J. M. Plavcan) andadapted with artist’s permission.

Precision studiesData used in this study were collected bythree observers working under a commondata collection protocol. To examine theeffects of intra- and interobserver error, eachobserver collected multiple observations of asingle specimen, a male P. hamadryas ursinuscranium. Raw landmark data were subjectedto generalized Procrustes analysis (GPA)using GRF-ND (Slice, 1999) in order to

place all observations in a common frame ofreference, and the Euclidean distance ofeach landmark to its respective centroid wascomputed. For each observer, landmarkdeviations were calculated relative to theobserver landmark mean. Mean deviationsand percentage errors were calculated forindividual landmarks and subsequentlyaveraged to give a mean deviation and per-centage error for each observer across alllandmarks (Table 3). One-way analysis ofvariance (ANOVA) was conducted for eachlandmark by observer, and the root meansquares were examined (Table 4). In thecontext of this analysis, the root of thewithin-groups mean squares (root meansquare error) corresponds to intraobservererror (Sokal & Rohlf, 1981), while the rootof between-groups mean squares corre-sponds to interobserver error. Intraobservererror does not exceed 0·33 mm or 2%.Interobserver error averages approximately1 mm, again less than 2%, and even the

specimen scale. This procedure orientedeach specimen in a coordinate system withthe Z-coordinate axis perpendicular to thespecimen mid-sagittal plane and the originlocated on the specimen midline at inion.Coordinates for each missing landmark werethen taken from those of its antimere byreversing the sign of the Z-coordinate. Thelarge number of specimens with damage inthe region of Basion, an unpaired landmark,necessitated its exclusion, leaving a total of44 landmarks for analysis.

554 .

Table 2 Osteological landmarks

Landmark Abbreviation

MidsagittalInion INBregma BRGlabella GLNasion NARhinion RHNasospinale NSProsthion PROpisthion OPBasion BAStaphylion STIncisivion IV

BilateralProsthion-2 PR2Mesial P3 MP3M1–M2 Contact M12Distal M3 DM3Premaxillary suture–inferior PMIPremaxillary suture–superior PMSZygomaxillare inferior ZMIZygomaxillare superior ZMSDacryon DACMidtorus inferior MTIMidtorus superior MTSFrontomalare orbitale FMOFrontomalare temporale FMTZygotemporale superior ZTSZygotemporale inferior ZTIPorion PORPostglenion PGL

Osteological landmarks used in the present study.Landmarks are defined following Frost et al. (in prep.)and are illustrated in Figure 2.

most imprecise observations do not exceed5% error. While average interobserver errorexceeds intraobserver error by approxi-mately 0·66 mm, average percentage errorsare comparable within and between ob-servers. Given the relatively large size of theprimate crania in this study, these marginsof error were considered acceptable.

landmark configurations are scaled to unitcentroid size and optimally superimposed soas to minimize summed squared distancesacross all landmarks and specimens relativeto a reference configuration, the Procrustesmean (Slice et al., 1996). Figure 3 shows theProcrustes superimposition of all 238 speci-mens in dorsal view. Differences betweenforms, indicated by the point scatters at eachlandmark, represent pure shape variation. Itshould be noted that, by scaling all speci-mens to unit centroid size, the Procrustesprocedure corrects for gross size effects, i.e.,scale, in a manner analogous to traditionalbivariate ratios. But, as with ratios, Pro-crustes analysis does not correct for allo-metric effects. Therefore, the shape variationsummarized by the Procrustes-aligned co-ordinates may include both size-correlated,i.e., allometric, and size-uncorrelatedcomponents.

Procrustes-aligned specimens may berepresented as points in a curvilinear mor-phospace of dimension 3p–7 for p three-dimensional landmarks (Dryden & Mardia,1998; Rohlf, 1999a). Because this morpho-space is non-Euclidean, statistical analysis isconducted using the orthogonal projectionof points onto a Euclidean space set tangentto shape space at the Procrustes mean(Dryden & Mardia, 1998; Rohlf, 1999a). Inpractice, if shape variation is sufficientlysmall in the vicinity of the referenceform, the Procrustes-aligned coordinates area reasonable approximation of tangentspace coordinates (Dryden & Mardia, 1998;Rohlf, 1999a). This assumption may betested using tpsSmall (Rohlf, 1999b), soft-ware which computes the ordinary leastsquares regression (through the origin) ofEuclidean (tangent space) distances versusthe corresponding Procrustes (morpho-space) distances for all possible specimenpairs. Analysis of the present data setshowed a virtual one to one correspondencebetween the Procrustes and tangentspaces distances (r=0·9999, slope=0·9967)

Procrustes analysisRaw landmark data were subjected to ageneralized Procrustes analysis (GPA) usingthe Morpheus et al. morphometrics package(Slice, 1998). GPA is an iterative leastsquares procedure in which individual

555

indicating that the Procrustes-aligned co-ordinates are a good approximation fortangent space coordinates.

Principal components analysisLandmark-based morphometrics involveslarge numbers of statistically interdependentand partially redundant variables. Pro-crustes superimposition—through the con-straints of translation, rotation, and unitscaling—results in the loss of additionaldegrees of freedom (Dryden & Mardia,1998). Principal components analysis(PCA) of the Procrustes-aligned coordinatescorrects for these spurious dimensions byreducing the data to a full-rank matrix,namely the principal component scores foraxes corresponding to nonzero eigenvectors(Rohlf, 1999a). This procedure both gen-erates a reduced number of statisticallyuncorrelated summary shape variables andordinates specimens relative to the majoraxes of shape variation (Dryden & Mardia,

Observer mean deviation and mean percentage error across landmarks

Observer nMean deviations % Errors

Min Max Mean Min Max Mean

1 10 0·07 0·54 0·24 0·07 1·00 0·422 8 0·06 0·55 0·16 0·06 1·04 0·263 6 0·05 1·42 0·25 0·07 1·77 0·40

Landmark mean deviations are computed as the mean of the absolute deviations ofobservations from the landmark mean [ABS(xi� x)]; values are reported in milli-meters. Landmark percentage error is calculated as the landmark deviation expressedas a percentage of the landmark mean [ABS(xi� x)/x]�100. Observer mean devi-ations and observer mean percentage errors are calculated as the mean value across alllandmarks by observer.

Table 3

Figure 3. Dorsal view of Procrustes superimposition of238 papionin crania. Specimens have been scaled tounit centroid size and optimally superimposed. Intheory, all remaining differences between forms, indi-cated by point scatters at each landmark, are due topure shape variation.

Inter- and intraobserver mean errors across landmarks

Level

RMSE % RMSE

Min Max Mean Min Max Mean

Intraobserver 0·10 1·10 0·31 0·10 1·47 0·51Interobserver 0·04 2·81 1·05 0·32 4·70 1·70

Root mean square error (RMSE) is calculated as the root of the within-groups(intraobserver) or between-groups (interobserver) mean sums of squares for ANOVAof landmark distances by observer; values are reported in millimeters. Percentage rootmean square errors (% RMSE) are calculated relative to the landmark grand mean as[RMSE/Xz ]�100. Mean RMSE and % errors are calculated as the mean value acrossall landmarks.

Table 4

556 .

Table 5 PCA of Procrustes-aligned coordinates

Component Eigenvalue Proportion Cumulative

1 0·01090 0·667 0·6672 0·00121 0·074 0·7413 0·00088 0·054 0·7944 0·00050 0·030 0·8255 0·00040 0·024 0·8496 0·00022 0·014 0·8627 0·00018 0·011 0·8748 0·00016 0·010 0·8849 0·00014 0·009 0·892

10 0·00011 0·007 0·899

Principal Components 1–10 account for approxi-mately 90% of total variance. Principal components 11and higher each account for less than 1% of totalvariance.

pP6Itflatc

odMdmdGsmwc

prhors

1998). Principal components analysis waserformed on the covariance matrix of therocrustes-aligned coordinates using SAS.12 for Windows (SAS Institute, Cary, NC).nspection of PCA results (Table 5) showedhe first principal component to account forully 67% of total shape variance. Cumu-atively, the first ten principal componentsccounted for 90% of total variance, whilehe contributions of individual higher orderomponents were negligible.Because principal components analysis

rients successive components in theirection of maximum variation (Neff &arcus, 1980), a single divergent form can

efine the extremes of shape variation onultiple components, obscuring possible

iferences among the remaining taxa.iven the focus of this study on relation-

hips among the mangabeys, baboons, andandrills, the inclusion of Theropithecus,hich exhibits a unique and highly derivedranial morphology (Szalay & Delson, 1979;

Fleagle, 1999) was found to be counter-roductive, and principal components wereecomputed excluding Theropithecus. Thisad no effect on principal componentrdinations, but it did result in a slighteordering of the principal component axesuch that Principal Component 3, which

Regression analysisTo identify allometric effects, principalcomponent scores were tested for significantlinear relationships with size, representedhere by cranial centroid size. Centroid size isdefined as the square root of the sum ofsquared distances of a set of landmarks fromtheir centroid (Slice et al., 1996); compu-tationally, it falls in the class of Mosimannsize variables (Mosimann & Malley, 1979).Centroid size is the preferred size metric forProcrustes-based geometric morphometricanalysis (Bookstein, 1996), both because itis the size measure upon which the compu-tation of Procrustes distance is based andbecause it is approximately uncorrelatedwith all shape variables when assumptions ofhomogenous spherical landmark varianceare met (Slice et al., 1996). In the context ofthis study, cranial centroid size provides areasonable estimate of cranial size and, giventhe scarcity of individual body mass dataor associated postcrania, the best availablesurrogate for body size.

Principal component scores were plottedagainst log centroid size, Pearson productmoment correlations were computed, andregression analyses were performed. As bothdependent and independent variables werefunctions of landmark coordinates measuredwith error, a Type II regression analysis wasperformed (Sokal & Rohlf, 1981). Reducedmajor axis (RMA) regression equationswere calculated and pairwise tests forhomogeneity of slope and elevation wereperformed across genera using Cole’sNEWRMA program (Cole, 1996). Thisprogram calculates Clark’s (1980) test forhomogeneity of slope and performs thenonparametric quick test for differencesin elevation across all pairs of genera(Tsutakawa & Hewett, 1977). It also

separates males and females within genera(see below), corresponds to Principal Com-ponent 5 of the initial analysis including allgenera.

557

Influential landmarksIn classic principal components analysis,influential variables are identified by inspec-tion of eigenvectors, with large positive ornegative coefficients indicating greaterinfluence (Neff & Marcus, 1980). In three-dimensional morphometric studies, eachlandmark is represented by three variables,the X, Y, and Z coordinates, complicatingthe interpretation of eigenvectors and mak-ing it difficult to assess the influence ofindividual landmarks. To circumvent thisdifficulty, it was convenient to representeach landmark by a single variable. Thus,Procrustes-aligned coordinates were used tocompute the Euclidean distance of eachlandmark to the common centroid. Thisyielded 44 distance variables, one perlandmark, for each specimen. Principalcomponents analysis was performed on thecovariance matrix of these distance variablesusing SAS 6.12 for Windows (SAS Institute,Cary, NC). Principal components based ondistances accounted for virtually identicalproportions of the total shape variance andproduced ordinations similar to those basedon aligned coordinates. Having establishedthe comparability of the two analyses, eigen-vectors from the PCA of distances wereexamined. Inspection of eigenvectorsshowed coefficients in the range of �0·40,with most landmark coefficients in therange of �0·15. With few exceptions,paired landmarks showed coefficients ofthe same sign and similar magnitude, butvalues for paired landmarks (right and left)tended to diverge as the absolute magnitudeof coefficients decreased. Based on thispattern, landmarks or landmark pairs withcoefficient absolute values �0·15 wereidentified as influential for the component inquestion.

VisualizationVisualization of shape variation alongselected principal component axes was per-formed using Morphologika (O’Higgins &Jones, 1999). Mean male and mean femaleconfigurations were computed for eachtaxon using backscaled Procrustes-alignedcoordinates, thus summarizing averageshape and size for each taxon. Mean formswere realigned and subjected to principalcomponents analysis as above (O’Higgins &Jones, 1999). Inspection of the resultingprincipal component axes showed thatprincipal component ordinations of meanforms corresponded closely to those basedon individual specimens. The ‘‘ExploreShape Space’’ feature was then used tographically explore shape variation alongselected axes (O’Higgins & Jones, 1999).Shape variation along a principal com-ponent axis is visualized by adjusting thecoordinates of the Procrustes mean formby the coefficients of the correspondingeigenvector. By generating a series of suchforms, each corresponding to a differentpoint along the axis, Morphologika can‘‘morph’’ the mean form, displayed as awireframe diagram, along an axis, thusvisualizing shape variation summarized bythat component (O’Higgins & Jones, 1999).

Canonical variates analysisAs an alternate approach to evaluating thesize-uncorrelated component of papionincranial shape variation, a canonical variatesanalysis was employed. Procrustes-alignedcoordinates were regressed against log cen-troid size and the residual values, hereafterreferred to as ‘‘size-adjusted coordinates’’,were subjected to canonical discriminantanalysis using SAS 6.12 (SAS Institute,Cary, NC). Mahalanobis distances werecorrected for unequal sample sizes and thelarge number of variables using the formulafor unbiased Mahalanobis distance ofMarcus (1993). Similarly, significance levelsfor the Mahalanobis distances were adjusted

generates bootstrap confidence intervalsfor between-group differences in slope andelevation at the sample grand mean.

558 .

to account for multiple comparisons(Marcus, 1993). The resulting distancematrix was examined to assess the patternand magnitude of inter-group distances andthe canonical variates were plotted.

Results

0.2

0.1

0.0

�0.1

�0.2�0.2 �0.1 0.0

Principle Component 1�0.1 0.2

Pri

nci

ple

Com

pon

ent

2

CercocebusLophocebusMacacaMandrillusPapioTheropithecus

Figure 4. Plot of Principal Component 1 vs. Principal Component 2 based on PCA of Procrustes-alignedcoordinate data. Principal Component 1 separates specimens into large-bodied and small-bodied groups,while Principal Component 2 distinguishes genera within the two size groups.

Scaling relationshipsThe first principal component summarizes67% of total shape variation and distin-guishes large-bodied papionins—Papio,Mandrillus, and Theropithecus—from thesmaller taxa (Figure 4). It is common inbiological studies for the first principal com-ponent to be interpreted as ‘‘size’’ (Neff &Marcus, 1980); however, as previouslynoted, all specimens were scaled to unitsize during the initial Procrustes analysis.Nevertheless, PC 1 clearly summarizesdifferences in shape between large- andsmall-bodied papionins, respectively.

Principal Component 1 was found to behighly significantly correlated with logcentroid size across all papionins (Pearson’sr=0·96, P<0·0001), and Figure 5 demon-strates a strong linear relationship betweenthese variables across genera. This impliesthat PC 1 largely summarizes size-correlatedcranial shape variation and establishes thatallometric scaling is present.

Table 6 shows coefficients for thereduced major axis regression of PrincipalComponent 1 on log centroid size across allpapionins and by genus. Within-genus re-lationships are weaker than those for thepooled sample, with r2 values ranging from0·68 for Theropithecus to 0·87 for Papio.Inspection of Figure 5 suggests that thepooled sample regression line is a reasonableestimator of genus slopes but not elevations.To test this impression, the pooled sampleregression coefficients were treated asknown values and compared to those for

559

0.20

y = 0.399x -1.371

r2 = 0.92

0.05

�0.10

�0.15

�0.202.75 3.00 3.50

Log Centroid Size3.75 4.25

Pri

nci

ple

Com

pon

ent

1

CercocebusLophocebusMacacaMandrillusPapioTheropithecus

0.10

0.15

�0.05

0.00

3.25 4.00

Figure 5. Reduced major axis regression of Principal Component 1 on log centroid size. The strong linearrelationship between PC 1 and log centroid size confirms the presence of allometric scaling for aspects ofcranial shape summarized by this component. There is considerable homogeneity of slopes amongpapionin genera but significant differences in elevation are observed.

Table 6 Reduced major axis regression coefficients for PC 1 on log centroid size across papionins andby genus

Slope 95% CI† Intercept 95% CI† r2

All Papionins 0·399 0·385 to 0·413 �1·371 �1·422 to �1·321 0·918

Cercocebus 0·402 0·361 to 0·455 �1·410* �1·584 to �1·277 0·765Lophocebus 0·470 0·390 to 0·549 �1·628* �1·884 to �1·372 0·702Macaca 0·367 0·319 to 0·411 �1·236* �1·376 to �1·089 0·822Mandrillus 0·317 0·280 to 0·549 �1·062 �1·210 to �0·927 0·821Papio 0·322 0·292 to 0·347 �1·085 �1·175 to �0·974 0·872Theropithecus 0·313 0·228 to 0·445 �1·081 �1·557 to �0·776 0·682

Genus slopes are not significantly different from the pooled papionin slope using the single-sample test of Clarke(1980).

*Elevations significantly different from the pooled papionin line (Bonferroni adjusted P=0·05) using a modifiedquick test (Tsutakawa & Hewett, 1977) as follows: (1) under the null hypothesis that the genus elevation is equalto the pooled sample elevation, the prior probabilities of an individual case falling above or below the pooledregression line should be equal (p=q=0·5); (2) cases falling above and below the pooled sample line were tallied;and (3) the binomial probability of the resulting distribution was determined and the null hypothesis accepted orrejected accordingly.

†Bootstrap estimates of 95% confidence intervals generated by NEWRMA (Cole, 1996).

individual genera (see Table 6). Compari-sons of individual genus slopes to that forthe pooled sample found no statistically sig-

nificant differences from the commonregression slope (Bonferroni adjustedP=0·05). In comparison with the pooled

560 .

sample line, Cercocebus, Lophocebus, andMacaca had highly significantly differ-ent elevations (Bonferroni adjustedP<<0·001), while the remaining taxa werenot significantly different.

While the pooled sample regression linedescribes a general papionin allometrictrend, pairwise testing of regression coef-ficients reveals statistically significant differ-ences in scaling patterns between genera(Table 7). Four of the six papioningenera—Macaca, Mandrillus, Papio andTheropithecus—show a common slope forPrincipal Component 1 against log centroidsize. Macaca, Papio and Mandrillus alsoshare a common elevation from whichTheropithecus may be offset. The slope forCercocebus is intermediate between those ofLophocebus and Macaca and is not sig-nificantly different from either taxon; itselevation is likewise intermediate and indis-tinguishable from that of Lophocebus butdiffers significantly from Macaca. Results ofslope comparisons between the mangabeysand the remaining papionin genera arecontradictory. Under Clarke’s test, onlyLophocebus shows a significantly differentslope from any other papionin genus,namely Mandrillus. Bootstrap estimates findboth Cercocebus and Lophocebus to havesignificantly different slopes (unadjustedP=0·05) from the large-bodied taxa, Papio

and Mandrillus. Regression elevations for themangabeys are clearly significantly differentfrom those of Papio and Mandrillus. Insummary, the papionins show substantialuniformity in the scaling of PrincipalComponent 1 with respect to log centroidsize overall, but both mangabey genera dis-play significant differences in their scalingpatterns from the remaining papionins.

Principal Components 2 and higher are,by definition, uncorrelated with PC 1 (Neff

& Marcus, 1980), and were generally un-correlated with size. Only PC 5 (analysisincluding Theropithecus) was significantly,albeit weakly, correlated with log centroidsize (r=0·16, P=0·01); however, a bivariateplot of PC 5 against log centroid size (Figure6) revealed linear relationships betweenthese variables within genera and within-genus correlation coefficients greater than0·5 (range 0·52–0·90). Principal Compo-nent 5 separates sexes within genera (seebelow), and the strongest correlationsbetween PC 5 and size were observed in themost dimorphic genera. This could implythat PC 5 summarizes genus-specific allo-metric effects different from those sum-marized by PC 1. Unfortunately, in highlysize-dimorphic taxa, any shape differencebetween the sexes will be correlated withsize whether arising from allometric pro-cesses or not. Given the relative weakness of

Table 7 Pairwise comparisons of reduced major axis regression lines for PC 1 on log centroid size

Cercocebus Lophocebus Macaca Mandrillus Papio Theropithecus

Cercocebus — NS/NS NS/† NS/† NS/† NS/†Lophocebus NS/NS — NS/† †/† NS/† NS/NSMacaca NS/* */* — NS/NS NS/NS NS/†Mandrillus */* */* NS/NS — NS/NS NS/†Papio */* */* NS/NS NS/NS — NS/†Theropithecus NS/NS */NS NS/* NS/* NS/* —

Results of Clarke’s (1980) test for homogeneity of slope and the quick test for homogeneity of elevations(Tsutakawa & Hewett, 1977) are shown above the diagonal as slope/elevation, with significance levels adjusted formultiple comparisons. Results for bootstrap analysis (3000 replicates) of intergroup differences in slope andelevation at the pooled-sample grand mean are shown below the diagonal as slope/elevation; significance levels arenot adjusted for simultaneous multiple comparisons. NS not significant at Bonferoni adjusted P=0·05; NS notsignificant at unadjusted P=0·05; †significant at Bonferoni adjusted P=0·05; *significant at unadjusted P=0·05.

561

the observed correlations, it is likely thatPC 5 incorporates both size-correlated andsize-independent shape variation.

Because of concerns that within-taxoneffects might influence interpretations ofbetween-group shape variation, separateprincipal components analyses were con-ducted for males and females, respectively,and relationships of the resulting principalcomponents with centroid size were exam-ined. This procedure had the effect of gen-erating higher-order principal componentsuncorrelated with centroid size, thus achiev-ing a more rigorous statistical partition ofshape variation into size-correlated andsize-uncorrelated components. However,the resulting analyses produced principalcomponent ordinations virtually identicalto those for the combined-sex analysis.Between-genus comparisons of regressionlines for PC 1 on centroid size were likewise

similar to the combined-sex analysis withonly minor differences in significance levels,all easily attributable to the smaller single-sex sample sizes. In light of these results, itwas concluded that single-sex analysis con-tributed nothing to the understanding ofshape variation among genera and precludedthe examination of sexual shape dimorphismwithin genera. Thus, results for combined-sex analyses are presented below.

0.00

�0.06

�0.08

�0.102.75 3.00 3.50

Log Centroid Size3.75 4.25

Pri

nci

ple

Com

pon

ent

5

Cercocebus r2 = 0.34Lophocebus r2 = 0.28Macaca r2 = 0.58Mandrillus r2 = 0.81Papio r2 = 0.58Theropithecus r2 = 0.64

0.02

0.04

0.06

�0.04

�0.02

3.25 4.00

Figure 6. Plot of Principal Component 5 (analysis including Theropithecus) against log centroid size. PC 5is significantly correlated with log centroid size within genera and separates males and females. Asindicated by the r2 values, this effect is more pronounced in the larger-bodied and most highly sexuallydimorphic genera.

Shape trends and influential landmarksFigure 7 illustrates the shape differencessummarized by Principal Components 1–5(analysis excluding Theropithecus). It mustbe emphasized that, while shape trendsalong principal component axes frequentlycorrespond with observed morphologies,they do not represent the actual physicalappearance of specific animals. Rather, thewireframe diagrams illustrate morphometric

562 .

Figure 7. (a to c).

563

Figure 7. (d to f).

564 .

Figure 7. Visualization of shape variation along Principal Components 1–5 (a)–(e). All plots showPrincipal Component 1 on the X-axis. Wireframe diagrams (frontal and lateral views) illustrate theextremes of shape variation summarized by a given axis (target axis indicated by double-headed arrow).These diagrams represent the transformation of the Procrustes mean form to the maximum and minimumaxis values (indicated by single-headed arrows) with the other axis held constant at zero. Data pointsrepresent mean male and female forms for each taxon and indicate their positions relative to shape trendsdefined by each axis.

Influential landmarks for Principal Components 1–5

PC Positive coefficients Negative coefficients

1 IN, BR, OP, PGL RH, NS, PR, IV, PR2, MP3, PMI2 GL, NA, DAC, MTI, MTS IN, RH, OP, ST, IV, POR3 IN, BR, GL, NA, OP, IV, PMS, DAC FMT, ZTS, ZTI4 IN, BR ZMS, DAC5 BR, ZMI, MP3 RH, PMS, MTS, ZTI

Influential landmarks and landmark pairs (right and left) are defined as thosewith absolute coefficients �0·15 for the principal component in question (see text).Figure 8 maps influential landmarks for each component onto a representativepapionin skull. Landmark abbreviations follow Table 2.

Table 8

trends along a subset of all possible axes ofshape variation. In the following discussions,when a taxon is characterized as having aparticular trait, it is always relative to allother taxa and with reference to the shapecomponent in question. Table 8 summarizesthe influential landmarks which drive shapetrends along Principal Components 1–5,and Figure 8 displays the correspondinglandmark maps.

As previously noted, the first principalcomponent [Figure 7(a)] separates thepapionin genera by size, with Papio,Mandrillus, and Theropithecus (whereincluded) showing large positive scores,while Cercocebus, Lophocebus, and Macacahave negative scores. The large-bodied taxaare characterized by relatively small and lowneurocrania; relatively small orbits and cor-respondingly short upper faces; and long,somewhat more dorsally oriented muzzles.Small-bodied taxa exhibit relatively large,globular neurocrania; large orbits andrelatively tall upper faces; and shorter, moreventrally oriented muzzles. Consistent withthis pattern, Principal Component 1 wasfound to be most strongly influenced by

landmarks of the neurocranium and anteriorrostrum [Figure 8(a)]. It is tempting toconstrue this pattern as signifying ‘‘forwardmovement’’ of the anterior muzzle and‘‘contraction’’ of the neurocranium, but itis ill-advised to give these landmark mapssimple, mechanistic interpretations. Instead,each is best read conservatively as denotingregions of greatest morphological variationand, therefore, biological interest.

Principal Component 2 distinguishesMacaca and Papio, which exhibit relativelynegative scores, from all other papionins[Figure 7(b)]. Visualizations show theformer taxa to be relatively klinorhynch,showing prominence of the interorbital andsupraorbital regions and a ventral deflectionof the muzzle. By contrast, the remainingpapionins show a more dorsally orientedface, producing a straighter profile and lessprojecting superior orbital margins. Princi-pal Component 2 is influenced by rhinionand landmarks of the ventral neurocraniumand palate, which show strong negative coef-ficients, with positively weighted landmarksin the interorbital region and along thesupraorbital margins [Figure 8(b)].

565

Canonical variates analysisThe first three canonical variates accountfor 56, 25 and 19% of total among-groupvariance, respectively for a cumulative totalof over 99%. The fourth canonical variate,which separates males from females,accounts for less than 1% of total varianceand is not statistically significant. Table 9shows Mahalanobis distances betweengenera; all distances are highly significant(Bonferroni adjusted P=0·001). Papio con-sistently showed the smallest distances toother groups; Lophocebus, the largest.Lophocebus showed the smallest distanceto Papio; Cercocebus was next closest toLophocebus, followed by Mandrillus, andMacaca. Cercocebus, too, had its smallestdistance to Papio, with Mandrillus, Lopho-cebus, and Macaca increasingly more distant.Similarly, Macaca and Mandrillus were eachclosest to Papio, and markedly more distantfrom the mangabeys. Papio was least distantfrom Macaca and farthest from Lophocebus.Notably, Cercocebus and Lophocebus are quitedistant from each other, as are Papio andMandrillus.

Figure 9 shows a three-dimensional plotof the first three canonical axes. CanonicalVariate 1 separates Macaca from Lophocebus,with Papio, Mandrillus, and Cercocebusintermediate. Canonical Variate 2 contrastsPapio and Mandrillus. Canonical Variate 3separates Lophocebus from Cercocebus. Asmight be predicted from the Mahalanobisdistances, Papio and Macaca show the mostsimilar pattern of residual shape variationwith respect to the canonical axes. Cerco-cebus and Mandrillus show some similarities,

Principal Component 3 (analysis exclud-ing Theropithecus) separates males, withrelatively positive scores, from females,which exhibit relatively negative scores[Figure 7(c)]. This component correspondsto Principal Component 5 of the originalanalysis and is similarly correlated with logcentroid size. Relative to conspecific males,females show more globular neurocrania,relatively larger orbits, a more parabolicpalate outline, and narrow zygomatics.Males have lower neurocrania, relativelysmaller orbits, a more parallel-sided palate,and flaring zygomatics. Principal Com-ponent 3 is most strongly influenced bylandmarks of the neurocranium and inter-orbital and supraorbital regions, whichshow positive coefficients, with nega-tively weighted landmarks bracketing thezygomatics [Figure 8(c)].

Principal Component 4 distinguishesMacaca and female Mandrillus from Papio,with the remaining groups occupyingintermediate positions along this axis[Figure 7(d)]. Visualizations show theformer taxa to be characterized by relativelytall orbits and posteriorly positioned zygo-matic roots. At the opposite extreme, Papioexhibits relatively short orbits, a deeperzygomatic, and anteriorly positioned zygo-matic roots. Only four influential landmarkswere identified for Principal Component 4,with inion and bregma having large posi-tive coefficients and dacryon and zygo-maxillare superior large negative coefficients[Figure 8(d)].

Principal Component 5 separates Lopho-cebus and Cercocebus, while all other taxafall close to the origin. Lophocebus ischaracterized by a relatively low and longneurocranium; supraorbital margins highand arched; elongated and narrow nasals; arelatively narrow malar region; and a narrowpalate [Figure 7(e)]. By contrast, Cercocebusshows a shorter and higher neurocranium;supraorbital margins low and relativelystraight; shorter, broader nasals; a broader

malar region; and a broader and somewhatmore parabolic palate. Principal Com-ponent 5 is most strongly influenced bybregma and landmarks bracketing themaxilla with positive coefficients, whilelandmarks of the anterior nasal region,supraorbital margin and zygomatic archshow negative coefficients [Figure 8(e)].

566 .

Figure 8. (a to c).

567

Figure 8. (d and e).

Figure 8. Maps of influential landmarks for Principal Components 1–5 (a)–(e). Maps based onPCA of landmark distances (excluding Theropithecus). Influential landmarks (Table 8) are defined asthose with absolute coefficients �0·15 for the principal component in question (see text). Theselandmarks identify regions of greatest morphological variability and contribute disproportionately toshape trends summarized by the corresponding principal component (see Figure 7). Line drawingsof female Macaca fascicularis skull courtesy of J. Michael Plavcan (� J. M. Plavcan) and adapted withartist’s permission.

both relative to the canonical axes and toPapio, but superimposition of a minimumspanning tree on the canonical clusters (not

shown) confirms that Mandrillus is mostsimilar to Macaca. Lophocebus is widelyseparated from all other groups.

568 .

CercocebusLophocebusMacacaMandrillusPapio

15

10

5

0

�5

�10

�15

�2015

105

0�5

�10�15 �15

�10�5

05

1015

20

Can

onic

al V

aria

te 2

Canonical Variate 3Canonical Variate 1

Figure 9. Three-dimensional canonical plot based on CVA analysis of size-adjusted landmark coordinates.Macaca and Papio show the most similar pattern of residual shape variation with respect to the canonicalaxes. Lophocebus is widely separated from all other taxa, including Cercocebus.

Mahalanobis distances between papionin genera

Cercocebus Lophocebus Macaca Mandrillus Papio

Cercocebus — 16·64 20·99 15·00 14·07Lophocebus 10·01 — 24·51 17·79 15·85Macaca 12·85 15·07 — 13·45 11·60Mandrillus 9·02 10·80 8·04 — 14·31Papio 8·46 9·59 6·88 8·66 —

Mahalanobis distance values (D) are shown above the diagonal; unbiasedMahalanobis distance values (Du)—corrected for unequal sample sizes and the largenumber of variables—are shown below the diagonal. All distances are significant atBonferroni adjusted P=0·001. The unbiased squared Mahalanobis distance D2

u iscomputed as:

where N=total sample size; g=number of groups; p=number of variables; and n1 andn2 are the sample sizes of the groups under comparison (Marcus, 1993).

Table 9

Cnv

Discussion and conclusions

ombining geometric morphometric tech-iques and traditional bivariate and multi-ariate statistics, the present study takes

advantage of the complementary infor-mation these methods provide. Principalcomponents analysis of aligned coordinatedata serves the prosaic functions of data

569

Allometric effectsPrincipal Component 1, which summarizes67% of total cranial shape variation, ishighly significantly correlated with cranialsize. Thus, craniofacial allometry is presentin the Papionini; however the precise natureof scaling relationships among papionin taxais complex. Parametric and nonparametrictests for homogeneity of slopes and inter-cepts produced broadly similar results, butinconsistencies were observed betweendifferent tests and significance levels wereoften borderline making some results incon-clusive. Pairwise comparisons of regressioncoefficients showed that Macaca, Papio,Mandrillus, and Theropithecus share a com-mon slope for PC 1 on log centroid size.Macaca, Papio and Mandrillus also share acommon elevation, and thus fall on a com-mon line from which Theropithecus may ormay not be vertically transposed. In thelatter case, results of the quick test andbootstrap were contradictory, due in part tothe small sample size and relatively low r2

values for Theropithecus. Larger samples willbe required to adequately assess scaling pat-terns for this taxon. Cercocebus also sharesthe common papionin slope; however, itselevation differs from the large-bodied

animals and possibly Macaca. Results forLophocebus are the most ambiguous. Itclearly shares a common slope and elevationwith Cercocebus. Like Cercocebus, it may alsoshare a common slope with Macaca, Papio,and Theropithecus, but not Mandrillus. In anycase, the mangabeys share a common scal-ing pattern to the exclusion of the remainingpapionins.

Strictly defined, ontogenetic scalingoccurs when observed differences in shapeare ‘‘produced or accompanied by extensionor truncation of common (or ancestral)growth allometries’’ (Shea, 1985:179). Afinding that adults of closely related speciesshare common allometric patterns is oftenindicative of the presence of ontogeneticscaling but never, in itself, conclusive. Thepresent finding of a shared adult allometricscaling pattern for Macaca and Papio iswholly consistent with previous growthstudies reporting ontogenetic scaling forthese taxa (Swindler & Sirianni, 1973;Swindler et al., 1973; Profant, 1995). Thepresence of the same allometric pattern inMandrillus and possibly Theropithecus raisesthe possibility that the aspects of cranio-facial shape summarized by PrincipalComponent 1 have a common ontogeneticbasis in all large-bodied papionins as well asmacaques. Although similar adult inter-specific scaling patterns may be producedby functional allometry (Jim Cheverud,personal communication), the prevalence ofontogenetic scaling among catarrhine pri-mates renders it the most plausible expla-nation for the observed similarities amongadult static allometries.

While cranial developmental patterns ofMacaca and Papio have been intensivelystudied, ontogenetic trends among theremaining papionin taxa are less well estab-lished. Profant (1995) found that papionins,including mangabeys, conformed closelyto ontogenetic trajectories established forMacaca. However, Shah & Leigh (1995)reported incongruent ontogenetic scaling

reduction and ordination of specimens andpermits a rough partition of total shapevariation into size-correlated and size-uncorrelated moieties. Traditional regres-sion analysis of morphometrically-derivedsize and shape variables simplifies the com-parison of multivariate allometric relation-ships across and among groups, whileMahalanobis distances and canonical plotssummarize patterns of overall morphologicalsimilarity when size effects are controlled.Direct visualization of shape componentsallows exploration of allometric shape trendsand patterns of residual shape variation,while the corresponding landmark mapslocalize shape variation to specific regions ofthe skull.

570 .

patterns among Cercocebus, Papio, andMandrillus, and Collard & O’Higgins (2000,2001) describe heterogeneity of multivariategrowth vectors among African papionins.The present study finds significant differ-ences in patterns of adult static allometrybetween the mangabeys and the large-bodied papionins in that Cercocebus andLophocebus share a common allometric linewhich is vertically transposed from that ofthe remaining taxa. The divergence of adultscaling patterns in Cercocebus and Lophocebusis inconsistent with models invoking simpletruncation of common growth vectors andargues against the presence of ontogeneticallometry, at least in its narrowest sense.This interpretation is supported by the workof Collard & O’Higgins (2001) who reportthat significant differences between theontogenetic trajectories of mangabeys andlarge-bodied papionins are already presentearly in postnatal development.

The apparent absence of classic ontogen-etic allometry notwithstanding, allometricscaling is a major determinant of papionincranial shape variation. Principal Compo-nent 1 summarizes a substantial portion oftotal shape variation and is strongly linearlyrelated with size across all papionins.Although the small-bodied taxa—Macaca,Cercocebus, and Lophocebus—are transposedabove and below the common RMA line,genus slopes are not significantly differentfrom the pooled slope. Despite observedheterogeneities among taxa, the pooled-sample regression line appears to be agood estimator of this general allometrictrend. Visualization of the shape variationsummarized by PC 1 shows positive facialallometry, characterized by increased facialprognathism, and negative neurocranialallometry as cranial size increases. The cor-responding landmark map supports thisfinding, identifying landmarks on the neuro-cranium, premaxilla, anterior maxilla,and anterior nasals as exerting particularinfluence on this component.

These results are not surprising giventhe extensive literature documentingmammalian cranial scaling patterns ingeneral, and primate cranial scaling in par-ticular. Cranial allometry characterized bypositive facial scaling and changes in facialorientation is observed across varying time-scales and taxonomic levels in mammaliangroups as diverse as equids (Reeve &Murray, 1942; Radinsky, 1984); domesticcanids (Weidenreich, 1941); New Worldanteaters (Reeve, 1939); carnivores(Radinsky, 1981); and titanotheres (Hersh,1934). Within Primates, positive facialallometry is well documented and suf-ficiently common to be accepted as a generaltrend among non-human catarrhines(Biegert, 1963; Vogel, 1968; Ravosa &Profant, 2000). The ontogenetic basis ofthis pattern has been most intensivelystudied in the extant great apes. In gorillas,chimpanzees, and orang-utans, cranialgrowth is dominated by relatively rapidforward growth of the face accompanied byincreased infraorbital facial depth, dorsalrotation of the alveolar process, anddecreased basicranial flexion (Krogman,1931a,b,c; Shea, 1983, 1985; Leutenegger& Masterson, 1989). Extension of theseontogenetic trajectories produces cranialsexual dimorphism within species anddifferences in cranial form between theAfrican apes (Krogman, 1931a,b,c; Giles,1956; Shea, 1983, 1985; Leutenegger &Masterson, 1989; O’Higgins et al., 1990;O’Higgins & Dryden, 1993). Among cerco-pithecoids, Shea (1992) documented similartrends for cercopithecins, where ontogeneticallometry appears to explain the craniofacialproportions of Miopithecus, the smallestextant catarrhine.

Within Papionini, comparative longi-tudinal growth studies of M. nemestrinaand P. (hamadryas) cynocephalus have alsodescribed facial ontogenies dominated bypositive forward growth of the rostrum anddorsal rotation of the maxilla and palate

571

(Swindler & Sirianni, 1973; Swindler et al.,1973; McNamara et al., 1976). Differ-ential growth along these shared ontogen-etic trajectories contributes to the shapedifferences observed between macaquesand baboons; differences among species andsubspecies of Papio and Macaca; andsexual shape dimorphism within species(Zuckerman, 1926; Freedman, 1962, 1963;Albrecht, 1980; Bookstein, 1985; Cochard,1985; Cheverud & Richtsmeier, 1986;Leigh & Cheverud, 1991; Ravosa, 1991;Richstmeier et al., 1993). In the presentstudy, Principal Component 1 summarizesa pattern of static multivariate cranialallometry across adult papionins whichclosely resembles established patterns ofcatarrhine cranial ontogeny as well as gen-eral trends in mammalian cranial allometry.These broader trends, irrespective of theirdevelopmental origins, appear sufficientto account for much of the homoplasy inpapionin cranial form.

If the pattern of cranial allometry withinthe Papionini is less than noteworthy, theextent to which allometric trends drive over-all cranial shape variation is remarkable.Profant (1995) found that allometric effectsaccounted for 98% of cranial shape variationamong Macaca fascicularis, Macaca nemes-trina, and Papio (hamadryas) cynocephalus,taxa known to share a common scalingtrajectory. Even allowing for a greater diver-sity of adult morphologies in the currentsample; observed heterogeneities of scalingpatterns among taxa; and the proportion ofPrincipal Component 1 variance not attribu-table to cranial size, allometric effects stillaccount for over 60% of total cranial shapevariation. This is consistent with the resultsof Profant & Shea (1994), who found thatallometric effects explained a preponderanceof craniometric variation within the Cerco-pithecinae, with papionins showing steeperallometric trajectories than cercopithecins.These slope differences resulted in greatermorphological diversification for papionins

over equivalent size ranges (Profant & Shea,1994; Ravosa & Profant, 2000). It is thismarked divergence of form that is reflectedin traditional morphological classifi-cations uniting taxa occupying the extremesof the papionin size ranges. An emphasis oncranial proportions in general, and relativefacial length in particular, resulted inpapionin phylogenies heavily influenced bysize-correlated features.

Residual shape variationResidual shape variation encompasses shapedifferences due to all factors other thanallometry, including phylogenetic effects.Canonical discriminant analysis of size-adjusted coordinates and visual explorationof size-uncorrelated shape componentsprovide complementary information aboutpatterns of residual shape variation. Theformer summarizes overall morphologicalsimilarity, while the latter gives insights intopatterns of cranial shape variation whichcontribute to these similarities.

The Mahalanobis distance matrix basedon size-adjusted coordinates (Table 9)shows little correspondence to commonconceptions of papionin cranial form orphylogenetic relationships. AlthoughLophocebus has its greatest similarity withPapio, the latter actually falls closest toMacaca. Cercocebus, too, has its strongestsimilarity with Papio, while Mandrillus isactually slightly closer to Macaca. In fact,the observed pattern of inter-genus dis-tances implies considerable morphologicalconservatism in papionin cranial form.Interestingly, Lophocebus is the most distinctof the non-Theropithecus African papionins,consistently showing the greatest pairwisedistances. The canonical plot (Figure 9)supports this interpretation. With respect tothe first three canonical axes, Macaca andPapio are most similar, while Lophocebus isclearly divergent from the other taxa andwidely separated from Cercocebus.

572 .

Visualization of shape variation alongPrincipal Component 5 (analysis excludingTheropithecus), whose extremes are occupiedby Cercocebus and Lophocebus, respectively,identifies the shape differences which con-tribute to the morphological distancebetween these taxa [Figure 7(e)]. In com-parison with Cercocebus, Lophocebus displaysa lower neurocranium; a narrower and moredorsally oriented muzzle with relatively pro-jecting nasals; and a narrower malar regionwith less flaring zygomatic arches. Thisagrees with Groves (1978) description of the‘‘stretched out’’ appearance of Lophocebusskulls. Groves (1978) also noted the rela-tively shorter and broader orbits of Cerco-cebus, which are reflected here in differencesin supraorbital configuration along PC 5[Figure 7(e)] and highlighted by the stronginfluence of the mid-torus superior land-mark [Figure 8(e)]. Interestingly, theobserved pattern of shape differencesbetween Cercocebus and Lophocebus repre-sents a reversal of the general papionin allo-metric trend. Larger papionins typicallyshow lower and narrower neurocrania andnarrower faces, but Lophocebus is smallerthan Cercocebus, both in body mass andcranial centroid size.

The second principal component [Figure7(b)] united Macaca and Papio to the exclu-sion of other papionins, a result congruentwith the canonical variates analysis andsmall Mahalanobis distance between thesetaxa. Visualization of shape trends along PC2 shows Macaca and Papio to be relativelyklinorhynch, with projecting interorbitaland supraorbital regions and procumbent(ventrally-deflected) faces relative to othertaxa. Supporting this, influential landmarksare found on the basicranium, anteroinferiorpalate, infraorbital region, and supraorbitalmargins [Figure 8(b)]. Relative prominenceof the supraorbital region is a discernablefeature of both Macaca and Papio, and Papiohas long been recognized as exhibiting pro-nounced facial procumbence (Zuckerman,

1926; Freedman, 1963), showing markedlygreater ‘‘downward bending’’ of the facethan Mandrillus (see illustrations in Hill,1974; Maier, 2000). While this feature is notordinarily associated with the relativelyorthognathic macaques, in comparison withthe similar-sized Cercocebus, and particularlyLophocebus, the macaque lower face is, infact, more ventrally oriented. Thus, Papioand Macaca appear to share a commonpattern of facial hafting when allometriceffects are controlled.

Principal Component 3 (analysis exclud-ing Theropithecus) separates males andfemales and is correlated with centroid sizewithin genera. Shape variation along thisaxis is similar in some respects to trendsalong Principal Component 1, probablyreflecting within-taxon allometric effects(see above). However, patterns of shapevariation along PC 3 also highlight differ-ences between sexes which are not directlyattributable to differences in cranial centroidsize [Figure 7(c)]. In particular, femaleneurocrania, when scaled to unit size, arehigher and more globular than those of theirmale counterparts but not appreciablylonger. Males exhibit a more parallel-sidedpalate, while females show a more parabolicpalatal outline, a difference clearly related tocanine sexual dimorphism. Finally, malesexhibit comparatively broad and flaringzygomatics in comparison with females,possibly reflecting differences in the relativesize and orientation of the temporalis andmasseter muscles (Swindler & Wood,1982).

Principal Component 4 separates Macacaand female Mandrillus from other papionins,most notably Papio. The canonical dis-criminant analysis likewise places Mandrillusslightly closer to Macaca than Papio. Thiscomponent appears to summarize differ-ences in the orientation of the zygomaticbodies and the relative contribution of theorbit to total facial height; however, thesedifferences are extremely subtle. Reference

573

to a small comparative sample found thatmale and female Macaca fascicularis doresemble female Mandrillus sphinx in thefeatures highlighted by PC 4. In both, theorbits constitute roughly half of total facialheight and the zygomatic bodies are some-what ‘‘swept back’’, giving the midface avisor-like appearance. This contrasts withmale (and to a lesser extent female) Papio,where the malar regions are more anteriorlyoriented and the orbits contribute a smallerproportion of total facial height. In contrastto other components, which show strongagreement between shape component visu-alizations and landmark maps, only fourinfluential landmarks were identified forPrincipal Component 4 [Figure 8(d)] andthese showed no clear correspondence tovisualized shape trends [Figure 7(d)]. Of allthe shape components considered, thepattern of variation along this component isleast robustly supported and its significance,if any, is unclear.

ConclusionsObserved patterns of cranial shape variationamong the Papionini indicate that the evo-lution of cranial form within this group isless straightforward than previously thought.The presence of a common allometric scal-ing pattern in Macaca, Papio, Mandrillus,and perhaps Theropithecus, suggests that thisis a shared feature and represents theprimitive condition for African papionins.Outgroup comparisons will be required todetermine whether this pattern is a derivedfeature for papionins or a retention froma more distant cercopithecine ancestor.The downward transposition of allometrictrajectories in Cercocebus and Lophocebusindicates that mangabeys are similar in pos-sessing enlarged neurocrania and reducedmuzzles in comparison with comparablysized macaques, mandrills, and baboons.Given the current understanding ofpapionin phylogeny, this shared mangabeyscaling pattern is most parsimoniously

interpreted as homoplastic and a product ofparallel evolution. However, it must benoted that Collard & O’Higgins (2001),applying similar methods to a somewhatdifferent dataset, have reached exactly theopposite conclusion concerning the polarityof papionin cranial allometries. Studiesincluding additional taxa occupying the fullrange of papionin body sizes may help clarifythe distribution and polarity of cranial scal-ing patterns, but detailed growth studiesin Cercocebus, Lophocebus, and the smallermacaques will be required to isolate thedevelopmental basis for and timing of theinitial divergence of allometric trajectories inthis group. At the same time, renewed atten-tion should be given to identifying ecologicaland functional selective forces contributingto variation in allometric scaling patternsamong papionin primates.

On first inspection, size-uncorrelatedshape components are not particularlyinformative concerning phylogenetic re-lationships among African papionins. Whilethe second principal component separatesthe outgroup Macaca and, interestingly,Papio from all other taxa, no componentunites members of the African molecularclades. Similarly, canonical analysis of size-adjusted coordinates shows no particularaffinity between clade members when sizeeffects are controlled. Thus, analysis of size-uncorrelated shape variation gives no directsupport to molecular phylogenies. Theresults are perhaps more significant forwhat is not observed. When size-correlatedshape variation is factored out, the pervasivesimilarities which have driven traditionalclassifications disappear. No shape com-ponent simultaneously unites the twomangabeys, on the one hand, and baboonsand mandrills, on the other. Rather,Principal Component 2 distinguishes Papioand Macaca from Mandrillus and themangabeys and Principal Component 5separates Cercocebus and Lophocebus. Simi-larly, both Mahalanobis distances and

574 .

canonical plots show Cercocebus andLophocebus to be clearly separated in canoni-cal space, while Papio has its strongestmorphometric affinities with Macaca.

As in the case of size-correlated shapevariation, Macaca and Papio share a com-mon pattern of residual shape variationwhich is inferred to be the ancestralpapionin condition. Further studies includ-ing cercopithecin outgroups are needed todetermine whether this pattern is unique tothe Papionini or part of a common cerco-pithecine heritage. The observed pattern ofMahalanobis distances suggests that Papiorepresents the primitive morphometricpattern for African papionins, from whichother taxa have diverged to varying degrees.Although Cercocebus and Mandrillus showcertain similarities in their overall pattern ofvariation with respect to the canonical axes,patterns of residual variation do not giveclear support to molecular phylogenies ofthe Papionini. Rather, they describe aremarkable level of conservatism in cranialform among non-Theropithecus papionins,with individual taxa exhibiting rela-tively minor variations on a commonmorphometric theme.

The Papionini are often cited as a particu-larly striking example of morphologicalhomoplasy (Disotell, 1996; Lockwood &Fleagle, 1999; Collard & Wood, 2000).The present study documents patterns ofhomoplastic similarity in papionin cranialshape due to allometric effects and providessome indication of the processes which maycontribute to this similarity. However, thisstudy raises as many questions as it answers.Results do not resolve the question ofmangabey cranial homoplasy, but ratherreframe it more narrowly in terms of parallelshifts in allometric scaling patterns. Onto-genetic studies will be required to explicatethe developmental origins of variationin papionin allometric trajectories. Theinference that Papio preserves the primitivemorphometric pattern for African papionins

implies that similarities in mangabey facialform result from parallel evolution. How-ever, broader taxonomic samples incor-porating additional papionin taxa andmultiple outgroups are needed to reconcileconflicting results (cf. Collard & O’Higgins,2001) and more rigorously test hypothesesof polarity for the shape trends observedhere. Analyses of residual (size-uncorrelated) shape variation reveal anabsence of morphometric affinity betweenCercocebus and Lophocebus when size effectsare controlled, but this lack of quantitativeresemblance is belied by similarities in keyqualitative features. Thus, it will be nec-essary to clarify the definition, phylogeneticdistribution, and functional significance ofdistinctive papionin traits, particularly thepresence of excavated facial fossae and thedevelopment of maxillary ridges (McGraw& Fleagle, 2000), before the issue ofpapionin craniofacial homoplasy can finallybe laid to rest. These uncertainties notwith-standing, the present study significantlyincreases our knowledge of patterns ofcraniometric variation in extant papioninprimates. It is hoped these results will pro-vide a more robust comparative frameworkwithin which to evaluate fossil papionintaxa and enhance our understanding ofthe evolutionary history of this fascinatinggroup.

Summary

Geometric morphometric analysis of craniallandmark data shows that allometricscaling of cranial shape is present acrossall papionins and accounts for certaincraniofacial similarities between like-sizedmembers of separate clades. Comparisonsof regression lines show considerablehomogeneity of scaling among papioningenera; however, results give no support tohypotheses of uniquely shared allometricscaling patterns either within papioninclades or across all African papionins.

575

Rather, Cercocebus and Lophocebus showminor differences in slope and highly sig-nificant negative displacement of theirallometric lines relative to other papionins.Nevertheless, allometric trends reflectinggeneral patterns of mammalian cranialallometry appear to be a major determinantof papionin cranial shape variation, con-tributing considerably to the markeddivergence of form which is reflected intraditional classifications of this group.Analysis of size-uncorrelated shape variationgives no clear support to molecular phylog-enies, but highlights the absence of morpho-metric similarity between the mangabeygenera when size effects are controlled.Patterns of allometric and residual shapevariation suggest conservatism of non-Theropithecus papionin cranial form, withPapio appearing to demonstrate a primitivemorphometric pattern from which otherAfrican papionins, most notably Lophocebus,have diverged. Outgroup comparisons withcercopithecin taxa will be required to deter-mine whether these trends are unique tothe Papionini or inherited from a more dis-tant cercopithecine ancestor. Descriptionsof craniometric variation among extantpapionins are a necessary basis for assess-ments of morphological and phylogeneticaffinities of fossil taxa, and the present studylays the groundwork for future studies ofextinct papionins.

Acknowledgements

This work was conducted under theauspices of the Morphometrics ResearchGroup of the New York Consortium inEvolutionary Primatology and supported byNSF Research & Training Grant # BIR9602234 (NYCEP) and NSF SpecialProgram Grant # ACI-9982351 (E. Delson,L. F. Marcus, D. P. Reddy, N. Tyson, andW. K. Barnett, Principal Investigators). Iespecially wish to thank Eric Delson for

his continued interest in and support ofthis work.

I am indebted to Steve Frost and TonyTosi for their substantial contributions tothe NYCEP-MRG primate cranial data-base, without which this research would nothave been possible. I also wish to thank theCurators and Collections Managers ofthe following institutions: the AmericanMuseum of Natural History, Departmentof Mammalogy; the National Museum ofNatural History, Department of Mam-malogy; the Field Museum of NaturalHistory, Division of Mammalogy; theBritish Museum of Natural History, Depart-ment of Zoology; the Powell–CottonMuseum; the Senckenberg Natural HistoryMuseum—Frankfurt; the Laboratory forHuman Evolutionary Studies—University ofCalifornia Berkeley; and the University ofCalifornia Museum of Vertebrate Zoology.

I am deeply grateful to Les Marcus, DavidReddy, Dennis Slice, and James Rohlf fortheir patient guidance as I ventured into thebrave new world of geometric morpho-metrics. Thanks are also due to Tim Cole,Dennis Slice, Richard Smith, Bill Jungers,Mike Plavcan, and John Hunter for adviceon technical and statistical matters, and toJim Cheverud for his thoughtful discussionof allometry and the meaning of Cheverud(1982). I wish to thank Dennis Slice andPaul O’Higgins for making beta copies oftheir morphometrics packages available andMike Plavcan for granting permission toadapt his illustration of the female Macacafascicularis skull.

Steve Frost, Les Marcus, Mike Plavcan,Steve Leigh, and two anonymous reviewersprovided comments on various drafts of thispaper which improved the final product con-siderably. Any errors of fact or interpretationare, of course, my own.

This paper is NYCEP MorphometricsContribution #3.

In Memoriam: This paper is dedicatedto the memory of Leslie Marcus, whose

576 .

contributions to the field of geometricmorphometrics were exceeded only by hisgenerosity to his friends and colleagues.

References

Albrecht, G. H. (1980). Longitudinal, taxonomic,sexual, and insular determinants of size variation inpigtail macaques, Macaca nemestrina. Int. J. Primatol.1, 141–152.

Barnicot, N. A. & Hewett-Emmett, D. (1972). Redcell and serum proteins of Cercocebus, Presbytis,Colobus and certain other species. Folia primatol. 17,442–457.

Biegert, J. (1963). The evaluation of characteristics ofthe skull, hands, and feet for primate taxonomy. In(S. L. Washburn, Ed.) Classification and HumanEvolution, pp. 116–145. Chicago: Aldine Press.

Bookstein, F. L. (1985). Modeling differences incranial form, with examples from primates. In (W. L.Jungers, Ed.) Size and Scaling in Primate Biology,pp. 207–229. New York: Plenum Press.

Bookstein, F. L. (1991). Morphometric Tools forLandmark Data. Cambridge: Cambridge UniversityPress.

Bookstein, F. L. (1996). Combining the tools of geo-metric morphometrics. In (L. F. Marcus, M. Corti,A. Loy, G. J. P. Naylor & D. E. Slice, Eds) Advancesin Morphometrics, pp. 131–151. New York: PlenumPress.

Burnett, G. T. (1828). Illustrations of the Manupeda orapes and their allies: being the arrangement of theQuadrumana or anthropomorphous beasts indicatedin the outline. Q. J. Sci. Lit. Art. 26, 300–307.

Cheverud, J. M. (1982). Relationship among onto-genetic, static, and evolutionary allometry. Am. J.phys. Anthrop. 59, 139–149.

Cheverud, J. M. & Richtsmeier, J. T. (1986). Finite-element scaling applied to sexual dimorphism inrhesus macaque (Macaca mulatta) facial growth. Syst.Zool. 35, 381–399.

Clarke, M. R. B. (1980). The reduced major axis of abivariate sample. Biometrika 67, 441–446.

Cochard, L. R. (1985). Ontogenetic allometry of theskull and dentition of the rhesus monkey (Macacamulatta). In (W. L. Jungers, Ed.) Size and Scaling inPrimate Biology, pp. 231–255. New York: PlenumPress.

Cole, T. M. (1996). NEWRMA. Department of BasicMedical Sciences, University of Missouri, KansasCity, MO.

Collard, M. & O’Higgins, P. (2000). Ontogeny andhomoplasy in the papionin face. Am. J. phys.Anthrop. Suppl. 30, 128.

Collard, M. & O’Higgins, P. (2001). Ontogeny andhomoplasy in the papionin monkey face. Evol. Dev.3, 322–331.

Collard, M. & Wood, B. (2000). How reliable arehuman phylogenetic hypotheses? Proc. natl. Acad.Sci. USA 97, 5003–5006.

Cronin, J. E. & Sarich, V. M. (1976). Molecularevidence for dual origin of mangabeys among OldWorld primates. Nature 260, 700–702.

Delson, E. (1973). Fossil colobine monkeys of theCircum-Mediterranean region and the evolutionaryhistory of the Cercopithecidae. Ph.D. Dissertation,Columbia University.

Delson, E. (1975a). Evolutionary history of theCercopithecidae. In (F. S. Szalay, Ed.) Contributionsto Primatology. Vol. 5: Approaches to PrimatePaleobiology, pp. 167–217. Basel: Karger.

Delson, E. (1975b). Paleoecology and zoogeography ofthe Old World monkeys. In (R. Tuttle, Ed.) PrimateFunctional Morphology and Evolution, pp. 37–64. TheHague: Mouton.

Delson, E. & Dean, D. (1993). Are Papio baringensisR. Leakey, 1969, and P. quadratirostris Iwamoto,1982, species of Papio or Theropithecus? In (N. G.Jablonski, Ed.) Theropithecus: The Rise and Fall of aPrimate Genus, pp. 125–156. Cambridge: CambridgeUniversity Press.

Delson, E., Terranova, C. J., Jungers, W. L., Sargis,E. J., Jablonski, N. G. & Dechow, P. C. (2000). Bodymass in Cercopithecidae (Primates, Mammalia): esti-mation and scaling in extinct and extant taxa. Anthr.Papers Am. Mus. Nat. Hist. 83, 1–159.

Disotell, T. R. (1994). Generic level relationships ofthe Papionini (Cercopithecoidea). Am. J. phys.Anthrop. 94, 47–57.

Disotell, T. R. (1996). The phylogeny of Old Worldmonkeys. Evol. Anthropol. 5, 18–24.

Disotell, T. R., Honeycutt, R. L. & Ruvolo, M.(1992). Mitochondrial DNA phylogeny of theOld-World monkey tribe Papionini. Mol. Biol. Evol.9, 1–13.

Dryden, I. L. & Mardia, K. V. (1998). Statistical ShapeAnalysis. New York: John Wiley.

Dutrillaux, B., Fosse, A. M. & Chauvier, G. (1979).Eutude cytogenetique de six especes ou sous-especesde mangabeys (Papiinae [sic], Cercopithecoidea).Ann. Genet. 22, 88–92.

Dutrillaux, B., Couturier, J., Muleris, M., Lombard,M. & Chauvier, G. (1982). Chromosomal phylo-geny of forty-two species or subspecies of Cerco-pithecoids (Primates, Catarrhini). Ann. Genet. 25,96–109.

Fleagle, J. G. (1999). Primate Adaptation and Evolution.2nd edn. New York: Academic Press.

Fleagle, J. G. & McGraw, W. S. (1999). Skeletal anddental morphology supports diphyletic origin ofbaboons and mandrills. Proc. natl. Acad. Sci. USA96, 1157–1161.

Freedman, L. (1962). Growth of muzzle length relativeto calvaria length in Papio. Growth 26, 117–128.

Freedman, L. (1963). A biometric study of Papiocynocephalus skulls from northern Rhodesia andNyasaland. J. Mammalogy 44, 24–43.

Frost, S. R., Marcus, L. F., Reddy, D. P., Bookstein, F.& Delson, E. (in prep.). Landmark comparisonamong large bodied papionins (Primates: Cerco-pithecinae), with extension of procrustes methods toridge curves.

577

Groves, C. P. (1978). Phylogenetic and popu-lation systematics of the mangabeys (Primates:Cercopithecoidea). Primates 19, 1–34.

Groves, C. P. (2000). The phylogeny of the Cerco-pithecoidea. In (P. F. Whitehead & C. J. Jolly, Eds)Old World Monkeys, pp. 77–98. New York:Cambridge University Press.

Harris, E. (2000). Molecular systematics of the OldWorld monkey tribe Papionini: analysis of the totalavailable genetic sequences. J. hum. Evol. 38, 235–256.

Harris, E. & Disotell, T. R. (1998). Nuclear gene treesand the phylogenetic relationships of the mangabeys(Primates: Papionini). Mol. Biol. Evol. 15, 892–900.

Hewett-Emmett, D., Cook, C. N. & Barnicot, N. A.(1976). Old World monkey hemoglobins: decipher-ing phylogeny from complex patterns of molecularevolution. In (M. Goodman & R. E. Tashian, Eds)Molecular Anthropology, pp. 257–275. New York:Plenum Press.

Hersh, A. H. (1934). Evolutionary relative growth inthe titanotheres. Am. Nat. 68, 537–561.

Hill, W. C. O. (1974). Primates: Comparative Anatomyand Taxonomy. Vol. 7: Catarrhini, Cercopithecinae,Cercocebus, Macaca, Cynopithecus. Edinburgh:Edinburgh University Press.

Jones, C. & Sabater Pi, J. (1968). Comparative ecologyof Cercocebus albigena (Gray) and Cercocebus torquatus(Kerr) in Rio Muni, West Africa. Folia primatol. 9,99–113.

Kingdon, J. (1997). The Kingdon Field Guide to AfricanMammals. San Diego: Academic Press.

Krogman, W. M. (1931a). Studies in growth changesin the skull and face of anthropoids. III. Growthchanges in the skull and face of the gorilla. Am. J.Anat. 47, 89–115.

Krogman, W. M. (1931b). Studies in growth changesin the skull and face of anthropoids. IV. Growthchanges in the skull and face of the chimpanzee. Am.J. Anat. 47, 325–342.

Krogman, W. M. (1931c). Studies in growth changes inthe skull and face of anthropoids. V. Growth changesin the skull and face of the orangutan. Am. J. Anat.47, 343–365.

Kuhn, H.-J. (1967). Zur systematik der Cercopitheci-dae. In (D. Starck, R. Schneider & H.-J. Kuhn, Eds)Neue Ergebnisse der Primatologie, pp. 25–46. Stuttgart:G. Fischer.

Leigh, S. R. & Cheverud, J. M. (1991). Sexual dimor-phism in the baboon facial skeleton. Am. J. phys.Anthrop. 84, 193–208.

Leutenegger, W. & Masterson, T. J. (1989). Theontogeny of sexual dimorphism in the cranium ofBornean orang-utans (Pongo pygmaeus pygmaeus): II.Allometry and heterochrony. Z. Morph. Anthrop. 78,15–24.

Lockwood, C. A. & Fleagle, J. G. (1999). The recog-nition and evaluation of homoplasy in primate andhuman evolution. Yearb. phys. Anthropol. 42, 189–232.

Maier, W. (2000). Ontogeny of the nasal capsule incercopithecoids: a contribution to the comparativeand evolutionary morphology of catarrhines. In(P. F. Whitehead & C. J. Jolly, Eds) Old World

Monkeys, pp. 237–268. New York: CambridgeUniversity Press.

Marcus, L. F. (1993). Some aspects of multivariatestatistics for morphometrics. In (L. F. Marcus, E.Bello & A. Garcıa-Valdecasas, Eds) Contributions toMorphometrics, pp. 99–130. Madrid: MonografiasMuseo Nacional de Ciencias Naturales.

McGraw, W. S. & Fleagle, J. G. (2000). Biogeographyand evolution of the Cercocebus–Mandrillus clade.Am. J. phys. Anthrop. Suppl. 30, 225.

McNamara, J. A., Riolo, M. L. & Enlow, D. H. (1976).Growth of the maxillary complex in the rhesusmonkey (Macaca mulatta). Am. J. phys. Anthrop. 44,15–26.

Morales, J. C. & Melnick, D. J. (1998). Phylogeneticrelationships of the macaques (Cercopithecidae:Macaca), as revealed by high resolution restrictionsite mapping of mitochondrial ribosomal genes.J. hum. Evol. 34, 1–23.

Mosimann, J. E. & Malley, J. D. (1979). Size andshape variables. In (L. Orioci, C. R. Rao & W. M.Stiteler, Eds) Multivariate Methods in EcologicalWork, pp. 175–189. Fairland, MD: InternationalCo-operative Publishing House.

Nakatsukasa, M. (1994). Morphology of the humerusand femur in African mangabeys and guenons: func-tional adaptation and implications for the evolutionof positional behavior. Afr. Study MonographsSuppl. 21, 1–61.

Nakatsukasa, M. (1996). Locomotor differentiationand different skeletal morphologies in mangabeys(Lophocebus and Cercocebus). Folia primatol. 66,15–24.

Neff, N. A. & Marcus, L. F. (1980). A Survey ofMultivariate Methods for Systematics. NumericalMethods in Systematic Mammalogy Workshop.American Society of Mammalogists.

O’Higgins, P. & Dryden, I. L. (1993). Sexual dimor-phism in hominoids: further studies of craniofacialshape differences in Pan, Gorilla, and Pongo. J. hum.Evol. 24, 183–205.

O’Higgins, P. & Jones, N. (1999). Morphologika.London: University College London.

O’Higgins, P., Moore, W. J., Johnson, D. R. &McAndrew, T. J. (1990). Patterns of cranial sexualdimorphism in certain groups of extant hominoids.J. Zool. 222, 399–420.

Page, S. L., Chiu, C. & Goodman, M. (1999).Molecular phylogeny of Old World monkeys (Cerco-pithecidae) as inferred from �-globin DNAsequences. Mol. Phyl. Evol. 13, 348–359.

Profant, L. (1995). Historical allometric inputs tointerspecific patterns of craniofacial diversity in thecercopithecine tribe Papionini. Am. J. phys. Anthrop.Suppl. 20, 175.

Profant, L. & Shea, B. (1994). Allometric basis ofmorphological diversity in the Cercopithecini vs.Papionini tribes of Cercopithecine monkeys. Am. J.phys. Anthrop. Suppl. 18, 162–163.

Radinsky, L. (1981). Evolution of skull shape incarnivores. 1. Representative modern carnivores.Biol. J. Linn. Soc. 15, 369–388.

578 .

Radinsky, L. (1984). Ontogeny and phylogeny in horseskull evolution. Evolution 38, 1–15.

Ravosa, M. J. (1991). The ontogeny of cranial sexualdimorphism in two Old World monkeys: Macacafascicularis (Cercopithecinae) and Nasalis larvatus(Colobinae). Int. J. Primatol. 12, 403–426.

Ravosa, M. J. & Profant, L. P. (2000). Evolutionarymorphology of the skull in Old World monkeys. In(P. F. Whitehead & C. J. Jolly, Eds) Old WorldMonkeys, pp. 237–268. New York: CambridgeUniversity Press.

Reeve, E. C. R. (1939). Relative growth in the snout ofanteaters. A study in the application of quantitativemethods to systematics. Proc. Zool. Soc. Lond., Ser. A110, 47–80.

Reeve, E. C. R. & Murray, P. D. F. (1942). Evolutionin the horse’s skull. Nature 150, 402–403.

Richtsmeier, J. T., Cheverud, J. M., Danahey, S. E.,Corner, B. D. & Lele, S. (1993). Sexual dimorphismof ontogeny in the crab-eating macaque (Macacafascicularis). J. hum. Evol. 25, 1–30.

Rohlf, F. J. (1999a). Shape statistics: Procrustessuperimpositions and tangent spaces. J. Classif. 16,197–223.

Rohlf, F. J. (1999b). tpsSmall v. 1.17. Department ofEcology and Evolution, State University of NewYork, Stony Brook, New York.

Rohlf, F. J. & Marcus, L. (1993). A revolution inmorphometrics. Trends Evol. Ecol. 8, 129–132.

Schwarz, E. (1928). The species of the genusCercocebus. Ann. Mag. Nat. Hist. 5, 664–670.

Shah, N. F. & Leigh, S. R. (1995). Cranial ontogeny inthree papionin genera. Am. J. phys. Anthrop. Suppl.20, 194.

Shea, B. T. (1983). Allometry and heterochrony in theAfrican apes. Am. J. phys. Anthrop. 62, 275–289.

Shea, B. T. (1985). Ontogenetic allometry and scaling:a discussion based on the growth and form of theskull in African apes. In (W. L. Jungers, Ed.) Size andScaling in Primate Biology, pp. 175–205. New York:Plenum Press.

Shea, B. T. (1992). Ontogenetic scaling skeletal pro-portions in the talapoin monkey. J. hum. Evol. 23,283–307.

Slice, D. E. (1998). Morpheus et al.: Software for Mor-phometric Research. Revision 01-30-98. Departmentof Ecology and Evolution, State University of NewYork, Stony Brook, New York.

Slice, D. E. (1999). GRF-ND: Generalized RotationalFitting of n-Dimensional Landmark Data. Revision12-15-99. Department of Ecology and Evolution,State University of New York, Stony Brook,New York.

Slice, D. E., Bookstein, F. L., Marcus, L. F. & Rohlf,F. J. (1996). Appendix I—a glossary for geometricmorphometrics. In (L. F. Marcus, M. Corti, A. Loy,G. J. P. Naylor & D. E. Slice, Eds) Advancesin Morphometrics, pp. 531–551. New York: PlenumPress.

Sokal, R. S. & Rohlf, F. J. (1981). Biometry. 2nd edn.New York: W. H. Freeman.

Strasser, E. & Delson, E. (1987). Cladistic analysisof cercopithecid relationships. J. hum. Evol. 16,81–99.

Swindler, D. R. & Sirianni, J. E. (1973). Palatal growthrates in Macaca nemestrina and Papio cynocephalus.Am. J. phys. Anthrop. 38, 83–92.

Swindler, D. R. & Wood, C. D. (1982). An Atlas ofPrimate Gross Anatomy: Baboon, Chimpanzee, andMan. Malabar, FL: Krieger Publishing.

Swindler, D. R., Sirianni, J. E. & Tarrant, L. H.(1973). A longitudinal study of cephalofacial growthin Papio cynocephalus and Macaca nemestrina fromthree months to three years. In Symp. IVth Int. Congr.Primat. vol. 3: Craniofacial Biology of the Primates,pp. 227–240. Basel: Karger.

Szalay, F. S. & Delson, E. (1979). Evolutionary Historyof the Primates. New York: Academic Press.

Thorington, R. W. & Groves, C. P. (1970). An anno-tated classification of the Cercopithecoidea. In (J. R.Napier & P. H. Napier, Eds) Old World Monkeys—Evolution, Systematics, and Behavior, pp. 629–648.New York: Academic Press.

Tsutakawa, R. K. & Hewett, J. E. (1977). Quick testfor comparing two populations with bivariate data.Biometrics 33, 215–219.

Van Der Kuyl, A. C., Kuiken, C. L., Dekker, J. T.,Perizonius, W. R. K. & Goudsmit, J. (1995). Phy-logeny of African monkeys based upon mito-chondrial 12S rRNA sequences. J. molec. Evol. 40,173–180.

Vogel, C. (1968). The phylogenetical evaluation ofsome characters and some morphological trends inthe evolution of the skull in catarrhine primates. In(A. B. Chiarelli, Ed.) Taxonomy and Phylogeny of OldWorld Primates with References to the Origin of Man,pp. 21–55. Torino: Rosenberg & Sellier.

Weidenreich, F. (1941). The brain and its role in thephylogenetic transformation of the human skull.Trans. Am. Phil. Soc. N.S. 31, 321–442.

Zuckerman, S. (1926). Growth-changes in the skull ofthe baboon, Papio porcarius. Proc. Zool. Soc. Lond. 55,843–873.