Biological journal ofthe Linncan So&& (1981), 16: 337-355. With 8 figures Evolution of skull shape in carnivores 2. Additional modern carnivores LEONARD B. RADINSKY Anatomy Department, The University of Chicago, Chicago, Illinois 60637, U.S.A. Accepted f o r publication Septmbcr 1981 Fifteen functionally significant aspects of skull morphology were measured on skulls of 36 additional species of carnivores to complete a survey of skull shape'in modern fissiped (land) carnivores that includes most of the living genera. The measurements were transformed to dimensionless variables based on the residuals from allometric equations, and were analysed singly and in a 10 variable principal components analysis. An initial study of 62 species of viverrids, canids, mustelids and felids had shown those families to be distinguished from each other by the functionally significant measurements. However, among the additional 36 species, some procyonids, ursids and mustelids display a range of diversity of skull morphology that overlaps that of the other families and diminishes the potential value of the measurements as taxonomic characters. Intraspecific variation is presented for 12 species, and is low enough to allow use ofsome features as species level diagnostic characters. The lack of correlation between diet and functionally significant aspects of skull morphology among omnivorous carnivores, and the absence of certain skull shapes among carnivores are discussed. KEY WORDS:-Carnivora - skulls - functional craniology - morphometrics. CONTENTS Introduction . . . . . . . . Materials and methods. . . . . . Results . . . . . . . . . Discussion . . . . . . . . . Significance for carnivore systematics . Functional significance and biological role Carnivore skull shape morphospace. . Acknowledgements. . . . . . . References. . . . . . . . . Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337 338 340 345 345 349 350 351 351 352 INTRODUCTION This paper is part of a study of changes in skull morphology that occurred in the origin and evolution of the modern carnivore families. The Order Carnivora first appeared around 60 million years ago (My), and its major evolutionary radiation, which gave rise to most of the modern families, took place around 35-40 My. It is hoped that a functionally oriented analysis of skull morphology will help provide insights into factors involved in that radiation-why it took place when it did 337 00224066/81/080337+19 s02.00/0 0 1981 The Linnean Society of London 338 L. B. RADINSKY (20-25 My after the origin of the Carnivora), and how early members of the modern families were differentiated from each other with respect to ecological niche. The purpose of this paper is to complete a survey of skull shape among modern land (fissiped) carnivores, to provide a background for analysis of the fossil carnivores. A previous paper (Radinsky, 1981), analysed skull shape in 62 species representing the four largest families of carnivores : viverrids, canids, mustelids and felids. That study showed that functionally significant aspects of skull morphology distinguished those four families from each other, and suggested that the anatomical differences are related to differences in killing behaviour. The present paper examines skull morphology in the smaller remaining families, procyonids, ursids and hyaenids, and in the genera of viverrids and mustelids that were not included in the first paper, to see to what degree modern procyonids, ursids and hyaenids are also distinguished by functionally significant aspects of cranial morphology, and if the distinctive features noted for viverrids and mustelids in the first paper also characterize the remaining genera in those families. Finally, this paper provides data on intraspecific variation in the cranial features under investigation, to allow estimation of the significance of differences observed among taxa at the species level. MATERIALS AND METHODS Sixty-two species of modern carnivores were covered in the first study, and this paper examines 36 additional species (listed in Appendix l ) , for a total of 98 species, including representatives of most of the living genera of fissiped carnivores. The original sample of viverrids was comprised mainly of viverrines and herpestines, and the six additional species measured for this paper are mainly paradoxurines and hemigalines. The original mustelid sample consisted primarily of mustelines, and the 13 additional species covered in this paper are badgers, skunks and otters. The seven species of procyonids, seven species of ursids, and three species of hyaenids represent all of the living genera in those families. Measurements of the 15 functionally significant aspects of cranial morphology used in the previous paper (Radinsky, 1981), were taken on one randomly selected skull of each of the 36 species. (See Appendix 2 for abbreviations, definitions and justifications of these measurements.) Bulla volume could not be estimated from external skull measurements of ursids and most of the mustelids, and so was omitted for those families. To allow comparisons of skull shape among species of different sizes while accounting for possible effects of allometry, all measurements were converted to logarithms (base lo), and were transformed into dimensionless variables based on the residuals from reduced major axis equations of each variable u. skull length (SL) or jaw length (3L).For example, if log Y = log a + b log SL, (which is equivalent to Y = a SLb) the transform of Yi=Yi/u $Lib. The dimensionless transforms of each measurement are indicated by a ' following the abbreviation. Thus, ZAW' is the transform of ZAW. For a fuller discussion of the procedures used in transforming the measurements to dimensionless variables, and for a list of the equations used for calculating the residuals, see Radinsky (1981). The data from the additional species of viverrids and mustelids were added to those of the original viverrid and mustelid samples, and for each transformed EVOLUTION OF SKULL SHAPE IN CARNIVORES 339 variable, t-tests were used to see if the family means of the expanded viverrid and mustelid samples still differed from each other and from the canid and felid family means as they had in the original study (Radinsky, 1981;Table 1). The procyonid, ursid and hyaenid samples each include relatively few species, and, for the first two families, a relatively large amount of morphological diversity. Therefore, the transformed variables of those families were not analysed by family means, but rather were examined on the level of individual species. In addition to univariate analysis, the transformed variables were also examined by a principal components analysis followed by varimax rotation. In the previous paper, that multivariate technique was shown to separate the 62 species of viverrids, canids, mustelids and felids into their respective families along the first two rotated axes of a 15 variable analysis (Radinsky, 1981: Fig. 3). Equally good separation of those families can be obtained along the first two rotated axes using only the ten highest loading variables of that analysis (i.e. the variables most important in determining the axes), and the ten variable analysis (Fig. 1 and Table 1) is used as a background for examining the 36 additional species in this paper. Because the present paper includes consideration of differences at the species r F E L. F A C 2 (27%) Figure 1 . Factor scores of the original four family, 62 species sample on the first two rotated a x a (FAC 1 and FAC 2) of the 10 variable principal components analysis. Polygons enclose all species of a given family: solid circles represent mustelids (MUS) and canids (CAN); open circles represent viverrids (VIV) and felids (FEL). See Table 1 for loadings of variables on those axes. L. B. RADINSKY 3 40 Table 1. Loadings of transformed variables on first two rotated axes ZAW OCPW‘ TFL‘ TRL‘ JL’ BRV’ ORBA MAT’ MFL‘ JXA’ nh‘ Variance FAC 1 FAC 2 0.323 0.749 0.874 -0.915 -0.865 0.170 -0.201 0.856 0.827 0.898 0.902 0.248 0.165 -0.171 0.266 0.852 0.916 -0.181 0.270 0.172 53.07 27.03 level in addition to analysis of differences at the family level, intraspecific variability of the transformed variables was examined. Intraspecific variation was estimated from measurements taken on six randomly selected skulls of each of three randomly selected species of each of the original four family samples (viverrids, canids, mustelids and felids), for a total of 12 species (listed in Appendix 4), for which intraspecific variation was measured. RESULTS The dimensionless variable transforms of the 36 additional species of modern carnivores are summarized in Table 2, and their factor scores on the first two rotated axes of the multivariate analysis are presented in Figs 2, 3 & 5. The addition of the six new species of viverrids to the original viverrid sample results in some small changes in means (by no more than two per cent), and in observed ranges for some of the variables, and the factor scores of the new species are within or close to the range of the original viverrid sample (see Fig. 2). The small differences in mean values for the transformed variables of the enlarged viverrid sample do not change the significance of differences from the canid, mustelid and felid means noted in the original analysis (see Appendix 3). The addition of the 13 new species of mustelids to the original mustelid sample changes the means and observed ranges for most of the transformed variables, but the differences that distinguished the original mustelid sample from that of the viverrids, canids and felids still hold for the enlarged mustelid sample (see Appendix 3), with the following minor changes: mean brain size (BRV’/3’)is now significantly higher than that of the viverrids: mean condyle to carnassial distance (COM1‘) is now significantly higher than that of the canids and felids as well as that of the viverrids; and mean moment arm of temporalis (MAT’) is still significantly higher than that of viverrids and canids, but no longer significantly higher than that of the felids. The biggest differences in skull shape among the new mustelids are seen in the otters and some of the badgers. The otters Enhydru (Enh), Aottyx (Aon) and Pteroneuru (Ptr) have unusualy high BRVlfl’, and Enhydru and Aonyx have relatively wide skulls (ZAW’) (see Table 2). Those features result in high factor scores on the second rotated axis (FAC 2) for those genera, and an extension of the mustelid 34 I EVOLUTION OF SKULL SHAPE IN CARNIVORES range in that area (see Fig. 3). Among the badgers, Arctonyx (Atx) has a relatively short temporal fossa length (TFL’), low MAT’ and low jaw cross section area (JXA1/4’); Myduus (Myd) has relatively low MAT’ and JXA114’; and Melogale (Mgl) has relatively low MAT’ and long jaw length UL’). As a result of those features, Arctonyx, Myduus and Melogale have relatively low factor scores on the first axis (FAC l ) , and are located below the original mustelid range on that axis (see Fig. 3). (Extremes of skull shape among mustelids are illustrated in Fig. 4.) Among the three new families, the hyaenids fall within the viverrid range for most transformed variables. A notable exception is COMl’, for which all three genera of hyaenids fall below the observed range of not only the viverrids but also all other carnivores except Ailuropodu (the giant panda). Relatively high ZAW’ in Hjuenu (Hyn) and relatively high BRVlP, and orbit area (ORBAlP’) in Crocutu place those genera outside of the viverrid range of factor scores along the second axis (FAC 2) in the multivariate analysis (see Fig. 2). The procyonids and ursids display a greater diversity in skull morphology than do most of the other families (see Fig. 5). Among the procyonids, the lesser panda, Ailurus (Ail), and the kinkajou, Potos (Pot), have relatively high values for most -2.0 - 2.0 L I -1.0 1 I 0 I I I 1.0 I 2.0 1 I 3.0 F A C 2 (27%) Figure 2. Factor scores of six additional species of vivemds ( 0 )and three species of hyaenids ( 0 )on the first two rotated axes of the 62 species principal components analysis. Polygons represent the ranges of factor scores of the 62 species sample shown in Fig. 1. Key to abbreviations of species names is in Appendix 1 . 1.71 1.89 1.54 1.39 1.59 1.60 1.67 1.45 1.58 215 1.59 1.63 1.89 Additional mustelids At% 1.24 Tax 1.29 Me1 1.20 1.13 Mgl Con 1.31 1.31 MPh 1.35 SPl 1.02 MYd Sul 1.11 Enh 157 pa 1.22 Lut 1.28 Aon 1.43 1.29 151 150 1.54 1.02 1.17 1.15 0.98 1.21 1.13 1.14 0.69 0.84 1.06 1.13 1.05 1.34 1.16 0.92 OCPW' 'IlW" 1.18 1.10 0.97 1.11 1.08 1.23 ZAW' Additional vivenids CYn 1.04 p"g 1.11 Hem 0.99 Par 1.18 1.1 1 451 Arc 1.15 Specit3 1.23 1.32 1.31 1.00 1.07 1.06 1.12 1.18 1.06 1.04 1.11 1.32 0.87 1.00 1.07 0.95 1.05 1.02 1.02 TFL' 1.00 0.94 0.91 0.97 0.92 0.91 0.98 0.86 0.90 0.84 0.81 0.87 0.89 1.10 0.89 1.08 0.98 0.94 0.89 ' I " 0.94 0.98 0.96 1.02 0.99 0.98 0.99 0.97 0.97 0.92 0.87 0.91 0.98 1.06 1.06 1.02 1.01 1.10 1.03 JL' 1.13 1.25 1.20 1.01 1.12 0.96 1.04 1.13 1.07 1.59 1.42 1.18 1.41 1.01 1.03 0.99 1.02 1.07 0.89 0.95 0.81 1.00 0.99 O M 1.13 0.H 0.67 1.02 1.04 0.88 0.93 0.84 0.80 0.78 0.83 0.84 0.74 0.84 1.00 0.92 0.87 0.87 0.99 0.97 1.09 131 1.26 1.05 1.a 1.23 1.25 1.19 1.30 1.40 1.26 1.22 1.10 0.99 0.88 0.85 0.92 0.82 0.89 1.09 1.08 1.16 1.08 1.22 123 1.13 1.14 1.13 1.11 1.13 1.13 1.18 1.14 095 1.09 1.13 1.12 a92 1.15 1.45 1.54 1.50 1.60. os3 0.85 1.42 1.26 1.12 1.24 1.41 1.27 1.22 0.90 1.27 1.26 1.13 0.88 BRV~P'BULV'P~ORBA~P~ o c p H COMI' MAT' Table 2. Transformed variables of 36 additional carnivorans* 1.05 0.99 1.14 1.01 1.06 1.27 1.35 1.24 0.89 0.96 la 1.08 0.19 1.18 om 0.93 1.10 0.97 1.20 MAM' 0.91 1.10 1.02 1.11 0.96 1.10 1.07 0.93 1.M 0.99 1.02 1.06 1.01 0.85 1.09 0.90 1.03 1.22 1.15 MFL' 1.16 1.53 1.21 1.31 1.33 091 1.31 1.13 1.06 1.15 I .29 1.18 o.!w 1.06 0.98 0.87 1.12 1.OO 1.07 JXAV. 13.62 11.95 10.95 8.54 8.23 7.77 5.19 8.67 7.51 13.29 15.14 10.92 8.59 10.94 11.64 10.60 10.41 9.41 13.06 SL 35 P r W P N 1.73 1.13 1.37 1.37 1.05 1.22 1.58 Ursids He1 MIS Sel TlT Thl UrS APa 1.32 1.22 1.38 1.15 1.26 1.19 1.34 1.61 1.47 1.24 1.43 1.82 205 1.28 1.17 1.03 1.78 0.74 1.18 1.10 a75 0.85 1.54 1.32 1.66 1.08 0.93 1.36 1.32 1.29 1.12 1.02 0.85 1.19 0.93 1.06 1.12 0.98 0.98 1.18 1.04 1.29 0.89 0.88 1.26 0.97 1.13 1.07 1.oo 0.89 0.82 0.78 0.93 0.93 0.84 0.92 1.14 1.07 1.11 1.18 1.06 0.89 1.08 1.03 0.97 0.99 1.06 I31 0.96 1.09 1.06 1.19 1.04 1.28 1.58 1.18 131 135 1.24 1.33 1.09 1.14 1.38 1.07 1.27 1.30 0.96 0.94 0.97 1.01 0.87 1.04 1.08 1.10 1.03 1.07 1.06 1.08 1.13 1.26 1.33 0.83 oa2 0.89 1.20 0.73 oa7 0.88 1.03 1.15 1.10 0.99 0.99 0.94 0.93 1.02 0.99 0.94 0.75 1.13 1.09 1.12 0.85 1.13 1.01 0.97 1.16 0.92 1.07 0.95 0.89 1.15 0.95 0.85 1.07 1.08 1.01 1.06 1.12 1.22 0.95 1.07 030 0.84 0.85 0.87 1.01 0.98 1.04 1.04 1.02 1.02 1.10 1.16 1.01 1.02 1.16 1.08 1.14 oai 1.08 1.23 1.46 1.09 1.17 1.31 1.23 1.27 1.53 1.29 1.25 1.39 0.69 oai 1.29 1.64 1.01 1.16 0.99 1.04 1.13 1.05 0.89 1.58 1.33 0.86 0.77 2.65 0.99 1.01 0.87 1.03 1.87 0.95 0.84 0.89 1.15 0.87 0.91 0.90 0.95 0.92 0.94 0.89 1.11 0.74 0.82 1.15 0.98 1.02 1.32 1.28 1.12 1.48 1.02 1.16 1.21 1.08 1.00 1.39 1.31 1.51 1.13 0.86 1.41 1.01 1.16 23.00 24.40 12.62 23.20 29.71 25.65 22.89 34.75 25.29 25.68 12.04 10.27 11.00 10.75 8.02 8.61 7.34 Boldface numbers denote relatively high or low values: for viverrids and hyaenids, values that fall outside of the observed range of the original viverrid sample; for mustelids, values that Bll outside of the observed range of the original mustelid sample; and for pmyonids and ursids, values that fall above or below the highest or lowest (respectively) family means of the vivemd, canid, mustelid and felid samples (see Appendix 3). *See Appendix 1 for full species names, and Appendix 2 for explanations of variables. Prt Hyaenids Hyn CrO 1.12 1.17 1.06 1.51 1.27 1.40 1.13 1.61 1.28 1.36 1.75 1.48 1.41 1.64 Procyonids Pro Ail Nsa Nsl Pot Bsc Bsn w e w f 2 z c ( m 9 w 8 $ 2 L. B. RADINSKY 344 F E L. -2.0 -2.0 , I -1.0 1 I I 0 I I 1.0 1 2.0 I 1 3.0 F A C 2 (27%) Figure 3. Factor scores of 13 additional species of mustelids on the h t two rotated axes of the 62 species principal components analysis. Polygons represent the ranges of factor scores of the 62 species sample shown in Fig. 1. Key to abbreviations of species names is in Appendix 1. transformed variables, excluding JL’ and tooth row length (TRL’), which results in their placement in the upper right quadrant of the factor score plots in Fig. 5. Potos is also unusual in having an extremely expanded angular process, which is reflected in the highest moment arm of masseter (MAM’) of any carnivore (see Fig. 6). The coatimundi, Nasuu (Nas), has relatively high JL’ and TRL’ values, and, together with the little coati, Naruellu (Nsl), has relatively low values for several other variables, which places those two procyonid genera in the lower left quadrant of the factor score plot (Fig. 5). Extremes of procyonid skull shape are illustrated in Fig. 6. Among the ursids, the Malayan sun bear, Helurctos (Hel), and the giant panda, Ailuropodu (Apa), stand out with relatively high values for many transformed variables, which places them in the upper right quadrant of the factor score plots of Fig. 5. Helurctos has the highest ZAW’ and second-highest occipital plate width (OCPW’) (after Enhydra) of the 98 species measured in this study. Extremes of ursid skull shape are illustrated in Fig. 7. Intraspecific variation in transformed variables is summarized in Appendix 4 & Fig. 8. As expected, in general intraspecific variation is less than interspecific variation with families, with the only exceptions in viverrid TRL’ and felid EVOLUTION OF SKULL SHAPE IN CARNIVORES 345 Figure 4. Extremes ofdiversity in skull shape in modern mustelids. All drawn to same scale. See text for discussion. masseteric fossa length (MFL’) and JXAU4’. The intraspecific standard deviations presented in Appendix 4 allow assessment of the significance of differences in transformed variables seen among species listed in Table 2, and the visual display of intraspecific variation in factor scores presented in Fig. 8 allows assessment of the significance of differences in placement of species on the factor score plots in Figs 2, 3 & 5. DISCUSSION Signijcance for carnivore systematics The previous study (Radinsky, 1981), showed that 62 species of viverrids, canids, mustelids and felids could be grouped into their respective families on the basis of functionally significant aspects of skull morphology. Those families are currently defined primarily on the basis of middle ear region and basicranial L. B. RADINSKY 946 2.1 MUS. - I.( c* 8 tc) Ln v - I VIV. 0 a LL -I. - 2.1 - .o 1 I -1.0 I I 0 I I I 1.0 I 2 .o I I 3.0 F A C 2 (27%) Figure 5. Factor scores of procyonids ( 0 )and unids ( 0 )on the first two rotated axes of the 62 species principal components analysis. Polygons represent the ranges of factor scores of the 62 species sample shown in Fig. 1. Key to abbreviations of species names is in Appendix I . anatomy (Hunt, 1974; Tedford, 1976), and it was hoped that the functionally significant aspects of skull morphology might provide additional characters for carnivore family diagnoses. The present study shows that additional species of viverrids fall largely within the range of morphology seen in the original viverrid sample, but that among mustelids, some species of badgers and otters fall far outside of the range of morphology of the original (mainly musteline) mustelid sample. The previous study had shown mustelids to be distinguished from viverrids by features of skull morphology reflecting relatively short, robust jaws and relatively large, powerful jaw muscles (variables important in determining FAC 1-see Fig. 1, Table 1). However, three of the badgers-Melogale, Myduur and Arctonyx-fall well within the viverrid range in those variables. Thus, although mustelid and viverrid family means are still significantly different from each other for many of the variables (see Appendix 3), there is now so much overlap that these features cannot be considered useful systematic characters for family level diagnoses. Of the new families examined in this study, hyaenids are distinguished only by their very low COM1’ values, which are lower in Hyaena and Crocutu than in any other carnivores in the study. The procyonids include relatively few taxa but are EVOLUTION OF SKULL SHAPE IN CARNIVORES 341 Figure 6. Extremes ofdiversity in skull shape in modern procyonids. All drawn to same scale. See text for discussion. extremely diverse in skull morphology. Most species have relatively small auditory but in other features procyonids overlap broadly with several bullae (BULV1’3’), other families. At one extreme, the coatimundis (Nusuu and Nusueflu) resemble canids in their long TRL and JL, and at the other extreme, the lesser panda (Ailurus) and the kinkajou (Potos) are similar to felids in having relatively hi h ZAW’ and BRV’l3’values, and to mustelids in having high MAT and JXA !/4, values. Ursids like procyonids, include relatively few taxa but a large degree of diversity in skull morphology. At one extreme, the polar bear (Thulurctos) and sloth bear (Melursus) are similar to viverrids in many aspects of skull shape, and at the other extreme, the giant panda (Ailuropodu) and Malayan sun bear (Helurctos) overlap the high end of mustelid and felid ranges in several transformed variables. The diversity in skull morphology displayed by the procyonids, ursids and some L.B. RADINSKY 348 6 cm 6 cm Figure 7. Extremes of diversity in skull shape in modem unids, drawn to diflercnt scales. See text for discussion. of the more unusual mustelids, is such that the original pattern of separation into family groups seen in the 62 species analysis (Fig. 1) is obscured with the addition of the new species (Figs 3, 5 ; Table 2, Appendix 3). The four large families (viverrids, canids, mustelids and felids) still differ in mean values of several of the transformed variables, but there is so much overlap across families that the characters cannot be considered useful for systematics at the family level. Further, the procyonids and ursids are so diverse (for the small number of taxa involved), that it is not even meaningful to characterize them by family means for the transformed variables. However, while these characters are not distinctive at the family level, the amount of intraspecific variation in the transformed variables is low enough that in many cases some of them can be used to distinguish species within families (cf. Table 2, Appendix 4). 349 EVOLUTION O F SKULL SHAPE IN CARNIVORES 9 8 \ 7 I \ I -2.0 -2.0 1 I / I -1.0 ? 4 I 0 I I I 1.0 I 2.0 I I 3.0 F A C 2 (27%) Figure 8. Intraspecific variation in factor scores of 12 species of modem carnivores on the first two rotated axes of the 62 species principal components analysis. Polygons represent the ranges of factor scores of the 62 species sample shown in Fig. 1. The circles represent species means, and the vertical and horizontal lines representftwo standard deviations (95% limits) on the FAC 1 and FAC 2 axes respectively. See Appendix 4 for list of species (indicated by numbers 1-12), and for the data upon which this graph is based. Functional signijcance and biological role The previous study (Radinsky, 1981) proposed the functional hypothesis that high loading FAC 1 variables reflect bite strength, and the biological role hypothesis that differences in bite strength are correlated with differences in killing behaviour. The present study covers mainly omnivorous carnivores, and if the functional hypothesis that FAC 1 variables reflect bite strength is correct, there is a surprising lack of correlation between inferred bite strength and diet. Among the ursids, Ailuropoda has been considered to represent an extreme of cranial modification for powerful chewing (Sicher, 1944; Davis, 1964), and its skull morphology was inferred to be an adaptation for chewing hard bamboo (which comprises an important part of its diet). However, Helurctos is even more derived (specialized) than Ailuropoda in cranial features related to strong jaws and powerful jaw muscles, and yet there are no unusually hard food items reported as important components of its omnivorous diet (Walker, 1964). Among the procyonids, the apparently powerful jaw apparatus of Aibrus may be causally 350 L.B. RADINSKY related to the reported presence of acorns as an important food item in its diet (Walker, 1964), but there are no obvious dietary correlates that account for the distribution of other procyonid species along the FAC 1 axis, from Nusua at one extreme to Potos at the other. Similarly, among the mustelids there are no obvious differences in reported diets to account for the differences in skull morphology among the omnivorous skunks and badgers, from Arctotip at one extreme to 7 m i d e a and Mephitis at the other. Finally, among the hyaenids, the skull morphology of Proteles, the aardwolf, suggests a weaker bite than in the hyaenas, Hyaena and Crocuta (see TFL’ and MAT’ in Table 2), but would not lead one to suspect that Proteles subsists largely on a diet of termites. The lack of correlation between diet and aspects of skull morphology related to bite strength may be accounted for by three kinds of explanations. First, the functional hypothesis that FAC 1 variables reflect bite strength may be too narrow : perhaps additional aspects of functional significance are involved also. Second, the biological role hypothesis that the observed dietary differences among modern carnivores will be causally related to differences in bite strength may be incorrect. Bite strength may be correlated with other biological roles, such as intraspecific combat or defence. Third, other factors besides strictly adaptational ones may be involved in determining differences in skull shape among modern carnivores (see Gould & Lewontin, 1979, for a discussion of such factors). Carnivore skull shape morphospuce The first two rotated axes of the multivariate analysis can be considered as framing a view into a 10-dimensional morphospace that describes skull shape in modern carnivores based on the variabIes listed in Table 1. The original four family analysis (Fig. 1) shows mustelids, viverrids and canids spread out along the first axis (FAC 1), determined primarily by variables reflecting relatively short, robust jaws and relatively large powerful jaw muscles at the mustelid end of the axis. Felids are segregated out on the second axis (FAC 2), determined primarily by relatively great skull width (ZAW) and large sensory capsules (ORBA and BRV) . The additional carnivore species covered by the present study fill in much of the empty morphospace of the first analysis. Hyaenids and various species of procyonids and ursids are located in the space between the original four family ranges (see Figs 2, 5), overlappifig the viverrid range on FAC 1 and the canid range along FAC 2. Otters and a few procyonid and ursid specks fall in the upper right quadrant, overlapping the mustelid range on FAC 1 and thefelid range on FAC 2 (see Figs 3, 5). The only region of the FAC 1-FAG 2 morphospace that is not filled by modern carnivores is the lower right quadrant, which would be occupied by species with relatively long, slender jaws, long tooth rows, and short temporal fossae, like canids, and relatively wide skulls with relativeiy large sensory capsules, like felids. The absence of modern species with those skull shape characteristics may in part be an artifact of how the original measdrements are transformed into dimensionless variables. Long jaws and long tooth rbw lengths add to skull length, and the variables that determine FAC 2-ZAW, ORBA and BRV-are transformed by regression against skull length. In other wofds, species with lbnger JL and TRL EVOLUTION OF SKULL SHAPE IN CARNIVORES 35 1 will have longer SL, and, for the same sized ZAW and sensory capsules, those species will have lower ZAW, ORBA’” and BRV’’3’than species with shorter SL. This problem of potential bias introduced by the process of transforming the measurements into dimensionless variables is discussed more hlly in the previous paper (Radinsky, 1981), particularly with respect to sensory capsules. On the other hand, the absence of species in the lower right quadrant of the FAC 1-FAC 2 morphospace may also reflect, at least in part, biomechanical constraints on carnivore skull shape. Relative zygomatic arch width (ZAW’) is an important determiner of FAC 2, and to the degree that ZAW reflects size ofjaw musculature, one would expect high ZAW’ to be associated with high values in other variables related to bite strength, and the latter would place species high on FAC 1. Thus one would not expect to find carnivore species with both high ZAW’ and low TFL’, MAT’ and JXAW, and this may in part account for the absence of species in the lower right quandrant of the FAC 1 v. FAC 2 factor score plot. The pattern of distribution of modern carnivore species in the FAC 1-FAC 2 morphospace raises additional questions with respect to the evolution of diversity of skull morphology among carnivores. For example, why are there no short-faced canids or long-jawed mustelids or felids. Or, why are the felids so conservative (i.e. limited in diversity) with respect to FAC 1 variables compared to the other families? The distributional patterns seen in Figs 1, 2, 3 & 5 may reflect biomechanical constraints on skull shape in the various carnivore families, but before that is assumed, the fossil representatives of those families should be surveyed, to see if they conform to the ranges of their modern relatives. Then, to approach the question of why a given family group has a particular range of skull morphology, more detailed biomechanical analyses of the functional significance of the morphology, and more extensive behavioral-ecological analyses of its biological roles must be carried out, to gain more understanding of the observed patterns. Work in those two areas is currently in progress. ACKNOWLEDGEMENTS For help and advice on this study, I am grateful to the people acknowledged in the previous paper (Radinsky, 1981). I thank S. Arnold for additional advice on statistics, and S. Emerson for critical comments on the manuscript. Measurements were taken on skulls from the Department of Mammalogy of the Field Museum of Natural History, Chicago, and I thank the curators of that collection for access to those materials. This work was supported in part by a grant from the National Science Foundation, DEB-7901584. REFERENCES DAVIS, D. D., 1964. The giant panda. A morphological study of evolutionary mechanisms. Fieldinma: (oologuul Memoirs, 3: 1-339. GOULD, S. J. & LEWONTIN, R. C., 1979. The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. Proceedings o f h Ryul Sock& of London (B), 205: 581-598. HUNT, R. M. 1974., The auditory bulla in Carnivora: an anatomical basis for reappraisal of carnivore evolution. Journal of Morphology, 143: 21-76. RADINSKY, L. B., 1981. Evolution of skull shape in carnivores. 1. Representative modern carnivores. Biologuul journal of the Linncun Socicg, 15: 369-388. L. B. RADINSKY 352 SICHER, H., 1944. Masticatory apparatus in the giant panda and the bears. Fialdiana: @logical SniCs, 29: 61-73. TEDFORD, R. H.,1976. Relationship of pinnipeds to other carnivores (Mammalia). Sysrcmatic &h~, 25: 363-374. WALKER, E. P., 1964. Mammals ofthe World.Baltimore: Johns Hopkins Prm. APPENDICES Appendix 1. Species measured and their abbreviations ~~~ Vivcnids Cyn Pag Hem Par Agl Arc sui Enh Ptr Lut Aon Cynogalebennetti Paguma larvata Hemigalus derbyanus Paradoxurus zelonensis Arctogalidea bivirgata Arcfictis binhtrong Procyonids Pro P n g v n cawiummrc Ail Ailuncs fulgens N m a nasua Nsa N w l h olivacea Nsl PolosJIamls Pot Bassariscus astutus BSC Ban Bassariyon allmi Arctonyx collaris Taxidea tam Meles meles Mehiale pnronahnn Ursids He1 Mls Sel Helarctos malayanus Melursus ursinus Seknarcl~slibeti Conefihrc semisl&tus TR rremarclos m l u r Mephitis nuphihi Spilogale putorius Mydaus jauanensis . . Suillotaxus marchei Enhydra luhic Phoneura brcyiliensis Lutra lutra Aonyx ciwea Thl Ul3 A P rhalarctos mariiimus Ursus americanus Ailuropoda melanoha Hyaenids Hyaena brunnea Hyn cro Crmtacronrta Prt Prolclcs EVOLUTION O F SKULL SHAPE IN CARNIVORES 353 Appendix 2. Measurements BRV BULV COMl JL JXA MAM MAT MFL OCPH OCPW ORBA SL TFL TFW TRL ZAW Brain volume, taken from endocranial capacity. Converted to BRV'p for comparisons with linear dimensions. Bulla volume, estimated from length, width and height measurements by the equation: V = 0.5236 L x W XH (from V = 1/2(4/3nP).Converted to BULV'I3 for comparisons with h e a r dimensions. Condyle to MI, measured from the back of the mandibular condyle to the MI carnassial notch. A measure of the moment arm of the resistance when biting at the carnassial. Jaw length, measured from the back of the condyle to the front of the median incisor alveolus. Resistance moment arm when biting with h n t teeth. Jaw cross-section area, calculated as the second moment of area (I), an estimator of resistance to bending. I = nabs/4, where a = 1/2 jaw width and b = 1/2 jaw height, measured beneath M I . Converted to JXA'P for comparisons with linear dimensions. Moment arm of masseter, measured from the dorsal surface of the condyle to the ventral border of the angular process. An estimator of the moment arm of the superficial masseter. Moment arm of temporalis, measured from the condyle to the apex of the coronoid pmess. An estimator of the moment arm of a portion of the temporalis. Masseteric fossa length, measured from the back of the condyle to the most anterior point of the masseteric fossa. An estimator of the size of the deep maSSeter and of the moment arm of the deep masseter. Occipital height, measured from the medventral border of the foramen magnum to the dorsal rim of the occiput. Occipital width, measured at the widest point of the occiput, across the mastoids. Orbit area, estimated from the circumference as measured around the orbital rim. A = C/12.5664, derived from A = ntl and C = 2nr. An estimator of eye size. Converted to ORBA'P for comparisons with linear dimensions. Skull length, measured from the back of the occipital condyles to the anterior tip of the prmaxilla. An estimator of body size with which most of the other measures are compared. Temporal fossa length, measured from the most posterior point ofthe lambdoidal crest to the back of the supraorbital process. An estimator of temporalis size. Temporal fossa width, calculated by subtracting width at the postorbital constriction from width across the zygomatic arches. An estimator of temporalis size. Tooth row length, measured parallel to the palatal midline, from a point level with the back of the last tooth to the front of the medial incisor alveolus. Zygomatic arch width, measured across the widest portion of the zygomatic arches. Maximum skull width, influenced by brain size and jaw muscle size. L. B. RADINSKY 354 Appendix 3. Means and observed ranges of transformed variables with amended samples of viverrids and mustelids Variable ZAW' OCPW' TFW' TFL TRL' JL' BRV'P' BULV1j3' ORBA'/*' OCPH COMI' MAT' MAM' MFL' JXAl/4' 23 Viverrids 16 Canids 28 Mustelids 14 Felids 1.09 (0.92-1.33) 1.10 (0.8s1.25; 1.38) 1.07 (0.87-1.45) 1.04 (0.92- 1.25) 0.99 (0.89-1.10) 1.02 (0.94- 1.10) 1.03 (0.97-1.13) 0.98 (0.84-1.14) 0.94 (0.78- 1.17) 0.98 (0.82-1.27) 1.04 (0.92- 1.14) 1.13 (0.86-1.42) I .07 (0.80; 0.W1.32) 1.02 (0.84-1.22) 1.07 (0.83-1.37) 1.17 (1.05-1.36) 1.11 (0.96-1.31) 1.01 (0.86-1.17) 068 (0.80-0.97) 1.16 (1.05-1.25) 1( I . W l . 13) 1.12 (0.99-1.28) 1.06 (0.89-1.16; 1.44) 1.07 (0.93-1.19) 1.oo (0.89-1.21) 1.04 (0.94-1.1 1) 1.02 (0.89-1.1 7) 1.07 (0.86-1.26) 0.84 (0.73-0.98) 0.94 (0.81-1.13) 1.24 (1.02-1.43 ; 1.57) 152 ( 1.3G.69) 1.36 ( I .25-1.54) 1.25 (1.06-1.46) 1.12 (0.96-1.20) 0.M (0.814-93) 0.99 (0.95-1.03) 1.29 ( I. 15-1.41 ) 1.29 (1.19-1.53) 158 ( 1.21-1.89 ; 2.15) 1.14 (0.69; 0.82-1.54) 1.13 (0.87; 1.00-1.32) Om (0.761.OO) w (0.86- 1.02) 1.16 (0.92-1.42; 1.59) 1.04 (0.83-1.21; 1.62)t 0.83 (0.64-1.13) 1.19 (0.96-1.48) 1.11 (1.0E.23) 1.39 (0.85-1.68) 1.07 (0.79-1.35) 1.06 (0.88-1.26) 1.26 (0.91-1.58) 1.a (1.2T.79) 1.15 1.02-1 29) I .05 (0.98-1.15) 1.17 (1.05-1.29) 1.25 ( 1.10-1.46) 1.14 (1.06-1.19) 1.25 (1.13-1.34) *Means significantly different (99"/0level) from vivmid aample are indicated in boldface. Underlined boldface means are also significantlydifferent from means between them and the viverrid sample mean. Semicolons denote discontinuously high or low values in observed ranges. ?Sample unamended (no additions to original mustelid sample). 0.032 0.023 0.017 0.029 0.066 0.047 0.030 0.015 0.123 0.069 0.063 0.141 0.089 0.061 0.010 0.024 0.060 0.022 0.012 0.024 0.139 0.042 0.022 0.059 0.054 0.015 0.014 0.027 0.077 0.017 0.036 0.039 0.076 0.025 0.042 0.020 0.028 0.018 0.012 0.008 0.026 0.013 0.015 0.019 0.044 0.018 0.014 0.020 0.044 0.023 0.017 0.033 0.045 0.012 0.019 0.025 0.092 0.041 0.030 0.047 0.080 0.041 0.078 0.044 0.166 0.038 0.036 0.070 0.104 0.029 0.050 0.102 0.048 0.021 0.019 0.016 0.108 0.023 0.018 0.029 0.069 0.040 0.046 0.050 0.045 0.028 0.020 0.017 0.097 0.053 0.053 0.039 0.109 0.057 0.042 0.021 0.094 0.060 0.060 0.077 0.122 0.068 0.073 0.085 0.096 0.033 0.088 0.044 0.163 0.065 0.109 0.047 0.130 0.056 0.044 0.052 0.219 0.033 0.113 0.107 0.145 0.078 0.079 0.082 0.118 0.030 0.042 0.037 0.051 0.024 0.075 0.065 0.055 0.064 0.101 0.041 0.071 0.032 0.047 0.047 0.042 0.065 0.034 0.086 0.061 0.100 0.083 0.035 0.163 0.046 0.063 0.090 0.092 0.029 0.057 0.027 0.128 0.070 0.055 0.083 0.160 0.116 0.129 0.132 0.524 0.132 0.173 0.254 0.135 0.093 0.090 0.372 0.486 0.203 0.185 0.157 BRV’/3’ORBA’/20CPH’COMl’ M A T MAM’ MFL’ JXA*14’ FACl 0.032 0.015 0.008 0.011 JL‘ 0.579 0.098 0.104 0.263 0.139 0.252 0.088 0.548 0.351 0.172 0.203 0.146 0.353 0.130 0.160 0.081 FAC2 * Intraspecific standard deviations calculated from six randomly selected individuals per species; intrafamily standard deviations calculated from the original samples of viverrids, canids, mustelids and felids. Numbers in parentheses after species names correspond to numbers in Fig. 8. 0.092 0.042 0.020 0.053 0.049 0.027 0.032 Felids (14 spp.) Fcfis contolor (10) Felis tigrim ( 1 1) Fefis bmgaLcnsis (12) 0.086 0.060 0.019 0.032 0.032 0.137 0.033 0.046 0.072 0.178 0.038 0.050 0.102 0.076 0.025 0.033 0.037 Mustelids (15 spp) Gulo gulo (7) Grison cuju (8) Must& frenu& (9) 0.080 0.043 0.048 0.012 0.022 0.019 0.102 0.055 0.091 0.031 0.021 0.028 0.077 0.047 0.048 0.024 Canids (16 spp.) Lycuon pichrr (4) Alopex lugopus (5) F m m zndu (6) 0.031 0.031 0.014 0.024 0.088 0.015 0.017 0.016 0.162 0.084 0.070 0.108 0.136 0.052 0.034 0.047 0.113 0.025 0.047 0.049 Vivemds (17 spp.) Viumu ribcthu ( 1) Genet& tigrim (2) Mungos mungos (3) TFL’ ZAW’ OCPW TFW TRL‘ Appendix 4. Intraspecific and intrafamily variation (in standard deviations) of transformed variables and factor scores* v, iEs 5
© Copyright 2026 Paperzz