Edited by Emiliano Bruner

Transcription

1 SHAPE MEETS FUNCTION: STRUCTURAL MODELS IN PRIMATOLOGY Edited by Emiliano Bruner Proceedings of the 20th Congress of the International Primatological Society Torino, Italy, August 2004 MORPHOLOGY AND MORPHOMETRICS

2 JASs Journal of Anthropological Sciences Vol. 82 (2004), pp The implications of variation in knuckle-walking features for models of African hominoid locomotor evolution Sandra E. Inouye 1, Brian T. Shea 2 1) Department of Anatomy, Midwestern University, st Street, Downers Grove, Illinois USA 2) Department of Cell and Molecular Biology, Northwestern University Feinberg School of Medicine, 303 East Chicago Avenue, Chicago, Illinois USA Summary Analysis of the range and patterning of variation in the dorsal metacarpal ridge (DMR), a key feature thought to be related to knuckle-walking in chimpanzees and gorillas, is shown to clarify the three primary competing models of African hominoid locomotor evolution. These models respectively view knuckle-walking and its associated musculoskeletal anatomy as (1) a synapomorphy shared by African apes; (2) a homoplasy in the two lineages; and (3) a symplesiomorphy of the African hominoids, thus requiring the human lineage to be derived from a knuckle-walking ancestor. We examined variation in the presence and relative size of the DMR in ontogenetic and adult samples of African apes. Results indicated a variable presence of the ridge, related to ray number and taxon/size. Analysis of relative size of the DMR revealed strong positive allometry of DMR height within the taxa, and ontogenetic scaling between the common chimpanzee and western lowland gorilla samples. We conclude that models supporting synapomorphy based on identical expression of knuckle-walking features in the African apes are not tenable. We further show that studies claiming knuckle-walking homoplasy in the African ape lineages based on minor differences in growth allometric trajectories are also not strongly supported. We believe the pattern of variation in the DMR supports a view that this feature is not biomechanically required for knuckle-walking behavior and that it may be a plastic response to loads sustained during terrestrial locomotion. Our data fit well with a model of homology of locomotor behavior in the African hominoids as a group, where early hominids are derived from ancestors which knuckle-walked in a manner similar to that seen today in chimpanzees and gorillas. Keywords Knuckle-walking, Dorsal Metacarpal Ridge, Variation, African Hominoids, Phylogeny. Knuckle-walking and hominoid phylogeny The extant African great apes typically engage in the unique form of locomotion known as knuckle-walking when moving on a terrestrial substrate. The kinematic patterning, biomechanics and associated musculo-skeletal anatomy of knuckle-walking have been ably discussed by a large number of authorities over the years (e.g., Schreibner, 1936; Straus, 1940; Napier, 1959; Washburn 1967, 1968a,b; Tuttle, 1967, 1969a,b,c; Susman, 1979; Corruccini, 1978; McHenry & Corruccini, 1983; Corruccini & McHenry, 2001; Lewis, 1972a,b, 1977, 1989; Preuschoft, 1973, 2004; Inouye 1992, 1994a,b; 2003; Sarmiento, 1988; Dainton & Macho, 1999a,b; Richmond & Strait, 2000, 2001; Richmond et al., 2001). Suffice it to say here that while purported skeletal features specifically linked to knuckle-walking have remained somewhat more debatable for the wrist and elbow regions of African apes, the position and anatomy at the metacarpophalangeal joint have attracted considerable anatomical interest due to the weight-bearing and hyperextension at this joint during digitigrade knuckle-walking. Figure 1 (from Tuttle, 1969a) illustrates the bony metacarpophalangeal joint during typical knuckle-walking in a chimpanzee hand. This figure

3 68 Variation in Knuckle-Walking also illustrates the dorsally expanded joint surface and the prominent bony transverse ridge which develops on this dorsal articular surface of the metacarpal heads. This dorsal metacarpal ridge (DMR) has been linked to maintaining the integrity of the metacarpophalangeal joints in hyperextension (Tuttle, 1967, 1969a), although others (Kimura, 1995; Preuschoft, 1973, 2004) have suggested the DMR is more likely related to loads sustained during below-branch suspension. We concur with Stern (1975, p. 63) that the most definitive skeletal sign of knuckle-walking is presented by the heads of the metacarpals. Our primary purpose in this paper is to examine qualitative and quantitative variation in the DMR in the African apes, both within and between species. We believe that a more complete analysis of this variation is central to understanding the biomechanics of knuckle-walking. We further maintain that such knowledge is prerequisite for attempts to (1) reconstruct knucklewalking behavior in extinct hominoids based on bony features; and (2) utilize knuckle-walking anatomy in reconstructing phylogenetic relationships and character evolution in the African hominoids, including humans. It is predominantly the second of these two emphases which we address in this paper Alternative scenarios: synapomorphy, homoplasy or symplesiomorphy? A full understanding of the evolution of knuckle-walking adaptations requires a highlycorroborated phylogenetic framework. Such a phylogenetic consensus has only recently emerged in the past several decades, built predominantly on the contributions of molecular systematics. This body of work initially supported Darwin s general contention that humankind s closest relatives were in fact the African great apes (e.g., Sarich & Wilson, 1967), thus helping to clarify a century or more of heated and interesting debates on the topic (see Tuttle, 1969a, 1974, 1975, for discussion). With Fig. 1 - (from Tuttle, 1969a) Schematic depiction of an exploded third ray in a chimpanzee adult during knuckle-walking. The metacarpophalangeal joint (MPH) separates the metacarpal (M) and proximal phalanx (p), with the dorsal metacarpal ridge clearly seen just superior to the arrowhead. Other symbols: L, lunate; C, capitate; F, flexor tendons; E, extensor tendons; m, middle phalanx; d, distal phalanx.

4 S.E. Inouye & B.T. Shea 69 Fig. 2 - Alternative phylogenetic relationships and character (knuckle-walking) evolution in the large-bodied hominoids. A: Gorilla-Pan as closest relatives, with knuckle-walking as a synapomorphy; B: Homo-Pan as closest relatives, with knuckle-walking a homoplasy in Gorilla and Pan ; C: Homo-Pan as closest relatives, with knuckle-walking a symplesiomorphy in Gorilla and Pan, lost in Homo. the accumulation of diverse studies on amino acid sequences, nucleotide sequences, and DNA- DNA hybridization, the African hominoid trichotomy began to resolve in favor of a humanchimpanzee clade to the exclusion of gorillas (Caccone & Powell, 1989; Pilbeam, 1996, 2000, 2002; Pilbeam & Young, 2001, 2004). By the late 1990 s, this Pan-Homo phylogenetic linkage was very well-established (e.g., Ruvolo, 1997a,b; Ruvolo et al., 1991, 1994). This resolved hominoid phylogeny provides a clear framework against which to assess the evolution of knuckle-walking behavior and anatomy. The importance of this scientific advance to anthropologists and evolutionary biologists cannot be overstated. As fascinating as the debates over alternative models of hominoid locomotor evolution have been - with champions for all the extant apes as key players, and almost all conceivable transformations advocated at one time or another (see Tuttle, 1974) - morphologists must now acknowledge that the list of viable alternative hypotheses has been trimmed dramatically due to the work in molecular systematics. Hypotheses linking humans to orangutans (e.g., Schwartz, 1984) are clearly now untenable. Those linking the African apes as sister taxa must present a compelling case to counter the growing weight of the molecular evidence over the past decade or more. With this in mind, we now briefly review the three primary competing hypotheses of character transformation for knuckle-walking features, with an emphasis on the DMR. (1) Knuckle-walking as African ape synapomorphy This view maintains that knuckle-walking behaviors and features are uniquely seen in African apes because they reflect a shared, derived complex present in the common ances-

5 70 Variation in Knuckle-Walking tor of chimpanzees and gorillas, which are therefore viewed as sister taxa to the exclusion of humans. Figure 2A depicts this scenario. This view has been argued strongly by Andrews (1987, 1992; Andrews and Martin, 1987) in the late 1980 s, as molecular data on the phylogenetic relationships of all the hominoids were accumulating, clearly suggesting orangutans as an outgroup to the African hominoids, including humans. Andrews drew heavily on Tuttle s landmark studies of knuckle-walking features, and his view was also strongly influenced by L. Martin s (1985) findings on enamel thickness and prism patterns in hominoids. What concerns us most in this discussion is the argument forwarded by Andrews and Martin (1987, p. 112) that the knuckle-walking features represent a compelling synapomorphy of the African great apes precisely because there are a large number of components to this complex, and they are all identical in chimpanzees and gorillas. In fact, they specifically rejected an alternative view of character transformation, independent development of knuckle-walking in the respective chimpanzee and gorilla lineages, again invoking the claim that all of the anatomical components were identical. Parenthetically, these authors also dismissed the possibility of the common ancestor of African apes and humans being a knuckle-walker. Clearly, however, the issue of morphological variation (or, more properly, lack thereof) in knuckle-walking features was central to the Andrews and Martin model of African ape synapomorphy. A number of other authorities maintained the view that knuckle-walking behavior and features represent a likely synapomorphy signaling the common ancestry of African great apes to the exclusion of humans or orangutans. Tuttle s (1969a, 1970, 1974, 1981) views were not as strongly influenced by the emerging genetic evidence for the outgroup status of Pongo among the extant large-bodied hominoids as were those of Andrews. Nevertheless, he rejected Washburn s troglodytian model, which favored a particularly close relationship between humans and chimpanzees and reconstructed knucklewalking as the common ancestral condition for the African hominoid clade, and therefore subsequently lost and transformed in early hominids (Tuttle, 1981, p. 90). Tuttle viewed as most parsimonious the scenario of chimpanzee-gorilla synapomorphy of knuckle-walking behaviors and morphology (Tuttle, 1970, p. 247), likely developing subsequent in time to the earlier branching of the hominid lineage (Tuttle & Basmajian, 1974, p. 341; Tuttle, 1967, p. 203). He further logically and taxonomically enunciated this view by advocating that chimpanzees and gorillas should be placed together in the genus Pan (Tuttle, 1967, p ). Finally, it should be stressed that both Andrews and Tuttle were using knuckle-walking features as key (morphological) elements to build a phylogeny for the large-bodied hominoids in these publications cited. They were not merely mapping these features onto the molecular phylogeny we now currently utilize, so it is not at all clear which alternative scenario of character transformation they would have opted for were they forced to utilize a phylogeny linking humans and chimpanzees to the exclusion of gorillas. (2) Knuckle-walking as African ape homoplasy Dainton and Macho (1999a,b; Dainton, 2001) have recently strongly advocated the view that knuckle-walking has evolved independently in the chimpanzee and gorilla lineages. This scenario is summarized in Figure 2B. Although this possibility has been entertained previously in phylogenetic scenarios of hominoid evolution, it is safe to say that many such frameworks were then closely tied to early divergence models advocated by various morphologists. For example, Simons and Pilbeam (1972) considered independent evolution of knuckle-walking behavior in the African great ape lineages quite likely, since they traced these respective lineages directly back to distinct and specific early-middle Miocene taxa which themselves demonstrated no evidence of knuckle-walking adaptations. By contrast, Begun (1992) recently discussed the possibility of parallel development of knucklewalking behaviors in chimpanzee and gorilla lineages, within the phylogenetic context of a chim-

6 S.E. Inouye & B.T. Shea 71 panzee-human clade. Dainton and Macho (1999a,b) have been explicit in favoring independent evolution of knuckle-walking within the current consensus phylogeny and timescale, requiring homoplastic evolution at some point in the period subsequent to the branching of the gorilla and chimpanzee-human lineages. What role does variation in knuckle-walking behavior and morphological adaptations play in the homoplasy scenario of Dainton and Macho (1999a,b)? It is central, since they opt for the independent-evolution scenario primarily on the basis of significant differences in the growth allometries of carpal dimensions in Pan troglodytes vs. Gorilla gorilla (see Fig. 3). These shifts in allometric growth trajectories, combined with previously-established variation in the fine details of locomotor kinematics of knuckle-walking in the two African great apes (including primary contributions by one of us - Inouye, 1992, 1994a,b, 2003), led these authors to conclude that it is parsimonious to suggest that knucklewalking has evolved in parallel in the two lineages (Dainton and Macho, 1999a, p. 171). Let us pause and juxtapose these two scenarios of synapomorphy and homoplasy in African ape knuckle-walking features, highlighting the role of variation. Whereas Andrews and Martin (1987) argued strongly for synapomorphy due to a presumed lack of variation in knuckle-walking features across the African ape species, Dainton and Macho (1999a,b) were so impressed at the degree of variation in their growth trajectories that they could not accept these locomotor adaptations as homologous in these taxa. Underlying each decision are notions and expectations of morphological variation, and its degrees and bases. To that, we of course must add that certain non-articulated expectations regarding the likelihood of a knuckle-walking ancestry for our own lineage were likely also at work in these preferred scenarios. (3) Knuckle-walking as African ape symplesiomorphy This scenario must be immediately distinguished from #1, where knuckle-walking adaptations are also viewed as homologous within the African apes, though of course in that model of Fig. 3 - (from Dainton and Macho, 1999a) An example of a significant transposition (vertical shift) in growth allometries for a carpal bone dimension in chimpanzees and gorillas. See text for discussion.

7 72 Variation in Knuckle-Walking Fig. 4 - Measurement of the dorsal metacarpal ridge (DMR) height. The DMR is calculated as the length of the line CD. This is the maximum perpendicular distance from the line AB, which is tangential to the articular surface just distal to the DMR, to the most dorsal point C on the articular surface of the metacarpal head. A DMR is considered to be present if the angle ABC is less than 180 degrees. synapomorphy they are uniquely homologous to those African great ape species. Here we instead depict the model promulgated by Washburn (1967; 1968a,b; 1972), wherein knuckle-walking is homologous for all the extant African hominoids, including humans, merely being lost in our own lineage once bipedality has evolved. This scenario is depicted in Figure 2C. Washburn was of course strongly influenced by the emerging biomolecular studies (e.g., Sarich & Wilson, 1967) suggesting a close relationship and possibly short evolutionary timespan between our own lineage and those of the African apes, particularly the chimpanzee. This resulted in Washburn s firm commitment to a large-bodied troglodytian model of human origins (Tuttle, 1969a, 1974). Morphological variation plays an indirect but key role in Washburn s views. For example, he noted many similarities between the wrist and hand of African apes and humans; and where there were clear differences, he emphasized that early hominid fossils often demonstrated an intermediate, transitional state (Washburn, 1972). Other advocates of this view, ourselves included (e.g., Shea & Inouye, 1993), have interpreted the lack of key knuckle-walking features such as the DMR in early hominids and other extinct hominoid fossils in light of the range of variation documented for all the extant hominoids. It is to these data which we now turn, and we focus on the implications of the qualitative and quantitative variation in such skeletal features for the alternative scenarios posed immediately above. Morphological variation in the DMR and other features The results we summarize here represent a portion of a larger ongoing study (Inouye and Shea, in prep.), in addition to some previously published material (Inouye, 1992, 1994a,b, 2003; Shea & Inouye, 1993). Our results on the presence and/or relative size of the DMR are based on the following museum samples of adults and non-adults: Pan paniscus (8 males, 9 females, 4 unknown; Museé Royale de l Afrique Centrale); Pan troglodytes (45 males, 47 females, 1 unknown; Powell-Cotton Museum); Gorilla gorilla gorilla (36 males, 48 females, 4 unknown; Powell-Cotton Museum); Gorilla gorilla beringei (4 males, 6 females, 6 unknown; Museé Royale de l Afrique Centrale & U.S. National Museum of Natural History). The ridges were digitized and measured from videotaped lateral views of the metacarpals using the software program MorphSys. The DMR was measured as illustrated in Figure 4 (see Inouye, 1994a, for additional details). The height of the DMR is calculated as the length of the line CD, a perpendicular to an

8 S.E. Inouye & B.T. Shea 73 extension of the line AB, which is fit tangential to the dorsal articular surface of the metacarpal head just distal to the ridge. Line CD extends to the most dorsal point on the articular surface of the metacarpal head (Point C). A metacarpal head is considered to be absent if the angle ABC is 180 degrees or greater. Our results on the presence/absence of the DMRs indicate that these ridges or tori are variably present in all African ape species, and in adults as well as non-adults. Table 1 presents data first for all specimens, then for only adults, for MC II MC V in the four taxa. Note that the only entry which reaches 100% present is for MC III in adult G. g. gorilla. All other entries exhibit varying percentages of ridge absence. Focusing on the variation across the digits, in all groups (save one entry for the poorly-sampled bonobos) MC III and MC IV exhibit the highest percentages of ridge presence. This fits with previous observations (Inouye, 1994b) that African apes typically load these central rays during knuckle-walking, more variably using MC II and MC V (this is differentially true for the less stereotyped knuckle-walking seen in chimpanzees). The data in Table 1 also demonstrate a higher proportion of specimens with DMRs in the adult stages (excluding the poorly-sampled adult bonobos and mountain gorillas), suggesting a general correlation with age and/or size. If we now emphasize taxonomic variation (and leave aside the low samples for bonobos and mountain gorillas), it is clear that lowland gorillas show a higher percentage of ridge presence for all metacarpals than do P. troglodytes. This is true for both the adult and mixed-age samples, and again suggests an association with overall body size. Turning now to those specimens where DMRs are present, we found that the height of the ridge exhibits a marked degree of variation within each of the species (Fig. 5). Table 2 lists the average heights for MC II-V for the adults in each species. These means are larger for MC III and MC IV than in MC II or MC V in all four taxa. The mean DMR is approximately twice as large in G. g. gorilla as in P. troglodytes, supplementing the previous results that they are also more commonly present in gorillas. Regression analysis of ridge height vs. a body size surrogate (geometric mean of a series of postcranial skeletal dimensions) during growth demonstrated significant correlations for the two well-sampled species, P. troglodytes and G. g. gorilla (Tab. 3). The bonobos and mountain gorillas had insignificant correlations for the majority of the metacarpals, likely due to their small sample sizes. The slope values of the logarithmic growth regressions are strongly positively allometric in P. troglodytes and G. g. gorilla, indicating a marked relative increase in the ridge height as body size increases during growth. This strong positive allometry also contributes to increasing shape variance (relative ridge height) during growth and size increase in the African ape species. We undertook analyses of covariance in order to test whether the DMR heights are ontogenet- Tab. 1 - Percentage and number of specimens that exhibit metacarpal ridges based on the measured angle and ridge height. Ontogenetic (Subadult +Adults) and Adult samples.

9 74 Variation in Knuckle-Walking Tab. 2 - Average heights and standard deviations of metacarpal ridges (in mm). Adults only. Fig. 5 - Box-and-whisker plots of adult DMR heights. Species 1 = Pan paniscus; 2 = Pan troglodytes; 3 = Gorilla gorilla gorilla; and 4 = Gorilla gorilla beringei. The center vertical line marks the median of the sample and the length of each box shows the range within which the central 50% of the values fall, with the box edges (called hinges) at the first and third quartiles. The whiskers show the range of observed values that fall within the +/- of 1.5 times the midrange of the interquartile range.

10 S.E. Inouye & B.T. Shea 75 Tab. 3 - Regressions statistics of Dorsal Metacarpal Ridge Heights versus Geometric Mean for Pan troglodytes troglodytes and Gorilla gorilla gorilla. Ontogenetic and Adult samples. vs. LGM Species Group N R P Slope Y-int Ridge 2 Ht. P. t. troglodytes Ontogeny Adult G. g. gorilla Ontogeny Adult Ridge 3 Ht. P. t. troglodytes Ontogeny Adult G. g. gorilla Ontogeny Adult Ridge 4 Ht. P. t. troglodytes Ontogeny Adult G. g. gorilla Ontogeny Adult Ridge 5 Ht. P. t. troglodytes Ontogeny Adult G. g. gorilla Ontogeny Adult Ridge 3+4 Hts. P. t. troglodytes Ontogeny Adult G. g. gorilla Ontogeny Adult ically scaled in P. troglodytes and G. g. gorilla. Ontogenetic scaling here refers to coincidence of trajectories of growth allometry, or sharing of common patterns of relative growth (e.g., Gould, 1975; Shea, 1981,1995). In such cases, the shape differences between adult endpoints are direct consequences of the overall size differences. Table 4 summarizes the results of the analyses of covariance for MCs II through V, plus a summary variable for MC III + MC IV (the two largest ridge heights). There are no significant slope or position differences for any of these comparisons, indicating that relative ridge height in gorillas is what would be expected in a chimpanzee allometrically grown to comparable size. A representative example is illustrated in Figure 6, for log MC IV ridge height plotted against log geometric mean size. Chimpanzees and gorillas of similar body sizes exhibit comparable ridge heights, and the large adult gorillas fall along an extrapolation of the chimpanzee trajectory. In summation, the essential point we emphasize here is that the DMRs are not invariable and identical in their expression in the African apes. In fact, they exhibit considerable variance in their very presence, as well as in their absolute and relative size when they are indeed present (Tab. 1). Both Napier (1959) and Susman (1979; Susman & Creel, 1979) have previously noted aspects of the variable presence of these dorsal metacarpal ridges in the African apes. They each suggested that the lower percentage in chimpanzees and higher frequency in gorillas might be taken as a direct reflection of locomotor preferences, with the chimpanzees more arboreal and the gorillas more terrestrial. But we stress that of course this pattern is also consonant with a more simple and direct explanation, i.e., the presence/absence of the DMRs is related to overall body size, even if size itself also plays a role in the changing locomotor proclivities of African apes as they get larger and heavier ontogenetically, as dimorphic adults, and across the species means. We believe this simpler explanation is likely at work here, in part because the absence of the DMRs in many knuckle-walking African apes means that the feature is clearly not biomechanically required for the specific locomotor behavior itself (Shea & Inouye, 1993; Inouye, 2003). Furthermore, the more terrestrially-inclined mountain gorillas exhibit neither higher percentages of ridges nor relatively larger

11 76 Variation in Knuckle-Walking Tab. 4 - ANCOVA of Dorsal Metacarpal Ridge Heights vs. Geometric Mean. P. t. troglodytes vs. G. g. gorilla. NS = not significant for p<.05. Fig. 6 - A representative example of ontogenetic scaling of allometric growth trajectories of the DMR in P. troglodytes (open circles) and G. g. gorilla (filled circles). This scatter depicts DMR height of metacarpal IV vs. geometric mean size. See Table 4 for statistical comparisons.

12 S.E. Inouye & B.T. Shea 77 Fig. 7 - Two simple schematic path diagrams illustrating alternative inputs to variation in DMR presence and relative size. Body size and locomotor behavior (percentage of arboreal vs. terrestrial locomotion) represent the other two variables. In A, locomotor behavior has a direct effect on DMR, while body size effects locomotor behavior; any correlation between DMR and size is indirect. In B, size has a direct effect on DMR and also on locomotor behavior; any correlation between DMR and locomotor behavior is indirect. Our analyses support the model depicted in B. ridges than the somewhat more arboreal lowland gorillas (Inouye, 2003). Inouye (1994b) demonstrated some minor differences in hand postures during knuckle-walking in gorillas as compared to chimpanzees, even controlling for body size, yet our data on the DMRs shows these to be ontogenetically scaled. Our view would therefore be most appropriately modeled as a path diagram, with size influencing both the DMRs and the percentage of arboreal/terrestrial behavior, but with no clear or necessary independent causal link between locomotor variance and the DMR s themselves (see Fig. 7). Alternative models in light of variation The alternative models of African ape synapomorphy, homoplasy and symplesiomorphy can be reconsidered in light of our results, integrated with other related findings. To the extent that arguments for African ape synapomorphy have been buttressed by claims of invariant and identical expression of knuckle-walking features in these taxa (e.g., Andrews, 1987; Andrews and Martin, 1987), our data lead us to fundamentally question these assessments. Certainly the DMRs vary considerably in their presence & absolute and relative size. Moreover, other features of the elbow, wrist, and hands, some of which have been linked to knucklewalking by various authorities, also demonstrate considerable variation among African apes (e.g., Lewis, 1972a,b; Inouye, 1992, 1994b, 2003; Corruccini, 1978; Corruccini & McHenry, 2001; Sarmiento, 1988; Dainton & Macho, 1999a,b). Kinematic patterns of knuckle-walking behavior itself exhibit minor but significant differences across age/size and taxonomic categories (Tuttle & Watts, 1985; Inouye, 1994b), as

13 78 Variation in Knuckle-Walking Fig. 8 - A scatterplot of metacarpal III length against geometric mean size, illustrating successive downward transpositions from P. paniscus (black circles) and P. troglodytes (grey circles) to G. g. gorilla (black triangles) to G. g. beringei (grey triangles). See text for discussion. do other locomotor frequency data (e.g., Doran, 1997). There is of course no a priori reason to expect that shared derived features in general would exhibit identical morphological expression on the level of minor quantitative comparisons. In fact, one of the primary advocates for knuckle-walking as a derived adaptation uniquely linking chimpanzees and gorillas was also one of the primary catalogers of such variation within and the between the taxa (Tuttle, 1967,1969a,b, 1970; Tuttle & Watts, 1985). But it was the strong advocacy of some (e.g., Andrews, 1987; Andrews & Martin, 1987) in linking the chimpanzees with gorillas to the exclusion of humans in the face of the emerging genetic data which revealed a very close relationship among the three taxa, that led to overzealous statements regarding the supposed uniformity of the knuckle-walking features. Our own and other studies have revealed some of the variation in the DMR and other features argued to be associated with knuckle-walking, but is this variation between the taxa perhaps then indicative of independent evolution of knuckle-walking in the chimpanzee and gorilla lineages? A number of authorities have raised

14 S.E. Inouye & B.T. Shea 79 this possibility (e.g., Simons & Pilbeam, 1972; Begun, 1992), but the strongest recent advocates for this position of knuckle-walking homoplasy have been Dainton and Macho (1999a,b; Dainton, 2001). These authors have rejected a hypothesis of homology in African apes because approximately one-half of their wrist bone dimensions yielded significant differences in growth allometries between chimpanzees and gorillas. Figure 3 illustrates one such growth trajectory difference noted by Dainton & Macho (1999a) as supporting homoplasy in knucklewalking in the chimpanzees and gorillas. We do not share their interpretation of these differences, however, for a number of reasons. First, such minor divergences in growth allometries are typically reflective of variations on a homologous theme in many morphological systems, particularly given any substantial time since divergence, the evolution of substantial body size differences, and so forth. Dainton and Macho (1999b) invoke heterochrony to justify homoplasy as a basis of their differences in wrist allometries, as if tagging such minor differences with this label somehow substantiates morphological and phylogenetic distance. Yet, heterochrony is by definition constrained to minor shifts in timing and rates of development of homologous features seen in ancestral and descendant forms (Gould, 1977, 2000; Shea, 2000). An analysis of morphological variation in terms of heterochronic categories not only bears no necessary connection to homoplasy, it implies that the features under consideration are in fact homologies. Second, controlled and experimental studies reveal that such minor quantitative shifts may result from plasticity in the face of environmental differences (e.g., Huxley, 1932; Bernays, 1986; Schlichting & Pigliucci, 1998), or they may be rapidly produced by drift or artificial or natural selection within populations (e.g., Kidwell et al., 1979; Grant, 1985; Emlen, 1996). In such cases, the minor shifts in growth allometries and quantitative expression of features do not negate the fundamental homology of the system in question. Myriad examples of minor quantitative shifts in proportions (reflected in dissociated allometries) within a clearly homologous feature complex during primate evolution can also be cited. Certain details of limb proportions and joint anatomy in the gibbons and siamangs exhibit minor quantitative and allometric differences (Lumer, 1939; Shea, 1981), yet these do not invalidate other fundamental similarities and question the basic homology of brachiating adaptations in these closely related genera. Similarly, various species of early hominids may exhibit minor specializations in the morphology associated with bipedal adaptations (e.g., Susman et al., 1984; Berge, 1984), but this certainly does not require that we reject a fundamental synapomorphy of bipedality across these taxa. Scapula form and other aspects of thoracic anatomy differ in minor ways in chimpanzees and gorillas (Shea, 1986; Jungers &s Susman, 1984; Taylor, 1997), but this does not incline us to question the homology of the upper body complex in these species (the same could be said for all the large-bodied hominoids, and perhaps all extant hominoids). The high intermembral indices of chimpanzees and gorillas exhibit differences in growth allometries and terminal adult shapes (Shea, 1981), but this does not lead us to conclude that these high indices are independently derived from ancestral conditions of an index close to 1.0. Mountain gorillas exhibit significant differences from eastern and western lowland populations in mandibular allometries (Vogel, 1966) and adult morphologies (Groves, 1970), but we don t conclude that the shared basic folivorous dietary specialization of these closely related forms reflects homoplasy. Finally, Shea (1982, 1983, 1985) found that for a set of 65 bivariate cranial allometries in the African apes, P. paniscus actually exhibited fewer statistically significant differences from G. g. gorilla than from its congener P. troglodytes (though the degree of difference between the gorillas and either of the chimpanzee species was on average substantially greater). Jungers & Hartman (1988) found a similar dissociation between phylogenetic proximity and overall similarity of postcranial skeletal growth allometries in the large-bodied hominoids. Obviously, these minor growth allometric dissociations bear no

15 80 Variation in Knuckle-Walking direct and simple relation to phylogenetic propinquity (see below). One could go on with such general examples at much greater length, but we have a case of even more central relevance to the present discussion. Inouye (1992, 2003) has demonstrated that gorillas exhibit a clear downward transposition in metacarpal (and phalangeal) lengths relative to chimpanzees and bonobos of comparable body size. In other words, gorillas have relatively shorter metacarpals and rays than would be found in chimpanzees and bonobos ontogenetically scaled to their body sizes. Inouye (1992, 2003) has interpreted these differences in light of proportion shifts required to maintain functional comparability at the larger size of gorilla adults. Of greatest significance to our discussion here, however, is the fact that mountain gorillas are also transposed relative to P. troglodytes/p. paniscus and G. g. gorilla in plots of metacarpal length vs. geometric mean size (Fig. 8). If we were to rigidly hew to the criterion for homoplasy proposed by Dainton & Macho (1999a), we would have to conclude that knuckle-walking (or at least ray shortening, whatever its biomechanical basis) has evolved three times independently in these taxa alone! Add to this Sarmiento s (1985) finding that captive (and more terrestrial) orangutans exhibit discernible quantitative differences in carpal morphology compared to wildshot (and more arboreal) specimens, and we must reject the claim of homoplasy advanced by Dainton and Macho (1999a,b; Dainton, 2001) simply on the basis of the fact that not all their ontogenetic allometric trajectories coincide precisely. A few additional theoretical and methodological points must be made regarding Dainton & Macho s (1999a,b) analyses. As is often the case with joint surface widths and ontogenetic trajectories, their data scatters evidence considerable curvilinearity, yet they used linear statistics in comparing the patterns. Although they argue that adults should not be included within the scatter used to estimate cross-sectional growth allometries, this terminal stage of ontogeny is in fact a necessary component of the entire trajectory. The issue that there is little or no growth within the adult stage is also true for any other static stage in the ontogenetic sequence (Cock, 1966). This is not to say that differential sampling of one stage vs. another cannot have an effect on the estimation of the trajectory, however. A more important factor not adequately considered by Dainton & Macho (1999a) is a basic observation of long-standing importance in studies of growth allometry across species. I refer to the common observation that as size increases in a related series of animals, markedly positively allometric growth trajectories are usually downtransposed (decreasing position or y-intercept values), while strongly negatively allometric growth trajectories are typically up-transposed (e.g., Gould, 1966, 1971, 1977; White & Gould, 1965; Shea, 1981, 1983, 1985; Emerson & Bramble, 1993). The adaptive basis for such patterning is thought to relate to the need to avoid extreme shape changes and functional compromise when body size differentiation is considerable. In this light, we point out that many of the carpal measurements discussed by Dainton & Macho (1999a,b) are in fact positively allometric in slope, with the transposition differences following the predicted pattern. It is likely that their statistical differences reflect dissociation and repatterning of the growth allometries in the large-bodied gorilla lineage, as Inouye (1992) has previously argued for hand dimensions. Combining this observation with the fact that these authors chose to downplay the fact that fully one half of their measurements were clearly ontogenetically scaled in the two taxa, we conclude that their argument for homoplasy based on these data is very weak indeed. The homoplasy scenario in the African apes falls for the same reason as did that related to synapomorphy, since the more parsimonious alternative of simple homology (with humans included with their two sister taxa) does not require identical allometric patterning across all conceivable comparisons. At this juncture we discuss our own preferred view, that knuckle-walking adaptations are homologous in African apes, having characterized the last common ancestor of the extant African hominoids (including humans), and sub-

16 S.E. Inouye & B.T. Shea 81 sequently having been lost in our own lineage as our locomotor behavior was transformed to bipedality. This scenario was favored in its essentials many years ago by Keith (1923), and of course most recently by Washburn (1968b, 1972). A number of other recent authors have advocated similar scenarios (e.g., Gebo, 1992, 1996; Richmond & Strait, 2000), though there is no question that the majority of paleontologists and morphologists over the past quarter century have not favored this view. This is undoubtedly due in part to a preference for deep phylogenies linking specific ancient fossils with living taxa, as well as the related rejection of recent molecular timescales favored by Wilson, Sarich, Washburn and their followers (Sarich and Wilson, 1967). A number of morphological features might be cited in support of at least generalized terrestrial quadrupedalism (and most likely knucklewalking, given the phylogenetic and temporal context) as directly antecedent to the locomotor behaviors of extinct early humans. These are predominantly features of the wrist and they have been stressed by Corruccini (1978), Corruccini & McHenry (2001), Lewis (1972a,b; 1989), Miyashiro (1985), Sarmiento (1988), Pilbeam (1996), and Richmond & Strait (2000, 2001), among others. Napier (1959, 1962) previously noted certain similarities between hand bones from Swartkrans and Olduvai to those of African apes. Nevertheless, there are no documented cases of DMRs in early hominid fossils, in contrast to the morphology of the scaphoid ridge on the distal radius of some early hominid fossils, as recently argued by Richmond & Strait (2000, 2001). Three possibilities seem most likely. First, if one or more species of early hominid just subsequent to the bifurcation of the chimpanzee and human lineages still engaged in knuckle-walking, then issues of preservation, sampling and/or small body size might account for the lack to date of such features in the hominid fossil record. Begun (1994, 1999) suggested such an explanation. Second, if the ridges are evolutionary adaptations, they may have been quickly selected against and lost due to the adoption of a novel form of locomotion, freeing the forelimb from any supportive role and perhaps characterized by new selective pressures for different functions and biomechanics of the early hominid hand. Third, if the ridges are plastic responses to the loading history of the joint head during knuckle-walking, then we would not expect to find them in any early hominid species which even facultatively had abandoned knuckle-walking. We favor this third alternative. Our preference for this scenario is based primarily on our own studies of variation in the DMRs, since the general correlation of presence and relative size of the ridges with overall body size fits well with a plastic response loading model (e.g., Frost, 1987; Hamrick, 1999). This interpretation would be rejected if small-bodied species clearly phylogenetically linked to chimpanzees or gorillas were discovered with ridges as strongly (or moreso) developed as in extant Gorilla. Additional experimental or comparative studies of the response of small joint surfaces to large weight support roles would also contribute important information. The well-known changes occurring in the distal hallucial ray in ballet dancers offers some indirect but suggestive support (e.g., Mann, 1995). Sarmiento s (1985) investigation of captive vs. wild orangutans noted above is also very relevant here. The recent claims by Richmond & Strait (2000, 2001) that osseous features associated with knuckle-walking can be seen in early hominids fit well with Washburn s (1968a,b; 1972) troglodytian scenario of human evolution and our own conclusions here. In particular, they emphasize the presence of a scaphoid ridge on the distal radius in a specimen of Australopithecus anamensis as evidence supporting a knuckle-walking ancestry for early hominids. Richmond & Strait do not see such features as indicating knuckle-walking behaviors in these early hominids themselves, but rather as retentions from antecedent phylogenetic stages. Such arguments always raise intriguing and problematic issues in our interpretations of adaptations in extinct forms (e.g., see Kay and Cartmill, 1977), and this case is no different. Moreover, one wonders why a scaphoid ridge reflecting antecedent knuckle-walking stages

17 82 Variation in Knuckle-Walking would be retained in early hominids, while the DMR is not. Perhaps this provides additional indirect support of the view of the DMR as a plastic response to habitual weight transfer. But clearly we need much greater understanding of the loading and function of the wrist and ray joints during knuckle-walking and other locomotor behaviors. In spite of the fact that the Richmond & Strait (2000, 2001) perspective agrees with our own preferred scenario, we agree with many of the critical issues raised in response to their paper (e.g., Corruccini & McHenry, 2001). In particular, we feel that their multivari ate discriminant analysis does not necessarily succeed in identifying bony features associated specifically with knuckle-walking. We maintain that detailed studies of the patterning of variation exhibited by all such purported knucklewalking features is required, much as we have done in the present case with the DMR. The role of growth allometries in systematic studies Important general issues are raised herein regarding variation and the role that concordant and discordant growth allometries should play in systematic investigations. It has recently been argue by Pilbeam & Young (2001, p. 354) that allometric analyses, particularly those involving ontogenetic allometry, strike us as providing very suitable characters, in that they can be identified as probably homologies. While those of us working in the realm of growth allometry would be heartened if such data could make substantial contributions to systematic investigations, we caution that any direct reading of phylogeny from such data is likely to be quite problematic, for precisely the reasons noted above (see Inouye & Shea, in prep., for additional discussion). We believe there is unfortunately a clear asymmetry associated with growth allometric data for systematics. While broad-based ontogenetic scaling may indeed be strongly indicative of homology (it being quite unlikely that homoplastic similarities would come to share precisely the same growth trajectories), it does not follow that dissociations in ontogenetic allometries are necessarily evidence of derived novelty and homoplasy (see discussion above). This is particularly true of minor quantitative shifts in growth allometries. Strongly divergent growth allometries combined with other data suggesting novel features and functions may certainly reflect non-homology, but in such situations the allometric data themselves are likely to be redundant and insufficient. Considerable research needs to be done before resolving how and when such growth allometric data can be mined for phylogenetic signals, and we would suggest an approach that combines information on genetic bases, novel function and the degree of interspecific allometric discordance relative to known environmental influences and intraspecific variance. A framework carefully integrating the effect size as advocated by Cohen (1988) would be a useful start, helping sort minor quantitative shifts from more fundamental repatterning and novel configuration. On the whole, we view the cladistic and systematic issues associated with growth allometries to be similar to those raised generally for quantitative features (e.g., Pimental & Riggins, 1987; Thiele, 1993; Wiens & Servedio, 1997; Zelditch et al., 1995). These issues gain even broader relevance when it is realized that most dichotomous qualitative features in studies of hominid phylogeny are actually disguised continuously-variable features (e.g., larger, smaller, more robust, etc.). Conclusions We have demonstrated in this paper that analysis of the variation in the dorsal metacarpal ridge (DMR), a key feature in the hands of knuckle-walking African apes, clarifies our understanding of competing scenarios of phylogeny and character transformation in hominoid evolution. We find that models claiming sister relationship between chimpanzees and gorillas on the basis of invariant expression of knuckle-walking features are not tenable. Considering that molecular data have essentially already rejected this phylogeny, our primary contribution here is in demonstrating why the knuckle-walking features never truly were as compelling a morphological synapomorphy of

18 S.E. Inouye & B.T. Shea 83 chimpanzees and gorillas as was frequently asserted. We also conclude that models requiring the independent evolution of knuckle-walking behavior in the respective chimpanzee and gorilla lineages, based on minor differences in patterns of growth allometry, are not strongly supported. Our analyses have demonstrated that there is considerable variation in the presence and relative size of the DMRs in chimpanzees and gorillas, and that overall body size and allometric factors are the primary inputs to this variance. DMRs grow with strong positive allometry within the species, and the differences in relative size and shape of the DMRs between adult chimpanzees and gorillas result from ontogenetic scaling, or common patterns of relative growth terminating at different adult sizes (Gould, 1975). This is a pattern observed in many, though certainly not all, features of the bony skeleton in comparisons of the closely related African apes. Furthermore, we see no evidence of any locomotor influence on the presence or relative size of the DMR, independent of body size and its general association with frequencies of arboreal and terrestrial locomotion in the African apes. In our view, the DMR of African apes cannot be interpreted as biomechanically necessitated by knuckle-walking behavior per se, but rather is likely to be a plastic response to loads sustained at the hyperextended metacarpophalangeal joint during knuckle-walking. The nature and basis of patterns of variation in other bony hand, wrist and elbow features purportedly associated with knuckle-walking need comparable analysis before the evolutionary transformations in these morphologies can be fully understood. In the final analysis, whether the DMRs are plastic physiological adaptations or true evolutionary adaptations, the absence of such bony features in early hominids poses no difficulties for models which hold that these species are descended from forms engaging in knucklewalking behaviors essentially identical to those seen in the extant African apes. We believe this troglodytian model favored by Keith (1923) and Washburn (1968a,b; 1972), and more recently advocated by Gebo (1992, 1996), Pilbeam (1996, 2002) and others, is the most likely model of African hominoid locomotor evolution over the past eight-plus million years. Acknowledgments We thank Dr. Emiliano Bruner for his kind invitation to publish our paper in this special edition of the Journal. SEI gratefully acknowledges Midwestern University (Downers Grove, IL, USA), and BTS thanks Northwestern University (Chicago, IL, USA), for support and travel funds. References Andrews P Aspects of hominoid phylogeny. In Patterson C. (ed), Molecules and Morphology in Evolution: Conflict or Compromise?; pp Cambridge University Press, Cambridge, U.K. Andrews P Evolution and environment in the Hominoidea. Nature, 360: Andrews P. & Martin L Cladistic relationships of extant and fossil hominoids. J. Hum. Evol., 16: Begun D.R Miocene fossil hominids and the chimp-human clade. Science, 257: Begun D.R Relations among the great apes and humans: new interpretations based on the fossil great ape Dryopithecus. Yrbk. Phys. Anthropol. 37: Begun D.R Hominid family values: morphological and molecular data on relations among the great apes and humans. In Parker S.T., Mitchell R.W. & Miles M.T. (eds), The Mentalities of Gorillas and Orangutans; pp Cambridge University Press, Cambridge, U.K. Berge C Multivariate analysis of the pelvis for hominids and other extant primates: implications for locomotion and systematics of the different species of australopithecines. J. Hum. Evol., 13: Bernays E.A Diet-induced head allometry among foliage-chewing insects and its importance for graminivores. Science, 231: