Interpreting the Posture and Locomotion of Australopithecus afarensis: Where Do We Stand?

Transcription

1 YEARBOOK OF PHYSICAL ANTHROPOLOGY 45: (2002) Interpreting the Posture and Locomotion of Australopithecus afarensis: Where Do We Stand? CAROL V. WARD* Department of Anthropology, and Department of Pathology and Anatomical Sciences, University of Missouri, Columbia, Missouri KEY WORDS bipedality; early hominins; interpretation ABSTRACT Reconstructing the transition to bipedality is key to understanding early hominin evolution. Because it is the best-known early hominin species, Australopithecus afarensis forms a baseline for interpreting locomotion in all early hominins. While most researchers agree that A. afarensis individuals were habitual bipeds, they disagree over the importance of arboreality for them. There are two main reasons for the disagreement. First, there are divergent perspectives on how to interpret primitive characters. Primitive traits may be retained by stabilizing selection, pleiotropy, or other ontogenetic mechanisms. Alternately, they could be in the process of being reduced, or they simply could be selectively neutral. Second, researchers are asking fundamentally different questions about the fossils. Some are interested in reconstructing the history of selection that shaped A. afarensis, while others are interested in reconstructing A. afarensis behavior. By explicitly outlining whether we are interested in reconstructing selective history or behavior, we can develop testable hypotheses to govern our investigations of the fossils. To infer the selective history that shaped a taxon, we must first consider character polarity. Derived traits that enhance a particular function, are found to be associated with that function in extant homologs, and that epigenetically sensitive data indicate were actually being used for that function, can be interpreted as adaptations. The null hypothesis to explain the retention of primitive traits is that of selective neutrality, or nonaptation. Disproving this requires demonstration of active stabilizing or negative selection (disaptation). Stabilizing selection can be inferred when primitive traits compromise a derived function clearly of adaptive value. Prolonged stasis, continued use of the trait for a particular function, or no change in variability in the trait are evidence that can support a hypothesis of adaptation for primitive traits, but still do not falsify the null hypothesis. Disaptation, or negative selection, should result in a trait being reduced or lost. To infer the behaviors of a fossil species, we must first determine its adaptations, use this to make hypotheses about its behavior, and test these hypotheses using epigenetically sensitive traits that are modified by an individual s activity pattern. When the A. afarensis data are evaluated using this framework, it is clear that these hominins had undergone selection for habitual bipedality, but the null hypothesis of nonaptation to explain the retention of primitive, ape-like characters cannot be falsified at present. The apparent stasis in Australopithecus postcranial form is currently the strongest evidence for stabilizing selection maintaining its primitive features. Evidence from features affected by individual behaviors during ontogeny shows that A. afarensis individuals were habitually traveling bipedally, but evidence presented for arboreal behavior so far is not conclusive. By clearly identifying the questions we are asking about early hominin fossils, refining our knowledge about character polarities, and elucidating the factors influencing morphology, we will be able to progress in our understanding of the posture and locomotion of A. afarensis and all early hominins. Am J Phys Anthropol 45: , Wiley-Liss, Inc. Grant sponsor: NSF; Grant number: SBR ; Grant sponsor: University of Missouri Research Council and Research Board. *Correspondence to: Carol V. Ward, Department of Anthropology, and Department of Pathology and Anatomical Sciences, 107 Swallow Hall, University of Missouri, Columbia, MO WardCV@missouri.edu DOI /ajpa Published online in Wiley InterScience ( com). TABLE OF CONTENTS Introduction Current Hypotheses of Australopithecus afarensis Locomotor Behavior Reasons for the Debate Empirical Frameworks Hypotheses about adaptation Hypotheses about behavior WILEY-LISS, INC.

2 186 YEARBOOK OF PHYSICAL ANTHROPOLOGY [Vol. 45, 2002 Reexamination of the Fossil Evidence Using This Framework Character polarities and vectors of morphological change in early hominins Do the derived traits of A. afarensis reflect bipedality? Did primitive traits compromise bipedality? Continuing morphological refinement or stasis? Did A. afarensis actually climb trees? The Many Influences on Morphology Implications for Testing Hypotheses About Why Hominin Bipedality Evolved Summary and Conclusions Acknowledgments Literature Cited INTRODUCTION Bipedality is the hallmark of the human lineage. Whether or not our unique locomotor mode evolved in the original members of our clade, it represents a fundamental adaptive shift away from the apes. Bipedality has served as a significant preadaptation to the acquisition of other key human characteristics, such as tool-making, altriciality of infants, and arguably even intelligence. Understanding the nature and timing of the transition to terrestrial bipedal locomotion is key to an accurate interpretation of how and why humans evolved. Over the past 78 years, our knowledge of the nature of early hominin (which I define as members of the human clade more closely related to us than to Pan; see Richmond et al., 2001) bipedality has improved dramatically. Dart (1925) correctly inferred that the Taung child had upright posture, and that bipedality characterized the earliest part of the human lineage. This interpretation was bolstered by further early hominin fossil finds, particularly in South Africa, over the next few decades (e.g., Broom, 1943; Broom and Schepers, 1946; Le Gros Clark, 1947, 1955; Straus, 1948; Broom and Robinson, 1949, 1950; Dart, 1949a,b, 1958; Kern and Straus, 1949; Broom et al., 1950; Mednick, 1955; Chopra, 1962; Napier, 1964, 1967; Day and Wood, 1969; Lovejoy and Heiple, 1970), with interpretations culminating in Early Hominid Posture and Locomotion (Robinson, 1972). With the discovery of earlier and more numerous Australopithecus afarensis 1 fossils in the 1970s (Taieb et al., 1974, 1975; Johanson et al., 1982b and references therein) and footprints at Laetoli (Leakey and Hay, 1979), research on Australopithecus locomotion intensified, and more detailed questions began to be asked about the functional morphology of these hominins. 1 Some researchers have suggested that because the genus Australopithecus as traditionally defined may be paraphyletic or polyphyletic, some species should be removed from Australopithecus and placed in Praeanthropus or Paranthropus (Senut, 1996, 1999; references in Collard and Wood, 2000). Because this paper does not consider the alpha taxonomy of hominins, and there is no general consensus about hominin taxonomy, I will simply refer to all hominin species originally attributed to Australopithecus (except Ardipithecus ramidus) as a single genus, and focus my discussions on functional interpretation of the fossils. The debate over hominin locomotion became intense in the early 1980s, after the publication of the Hadar fossil descriptions, and it continues to this day. Some researchers have presented data supporting the hypothesis that Australopithecus afarensis individuals were obligate bipeds for whom arboreality was adaptively insignificant (Lovejoy et al., 1973; Lovejoy, 1975, 1978, 1988; Day and Wickens, 1980; White, 1980; Latimer, 1983, 1991; Ohman, 1986; Latimer et al., 1987; Latimer and Lovejoy, 1989, 1990a, b; Crompton et al., 1998; Kramer, 1999). Others contend that A. afarensis was primarily bipedal, yet retained significant adaptations to arboreality and thus was partly arboreal (Senut, 1980; Stern and Susman, 1981, 1983, 1991; Feldesman, 1982; Jungers, 1982, 1991; Jungers and Stern, 1983; Schmid, 1983; Rose, 1984, 1991; Susman et al., 1984; Deloison, 1985, 1991, 1992; Tardieu 1986a, b; Susman and Stern, 1991; Duncan et al., 1994; Stern, 2000), perhaps with a compromised form of bipedal progression stemming from these retained arboreal characters (Susman et al., 1984; Preuschoft and Witte, 1991; Rak, 1991; Cartmill and Schmitt, 1996; MacLatchy, 1996; Schmitt et al., 1996, 1999; Ruff, 1998; Stern, 1999). At least one researcher argued that A. afarensis was not bipedal at all, but was a palmigrade-plantigrade quadruped (Sarmiento, 1987, 1994, 1998). The debate is polarized and polarizing, and is to the point that it impedes our further understanding of the posture and locomotion of early hominins. It serves little further purpose as it has been framed thus far. Recently, even earlier hominins have been discovered. Australopithecus anamensis is now known to have been a committed biped at 4.2 mya, but is poorly known postcranially (Leakey et al., 1995, 1998; Ward et al., 1999a, 2001). Putative hominins Sahelanthropus tchadensis (Brunet et al., 2002) existed 6 7 mya, Orrorin tugenensis about 6 mya (Senut et al., 2001), and Ardipithecus ramidus from mya (White et al., 1994, 1995; Haile- Selassie, 2001). From the limited data published so far, the nature of these hominins locomotor adaptations cannot be fully ascertained. Regardless of what we learn about Sahelanthropus, Orrorin, and Ardipithecus in the coming years, Australopithecus, and in particular A. afarensis, will remain the basis for our understanding of the origins and early evolution of hominin locomotion. It is the model to

3 Ward] which we will compare the new fossils in order to assess the trajectory of early hominin locomotion, and is central to our understanding of early hominin evolution. So, resolving the debate over its locomotor behavior, and its biology, is still a critical concern of paleoanthropology. CURRENT HYPOTHESES OF AUSTRALOPITHECUS AFARENSIS LOCOMOTOR BEHAVIOR The first of three hypotheses suggests that A. afarensis individuals were palmigrade-plantigrade quadrupeds (Sarmiento, 1987, 1994, 1998). This idea is based primarily on a suite of morphological similarities between the hands and feet of A. afarensis and gorillas, in particular lowland gorillas, which Sarmiento (1987, 1994, 1998) argued are adaptations to weight transfer through the fore- and hindlimbs during quadrupedal progression. For example, Australopithecus and Gorilla share shorter fingers and toes than typical for other great apes, a similar morphology of the hamate-triquetral facet, a more palmarly directed hamulus and pisiform, broad tibialis posterior insertion, and large plantar aponeurosis (Sarmiento, 1994). While these similarities might seem to support this hypothesis, the overwhelming suite of skeletal modifications that enhance bipedal posture and locomotion seen in Australopithecus, including the sinusoidal vertebral curvatures (Robinson, 1972; Ward and Latimer, 1991; but see Sarmiento, 1998), short pelvis with laterally-rotated iliac blades (Johanson et al., 1982a; Tague and Lovejoy, 1986; Lovejoy et al., 1999), broad sacrum (Robinson, 1972; Leutenegger and Kelly, 1977), and femoral neck structure and condyle shape (Lovejoy et al., 1973; Johanson et al., 1982a), are not shared with gorillas, and do not support the hypothesis of quadrupedal posture or locomotion in Australopithecus. In particular, the valgus angle of the knee of hominids is an epigenetically labile trait (Tardieu and Trinkaus, 1994; Duren and Ward, 1995; Duren, 1999) and demonstrates that A. afarensis walked upright not quadrupedally. In a recent study of the development of a valgus knee in humans in normal and nonambulatory myelodysplastic children, Duren (1999) found that a bicondylar angle was the direct result of a habitual bipedal gait. It does not form in these affected individuals who engage only in intermittent, or facultative, bipedality. The presence of a valgus knee in Australopithecus demonstrates that it was a habitual biped, falsifying the hypothesis of Sarmiento (1987, 1994, 1998). Many shared similarities of early hominins and gorillas may be primitive, and do not necessarily signify equivalent adaptations to a particular shared locomotor repertoire in these taxa. Alternately, they could be homoplasies evolved for different purposes in hominins and gorillas. Therefore, this hypothesis will not be dealt with further here. INTERPRETING A. AFARENSIS LOCOMOTION 187 The second hypothesis is the first of two that have received much broader support by data and by the paleoanthropological community. This position has been articulated most clearly and extensively by Stern, Susman, and Jungers (summarized by Stern, 2000), and states that A. afarensis individuals were primarily terrestrial bipeds, but they also climbed trees (e.g., Senut, 1980; Berge and Ponge, 1983; Jungers and Stern, 1983; Schmid, 1983; Stern and Susman, 1983, 1991; Tardieu, 1986a, b, 1999; Berge, 1984, 1991, 1994; Rose, 1984; Susman et al., 1984; Berge and Kazmierczak, 1986; Duncan et al., 1994; Hunt, 1994; Schmitt et al, 1996; Stern, 1999, 2000). These researchers based their conclusions on the mosaic of morphological features found in the A. afarensis skeleton (see McHenry, 1994; Stern, 2000). They acknowledged the suite of features reflecting upright bipedal progression, but also emphasized numerous ape-like characters. They argued that when reconstructing the probable positional behaviors of A. afarensis, as for any fossil taxon, one must consider the total morphological pattern evident in a skeleton (Le Gros Clark, 1955; Stern and Susman, 1991). From this perspective, all features of an organism are valuable and used, and if there are any primitive traits retained in the skeleton, they must be there for a functional reason. Some researchers have gone one step further, suggesting that A. afarensis exhibits a compromise morphology, and that these primitive traits would have affected the nature of bipedal progression in Australopithecus (Stern and Susman, 1983; Jungers, 1982; Jungers and Stern, 1983; Rose, 1984, 1991; Susman et al., 1984; Schmitt et al, 1996, 1999; Stern, 1999), resulting in a bent-knee, bent-hip bipedal progression. The third hypothesis has also received widespread support. The most widely published proponents of this idea are Latimer, Lovejoy, and Ohman, and the clearest articulation of their perspective can be found in Latimer (1991). These researchers also interpret the numerous derived features of the A. afarensis skeleton as clear evidence of selection for habitual terrestrial bipedality. They note that because A. afarensis had not only acquired adaptations for bipedality, it had, in the process, reduced its arboreal efficiency by sacrificing traits such as relatively long forelimbs, fingers, and toes, and perhaps most importantly, a grasping foot (Latimer and Lovejoy, 1989; Latimer, 1991). They cite the clear vector of morphological change (defined as the magnitude and direction of morphological character transformation throughout a lineage; Simpson, 1953) that was towards anatomy that enhanced terrestrial bipedality and diminished arboreal competence (Latimer, 1991). Thus, the shift away from features enhancing arboreal ability must derive from a behavioral repertoire in which arboreality was not favored by selection (Latimer, 1991). Hence, arboreality must have been relatively unimportant for survival and reproduction in A. afarensis and its immediate ancestors. They argue that if A. afarensis

4 188 YEARBOOK OF PHYSICAL ANTHROPOLOGY [Vol. 45, 2002 individuals did climb trees, arboreal agility had little adaptive significance for them. Although at one point Lovejoy (1988) went one step further to add that A. afarensis not only was not adapted to climb trees, but could not have done so effectively without adverse consequences for their fitness, he later returned to the argument that that it simply is not possible to determine the fitness benefits of arboreality in A. afarensis (Ohman et al., 1997). In summary, these latter two hypotheses about early hominin locomotion are in complete accord over the fact that the recent ancestors of A. afarensis had undergone selection for terrestrial bipedal locomotion. They differ substantially only over the relative importance of arboreality for A. afarensis. REASONS FOR THE DEBATE The first step necessary in unraveling the nature of early hominin bipedality is to understand the nature of the disagreements over interpretations of A. afarensis locomotion. The debate is not due to a lack of fossil evidence, because A. afarensis is known from most skeletal elements. There are a few disagreements over interpretations of available fossil evidence (summarized in Stern, 2000), but the debate has two more significant causes. First, researchers have different approaches to interpreting anatomy, and in particular primitive retentions (Latimer, 1991; Coffing, 1999). Second, and equally importantly, they are asking fundamentally different questions of the data. One reason for the disagreement over Australopithecus locomotion is a fundamental philosophical difference in the interpretation of primitive characters (Coffing, 1999). Derived signals in morphology are relatively easy to interpret. The numerous derived traits of Australopithecus have been shown by biomechanical analyses to be accommodations for weight-bearing and movement in upright bipedal posture. These features, like sinusoidal vertebral curvatures or an adducted hallux, are unambiguous, and are considered adaptations to habitual upright posture and locomotion. Their unambiguous polarity, coupled with their clear affect on function (in the sense of Lauder, 1996: the use or action of phenotypic features), reveals their biological role (again as in Lauder, 1996: the role of phenotypic features in a specific environmental or ecological setting) for bipedal progression (Weishampel, 1995). These characters tell us that the ability to walk upright conferred significant fitness benefits on those individuals better designed to deal with the mechanical demands of terrestrial bipedality. The ape-like features of Australopithecus, onthe other hand, are generally assumed to be primitive retentions (lists in McHenry, 1994; Stern, 2000), and many probably are. Interpreting why these primitive features are retained in A. afarensis, or any taxon, is where perspectives diverge. Australopithecus had undergone numerous modifications that served to enhance terrestrial bipedal function at the expense of arboreal function, most notably the loss of a divergent hallux, reduction in intermembral index, and shorter fingers and toes. These changes imply that agile movement in the trees was of lesser reproductive value than was bipedality for the ancestors of A. afarensis. This is not to say that arboreality was not also selectively valuable, but perhaps it was to a lesser degree. Diminished arboreal competence did not result in a selective disadvantage to the immediate ancestors of A. afarensis, however. In this case, the retention of features ordinarily associated with arboreal progression, such as forelimbs longer relative to hindlimbs, or phalanges longer and more curved compared with those of humans, is inherently difficult to explain (see Gould and Lewontin, 1979; Bock, 1980, Baum and Larson, 1991; VanValkenburgh, 1994; Lauder, 1995, 1996). Such traits could have been actively retained by stabilizing selection because they continued to enhance the survival and reproductive success of their bearers by enabling them to be more agile tree climbers than they would otherwise be (adaptations; Gould and Vrba, 1982). Alternately, they may simply have been features that did not compromise arboreality and were neither selected for nor against (nonaptations; Gould and Vrba, 1982). They may even have been in the process of being selected against (disaptations; Gould and Vrba, 1982). Because any of these three mechanisms can occur, and undoubtedly does occur, these hypotheses are impossible to test without further information. Latimer, Lovejoy, and Ohman argue that because we cannot readily test these hypotheses, we cannot consider the adaptive value of retained primitive traits. Stern, Susman, and Jungers, on the other hand, argue that the stabilizing selection hypothesis is the only logical one to explain the maintenance of the numerous primitive features of A. afarensis. Biases towards one hypothesis can color researchers interpretations of the adaptive role of morphological structures, and thus the biology and behavior of extinct animals. The second, and equally important, reason for the disagreement over Australopithecus locomotion is that these two sets of researchers are asking different questions. Latimer, Lovejoy, and Ohman are interested in attempting to reconstruct the pattern of natural selection that produced A. afarensis: why did A. afarensis evolve? They would argue that it does not matter what actual activities individuals engaged in over the course of the day; the strength of selection on a particular trait that affects a behavior is not proportional to the time the individual engages in that behavior. Individuals sit, sleep, and engage in many other activities that do not impose adaptively significant consequences on those individuals that do not possess morphologies particularly suited to those activities. In other words, all activities do not have the same fitness effects with respect to the skeleton. They argue that we simply

5 Ward] cannot know, nor is it necessary to know, all of the behaviors in which the animals engaged. If we are interested in reconstructing adaptive history, then it is the vectors of morphological change (Simpson, 1953) that will reflect vectors of selection acting on a lineage, and provide the answers we seek (Latimer, 1991). Stabilizing selection is harder to discriminate from lack of selection for primitive structures, Latimer (1991) argues, so we must restrict basing our hypotheses about selective pressures that shaped a lineage to evidence from directional change within that lineage. Stern, Susman, and Jungers are asking a different question (Stern and Susman, 1983, 1991; Susman et al., 1984; Susman and Stern, 1991; summarized in Stern, 2000). They are attempting to infer the actual behavioral repertoires of A. afarensis individuals, and not the history of selection. These researchers note that individuals engage in varied locomotor activities over the course of days and lifetimes. Their approach seeks to identify what types of behaviors one might see A. afarensis engaged in if one traveled back in time to observe them: what was their locomotor repertoire? Stern (2000) chooses a quote from Duncan et al. (1994, p. 79) that summarizes their philosophy: understanding the overall functional pattern (emphasis mine) of the organism requires an equal consideration of all its anatomical features, regardless of whether they are apomorphies, plesiomorphies or homoplasies. This viewpoint serves to frame the fossils as once fully functional living organism. Stern, Susman, and Jungers argue that this approach is indeed important, because in order to understand our ancestors, we need to be able to reconstruct their behavior. These two approaches are different, and complementary rather than mutually exclusive. Thus, the apparent debate as it has been framed is not a true debate. It is important to keep in mind the distinction between reconstructing behaviors from reconstructing evolutionary pressures. Neither approach to understanding Australopithecus is better or worse than the other. They are intrinsically different, however. They address the data differently, and rely on different characters for different reasons. They should not be contrasted as opposite sides of a single argument. Both questions are worth pursuing. We need to understand the types of animals early hominins were, as well as the history of selection that shaped them. It is time to recognize the differences in these approaches, note their inherent limitations, and proceed. EMPIRICAL FRAMEWORKS To move forward in our understanding of Australopithecus locomotor behavior, we need to focus our approach to interpreting the fossils. The first step is to be explicit about what questions we are asking. Do we want to reconstruct the history of natural selection shaping the hominin lineage(s), or the locomotor/behavioral repertoires of fossil taxa? The INTERPRETING A. AFARENSIS LOCOMOTION 189 next step is to identify specific hypotheses and criteria for their potential falsification. To address the question of selective history, we first must understand the polarity of traits we are considering, and the vectors of morphological change apparent in the early hominin skeleton. To answer either question, we need to better understand morphology and its influences. Not all traits provide similar information about fossil taxa (e.g., Cartmill, 1994; Churchill, 1996; Lieberman, 1997; Lovejoy et al., 1999, 2000). Some will tell us more about genetic change within a species, and therefore the history of selection, and some will tell us more about an individual s behavior over the course of its lifetime. Hypotheses about adaptation To infer the selective regime acting on a species from morphology alone is notoriously difficult, and a substantial literature is devoted to this issue (e.g., Gans, 1966; Gould and Lewontin, 1979; Bock, 1980; Wake, 1982; Arnold, 1983; Mayr, 1983; Baum and Larson, 1991; Leroi et al., 1994; Koehl, 1996; papers and references in Thomason, 1995; Rose and Lauder, 1996). Without ecological data and data about which behaviors affect fitness in which ways, we lack critical information that would enable us to actually test hypotheses about the selective regimes acting on A. afarensis (Baum and Larson, 1991; Koehl, 1996; Larson and Losos, 1996). Instead, we must rely on phylogeny, biomechanical modeling, and comparative morphology to support or refute hypotheses about selective pressures as best we can. To construct hypotheses about the selective regime that shaped a taxon, one must first identify the character transformations that occurred in that lineage (Felsenstein, 1985; Brooks and MacLennan, 1991; Harvey and Pagel, 1991; Kay and Covert, 1984; Lauder, 1995, 1996; Weishampel, 1995; Witmer, 1995; Larson and Losos, 1996; Begun et al., 1997a,b). This depends on having a reliable phylogeny from which one can reconstruct ancestral character states. In the case of Australopithecus, molecular studies have produced robust hypotheses about extant hominoid relationships (see Ruvolo, 1997). This may not be enough to reconstruct the character states of the last common ancestor of apes and hominins, however. The burgeoning fossil record of Miocene and Pliocene apes, including hominins, slowly continues to add more data with which to develop and test phylogenetic hypotheses among living and fossil taxa. Because these fossil taxa do not resemble extant species in all features, they stand to alter the balance of parsimony for traits, and so could affect our interpretations of the primitive conditions from which Australopithecus or other early hominins evolved. Continual attention to phylogenetic analyses of hominoids will be critical to our ability to reconstruct character transformations that produced Australopithecus and other early hominins. While drift and other random forces may have lasting impacts on species over time (see Nitecki,

6 190 YEARBOOK OF PHYSICAL ANTHROPOLOGY [Vol. 45, ), the only evolutionary force that can produce long-term directional change is selection, particularly when there is a broad suite of features that all enhance the same function, e.g., bipedality in Australopithecus. Therefore, it is reasonable to infer that when long-term directional change is observed within a lineage, especially in numerous characters or character complexes, that selection for a behavior influenced by the observed morphology was the cause. Biomechanical modeling and comparative studies can help reveal the biological roles of the newly transformed structures by determining which function or functions they enhance (Baum and Larson, 1991; Weishampel, 1995; Lauder, 1995, 1996). Coupling phylogenetic data with biomechanical data allows paleontologists to form reasonable hypotheses about the behaviors on which selection acted. If phylogenetic analysis determines that a structure is apomorphic, and mechanical analysis determines the function it enhances, we can hypothesize that selection for that function produced the structure (Weishampel. 1995). Comparative morphological and behavioral studies can be used to support hypotheses of adaptation when similarly constructed modern animals are known to share a similar function. In the case of A. afarensis, there are numerous derived modifications of their skeletons that enhance terrestrial bipedal progression and resemble modern hominoid bipeds. Thus, we can hypothesize that these characters were shaped by selection for bipedal locomotor abilities. Testing hypotheses about the retention of primitive traits is more difficult (e.g., Frumhoff and Reeve, 1994). When a primitive trait remains unmodified in a taxon from the ancestral form, the null hypothesis is that it is selectively neutral, termed a secondary nonaptation (Lauder, 1996). To falsify the null hypothesis, one must be able to show that stabilizing selection or negative selection was at work. To demonstrate stabilizing selection for a primitive trait, one would need to show that the primitive structure compromised a derived function. In the case of Australopithecus, this would be whether its primitive traits compromised bipedality. Other lines of evidence might be invoked as weaker support for the hypothesis of stabilizing selection. If a trait was retained in its unmodified form for a substantial period of time (stasis), it is possible, but not certain, that stabilizing selection was involved. If one could demonstrate that an animal actually engaged in behaviors for which the primitive traits had been adaptations, it might suggest continued adaptive value of these traits, but again this would not be definitive. Even if these latter two criteria (stasis or behavior) were met, they would still not falsify the null hypothesis of nonaptation. They merely would be suggestive. If a trait was selectively neutral, variation should increase over time (e.g., Tague, 1997, 2002a). However, the lack of increased variability still might not signal stabilizing selection. Even if a trait was neutral with respect to selection, a trait could be pleiotropically or structurally linked to other traits that were under selective influence. Its maintenance could reflect lack of selection to change a developmental cascade influencing other structures that are themselves being maintained by stabilizing selection (discussions in Maynard Smith et al., 1985; Churchill, 1996). A final hypothesis for the retention of primitive traits could be that they were actively being selected against, but that selection had not had sufficient time to reduce or eliminate them. This hypothesis of disaptation would require that the trait would diminish or change over time, and not persist unaltered. A similar conundrum to the issue of arboreality in A. afarensis lies in the interpretation of the rudimentary forelimbs in Tyrannosaurus rex (Carpenter and Smith, 2001). Here also is a case where primitive features, in this case forelimbs, have been retained but altered in the light of a strong directional signal towards a new behavior, terrestrial bipedality. So, are these tiny forelimbs still present because they serve an adaptive function for some altered task, or are they retained for no reason? Some of the debate over their function takes the form of logical arguments, which are not necessarily empirical (Osborn, 1906; Newman, 1970; Paul, 1988). The most recent contribution uses forelimb bone robusticity and inferred muscle size and strength to argue that these are not useless, vestigial structures (Carpenter and Smith, 2001). Because both of these features are influenced by behavior, this evidence suggests that the forelimbs were still used for some purpose, which Carpenter and Smith (2001) hypothesize was manipulating food items during oral processing. This case is still different from that in A. afarensis, however, because A. afarensis had not lost the ability to use their primitive ape-like traits for their original function. It is, however, a useful example to explore how hypotheses about the functional significance of primitive traits, either adaptive and/or behavioral, can be tested. Unlike the Tyrannosaurus example, in which the question was if the structure was functional at all or merely vestigial, a more relevant example is the hindlimbs of the primitive whale Ambulocetus (Thewissen et al., 1994). The retention of functional hindlimbs in this taxon indicates that it could ambulate terrestrially, although there was a clear directional signal towards an aquatic lifestyle. The authors propose that the hindlimbs were propulsive in the water and on the land. The vector of morphological change clearly was towards improving swimming ability at the expense of terrestriality. When Ambulocetus was terrestrial, it would have been relatively awkward in its movements. Knowing that it could have moved about on land, and even that it may have done so, it is still difficult to assess the adaptive significance of terrestriality in these early whales. Again, here the fossil record reveals the

7 Ward] INTERPRETING A. AFARENSIS LOCOMOTION 191 Fig. 1. Molar shearing quotients in extant and early Miocene hominoids. Early Miocene apes have lower crests, but microwear analyses demonstrate that they had a similar range of diets to extant taxa. This illustrates a disjunction between behavior and anatomy in living and fossil forms. Similar effects could be operating in the postcranial skeleton, but may be more difficult to measure. Figure modified from Kay and Ungar(1997). vector of morphological change, but reconstructing the pattern of selection that produced this change relies on scientific inference. We also need to be careful not to assume immediately that because Australopithecus did not look exactly like humans that its behavior was necessarily different. The retention of primitive traits does not always mean that these traits were maintained for their original function. Examples of discrepancies in form-function relationships between extant and fossil taxa can and have been demonstrated between tooth morphology and diet, for which evidence of genetically based morphology (tooth form) can be compared to evidence of behavior (microwear and bone isotope analysis). One example of this phenomenon was found in the teeth of late Miocene/early Pliocene horses from Florida (MacFadden et al., 1999). All six species examined in this study retained inherited high-crowned molars, and based on tooth morphology would have been interpreted as grazers of abrasive C4 grasses. However, microwear and isotope data suggest that they had a range of diets from browsing to grazing on many types of C3 and C4 plants. Some species (particularly Dinohippus mexicanus and Astrohippus stockii) were almost exclusively browsers. Hypsodonty appears to have been retained, perhaps because of some sort of phylogenetic and/or developmental constraint, although behavior had changed. Another example comes from the primate literature. Kay and Ungar (1997) showed that although Miocene apes had a similar range of molar shearing crest lengths as do roughly the same taxonomic diversity of extant species, the molars of all early Miocene taxa were all more bunodont than those of extant apes (Fig. 1). These data seem to suggest that no Miocene apes were as folivorous as the most folivorous extant apes, i.e., gorillas and siamangs. This may be incorrect, as Kay and Ungar (1997) argued using microwear data. Their microwear

8 192 YEARBOOK OF PHYSICAL ANTHROPOLOGY [Vol. 45, 2002 analyses showed that, in fact, the Miocene taxa were eating the same range of diets seen in extant apes, and that even species such as Rangwapithecus, which had teeth with gross morphology most like those of modern frugivores, were eating as many leaves as do modern folivores. In effect, tooth morphology improved over time. This may well have happened in the evolution of hominin, and hominoid, postcranial anatomy, and it may be possible to uncover the extent to which it may have occurred by comparing traits with greater or less epigenetic variability resulting from individual behaviors. It underscores the difficulty of interpreting actual behaviors in fossil taxa. Hypotheses about behavior On a practical level, precise interpretation of locomotor repertoires may exceed the limit of resolution possible in the fossil record. For example, humans in many forager groups like the Yanamamo (Chagnon, 1997) and Achuar climb trees for fruit and other food resources (Descola, 1996a,b). This behavior is not reflected in genetically determined skeletal morphology, because anatomical variation that improves arboreal abilities apparently has not affected their fitness. We would not be able to see the arboreality in their skeletons, just as we cannot see behaviors such as lying down or sitting, because an individual s anatomy does not affect his fitness by allowing him to be more or less effective at these behaviors (Latimer, 1991). There is no selective disadvantage of having anatomy that is poorly designed for such tasks. This makes interpreting actual behaviors in extinct species particularly tricky. It may be possible at some level, but will always remain unsatisfying at another. If A. afarensis climbed trees, we have to be clever enough to find ways to read this in their morphology, potentially using ontogenetically labile characters that can be influenced by individual behaviors over the course of a lifetime. Still, it is worth the attempt, so that we can better understand what our extinct relatives were like. The simple presence of primitive features in the A. afarensis skeleton is not enough to identify the adaptive value of tree climbing. For example, a future paleoanthropologist studying skeletons of today s humans could argue that because we retained a grasping hand and mobile shoulder joint, features that evolved for an arboreal lifestyle, modern humans are partly arboreal. None of us, however, would argue that these traits were maintained by selection on humans primarily for the ability to climb trees well. Instead, these traits primarily represent exaptations for manipulative or throwing abilities, although they also are used in climbing rocks or trees. Behavior of an extinct animal can be examined by considering morphological features that are shaped by its activity pattern. One of the best examples of an environmentally determined trait is dental microwear, the pits and scratches on teeth caused by diet. Given a constant pattern of enamel microstructure and fracture behavior, which appears to be the case for all hominoids, microwear reveals individual activity, but not adaptive history (Teaford, 1988, 1994). In contrast, because tooth shape appears to be conservative within taxa, it is generally assumed to be tightly genetically controlled, and not influenced by an individual s diet (although fetal environment and nutrition may be involved in tooth formation, diet appears not to be an influence). Variations in molar shearing crest length among taxa, for example, are thus thought to reflect natural selection on crest length in response to diet (Kay and Hiiemae, 1974; Kay, 1975; Rosenberger and Kinzey, 1976; Strait, 1993a, b; Teaford, 2000). Tooth shape and features such as molar shearing crest length reflect a history of selection, whereas microwear reflects behavior. Unfortunately, postcranial anatomy has few, if any, such dichotomies. Examples of epigenetically plastic traits are physeal plate angles (Tardieu and Trinkaus, 1994; Duren, 1999) and diaphyseal crosssectional form (summarized in Martin et al., 1998; Ruff, 2000), and may also include long bone torsion (Martin and Saller, 1959; King et al., 1969; Sarmiento, 1985; Pieper, 1988) and perhaps even phalangeal curvature (Richmond, 1998) or aspects of joint conformation (Frost, 1979, 1994; Hamrick, 1999; but see Lieberman and Pearson, 2001). By carefully studying the influences of activity on bone form, either directly or via soft tissues, we will be better able to use morphological information reliably to reconstruct behavior. By identifying the developmental plasticity of morphology, we are better armed to reconstruct the behavioral repertoires of fossils, and to avoid misinterpreting characters when constructing phylogenies with morphological data. REEXAMINATION OF THE FOSSIL EVIDENCE USING THIS FRAMEWORK Armed with a more explicit identification of the questions we are asking about Australopithecus, we can begin to reevaluate the fossil record for early hominin locomotion using this empirical framework. For all phases of this process, it is important to elucidate the genetic and ontogenetic influences on morphology. Character polarities and vectors of morphological change in early hominins Reconstructing character transformations within hominoid lineages is notoriously difficult (Larson, 1998; Sarmiento, 1998; Begun, 1999; Richmond et al., 2001). Despite studies arguing that postcranial data are more informative about phylogeny than are craniodental data (Collard and Wood, 2000), it is not the case that we can assume hominins evolved from an ancestor that was similar to extant great apes in

9 Ward] INTERPRETING A. AFARENSIS LOCOMOTION 193 every aspect of its postcranial anatomy. Recent attention to the amount of apparent homoplasy in the hominin skeleton (e.g., Larson, 1996, 1988; Begun et al., 1997a, b; Rose, 1997; Ward CV, 1997; Lieberman, 1999; Coffing and McHenry, 2000; Collard and Wood, 2000; MacLatchy et al., 2000) highlights our need to remain cautious in making polarity determinations, especially based on extant hominoid data alone. Elucidating the genetic and ontogenetic influences on morphology will aid in the understanding of its phylogenetic signals. Until we better understand these effects, we can continue to invoke parsimony arguments if we do cautiously, understanding their inherent limitations. We can be most effective only by including all available data in constructing the phylogenies upon which we base our polarity determinations. Character selection determines our ability to make accurate phylogenetic hypotheses. This requires careful choices of features, and knowledge about the individual variability, phenotypic plasticity, and heritability of traits (e.g., Cartmill, 1994; Weishampel, 1995; Lieberman, 1997, 1999), as well about genetic and ontogenetic links among them. As the determinants of adult form become clearer, and we can identify behavioral or functional influences on morphology, we can use these inferences to make hypotheses about the behaviors of A. afarensis and its ancestors, allowing us to reconstruct the history of selection that led to bipedality (Richmond and Strait, 2000; Richmond et al., 2001 and references therein). Functional anatomy is not unrelated to phylogenetic reconstruction, as it both informs character selection and provides a venue in which to test phylogenetic hypotheses (Brooks and MacLennan, 1991; Harvey and Pagel, 1991; Weishampel, 1995; Witmer, 1995; Begun et al., 1997a; Ward et al., 1997). Functional and phylogenetic interpretations are codependent. Thus, our phylogenetic reconstructions are critical to reconstructing the vectors of change. Based on extant hominoid taxa alone, our perception of character transformation would be relatively straightforward. All four great ape species share a suite of morphologies that enhance their abilities to engage in vertical climbing and below-branch arboreal activities, given their relatively large body sizes. These include a strongly grasping hallux and pollex, a restructured torso involving laterally facing shoulder joints, a high intermembral index, long, curved phalanges, strong manual and pedal grasps, short lumbar vertebral columns, and craniocaudally elongate pelves. Modern great apes differ in comparatively minor ways. With no further information, it would seem clear that these shared features were indisputably homologous, and that hominins evolved from an ancestor that was postcranially much like extant great apes. However, extant hominoids are relics from an earlier and more diverse age. The vast majority of hominoid taxa that once existed are extinct. Only by including them to reconstruct character transformation through hominoids can we come close to uncovering actual patterns of change in hominoid evolution. By incorporating Miocene ape fossil data into our phylogenetic schemes, we may alter the balance of parsimony about any number of characteristics (e.g. Begun, 1994; Begun et al., 1997a; Rose, 1997; Ward CV, 1997). Determining the phylogenetic relations among fossil apes is problematic, of course. Furthermore, we can never hope to recover the diversity of hominoid species that must have existed. Nonetheless, without considering what we do know and making careful hypotheses about the relationships we have identified so far, we neglect vital information about the pattern of evolutionary history that produced living hominoids. This information may alter our perceptions of which hominin features are, in fact, derived, and which are primitive. A recently published phylogeny of hominoids based on 240 characters of craniofacial, dental, and postcranial skeletons is the most robust phylogeny of these taxa constructed to date (Fig. 2) (Begun et al., 1997a). Using this as a baseline, we can examine the likely polarities of the ape-like traits of A. afarensis. Not all features are known from the euhominoid (the clade including living hominoids; see Fig. 2) fossil record, such as proportions within and among pedal digits, or overall foot length. Still, the approach will allow us to test hypotheses of adaptation, stasis, and disaptation for most characters. One impact of fossil data on interpreting character transformations leading to hominins has been in the realization that extant great apes postcranial anatomy may not all be homologous. Sivapithecus is widely considered to be the sister taxon of Pongo (review in Ward S, 1997). Because humeral torsion is presumed to reflect shoulder joint orientation (Evans and Krahl, 1945; Le Gros Clark and Thomas, 1951; Martin and Saller, 1959; Napier and Davis, 1959; Larson, 1996, 1998; Pilbeam et al., 1990; Churchill, 1996), and therefore reorganization of torso structure (Benton, 1965, 1976; Ward, 1993; Ward et al., 1993; Churchill, 1996; but see Chan, 1997), this character has particular importance (although it can be epigenetically modified by behavior pattern; Martin and Saller, 1959; King et al., 1969; Sarmiento, 1985; Pieper, 1998). The large Sivapithecus parvada humerus from Sethi Nagri (GSP 30754) has less torsion than any extant hominoid, instead more closely resembling pronograde quadrupeds (Larson, 1996, 1998; Pilbeam et al., 1990; Madar, 1994; Ward S, 1997). 2 Furthermore, the Sivapithecus postcrania also reveal that members of this genus had less well-developed halluces than do extant orangutans or African apes, and shorter, less curved 2 Moyà-Solà and Köhler (1996) reconstructed the smaller Sivapithecus humerus from a cast, and inferred a different functional pattern. If their conclusion is supported by further analyses, this might indicate locomotor diversity within the genus.

10 194 YEARBOOK OF PHYSICAL ANTHROPOLOGY [Vol. 45, 2002 Fig. 2. Phylogenetic hypothesis of living and fossil hominoid relationships modified from Begun et al. (1997b). Diagram is a consensus tree based on 240 cranial and postcranial characters, and so is the most comprehensiveanalysis of these materials published to date. Euhominoids refer to clade including all living apes, and basal hominoids to earlier Miocene taxa. Kenyapithecus, Equatorius, and Equatorius were still considered one genus when the analysis was performed, but would all remain basal hominoids if considered separately. phalanges (Rose, 1986, 1997; Spoor et al., 1991; Begun, 1993). Their inferred torso structure, which reflects a ventrally oriented scapular glenoid fossae and an emphasis on flexion-extension abilities of the proximal forelimb at the expense of abduction-adduction movements, and appendicular characters perhaps indicating less well-developed manual and pedal grasping ability, are more like those of basal hominoids such as Proconsul (summarized in Ward S, 1997). These Sivapithecus postcranial fossils imply that either Pongo is convergent upon African apes in many of its more specialized climbing features, or that the many craniofacial similarities of Pongo and Sivapithecus are homoplasies (Pilbeam et al., 1990; Ward S, 1997). Either of these hypotheses is plausible, but they are mutually exclusive. The possibility of Asian and African great apes sharing postcranial homoplasies linked with below-branch arboreality is not unreasonable, given the apparent independent acquisition of some of these morphologies in Morotopithecus and atelines (Walker and Rose, 1968; Ward CV, 1993, 1997; Sanders and Bodenbender, 1994; MacLatchy, 1996; Rose, 1997; MacLatchy et al., 2000), suggesting a similar adaptive response in closely related organisms subjected to similar selective pressures (e.g., Jolly, 2001). The likelihood of independent acquisition of the Pongo and Sivapithecus craniofacial similarities is less certain at this point (Ward S, 1997). Hominins also share features with fossil basal hominoids (Begun and Kordos, 1997; Begun et al., 1997a,b) such as Sivapithecus, Proconsul, and/or Equatorius, and with Hylobates, that are not found among extant great apes, such as shorter pelves with longer lumbar vertebral columns, less curved forelimb bones, longer thumbs, shorter medial rays of the hand and foot, and less humeral torsion (e.g., Larson, 1996, 1998; Rose, 1997; Ward CV, 1997; MacLatchy et al., 2000). Foot length is not known from the fossil record after Proconsul, whose feet were not as long as those of extant apes (Walker and Teaford, 1988). Either hominins have reverted to the primitive condition in these features, or extant apes are homoplastic in these ways. Either scenario involves substantial homoplasy, making polarity determinations among hominoids difficult.

11 Ward] The likely Asian-African ape homoplasies open the possibility that chimps and gorillas may also be homoplastic in some features. This hypothesis is supported by some researchers studying kinematic and anatomical correlates of knuckle-walking in chimps and gorillas (Inouye, 1994; Dainton and Macho, 1999). If so, some of these hominin characters may be primitive, which would indicate less change from the last common ancestor than we now suppose. Recently, however, strong arguments were made in favor of the hypothesis that the last common ancestor was a knuckle-walker, and that African ape postcranial similarities are indeed homologous (Richmond et al., 2001). This perspective is the most widely accepted at present. Further research into the function and evolution of the forelimb is needed to continue to test these competing hypotheses. Most A. afarensis traits interpreted as primitive, based on comparison with extant taxa (McHenry, 1994; Stern, 2000), still appear to be primitive even when the fossil evidence is considered. A long pisiform, dorsoplantarly narrow navicular, and metatarsals lacking expanded dorsal margins of their head are known for Oreopithecus (Sarmiento, 1987) and extant apes. Australopithecus lacks the expanded vertebral bodies of humans, resembling all taxa for which there are data (e.g., McHenry, 1992): Proconsul (Napier and Davis, 1959; Walker and Pickford, 1983; Walker and Teaford, 1988; Ward et al., 1993), Nacholapithecus (Rose et al., 1996a,b; Nakatsukasa et al., 1998, 2000), Equatorius (Le Gros Clark and Leakey, 1951), Morotopithecus (Walker and Rose, 1968; Gebo et al., 1997; Mac- Latchy et al., 2000), Oreopithecus (Harrison, 1986; Sarmiento, 1987), Dryopithecus (Moyà-Solà and Köhler, 1996), and extant primates. Within the hand, a pronounced third metacarpal styloid process, found only in Homo and some gorillas, is absent in Proconsul (Napier and Davis, 1959) and Oreopithecus (Moyà-Solà et al., 1999). Also, the first metacarpal base highly is concavoconvex in Proconsul (personal observation), Sivapithecus (Spoor et al., 1991), Oreopithecus (Moyà-Solà et al., 1999), and extant apes. Small apical tufts are found on the distal phalanges of Afropithecus (Leakey et al., 1988), Proconsul (Napier and Davis, 1959; Begun, 1994), Nacholapithecus (Nakatsukasa et al., 1998), Oreopithecus (Moyà-Solà et al., 1999), Dryopithecus (Moyà-Solà and Köhler, 1996), and extant primates. The mediolaterally broad knee joint, typical of extant apes and seen to a lesser extant in A. afarensis (Tardieu, 1979, 1981), is also known for Proconsul (Walker and Pickford, 1983) and Oreopithecus (Harrison, 1986; Sarmiento, 1987). Many primitive traits, however, are derived toward a human-like condition from the ape-like one. Long, curved phalanges with pronounced flexor ridges are found in Oreopithecus (Harrison, 1986; Sarmiento, 1987; Moyà-Solà et al., 1999), Dryopithecus (Begun, 1992, 1993; Moyà-Solà and Köhler, INTERPRETING A. AFARENSIS LOCOMOTION ), Orrorin (Senut et al., 2001), and Ardipithecus (Haile-Selassie, 2001), and in extant apes, although less so in early hominoid taxa (Begun, 1994; Rose, 1997). Australopithecus appears to have had reduced manual and pedal phalanges that were relatively shorter than in any primate except Homo (Bush et al., 1982; White, 1994). 3 The hamate hamulus is only somewhat distally directed, intermediate between extant apes and humans (Sarmiento, 1994, 1998; Ward et al., 1999b). The large hamulus appears to be found only in African apes and hominins, and is not present in Sivapithecus (Spoor et al., 1991) or Proconsul (Beard et al., 1986). Hamulus form likely reflects morphology of the carpal tunnel rather than flexor carpi ulnaris function, as previously suggested (Sarmiento, 1998; Ward et al., 1999b). However, the pisohamate ligament inserts not on the distal end of the hamulus, but at its base, at least in chimpanzees and humans (personal observation). Thus, hamulus size and orientation appear not to reflect the flexor carpi ulnaris muscle or pisohamate ligament, but the palmar carpal ligament (flexor retinaculum), which forms the roof of the carpal tunnel. It is certainly possible that carpal tunnel length is related to overall hand length, and if so, hamulus distal projection may be related to hand length at least at some level. A. afarensis appears to have reduced the proximodistal length of its carpal tunnel relative to African apes. The high intermembral index (calculated as (humerus radius/femur tibia) * 100) of African apes appears to have been present in at least Oreopithecus (Hürzeler, 1960; Moyà-Solà and Köhler, 1996) and Dryopithecus (Moyà-Solà and Köhler, 1996), although not in Proconsul (Walker and Pickford, 1983; Walker and Teaford, 1988). Although Dryopithecus was interpreted to have an intermembral index higher than that of African apes and only like orangutans among extant species (Moyà-Solà and Köhler, 1996), it may not have been as extreme, as these data were based on extensive segment length reconstructions. Direct comparison of the reported ulna to femur lengths published with the fossil announcement (Moyà-Solà and Köhler, 1996) is 102, close to the African ape range (chimpanzees, 87 97; gorillas, ), and considerably less than that of orangutans ( ). Oreopithecus, on the other hand, is intermediate between extant African and Asian apes, with an ulna/femur ratio of 111. Australopithecus had an ulna (length estimate based on Drapeau, 2001) to femur length of about 80, showing an apparent reduction in forelimb to hindlimb proportions. This is evident in comparisons of humeral to femoral proportions as well (Jungers, 1982, 1994; Wolpoff, 1983; White et al., 1993; White, 1994). 3 In this paper, the term Homo is used to include Homo erectus/ ergaster and more recent members of the genus Homo only, following the grade-based approach of Wood and Collard (1999).

12 196 YEARBOOK OF PHYSICAL ANTHROPOLOGY [Vol. 45, 2002 Within the forelimb, A. afarensis also appears to be primitive in arm, forearm, and metacarpal segment lengths (Drapeau, 2001). By including the A. afarensis partial skeletons AL and AL 438-1, along with extant cercopithecids, Dryopithecus, and Oreopithecus, she demonstrated that the slightly higher brachial index of Australopithecus compared with extant apes (see Kimbel et al., 1994) appears to represent the primitive large-hominoid condition. A. afarensis does not appear to have shortened its upper limbs from the primitive condition, nor lengthened them, showing less directional change than often assumed (Jungers, 1982, 1994; Wolpoff, 1983; White et al., 1993; White, 1994). There is thus no evidence that A. afarensis is derived in forelimb proportions, supporting the hypothesis that their lower limbs are elongated rather than their forelimbs reduced (Wolpoff, 1983). Instead, Pan and Pongo, and to a lesser extent Gorilla, seem to have independently elongated their forearms from the likely ancestral condition. Another important finding by Drapeau (2001) is that chimpanzees and bonobos have uniquely long metacarpals relative to their forearm lengths among extant and fossil hominoids. Australopithecus metacarpals are the same length relative to the forearms as in all extant hominoids except Pan. This discovery highlights the potential error in assuming that Pan reflects the morphology of the last common ancestor. A number of primitive characters of A. afarensis appear to reflect soft-tissue differences from humans, and suggest a somewhat more extant great ape-like muscular configuration in these early hominins. Some of the Australopithecus hip and thigh muscle attachment sites appear to have differed from those of Homo. Instead, they more closely resemble those of extant apes, and as best as can be determined, fossil taxa as well. First, there is a large, roughened ovoid area on the anterolateral border of the greater trochanter (Fig. 3), which appears to have been for attachment of the anterior fibers of gluteus minimus, which on great apes is sometimes a separate muscle called the scansorius. In humans, this muscle attaches on the anterosuperior portion of the trochanter, and does not extend distally. Häusler (2001) reconstructs the gluteal muscles of A. afarensis and A. africanus as essentially human-like, although with a slightly expanded anterior portion of the gluteus medius (contra Berge, 1994). These observations are consistent with the hypothesis of Robinson (1972) that the large anterior lesser gluteals could compensate by the slightly less sagittal orientation of the iliac blades. The medial thigh muscles of A. afarensis also appear to have differed in their distal insertions. In humans, the sartorius, gracilis, and semitendinosus have a common fan-shaped insertion called the pes ansirinus that largely blends with the crural fascia along the medial side of the proximal tibia. In A. Fig. 3. Proximal femurs and tibias of Homo sapiens, Australopithecus afarensis AL 288-1, and Pan troglodytes. Note similar location of presumed muscle attachments in A. afarensis and in chimpanzee. Line drawings modified from Aiello and Dean (1990). afarensis, however, these muscles had an extant ape-like conformation, with a discrete bony insertion along the medial margin of the tibial tuberosity, leaving a roughened pit in the bone (Fig. 3). Although there is no bony evidence of it, on the lateral side of the knee in chimpanzees and gorillas, the biceps femoris extends to the lateral side of the tibial tuberosity but blends with the crural fascia, with no evident bony insertion (personal observation). It is at least possible that this would have been the case in Australopithecus, given the ape-like medial side of the knee. The conformation of the A. afarensis ischial tuberosity also differs from that of humans, being longer, with the adductor magnus origin site set at an angle to the rest of the tuberosity. This different conformation may reflect other primitive musculature. Latimer and Lovejoy (1990b) suggest that the convex first metatarsal head of A. afarensis, unlike the flatter configuration in humans, may reflect more idiosyncratic loading in A. afaernsis than in Homo. They argue this may be due to the fact that the triceps surae may not yet have undergone the changes in muscle belly length and pennation seen in humans, and so relatively more plantarflexion was accomplished with the peroneal muscles. The large peroneal muscles are suggested by the large peroneal groove on the fibula and peroneal trochlea on the calcaneus (Latimer and Lovejoy, 1989, 1990b). This hypothesis is difficult to test, as muscles do not fossilize. Given the lack of capacity for