Brian G. RichmondDepartment of Anthropology,George WashingtonUniversity, 2110 G St., NW,Washington, DC 20052,

DC 20052, U.S.A. E-mail:bwood@gwu.edu

Early hominin limb proportions

Recent analyses and new fossil discoveries suggest that the evolutionof hominin limb length proportions is complex, with evolutionaryreversals and a decoupling of proportions within and between limbs.

to2

eofoexo

c

he

BOU-VP-12/1 diers significantly from both OH 62 and AL 288-1,but not from KNM-WT 15000. Published length estimates, ifReceived 4 May 2001Revision received29 July 2002 andaccepted 1 August 2002

Keywords: limb proportions,postcrania, humanevolution,paleoanthropology,Australopithecus, Homo.

correct, suggest that the relative forearm length of BOU-VP-12/1 isunique among hominins, exceeding those of the African apes andresembling the proportions in Pongo.Evidence that A. afarensis exhibited a less apelike upper:lower limb

design than A. africanus (and possibly H. habilis) suggests that, if A.afarensis is broadly ancestral to A. africanus, the latter did not simplyinherit primitive morphology associated with arboreality, but isderived in this regard. The fact that the limb proportions of OH 62(and possibly KNM-ER 3735) are no more humanlike than those ofAL 288-1 underscores the primitive body design of H. habilis.

2002 Elsevier Science Ltd. All rights reserved.

Journal of Human Evolution (2002) 43, 529548doi:10.1006/jhev.2002.0594Available online at http://www.idealibrary.com on

Introduction

Variations in fore and hindlimb proportionsin living primates are correlated with signifi-cant dierences in positional behavior(Jungers, 1985; Fleagle, 1999). As a result,limb proportions are important in eorts toreconstruct locomotion in fossil taxa,including hominins. They also have rel-evance for those interested in recoveringphylogenetic relationships. Many primateclades show considerable diversity in limbproportions. For example, sympatric calli-

trichines dier in intermembral proportionsin ways that relate to significant dierencesin substrate use within their arboreal envi-ronments (Garber & Leigh, 2001). Large-bodied papionins apparently independentlyevolved high intermembral indices, probablyas a result of parallel increased commit-ments to terrestrial locomotion (Lockwood,1999). Extant and subfossil lemur clades areamong the most diverse in terms of limbdesigns (Jungers et al., 2001).Like other primate clades, the picture

emerging from the hominin fossil record and

00472484/02/100529+20$35.00/0 2002 Elsevier Science Ltd. All rights reserved.U.S.A. E-mail:brich@gwu.edu

Leslie C. AielloDepartment of Anthropology,University CollegeLondon, Gower Street,London WC1E 6BT, U.K.E-mail: l.aiello@ucl.ac.uk

Bernard A. WoodDepartment of Anthropology,George WashingtonUniversity, 2110 G St., NW,Washington,

This study takes innot the limb prop(AL 288-1, OH 6considered to berandomization mbetween pairs of fall possible pairs oPan troglodytes, Pdierence in hum288-1 does not eWhen humerofemdierence betweenspecies. This, inlengths estimates,H. (or Australopittions than thoseo account intraspecific variation to test whether orrtions of four early hominin associated skeletons, BOU-VP-12/1, and KNM-WT 15000) can besignificantly dierent from one another. Exactthods were used to compare the dierencesssil skeletons to the dierences observed betweenindividuals within large samples of Gorilla gorilla,ngo pygmaeus, and Homo sapiens. Although therofemoral proportions between OH 62 and ALceed variation in the extant samples, it is rare.ral midshaft circumferences are compared, theOH 62 and AL 288-1 is fairly common in extantombination with error associated with the limbsuggests that it may be premature to considercus) habilis as having more apelike limb propor-in A. afarensis. The humerofemoral index of

530 . . ET AL.analytical advances is one of complexity inthe evolution of limb design. A simplemodel of hominin evolution, with fossilhominins with primitive (i.e., apelike) limbproportions being succeeded by those withmore modern humanlike limb proportions,is no longer tenable. The 1974 discovery(Johanson & Taieb, 1976) of the Australop-ithecus afarensis (or Praeanthropus afarensis)partial skeleton AL 288-1 led to the realiza-tion that the humerofemoral proportions ofA. afarensis are intermediate between thoseof modern humans and extant African apes.Specifically, compared to modern humans,AL 288-1 has a short femur relative to bodymass (Jungers, 1982) but apparently notrelative to stature (Hens et al., 2000).Debates continue over whether, or how,

short legs relative to body mass would influ-ence locomotor kinematics and eciency(Jungers, 1982; Rak, 1991; Steudel, 1996;Kramer, 1999).The earliest unambiguous evidence of

more modern humanlike limb proportionsin the fossil record comes from the remark-ably complete skeleton KNM-WT 15000,attributed to Homo ergaster (Figure 1), orearly African H. erectus (Brown et al., 1985;Wood, 1992; Walker & Leakey, 1993b).Like modern humans, this skeleton has rela-tively long femora and short forearms (Ru& Walker, 1993).The discovery of a fragmentary partial

skeleton, OH 62, attributed to H. habilis(Figure 1), complicated the picture(Johanson et al., 1987). The femur of this

Figure 1. Hominin phylogram. Evidence for limb proportions is considered for the taxa (and the Bouriremains) shown in bold.

531 skeleton is comparable in size to, or smallerthan, that of AL 288-1, while the humerusappears to be substantially longer than thatof AL 288-1. In other words, if the limblength estimates are correct, OH 62 hasmore apelike upper-to-lower limb propor-tions compared to AL 288-1. This suggestsan evolutionary reversal in limb proportionsfrom the earlier (Figure 1) and generallymore primitive taxon A. afarensis, to themore recent and in some ways more derivedH. [or Australopithecus (Wood & Collard,1999)] habilis (Hartwig-Scherer & Martin,1991). However, the error associated withestimating the OH 62 humerus lengthunfortunately stands in the way of makingsecure conclusions about the skeletonshumerofemoral length proportions (Korey,1990).There is an unfortunate lack of associated

skeletal remains of A. africanus from whichrelative limb bone lengths can be deter-mined. However, there are enough un-associated upper and lower limb fossilfragments to ascertain that in A. africanusvarious joints of the upper limb are largerrelative to lower limb joints than is the casein either modern humans or A. afarensis(McHenry & Berger, 1998a). Thus, unlessthe joint:limb length relations dier betweenthese taxa, indirect evidence suggests thatA. africanus (Figure 1) has more apelikeupper to lower limb length proportions thanthose of the earlier, and craniodentally moreprimitive, taxon A. afarensis (McHenry &Berger, 1998b).The most recent factor complicating the

pattern of the evolution of the homininpostcranial skeleton is the partial skeletonfrom the Bouri deposits, possibly represent-ing A. garhi. The Bouri skeleton (Figure 1)reportedly combines modern human-likehumerofemoral proportions with a relativelylong, apelike forearm (Asfaw et al., 1999).The large, isolated Omo L40-19 ulna(Howell & Wood, 1974; Aiello et al., 1999)from the Shungura Formation (ca. 23 Ma;Feibel et al., 1989) also suggests the pres-ence of a hominin with substantial body sizeor forearm length.Therefore, if the limb length estimates are

correct for these various associated skel-etons, they suggest that the evolution of limbproportions within the hominin clade ismore complex than previously anticipated.Specifically, a purely temporal sequence(Figure 1) suggests that intermediatehumerofemoral, and possibly brachial, pro-portions are seen by at least 32 Ma in AL288-1. This would be followed by moreapelike humerofemoral proportions in A.africanus (McHenry & Berger, 1998a), thenby humanlike humerofemoral but apelikeupper limb proportions in the Bouriskeleton. More apelike humerofemoralproportions would follow in Homo (orA.) habilis. Finally, modern humanlikehumerofemoral and upper limb proportionsoccur in Homo ergaster and persist withrelatively minor changes in subsequent taxa.Because of their rarity, the known associ-

ated skeletons have been treated as rep-resentative for their respective taxa. How-ever, as with other morphological features,limb proportions vary within species. Arethe dierences in limb proportions amongthe associated skeletons reviewed above sosubstantial that they are unlikely to haveoccurred in two members of the samespecies? In other words, could the dier-ences between these associated skeletons bemerely a product of random samplingfrom species with essentially the same limbproportions?Eckhardt (2000) has recently argued that

the dierences in the humerofemoral pro-portions of OH 62 and AL 288-1 might notbe excessive when compared to the variationseen between, and possibly within, modernhuman populations. Specifically, he foundthe dierence in humerofemoral indexbetween OH 62 and AL 288-1 to match thedierence between a Japanese samplereported by Shapiro (1939) and a rough

apes and humans. The comparative dataconsist of mixed subspecies samples of

532 . . ET AL.Gorilla gorilla (30 males, 23 females), Pantroglodytes (19 males, 16 females), Pongopygmaeus (16 males, 16 females), a speciessample of Pan paniscus (nine males, eightfemales), and a sample of deliberately mixedpopulations of H. sapiens (46 males, 42females, 14 unknown). Measurements werecollected on specimens housed at the follow-ing institutions: the Natural HistoryMuseum (London), the SmithsonianInstitutions National Museum of NaturalHistory (Washington, D.C.), the PowellCotton Museum (Birchington), the MuseeRoyal de lAfrique Centrale (Tervuren), andthe Raymond Dart Collection of TheUniversity of Witswatersrand. The fossilassociated skeletons examined in this studyinclude AL 288-1 (A. afarensis), OH 62(H. habilis), the Nariokotome skeletonaverage from an unspecified modern humansample referred to in Leakey & Lewin(1992). However, the data on the Japanesesample were collected as external measure-ments on living subjects rather than osteo-logical measurements. Thus, these data arenot directly comparable and additionalanalyses are warranted.The goal of this study is to consider

hominin limb proportions in the context ofquantitative information about variationwithin relevant comparative taxa. This studytests the null hypothesis that the dierencesobserved among these associated skeletonsare not significantly dierent from oneanother. Specifically, it examines both thepossibilities and probabilities that thedierences between pairs of these homininspecimens could be found within extanthominoid species.

Methods

Length and midshaft circumference weremeasured on the humerus, radius, femur,and tibia of comparative samples of greatKNM-WT 15000 (H. ergaster, or earlyAfrican H. erectus), and the Bouri skeletonBOU-VP-12/1 (possibly A. garhi). Two ofthe original fossils (OH 62 and KNM-WT15000) were examined by BW and LA.However, length estimates and measure-ments for the fossils were taken fromthe literature in most cases (Table 1). Themidshaft circumference measurements ofKNM-WT 15000 were measurementstaken by BR on a cast (housed at theSmithsonian Institutions National Museumof Natural History). Published measure-ments (Walker & Leakey, 1993a) of theoriginal were compared with measurementsmade by BR on a cast, and the dierenceswere less than 5% between the two sets ofmeasurements. Although there appears tobe slight age-related variation in limb pro-portions between the ages of 10 and 18years, Ru & Walker (1993) concluded thatlimb proportions of 1112-year-old boys arevery similar to those of 18-year-old boys.Given the similarity in proportions at theseages, and the evidence that the growth pat-tern of KNM-WT 15000 may be similar tothat of modern humans (Clegg & Aiello,1999), we do not attempt to make adjust-ments for the age of KNM-WT 15000 (seeRu & Walker, 1993).Four limb proportion indices were exam-

ined, the humerofemoral index (100humerus length/femur length), the brachialindex (100radius length/humerus length),the crural index (100tibia length/femurlength), and the humerofemoral circumfer-ence index (100humerus midshaftcircumference/femur midshaft circumfer-ence) (Aiello & Dean, 1990). Two humero-femoral and brachial index estimates werecalculated for BOU-VP-12/1, one usinglimb length estimates based on anatomicalgrounds and preferred by Asfaw et al.(1999), and a second using the limb lengthestimates derived from regressions (Asfawet al., 1999). Two brachial indices werealso computed for AL 288-1 and OH 62,

remains, we also compared measures of from Hartwig-Scherer & Martin, 1991).

2

i

tsfs,o

3nM1

533 humerus and femur size that could bedirectly taken on the fossils, namely mid-shaft circumferences. These are not proxiesfor length, but provide an independent andmore direct assessment of limb bone size. Asnoted, published midshaft circumferenceswere employed for AL 288-1 and OH 62(Hartwig-Scherer & Martin, 1991), andthose for KNM-WT 15000 were taken onthe NMNH cast. Because of the uncertaintyin identifying the precise location of mid-shaft without knowing the exact lengths ofthe OH 62 femur and humerus, circumfer-ence measurements were collected on a cast(courtesy of C. Lockwood, Institute forHuman Origins, Arizona State University)at 1 and 2 cm above and below the locationidentified as midshaft. The highest, and thusmost apelike, humerofemoral circumference

The lowest, most humanlike humerofemoralcircumference index was calculated by com-bining the smallest humeral circa-midshaftcircumference (520 mm, Hartwig-Scherer& Martin, 1991) with the largest femoralcirca-midshaft circumference (62 mm, 2 cmabove midshaft, C. Lockwood, personalcommunication). In this manner, weassessed the potential eect on humero-femoral circumference proportions of meas-uring circumference above or below the truelocation of the midshafts of OH 62.Exact randomization was used to compare

the dierences between pairs of fossil skel-etons with the dierences among all possiblepairings of individuals in the referencesamples (Grine et al., 1993; Richmond &Jungers, 1995; Lockwood et al., 1996). Thehuman and gorilla samples were limited tobased on the highest and lowest publishedestimates of the radius lengths of each speci-men (Table 1; Schmid, 1983; Hartwig-Scherer & Martin, 1991; Asfaw et al., 1999).Because of the problems involved in esti-

mating limb lengths from fragmentary

Table 1 Long bone dimensions (mm) of early hom

Measurement AL 288-1

Humerus length 237, 239*Humerus midshaft circumference 555 52Radius length 174, 215**Radius midshaft circumference 34Femur length 2805Femur midshaft circumference 67Tibia length

Dimensions of AL 288-1 and OH 62 taken from HartwAsfaw et al. (1999), and KNM-WT 15000 from Walker*From Jungers (1982). Used in computation of lowesBased on regression of African apes and humans (ABased on anatomical grounds, and preferred by AValue taken from Hartwig-Scherer & Martin (1991)

values from Hartwig-Scherer & Martin (1991) and fr(Lockwood, personal communication).Collected by BR on a cast.Based on the humanlike estimate from Schmid (198**Based on regression limited to chimpanzees and boConservatively low estimate by Hartwig-Scherer &Maximum estimate of Hartwig-Scherer & Martin (index was computed by combining the larg-est humeral circa-midshaft circumference(555 mm, taken 2 cm above midshaft, C.Lockwood, personal communication) withthe smallest in the range of values for fem-oral circa-midshaft circumference (60 mm,

inin associated skeletons

OH 62 BOU-VP-12/1 KNM-WT 15000

264 226, 236 3195 (520555) 5510, 246 231 255

38280 335, 348 432

60 (6062) 71380

g-Scherer &Martin (1991), those of BOU-VP-12/1 from& Leakey (1993a), unless indicated otherwise.Brachial Index estimate for AL 288-1.aw et al., 1999).faw et al. (1999).values in parentheses represent ranges of the combinedm measurements taken at, above and below midshaft

).obos (Asfaw et al., 1999).artin (1991).

991).

The dierence in limb proportionsbetween individuals was calculated as the

534 . . ET AL.arithmetic dierence in limb proportionsbetween the two fossils in question. Arith-metic dierence was employed because thelimb proportion indices are ratios and thusreflect shape (Jungers et al., 1995). Thesame computation was performed for eachextant species, with dierences computedfor every possible pair of individuals. Thus,the sample of 50 gorillas generated 1225unique pairwise comparisons. The final stepwas a comparison with the fossil pair dier-ences in order to assess the probability ofsampling such a dierence within a singlespecies.

Results

The dierences in limb proportions betweenthe great apes and modern humans foundhere are consistent with those reported inprevious studies (e.g. Schultz, 1930; Shea,1981; Jungers, 1985; Aiello & Dean, 1990;Lovejoy, 1993). Orang-utans exhibit thehighest humerofemoral index, followed bygorillas, common chimpanzees, bonobos,and finally modern humans (Figure 2). Thebivariate relationship between humerus andfemur lengths shows that intertaxonomicdierences are not due to size alone (Figure3; Jungers, 1985). Both the Nariokotomefifty individuals for the exact randomizationcalculations. This balanced-sex humansample comprises individuals of known sex,randomly selected from a larger dataset,from the following museum collections:Dart, University of the Witwatersrand,southern African blacks (n=12); CanadianMuseum of Civilization, Inuit (n=12);Terry, Smithsonian Institution, NorthAmerican blacks (n=12); and the NaturalHistory Museum (London), various British(n=14). The limited availability of thePan paniscus sample precluded the useof this species in the exact randomizationcomputations.and Bouri skeletons resemble the modernhuman sample in their humerofemoral pro-portions, and AL 288-1 is intermediatebetween Pan and Homo. If the publishedlength estimates are correct, OH 62 has arelatively longer humerus than AL 288-1and the other associated fossil homininskeleton considered here.The dierence in humerofemoral propor-

tions between the Nariokotome and Bouriskeletons is small when compared to thedierences observed between pairs of indi-viduals within any of the reference taxa(Figure 4; Table 2). The anatomicalgrounds length estimates (those preferredby Asfaw et al., 1999, as opposed to theirregression-based estimates; see above) forBouri result in humerofemoral proportionsthat are very similar to those of theNariokotome skeleton (Figure 2; Table 2).Limb length estimates for the Bouri skeletonbased on an African ape/human regressionyield a humerofemoral index less similar tothe Nariokotome skeleton (Figure 2; Table2). The dierence between Nariokotomeand this estimate of Bouri is observed in lessthan 5% of human (01%) and chimpanzee(36%) samples, and is rare in orang-utan(50%) and gorillas (71%) (Table 2). Thedegree of dierence between OH 62 and theBouri skeleton is never observed within anyof the comparative taxa, regardless of whichBouri length estimates are used (Table 2).The dierences between AL 288-1 and theNariokotome skeleton are significant(P

535 between AL 288-1 and OH 62 becomenonsignificant compared with the gorillasample. Thus, if the length of the AL288-1m (right humerus) is 239 mm(Jungers, 1982), as opposed to 237 mm(Johanson et al., 1982; Hartwig-Scherer &Martin, 1991), the humerofemoral indexdierence between these specimens are notsignificant compared to the variation inorang-utan (P>005) and gorilla (P>006)samples (human, P

io

Homo, and KNM-WT 15000 and BOU-VP 12/1 most closely resemble modern humans. The two values

536 . . ET AL.(Figure 2) in terms of their interspecificdistributions. Note also that intraspecificvariation in circumference proportions(Figure 5) is greater than the variation inlength-based proportions (Figure 2). Aswith the humerofemoral length proportions,orang-utans have the largest humerofemoralcircumference ratios, humans have thelowest ratios, and the African apes areintermediate (Figure 5). However, whilegorillas have relatively longer humeri thanchimpanzees and bonobos, they do not haverelatively larger humeral circumferences. Inthis measure, too, OH 62 is more apelikethan AL 288-1 (Figure 5). However, unlikethe results based on limb length estimates,the dierence in humerofemoral shaft cir-cumference between OH 62 and AL 288-1is easily matched in extant species (Figure 6;Table 3). This pattern holds when maxi-mum and minimum circumferences taken

near midshaft are combined to produceextreme ranges of humerofemoral circum-ference indices for OH 62 (Figure 5). It issomewhat surprising that the dierencesbetween the Nariokotome skeleton and AL288-1 are also not significant (Figure 6;Table 3). Thus, individuals with humero-femoral circumferences as dierent as thatobserved between AL 288-1 and KNM-WT15000 could readily be drawn from any ofthe extant species. Even the dierencesbetween OH 62 and KNM-WT 15000 canbe accommodated in 714% of pairwisecomparisons within the great ape andhuman samples (Figure 6; Table 3).The discrepancy between results in limb

length proportions and limb shaft circum-ference proportions due to the greaterintraspecific variability and interspecificoverlap in circumference proportions(compare Figures 2 and 5). The greater

shown for BOU-VP 12/1 represent the combination of humerus and femur length estimates (Asfaw et al.1999) resulting in high and low WF indices. Size alone does not account for the dierences in homininlimb proportions.Figure 3. Bivariate plot of humerus length and femExtant genera are fairly distinct from one anotherproportions of OH 62, if correct, closely resemble thur length in extant great apes and fossil hominins.n humerofemoral proportions. The humerofemoralse of Pan. AL 288-1 is intermediate between Pan and

537 variability is undoubtedly a product of cor-tical modeling and remodeling sensitivity toactivity that does not appear to influencelength (Lanyon, 1980; Trinkaus et al., 1994;Sumner & Andriacchi, 1996). Humeral andfemoral cross-sectional properties also varywith length during growth as a result of thedierent loading demands on the upper and

lower limbs (Ru et al., 1994; Sumner &Andriacchi, 1996). Periosteal surface mod-eling is most sensitive to mechanical activityprior to mid-adolescence, after which theendosteal surface becomes more sensitive(Ru et al., 1994). The femur:humerus ratioof cross-sectional properties reaches adultlevels after about 10 years of age (Sumner &

Figure 4. Dierences in humerofemoral (HF) index between fossil pairs and between all possible pairs ineach extant taxon. Note that dierence in HF proportions between OH 62 (estimate) and KNM-WT15000 is never observed in any extant sample, whereas the dierence between BOU-VP 12/1 (estimate)and KNM-WT 15000 is comparable to intraspecific dierences. The dierences between AL 288-1 andOH 62 (estimate), and between AL 288-1 and KNM-WT 15000, are rare but possible in most extantsamples. Given the uncertainties in the limb length estimates of OH 62, the HF index may or may not besignificantly dierent from that of AL 288-1.

The problems of working with estimatedlimb lengths from associated skeletons are

skeleton is never observed in any of theextant samples.

ierences in humerofemoral length indices

gtor

538 . . ET AL.even more pronounced when investigatingvariation in length proportions within theupper limb. Depending on the method cho-sen to reconstruct length, or the referencesample used in a regression analysis, thelimb length estimates for AL 288-1 (Table1) yield brachial indices that range betweenvalues that are low for modern humans tothose typical of chimpanzees (Figure 7). Infact, the range of estimates for AL 288-1 isgreater than the ranges of actual values foreach of the extant taxa (Figure 7). The rangein estimates for OH 62 is also very wide(Figure 7). The relative forearm length ofKNM-WT 15000 resembles that of gorillasand humans, and Ru & Walker (1993)

Length proportions within the lower limbcan only be computed for the Nariokotomeskeleton. The crural index of KNM-WT15000 is high. It is greater than thatobserved in gorillas and bonobos, falls at theupper range of the human and chimpanzeesamples, and best approximates the orang-utan values (Figure 8). However, Ru &Walker (1993) point out that the elongateddistal limb segments of the Nariokotomeskeleton resembles those seen in modernhumans from low latitudes. Apart fromslightly lower values in gorillas, there isvery little dierence among the African apesand humans in crural indices (Schultz,1930).Andriacchi, 1996), suggesting that theseage-related changes do not interfere withcomparisons between the adult homininskeletons and KNM-WT 15000, who wasabout 1114 years old at death (Ru &Walker, 1993). However, as a result ofgreater variability in diaphyseal dimensions,circumference proportions have far lesspower in discriminating species-leveldierences in skeletal form.

Table 2 Probabilities (%) of sampling pairwise d

Fossil pairsHomosapiens

AL 288-1/OH 62 00AL 288-1/BOU-VP-12/11 00AL 288-1/BOU-VP-12/12 00AL 288-1/WT 15000 00OH 62/BOU-VP-12/11 00OH 62/BOU-VP-12/12 00OH 62/WT 15000 00WT 15000/BOU-VP-12/11 209WT 15000/BOU-VP-12/12 01

Probabilities (%) of sampling, from extant hominoidthose observed between pairs of fossils (i.e., probabilityextant samples) * Denotes a probability of 5%, and

1Humerofemoral index (70) based on anatomical gr2Humerofemoral index (65) based on estimates of Afshow that the brachial index of the Narioko-tome skeleton most closely resembles peoplefrom warm climates. Interestingly, if thelength estimates for the Bouri skeleton arecorrect, then this skeleton is unique amonghominins in having a relative forearm lengththat can only reasonably be matched inextant orang-utans (Figure 7). The degreeof dierence in brachial proportionsbetween KNM-WT 15000 and the Bouri

Pantroglodytes

Gorillagorilla

Pongopygmaeus

25* 46* 41*00 03 0200 00 0014* 29* 14*00 00 0000 00 0000 00 00

426 452 47036* 71 50

roups, dierences in humerofemoral indices as great ashat dierences between fossil pairs could be drawn fromindicates a probability of 1%.unds estimates, preferred by Asfaw et al. (1999).ican ape and human regression (Asfaw et al., 1999).

Figure 5. Box plot of humerofemoral circumference (HFC) index (100humerus midshaftcircumference/femur midshaft circumference) in extant great apes and fossil hominins. With the exception

539 DiscussionCurrent data on associated skeletons suggestthat the evolution of hominin limb propor-tions is far from parsimonious. Despite dis-agreements on hominin phylogeny, mostpublished phylogenies (based on cranioden-tal evidence) agree that H. ergaster is moreclosely related to modern humans than areeither A. afarensis or H. habilis, and that H.habilis is more closely related to the H.ergasterH. sapiens clade than is A. afarensis(e.g., Chamberlain & Wood, 1987; Skelton& McHenry, 1992; Lieberman et al., 1996;Strait et al., 1997; Wood & Collard, 1999).If these phylogenetic hypotheses are correct,then the presumably apelike humerofemoralproportions of the Pan/Homo commonancestor are followed by intermediate pro-portions in A. afarensis, followed by more

apelike proportions in A. africanus (depend-ing on its phylogenetic position) and poss-ibly in H. habilis, and finally by humanlikelimb proportions in H. ergaster (Figure 9).Unless the Bouri skeleton samples the sistertaxon of the H. ergasterH. sapiens clade(Asfaw et al., 1999), then this skeletonserves as yet another example of potentialhomoplasy in hominin evolution. Only ifcurrent limb length estimates are grossly inerror could the dierences between OH 62and BOU-VP 12/1 be encompassed byintraspecific variation. The only phylo-genetic scenario involving no humerofemo-ral proportion reversals is one in which A.afarensis is more closely related to a BouriskeletonH. ergasterH. sapiens clade than itis to A. africanus. The position of H. habilisin this scenario depends on whether its limb

of the relatively low HFC index of gorillas, the intertaxonomic pattern mimics the pattern in humero-femoral (length) index. Note, however, that variation is greater within taxa, and greater overlap existsbetween taxa. Of the fossil hominin partial skeletons, OH 62 has the highest and most apelike HFproportions, followed by lower HF proportions in AL 288-1, and KNM-WT 15000. In each box plot, thesmall square represents the median value, the box represents the inter-quartile range (central 50% of thedata), the bars indicate the non-outlier range, and open circles denote outliers.

540 . . ET AL.proportions more closely resemble those ofA. afarensis or A. africanus. The scenariobecomes even more complex when intra-limb proportions are considered because thebrachial index of the Bouri skeleton does notfollow the humerofemoral index in beingmodern humanlike, and may represent areversal from a less extreme brachial index.The results presented here support con-

clusions by Korey (1990) and Eckhardt

(2000) that it may be premature to considerthat the species represented by OH 62 hassubstantially more apelike humerofemoralproportions than does A. afarensis. If limblength estimates are correct for these associ-ated fossil skeletons, then their dierencesare very close to what we would accept assignificant. When compared to the vari-ation in our skeletally-derived mixed-population human sample, humerofemoral

Figure 6. Dierences in humerofemoral circumference (HFC) index between fossil pairs and between allpossible pairs in each extant taxon. Note that, while not common, the dierence in HFC proportionsbetween OH 62 (estimate) and KNM-WT 15000 is matched in more than 5% of every extant sample.Pairs as dierent in HFC proportions as AL 288-1/KNM-WT 15000 and AL 288-1/OH 62 are easilyfound in each extant sample.

62 and AL 288-1, then there is a low prob-ability that these individuals could be drawn

unfortunately too fragmentary to estimatelimb lengths with confidence. However,

rolusmniti

541 from the same species. However, very smallalterations (mm or less) in limb length esti-mates result in profoundly dierent inter-pretations. Korey (1990) has shown that thedegree of error involved in these lengthestimates substantially exceeds that neededto conclude that OH 62 has a humerofemo-ral index that is no more apelike than that ofAL 288-1. Our results based on measure-ments of humerofemoral circumference,although far more conservative, support thenotion that AL 288-1 and OH 62 may notbe significantly dierent in humerofemoralproportions. Therefore, with the evidencecurrently at hand, it is premature to con-clude that AL 288-1 and OH 62 have sig-nificantly dierent limb proportions (Figure

Leakey et al. (1989) compared the sizes ofthe upper and lower limb elements (e.g.,joint surfaces) of KNM-ER 3735 with thoseof AL 288-1 and H. sapiens and P. troglo-dytes. In this comparison, the lower limbelements of KNM-ER 3735 were onlyslightly larger than those of AL 288-1, butthe upper limb elements of KNM-ER 3735were much larger than those of AL 288-1. Inother words, KNM-ER 3735 exhibits ahigher ratio of upper to lower limb elementsizes than that of AL 288-1 and thereforeappears more like chimpanzees in thisrespect (Leakey et al., 1989). Furtherresearch is needed to determine how muchmore apelike this skeleton is compared toAL 288-1, and thus to assess precisely howdierences between OH 62 and AL 288-1are highly unusual (Figure 4; Table 2).This result contrasts with Eckhardts(2000) conclusion that the dierencebetween OH 62 and AL 288-1 is most likelyless than the variation within a single mod-ern human sample. Therefore, if estimatesare correct for the humeri and femora of OH

Table 3 Probabilities (%) of sampling pairwise diindices

Fossil pairsHomosapiens

AL 288-1/OH 62 454AL 288-1/WT 15000 382OH 62/WT 15000 118OH 62-A/AL 288-1 132OH 62-A/WT 15000 24*OH 62-B/AL 288-1 876OH 62-B/WT 15000 298

Probabilities (%) of sampling, from extant hominoid gence indices as great as those observed between pairs of fcould be drawn from extant samples). Circumference va(1991). The largest and smallest circumference valuemidshaft were used to construct maximum (OH 62-A;humerofemoral circumference indices. Only in the codierence so great that there is less than a 5% probabilhuman sample. * Denotes a probability of 5%, and 9). Additional remains of A. afarensis and H.habilis are needed to gain a better assessmentabout intraspecific variation in these taxaand to test the hypothesis that H. habilis hadrelatively apelike limb proportions.It has been suggested that another partial

skeleton, KNM-ER 3735, may be attribu-table to H. habilis (Leakey et al., 1989). It is

erences in humerofemoral midshaft circumference

Pantroglodytes

Gorillagorilla

Pongopygmaeus

483 421 445417 347 374134 73 14083 150 15005 20* 50*

872 891 877265 342 303

oups, dierences in humerofemoral midshaft circumfer-ssils (i.e., probability that dierences between fossil pairses for OH 62 are taken from Hartwig-Scherer & Martinat midshaft and 1 cm and 2 cm above and belowost apelike) and minimum (OH 62-B; most humanlike)

trast between OH 62-A and KNM-WT 15000 is they of being randomly drawn from an extant great ape orndicates a probability of 1%.

ehominins. Of the fossil hominin partial skeletons, BOU-VP-12/1 has the highest brachial proportions,followed by a lower index in the other three partial skeletons. The range of values for BOU-VP-12/1 results

542 . . ET AL.KNM-ER 3735 aects the claim that H.habilis had significantly more apelike limbproportions than A. afarensis.Lovejoy (1993) raises the possibility that

the AL 333-109 humerus and AL 333-3femur fragments might belong to the sameindividual. If so, it would provide a secondindividual to compare with AL 288-1 andallow researchers to gain a better idea of thehumerofemoral proportions typical of A.afarensis. Aside from deriving from the samefossil-rich locality, there is no clear evidencethat AL 333-109 and AL 333-3 belong tothe same individual (Lovejoy et al.,1982a,b). These two specimens are unfortu-nately much more fragmentary than thehumeri and femora of OH 62 and AL 288-1,making length estimates of both bones evenless reliable. At present, then, these AL 333fossils do not oer sound evidence regarding

limb length proportions. Lovejoy (1993)also makes the provocative argument (notcontingent on the AL 333 fossils represent-ing one individual) that AL 288-1s inter-mediate humerofemoral index is morehumanlike than it appears due to negativeallometry in human humerofemoral propor-tions (i.e., a relatively high humerofemoralindex might be expected for smallerhumans). This suggestion was tentativelysupported by the appearance of slight nega-tive allometry in the humerofemoral index ofa large sample of Caucasian males. How-ever, the significance (i.e., whether the linediers significantly from isometry) of therelation was not reported and, unfortu-nately, AL 288-1 is so much smaller thanthe human sample that a great deal ofextrapolation was necessary. Furthermore, itwould be appropriate to include females in a

from the lowest and highest estimates of humerus lengths (see Table 1) provided by Asfaw et al. (1999).Both, including the lower brachial index obtained from their preferred anatomical grounds estimates,exceed brachial proportions typical of Pan and resembles Pongo. In each box plot, the small squarerepresents the median value, the box represents the interquartile range (central 50% of the data), the barsindicate the nonoutlier range, and open circles denote outliers.Figure 7. Box plot of brachial index (100radius l ngth/humerus length) in extant great apes and fossil

hh

and overlaps with Pan and Pongo. However, the high crural proportions in KNM-WT 15000 may beinfluenced by climatic adaptations and possibly pathology. Note the overall similarity in extant hominoid

543 regression analysis given the likelihood thatAL 288-1s gender was female (Johansonet al., 1982; Wood & Quinney, 1996; Tague& Lovejoy, 1998). Jungers (1982) foundnearly isometric relations for femur andhumerus lengths relative to body mass in alimited sample that included small- andlarge-bodied humans. The conclusion thatLucys femur is relatively short holds whencompared to a greater sample of pygmyhumans of comparable body mass (Jungers,1991).The associated skeletons AL 288-1 and

OH 62 are unfortunately too fragmentary tomake reliable length estimates for the radius(Asfaw et al., 1999), thus rendering anyestimates of the brachial indices equallyunreliable for these fossils. However, evenwith the most generous radius length esti-mates, no one has proposed that the brachial

indices of these skeletons would haveexceeded those of African apes (Figure 7).On the other hand, if the radius lengthestimates are correct for the Bouri skeleton,this individual has a brachial index substan-tially higher than any proposed for AL 288-1and OH 62. The long ulna Omo L40-19,dated to ca. 23 Ma (Feibel et al., 1989),may be further evidence of a hominin witheither high brachial proportions, or largebody mass, around this time (Aiello et al.,1999). If their chronological sequencereflects a simple ancestor-descendant rela-tionship, this would suggest a reversal inbrachial proportions in the Bouri skeleton(more apelike) from the A. afarensis con-dition, followed by a second reversal in H.ergaster (more humanlike). As noted in theoriginal description (Asfaw et al., 1999),the Bouri skeleton is especially intriguing

crural proportions compared to humerofemoral and brachial proportions. In each box plot, the smallsquare represents the median value, the box represents the interquartile range (central 50% of the data),the bars indicate the nonoutlier range, and open circles denote outliers.Figure 8. Box plot of crural index (100tibia lengt15000. The crural index of KNM-WT 15000 is at t/femur length) in extant hominoids and KNM-WTe upper range of the modern humans sampled here,

humanlike humerofemoral and apelike (Latimer & Lovejoy, 1990). Latimer (1991)

etiytidneosA9c

544 . . ET AL.(Pongo-like) brachial proportions, demon-strating that intra- and interlimb propor-tions can be dissociated, and in fossil formscan assume combinations not seen in extanthominids (Wood & Richmond, 2000).Determining whether limb proportions

are significantly dierent between homininspecies is of obvious importance to ourunderstanding of the evolutionary history ofhominin skeletal design. It also influencesour interpretations of locomotor adaptationswithin the hominin clade. To those whoview australopiths as partly arboreal, evi-dence of a higher humerofemoral index inH. habilis would suggest that this taxonwas even more committed than A. afarensisto arboreal postures and behaviors.Researchers that reconstruct early homininlocomotion without a significant arborealcomponent generally argue that primitively-retained arboreal characteristics are not

has oered the most explicit rationale. Heargued that, because derived, functionallyrelevant features are the product of naturalselection, they provide unambiguous evi-dence for functional behavior (see also Wardet al., 1997). Few would disagree with thisstatement. Within this framework, evidencethat H. habilis had more apelike limb pro-portions than that seen in A. afarensis wouldprovide what most researchers would con-sider as unambiguous evidence of arborealbehavior in the former because the derivedlimb proportions of H. habilis would bethought to be the product of natural selec-tion. This would be consistent with primi-tive arboreal features retained in the skel-eton of H. habilis (Susman & Stern, 1982).The relatively larger upper limbs of A. afri-canus compared to those of A. afarensis(McHenry & Berger, 1998a) suggest thatlarger upper limbs are derived. Therefore,because, if present length estimates arecorrect, the skeleton uniquely combines

Figure 9. Schematic diagram of limb proportions inand fossil taxa. Two schematic individuals, representaxon. Each fossil taxon (and Bouri) is represented bwhether or not these specimens are truly representaH. (A.) habilis (OH 62), Bouri (BOU-VP 12/1), anlimbs are not currently available for A. africanus; iA. africanus is provisionally given limb proportionslength, relative to body mass, does not dier amschematic representations were given equal humerushown relative to this standardized humerus length.dierences in crural proportions (Aiello & Dean, 1indices, with the exception of H. ergaster, for whichshown for illustrative purposes only and are not drawdirect evidence is available.informative, and instead base locomotorreconstruction on derived morphology

xtant taxa (P. torglodytes, far left; H. sapiens, far right)ng extremes in air sample, are shown for each extantan individual skeleton (it remains to be determinedve of their respective taxa): A. afarensis (AL 288-1),H. ergaster (KNM-WT 15000). Data on associatedlight of studies by McHenry & Berger (1998a,b),

quivalent to those of H. (A.) habilis. Since humerusng African apes and humans (Jungers, 1994), thelengths. Intermembral and brachial proportions arefrican apes and humans also show little interspecific90), so all schematics were given equivalent cruralrural proportions are known. Trunks and heads aren to scale. Lines are shown in grey where little or no

never observed in the extant samples. Theextremely long forearm of the Bouri skeleton

545 regardless of ones philosophy on the impor-tance of primitive retentions in reconstruct-ing function, relatively (compared to A. afa-rensis) large upper limbs in A. africanus (orH. habilis) provide strong evidence thatarboreal positional behaviors were import-ant components of the positional repertoire.The approach of interpreting behavior

from derived morphology is, however,dependent upon a reliable phylogenybecause it is necessary to determineforevery functional traitwhether the featuresare derived or primitive. For example, therelatively apelike limb size proportions of A.africanus would not be considered secondar-ily derived in alternative phylogenetic sce-narios, including one in which A. africanusdid not evolve from an A. afarensis-likeancestor, or if A. africanus retains the primi-tive condition from which A. afarensis isderived. Only in these phylogenetic sce-narios would the relatively apelike limb sizeproportions of A. africanus be consideredprimitive and thus ambiguous to some intheir relevance to functional reconstruc-tions. The approach of using only derivedtraits to interpret function is only as reliableas the phylogeny upon which it is based.The limb proportions of H. habilis also

bear on the issue of its taxonomy. No matterwhether this taxon had limb proportions likethat of A. afarensis, or was more apelike, H.habilis diered from later members of thegenus Homo in this fundamental measure ofskeletal design. Furthermore, if the Bouriskeleton represents an ancestor of Homo,then OH 62 either represents a reversal to amore apelike skeletal design, or OH 62 isnot as closely related to later Homo as is theolder Bouri material. If H. habilis is a sistertaxon to a clade comprising Bouri and laterHomo, and Bouri postcrania belong to A.garhi, then Homo would be paraphyletic.The primitive nature of the OH 62 limbproportions, and the large upper:lower limbsizes of KNM-ER 3735, support thehypothesis that H. habilis was more similargreatly exceeds estimates for AL 288-1 andappears to be derived relative to a moreAfrican apelike forearm.Very small changes in estimated limb

lengths (e.g., the 2 mm dierence in lengthsto the australopiths in limb design than tothe derived skeletal proportions and com-mitted bipedal locomotor pattern of laterHomo (i.e., members of the H. ergasterH.sapiens clade) (Wood & Collard, 1999;McHenry & Cong, 2000).

Conclusion

In comparison with intraspecific variation inextant great apes and humans, the human-like humerofemoral indices of KNM-WT15000 and the Bouri skeleton (BOU-VP-12/1) are significantly dierent from bothOH 62 and AL 288-1, but not from oneanother. Dierences in humerofemoral indi-ces as great as that between AL 288-1 andOH 62 are almost never observed inhumans, and are rare in extant great apes.Dierences in humerofemoral midshaft cir-cumferences between the two fossils arecommonly matched within extant species.However, intraspecific variation in midshaftdimensions is very high, and therefore veryconservative. In all but one fossil pair con-trast (KNM-WT 15000 vs. most apelikeestimate of OH 62), the degree of dier-ences in humerofemoral circumferenceindex are fairly commonly observed withinextant species.Estimates of the brachial indices of

AL 288-1 and OH 62 are too variable tocontribute much to our understanding ofhominin limb evolution. The brachial indexof KNM-WT 15000 resembles those ofmodern humans and gorillas, while that ofBOU-VP-12/1 (if estimates are correct) iscomparable to the high brachial index ofmodern orang-utans. The dierence inbrachial index between these skeletons is

Therefore, given current evidence, it is pre-mature to consider OH 62s humerofemoral

21, 439450.Hausler, M. & Schmid, P. (1995). Comparison of the

546 . . ET AL.length proportions as significantly dierent,or more apelike, than that of AL 288-1.If the upper:lower limb joint size compari-

sons of fossils attributed to H. habilis(Leakey et al., 1989) and A. africanus(McHenry & Berger, 1998a) prove to besignificantly more apelike than those of A.afarensis, it would indicate a reversal (inmost phylogenetic models) in the evolutionof limb joint size. Having derived traitsfunctionally related to arboreal locomotionwould lend strong support to the hypothesesthat A. africanus and H. habilis practicedarboreal positional behaviors in conjunctionwith bipedalism. The primitive limb length,circumference, and joint size proportions ofH. habilis (OH 62 and KNM-ER 3735)support the hypothesis that this taxon ismore similar in body design to the australo-piths than to later members of the genusHomo (and to the older Bouri skeleton),which may be relevant to decisions overwhether or not H. habilis belongs in thegenus Homo (Wood & Collard, 1999).

Acknowledgements

We thank Richard Thorington and LindaGordon of the Smithsonian InstitutionsNational Museum of Natural History, andthe curatorial sta at the other museumsvisited for their assistance and for provid-ing access to specimens in their care.We also thank C. Lockwood for cir-cumference measurements for OH 62. Thisof the AL 288-1m humerus estimated byJungers, 1982, and Johanson et al., 1982)make the dierence in humerofemoral indexbetween AL 288-1 and OH 62 insignificantcompared to the variation within extantgreat apes and humans. The error associatedwith estimating the length of the OH 62humerus alone (Korey, 1990) is muchgreater than that needed to make the dier-ences between these fossils insignificant.pelves of Sts 14 and AL 288-1: implications for birthand sexual dimorphism in australopithecines. J. hum.Evol. 29, 363383.

Hausler, M. & Schmid, P. (1997). Assessing the pelvisof AL 288-1: a reply to Wood and Quinney. J. hum.Evol. 32, 99102.

Hens, S. M., Konigsberg, L. W. & Jungers, W. L.(2000). Estimating stature in fossil hominids: whichpaper benefited from the comments andsuggestions of R. Bernstein, G. Blomquist,P. Garber, W. Jungers, M. Kowalewski,S. Leigh, C. Ru, and one anonymousreviewer. This research was supported byThe Henry Luce Foundation and by a grantfrom the Leverhulme Trust (LA).

References

Aiello, L. & Dean, C. (1990). An Introduction to HumanEvolutionary Anatomy. London: Academic Press.

Aiello, L. C., Wood, B. A., Key, C. & Lewis, M.(1999). Morphological and taxonomic anities ofthe Olduvai Ulna (OH 36). Am. J. phys. Anthrop.109, 89110.

Asfaw, B., White, T., Lovejoy, O., Latimer, B.,Simpson, S. & Suwa, G. (1999). Australopithecusgarhi: A new species of early hominid from Ethiopia.Science 284, 629635.

Brown, F. H., Harris, J. M., Leakey, R. E. & Walker,A. C. (1985). Early Homo erectus skeleton from WestLake Turkana. Nature 316, 788792.

Chamberlain, A. T. & Wood, B. A. (1987). Earlyhominid phylogeny. J. hum. Evol. 16, 119133.

Clegg, M. & Aiello, L. C. (1999). A comparison of theNariokotome Homo erectus with juveniles from amodern human population. Am. J. phys. Anthrop.110, 8193.

Eckhardt, R. B. (2000). Human Paleobiology.Cambridge: Cambridge University Press.

Feibel, C. S., Brown, F. H. & McDougall, I. (1989).Stratigraphic context of fossil hominids from theOmo group deposits: northern Turkana Basin,Kenya and Ethiopia. Am. J. phys. Anthrop. 78, 595622.

Fleagle, J. G. (1999). Primate Adaptation and Evolution.San Diego: Academic Press.

Garber, P. A. & Leigh, S. R. (2001). Patterns ofpositional behavior in mixed-species troops of Cal-limico goeldii, Saguinus labiatus, and Saguinus fuscicollisin northwestern Brazil. Am. J. Primatol. 54, 1731.

Grine, F. E., Demes, B., Jungers, W. L. & Cole, T. M.II. (1993). Taxonomic anity of the early Homocranium from Swartkrans, South Africa. Am. J. phys.Anthrop. 92, 411426.

Hartwig-Scherer, S. & Martin, R. D. (1991). WasLucy more human than her child? Observationson early hominid postcranial skeleton. J. hum. Evol.

regression model and reference sample to use? Leakey, R. E. F. & Lewin, R. (1992). Origins Reconsid-

547 J. hum. Evol. 38, 76784.Howell, F. C. & Wood, B. A. (1974). Early hominidulna from the Omo basin, Ethiopia. Nature 249,174176.

Johanson, D. C., Lovejoy, C. O., Kimbel, W. H.,White, T. D., Ward, S. C., Bush, M. E., Latimer,B. M. & Coppens, Y. (1982). Morphology of thePliocene partial hominid skeleton (AL 288-1) fromthe Hadar Formation, Ethiopia. Am. J. phys.Anthrop. 57, 403451.

Johanson, D. C. & Taieb, M. (1976). Plio-Pleistocenehominid discoveries in Hadar, Ethiopia. Nature 260,293297.

Johanson, D. C., Masao, F. T., Eck, G., White, T. D.,Walter, R. C., Kimbel, W. H., Asfaw, B., Manega,P., Ndessokia, P. & Suwa, G. (1987). New partialskeleton of Homo habilis from Olduvai Gorge,Tanzania. Nature 327, 205209.

Jungers, W. L. (1982). Lucys limbs: skeletal allometryand locomotion in Australopithecus afarensis. Nature297, 676678.

Jungers, W. L. (1985). Body size and scaling of limbproportions in primates. In (W. L. Jungers, Ed.) Sizeand Scaling in Primate Biology, pp. 345381. NewYork: Plenum.

Jungers, W. L. (1991). A pygmy perspective on bodysize and shape in Australopithecus afarensis (AL 288-1,Lucy). In (B. Senut & Y. Coppens, Eds) Origine(s)de la Bipedie Chez les Hominides, pp. 215224. Paris:Cahiers de Paleoanthropologie.

Jungers, W. L. (1994). Ape and hominid limb length.Nature 369, 194.

Jungers, W. L., Falsetti, A. B. & Wall, C. E. (1995).Shape, relative size, and size-adjustments in morpho-metrics. Yearb. phys. Anthrop. 38, 137161.

Jungers, W. L., Godfrey, L. R., Simons, E. L.,Wunderlich, R. E., Richmond, B. G., Chatrath, P. S.& Rakotosamimanana, B. (2001). Ecomorphologyand behavior of giant extinct lemurs fromMadagascar. In (J. M. Plavcan, R. F. Kay, C. P.Van Schaik & W. L. Jungers, Eds) ReconstructingBehavior in the Primate Fossil Record, pp. 371411.New York: Kluwer Academic/Plenum Publishers.

Korey, K. A. (1990). Deconstructing reconstruction:the OH 62 humerofemoral index. Am. J. phys.Anthrop. 83, 2533.

Kramer, P. A. (1999). Modelling the locomotor ener-getics of extinct hominids. J. exp. Biol. 202, 28072818.

Lanyon, L. E. (1980). The influence of function on thedevelopment of bone curvature. An experimentalstudy on the rat tibia. J. Zool., Lond. 192,457466.

Latimer, B. M. (1991). Locomotor adaptations inAustralopithecus afarensis: the issue of arboreality. In(Y. Coppens & B. Senut, Eds) Origine(s) de la Bipediechez les Hominides, pp. 169176. Paris: CNRS.

Latimer, B. M. & Lovejoy, C. O. (1990). Hallucaltarsometatarsal joint in Australopithecus afarensis. Am.J. phys. Anthrop. 82, 125133.ered: In Search of What Makes Us Human. New York:Doubleday.

Leakey, R. E. F., Walker, A., Ward, C. V. & Grausz,H. M. (1989). A partial skeleton of a gracile hominidfrom the Upper Burgi Member of the Koobi ForaFormation, East Lake Turkana, Kenya. In (G.Giacobini, Ed.)Hominidae, pp. 167173. Milan: JakaBooks.

Lieberman, D. E., Wood, B. A. & Pilbeam, D. R.(1996). Homoplasy and early Homo: an analysisof the evolutionary relationships of H. habilissensu stricto and H. rudolfensis. J. hum. Evol. 30,97120.

Lockwood, C. A. (1999). Homoplasy and adaptation inthe atelid postcranium. Am. J. phys. Anthrop. 108,459482.

Lockwood, C. A., Richmond, B. G., Jungers, W. L. &Kimbel, W. H. (1996). Randomization proceduresand sexual dimorphism in Australopithecus afarensis.J. hum. Evol. 31, 537548.

Lovejoy, C. O. (1993). Modeling human origins: arewe sexy because were smart, or smart because weresexy? In (D. T. Rasmussen, Ed.) The Origin andEvolution of Humans and Humanness, pp. 128.Boston: Jones and Bartlett Publisher.

Lovejoy, C. O., Johanson, D. C. & Coppens, Y.(1982a). Hominid lower limb bones recovered fromthe Hadar Formation: 19741977 collections. Am. J.phys. Anthrop. 57, 679700.

Lovejoy, C. O., Johanson, D. C. & Coppens, Y.(1982b). Hominid upper limb bones recovered fromthe Hadar formation: 19741977 collections. Am. J.phys. Anthrop. 57, 637649.

McHenry, H. M. & Berger, L. R. (1998a). Bodyproportions in Australopithecus afarensis and A. africa-nus and the origin of the genus Homo. J. hum. Evol.35, 122.

McHenry, H. M. & Berger, L. R. (1998b). Limblengths in Australopithecus and the origin of the genusHomo. South Afr. J. Sci. 94, 447450.

McHenry, H. M. & Cong, K. E. (2000). Australo-pithecus to Homo: transformations in body and mind.Ann. Rev. Anthrop. 29, 125146.

Rak, Y. (1991). Lucys pelvic anatomy: its role inbipedal gait. J. hum. Evol. 20, 283291.

Richmond, B. G. & Jungers, W. L. (1995). Size vari-ation and sexual dimorphism in Australopithecus afa-rensis and living hominoids. J. hum. Evol. 29, 229245.

Ru, C. B. &Walker, A. C. (1993). Body size and bodyshape. In (A. C. Walker & R. E. Leakey, Eds) TheNariokotome Homo erectus Skeleton, pp. 234265.Cambridge: Harvard University Press.

Ru, C. B., Walker, A. C. & Trinkaus, E. (1994).Postcranial robusticity in Homo. III: Ontogeny. Am.J. phys. Anthrop. 93, 3554.

Schmid, P. (1983). Eine Rekonstruktion des Skelettesvon A.L. 288-1 (Hadar) und deren konsequenzen.Folia Primatol. 40, 282306.

Schultz, A. H. (1930). The skeleton of the trunk andlimbs of higher primates. Hum. Biol. 2, 303438.

Shapiro, H. S. (1939).Migration and Environment. NewYork: Oxford University Press.

Shea, B. T. (1981). Relative growth of the limbs andtrunk in the African apes. Am. J. phys. Anthrop. 56,179201.

Skelton, R. R. & McHenry, H. M. (1992). Evolution-ary relationships among early hominids. J. hum. Evol.23, 309349.

Steudel, K. (1996). Limb morphology, bipedal gait,and the energetics of hominid locomotion. Am. J.phys. Anthrop. 99, 345355.

Strait, D. S., Grine, F. E. & Moniz, M. A. (1997). Areappraisal of early hominid phylogeny. J. hum. Evol.32, 1782.

Sumner, D. R. & Andriacchi, T. P. (1996). Adaptationto dierential loading: comparison of growth-relatedchanges in cross-sectional properties of the humanfemur and humerus. Bone 19, 121126.

Susman, R. L. & Stern, J. T. Jr. (1982). Functionalmorphology of Homo habilis. Science 217, 931934.

Tague, R. G. & Lovejoy, C. O. (1998). AL 288-1Lucy or Lucifer: gender confusion in thePliocene. J. hum. Evol. 35, 7594.

Trinkaus, E., Churchill, S. E. & Ru, C. B. (1994).Postcranial robusticity in Homo. II: bilateral asym-

metry and bone plasticity. Am. J. phys. Anthrop. 93,134.

Walker, A. C. & Leakey, R. E. F. (1993a). Thepostcranial bones. In (A. C. Walker & R. E. F.Leakey, Eds) The Nariokotome Homo erectusSkeleton, pp. 95160. Cambridge, MA: HarvardUniversity Press.

Walker, A. C. & Leakey, R. E. F. (Eds) (1993b). TheNariokotome Homo erectus Skeleton. Cambridge,MA: Harvard University Press.

Ward, C. V., Begun, D. R. & Rose, M. D. (1997).Function and phylogeny in Miocene hominoids. In(D. R. Begun, C. V. Ward & M. D. Rose, Eds)Function, Phylogeny, and Fossils: Miocene HominoidEvolution and Adaptations, pp. 112. New York:Plenum Press.

Wood, B. A. (1992). Origin and evolution of the genusHomo. Nature 355, 783790.

Wood, B. A. & Quinney, P. S. (1996). Assessing thepelvis of AL 288-1. J. hum. Evol. 31, 563568.

Wood, B. A. & Collard, M. C. (1999). The humangenus. Science 284, 6571.

Wood, B. A. & Richmond, B. G. (2000). Humanevolution: taxonomy and paleobiology. J. Anat. 196,1960.

548 . . ET AL.

Early hominin limb proportionsIntroductionFigure 1

MethodsTable 1

ResultsFigure 2Figure 3Figure 4Table 2Figure 5

DiscussionFigure 6Table 3Figure 7Figure 8Figure 9

ConclusionAcknowledgementsReferences