Journal of Human Evolution 46 (2004) 433–465

Embed Size (px)

Citation preview

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    1/33

    Body proportions of Homo habilis reviewed

    Martin Haeuslera,b,*, Henry M. McHenrya

    aDepartment of Anthropology, University of California, Davis, CA 95616, USAbAnthropologisches Institut und Museum, Universitaet Zuerich-Irchel, 8057 Zuerich, Switzerland

    Received 15 October 2002; accepted 21 January 2004

    Abstract

    The ratio of fore- to hindlimb size plays an important role in our understanding of human evolution. Although

    Homo habilis was relatively modern craniodentally, its body proportions are commonly believed to have been more

    apelike than in the earlier Australopithecus afarensis. The evidence for this, however, rests, on two fragmentary

    skeletons, OH 62 and KNM-ER 3735. The upper limb of the better-preserved OH 62 from Olduvai Gorge is long and

    slender, but its hindlimb is represented mainly by the proximal portion of a thin femur of uncertain length.

    The present analysis shows that upper-to-lower limb shaft proportions of both OH 62 and AL 288-1 ( A. afarensis)

    fall in the modern human range of variation, although OH 62 also falls inside that of chimpanzees due to their overlap

    in small individuals. Despite being more fragmentary, the larger-bodied KNM-ER 3735 lies outside the chimpanzeerange and close to the human mean. Because the differences between any of the three individuals are compatible with

    the range of variation seen in extant hominoid groups, it is not legitimate to infer more primitive upper-to-lower limb

    shaft proportions for either H. habilis or A. afarensis.

    Femur length of OH 62 can only be estimated by comparison. Its closest match in size and morphology is with the

    gracile OH 34 specimen, which therefore provides a better analogue for the reconstruction of OH 62 than the stocky AL

    288-1 femur that is traditionally used. OH 34s slender proportions are hardly due to abrasion, but match those of a

    modern human of that body-size, suggesting that the relative length of OH 62s leg may have been human-like. Brachial

    proportions, however, remained primitive. Long legs may imply long distance terrestrial travel. Perhaps this adaptation

    evolved early in the genus Homo, with H. habilis providing an early representative of this important change.

    2004 Elsevier Ltd. All rights reserved.

    Keywords: Early hominids; Limb length proportions; Humero-femoral index; Brachial index; Locomotion; Human evolution

    Introduction

    Homo habilis has been considered to be the

    earliest member of the genus to which modern

    humans belong (Leakey et al., 1964). Having a

    slightly larger brain size and smaller teeth com-

    pared to the australopithecines, it was thought to

    * Corresponding author. Tel.: +41-1-635-54-33;

    fax: +41-1-635-68-04

    E-mail addresses: [email protected] (M. Haeusler),

    [email protected] (H.M. McHenry).

    Journal of Human Evolution 46 (2004) 433465

    0047-2484/04/$ - see front matter 2004 Elsevier Ltd. All rights reserved.doi:10.1016/j.jhevol.2004.01.004

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    2/33

    be able to make the first stone tools. The validity

    and phylogenetic position of this taxon, however,

    and whether its hypodigm contains one or two

    species, have remained contentious since its dis-

    covery (e.g., Robinson, 1965, 1966, 1972; Stringer,

    1986; Wood, 1987, 1992; Lieberman et al., 1988;

    Kramer et al., 1995; Wood and Collard, 1999;

    Miller, 2000; Blumenschine et al., 2003; Tobias,2003). Nevertheless, phylogenetic analyses have

    repeatedly associated the cranial evidence of H.

    habilis, in particular that of the KNM-ER 1813

    subset, referred to as H. habilis sensu stricto, with

    later Homo (e.g., Wood, 1992; Lieberman et al.,

    1996; Strait et al., 1997). The many intermediate

    features of a recently discovered skull from

    Dmanisi, Georgia (Vekua et al., 2002), further

    strengthen the relationship between H. habilis and

    H. erectus.

    Important aspects of the postcranial mor-

    phology ofHomo habilis sensu stricto, however, do

    not appear to resemble later species of Homo

    (McHenry and Coffing, 2000, and references

    therein). Particularly conspicuous is the size of

    the forelimbs relative to the hindlimbs in the

    w1.8 million year old partial skeleton from

    Olduvai Gorge, Tanzania (OH 62; Johansonet al., 1987; Johanson, 1989). The humerus, radius,

    and ulna of this specimen seem to be long and

    relatively robust compared to the neck and shaft

    breadth of the femur (see Fig. 1). When the

    surviving portions of the OH 62 limbs are com-

    pared to the best preserved Australopithecus

    afarensis specimen, AL 288-1 (Lucy), H. habilis

    appears more primitive than its purported ancestor

    (Johanson et al., 1987; Hartwig-Scherer and

    Martin, 1991). Its reconstructed humerus length

    Fig. 1. The partial skeletons of AL 288-1 (A. afarensis), OH 62, and KNM-ER 3735 (H. habilis) superimposed on the drawings byJ. Gurche of the A. afarensis skeleton (see Berger et al., 1998). The skeletal drawings are arbitrarily of identical size, although actualbody size might have differed considerably between the three specimens. Note the longer arm skeleton and more gracile proximalfemur of OH 62 compared to AL 288-1.

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465434

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    3/33

    considerably exceeds that of AL 288-1, although

    humerus shaft diameters are roughly similar and

    femur shaft diameters are less than those of AL

    288-1. Johanson and colleagues (Johanson et al.,1987) speculated, therefore, that its humero-

    femoral ratio might have been 95% or more, as

    compared to values of 85% for AL 288-1, 72% for

    Homo, and 100% for Pan. In contrast to the

    humerus, which lacks its proximal and distal

    articular ends, only a small portion of the femoral

    neck and proximal shaft is preserved, and even less

    is known of the proximal tibia. Korey (1990)

    therefore cautioned that assuming a greater index

    than that of AL 288-1 is not warranted, and also

    that a value close to the human range cannot be

    excluded. Yet, a second H. habilis partial skeleton,

    KNM-ER 3735, despite being very fragmentary, is

    also reported to possess upper-to-lower limb pro-

    portions that possibly approximate those of

    chimpanzees (Leakey et al., 1989).

    This unexpected twist in hominid phylogeny is

    further complicated by the observation that fore-

    to-hindlimb joint size proportions seem to be

    more humanlike in the earlier and craniodentally

    more primitive A. afarensis than in the later and

    craniodentally more humanlike A. africanus and

    H. habilis (McHenry and Berger, 1998a).The primitive A. africanus-like body propor-

    tions of H. habilis were an important factor in

    Wood and Collards (1999) transferring this

    taxon to the australopithecines as Australopithecus

    habilis, thus concurring with other scholars who

    doubted that the craniodental features of some

    H. habilis specimens differed markedly from

    Australopithecus , especially from A. africanus (e.g.,

    Robinson, 1965; see also Blumenschine et al.,

    2003). At first glance, the recent discovery of

    2.5 million year old associated upper and lowerlimb bones contemporary with A. garhi with

    an apparently humanlike humero-femoral ratio,

    but apelike brachial length proportion, made it

    seem even more unlikely that H. habilis was an

    ancestor of later Homo (Asfaw et al., 1999). Their

    taxonomic association is, however, not established

    (DeGusta, 2003), but if they were conspecific with

    A. garhi, the marked megadontia of that species

    would raise further difficulties for it having been a

    direct ancestor of H. habilis, whose oldest known

    fossil is dated to only 0.2 million years later in time

    (Kimbel et al., 1997).

    The incongruencies between the cranial and

    postcranial data ofHomo habilis rests, on the otherhand, almost exclusively on the length estimation

    of OH 62s femur, using that of AL 288-1 as an

    analogue (Johanson et al., 1987), and on the

    comparison of KNM-ER 3735s shaft dimensions

    with that of a single modern human and chimpan-

    zee skeleton (Leakey et al., 1989). Yet, there are

    many striking morphological differences between

    OH 62 and AL 288-1. Not only is the OH 62

    humerus longer than that of AL 288-1, it is also

    slimmer, particularly distally, and even more so if

    adjusted for its greater length. The same is true

    for the ulna and most markedly for the femur

    (Johanson et al., 1987). The overall impression,

    therefore, is not of a robust skeleton, but of a

    surprisingly gracile and slender one compared to

    that of AL 288-1. As the craniodental characteris-

    tics of OH 62 are also described as differing

    considerably from those of AL 288-1 (Johanson

    et al., 1987), it is doubtful whether the AL 288-1

    femur is indeed a good model to estimate femur

    length in this specimen.

    Of all known early Pleistocene femora, the

    slender OH 34 specimen from Olduvai Gorge (Dayand Molleson, 1976) is the only one to show a

    remarkable morphological similarity to that of OH

    62. Day and Molleson (1976: 455) wrote in the

    abstract: their strange appearance resulted in

    their neglect for many years. This neglect has

    apparently continued.

    After re-analysing shaft-proportions of the OH

    62 and KNM-ER 3735 partial skeletons, the

    present study will explore the usefulness of the OH

    34 femur, instead of AL 288-1, as a basis to

    estimate OH 62s femur length. In this context,another look will be taken at the issue of possible

    abrasion of the OH 34 femur shaft.

    Material and methods

    Modern comparative sample

    One set of our modern comparative materials

    includes 139 modern humans with an emphasis on

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 435

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    4/33

    small-bodied individuals: 8 Neolithics from

    Switzerland (Anthropologisches Institut der

    Universitat Zurich and Universite de Geneve),

    6 Pygmies from Ituri (Universite de Geneve),7 Bushmen, and 6 Hottentots (University of the

    Witwatersrand, Johannesburg). The remainder

    is housed at the Anthropologisches Institut

    der Universitat Zurich (composed mainly of

    European, African, and Asian individuals) and the

    Hearst Museum of the University of Berkeley,

    California (prehistoric Californian Indians). The

    sample of 79 Pan troglodytes (28 males, 40 females,

    11 of unknown sex), 46 Gorilla gorilla (28 males, 18

    females), and 32 Pongo pygmaeus (15 males, 17

    females) comes from the Anthropologisches

    Institut der Universitat Zurich, which also includes

    the Adolph H. Schultz-Sammlung and the Boesch

    Collection of Ta-chimpanzees; the Zoologisches

    Museum Zurich; the Naturhistorisches Museum

    Basel; the Museum de lhistoire naturelle de

    Geneve; the University of the Witwatersrand,

    Johannesburg; and the Transvaal Museum,

    Pretoria.

    The chimpanzee sample consists of a large por-

    tion of captive specimens (39 versus 35 wild-

    collected and 5 of unknown origin). Regression

    analyses of the captive and wild-collected samplesshowed that, on average, captive chimpanzees

    have slightly larger cross-sections of the arm skel-

    eton relative to femur shaft dimensions than do

    wild animals. The lower boundaries of the ranges

    of variation are, however, almost identical in the

    two groups. Thus, for the purpose of this study,

    combining the two groups probably has no effect

    on the relative position of the fossil hominids.

    Additional data for long bone lengths of 36 Pan

    troglodytes, 10 Pan paniscus, 11 Gorilla gorilla, 47

    Pongo pygmaeus, and 194 Hylobates sp. werederived from the records of Adolph H. Schultz,

    Anthropologisches Institut der Universitat Zurich.

    A further 120 Homo sapiens, 74 Pan troglodytes (22

    males, 54 females), 17 Pan paniscus (9 males,

    8 females), 87 Gorilla (42 males, 45 females), and

    41 Pongo (17 males, 24 females) stem from

    McHenrys data set (McHenry, 1972, 1974, 1978,

    1992; McHenry and Corruccini, 1975, 1978;

    McHenry et al., 1976). Before combining these

    data sets, they were tested for equality of the slopes

    and intercepts of the regression lines with an

    analysis of covariance (see Sokal and Rohlf, 1995).

    Fossil material

    The OH 62 skeleton was studied by M.H. at the

    National Museum of Tanzania in Dar es Salaam

    and by H.M.M. at the Institute of Human Origins.

    It is dated to w1.8 Ma and includes several

    hundred skull fragments, parts of the right arm,

    the proximal left femur, and a right tibial tuber-

    osity (Fig. 1; Johanson et al., 1987). Measurements

    of the KNM-ER 3735 skeleton were taken by both

    authors at the National Museum of Kenya, Nari-

    obi. KNM-ER 3735 comes from the Upper Burgi

    Member of the Koobi Fora Formation and is

    dated to 1.88 to 1.91 Ma (Feibel et al., 1989),

    making it slightly older than OH 62. Besides more

    than 50 pieces of the skull and several short

    fragments of limb bones, this associated skeleton

    consists of a weathered partial sacrum, distal right

    humerus, proximal right radius, distal left (right,

    according to Leakey et al., 1989) femoral shaft, a

    tibial mid-shaft fragment, and two partial hand

    phalanges.

    Comparative fossil material consists of all avail-

    able East and South African Plio-Pleistocenehominid long bones that are attributed to

    A. anamensis, A. afarensis (including the AL 288-1

    partial skeleton), A. africanus, H. habilis,

    Paranthropus robustus, P. boisei, Homo sp., and

    H. erectus/ergaster (called henceforth simply

    H. erectus).

    The OH 34 femur

    The OH 34 femur is much better preserved than

    that of OH 62, but it, too, has an abraded head.The trochantera major and minor, as well as the

    condyles, are lacking, and the shaft is broken in the

    middle (Fig. 2; Day and Molleson, 1976). It was

    found by M. Kleindienst in situ in an ancient

    stream channel during the 1962 excavation at

    JK2 West, Olduvai Gorge, together with a frag-

    ment of a slender tibia midshaft (Kleindienst,

    1973). In addition, the femur is remarkably slender

    and possesses a sharp pilaster and an antero-

    posteriorly compressed femoral neck, which

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465436

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    5/33

    correspond well to OH 62 (due to its fragmentarycondition, only the beginning of the pilaster can be

    seen).

    Another peculiarity of the OH 34 femur is its

    huge foramen nutricium. Together with its slender-

    ness, this has raised speculations of a possible

    pathology associated with this femur, but Day and

    Molleson (1976) could not substantiate this claim.

    Atrophy following a neurological disease with

    paralysis of the leg could have led to a comparably

    thin bone. However, this can be excluded because

    a radiological analysis failed to show any anomalyof the trabecula, nor is the microscopic bone

    structure pathological (Day and Molleson, 1976).

    Day and Molleson (1976) therefore preferred an

    alternative explanation for the slenderness of OH

    34: fluvial erosion unmistakably shaped the

    femoral head and superior part of the neck and the

    shaft bears numerous scratch marks. Accordingly,

    Day and Molleson (1976) could not observe any

    outer circumferential lamellar bone in a micro-

    scopic study. Based on a regression formula of

    shaft thickness on medullary cavity diameter thatwas derived from a British archaeological sample,

    they speculated that the total thickness of the OH

    34 femur at mid-shaft could have been some 6 mm

    greater than it is at present. But they cautioned:

    Whether this clear evidence of surface chemical

    erosion and abrasion was sufficient to account for

    the bizarre appearance of the bone was a different

    matter. This particular issue was clouded by the

    clear retention of such superficial muscle markings

    as the spiral line and the supracondylar lines (Day

    and Molleson, 1976: 463). Indeed, if abrasion hadconsiderably affected the circumference of the fem-

    oral shaft, all projecting structures and edges

    would be expected to be ground off; abrasion never

    enhances these features. As on modern human

    femora, the above-mentioned muscle markings

    usually do not project more than fractions of a

    millimeter, and thus the actual shaft diameters and

    circumferences probably do not differ greatly from

    the original measured dimensions. Heavy abrasion

    is also contradicted by the sharpness of the lip of

    Fig. 2. Anterior aspect and cross-sections of the femur fragment of OH 62 (middle) compared with that of AL 288-1 (top) and OH34 (bottom). The restored parts of OH 34 are transparent. Minimum (dashed) and maximum (continuous) estimates are shown, as wellas the maximum length and femur head estimates of Day and Molleson (1976) (dotted outlines).

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 437

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    6/33

    the pilaster and the depression lateral to it, as well

    as by the sharp-edged triangular cross-section of

    the associated tibia mid-shaft fragment (every

    object becomes spherical with time when exposedto the forces of moving water and sand particles,

    be it wood, stone, or bone; e.g., Nawrocki et al.,

    1997; Haglund and Sorg, 2002, and references

    therein). Likewise, the sharpness of the edges of

    the femoral foramen nutricium refutes the idea

    that abrasive processes enlarged its diameterthe

    edges would at least have been rounded off. More-

    over, save for several grooves, pits, and micro-

    scopic scratch marks, which are mainly on the

    anterior aspect, the surfaces of OH 34s femur and

    tibia shaft are surprisingly smooth macroscopically

    compared to the femoral head and neck and other

    bones subjected to extensive fluvial reworking that

    we have studied. Their surfaces look as if they were

    rubbed with coarse sand paper, and often lack

    numerous large flakes. On the other hand, assum-

    ing that polishing was responsible for the smooth

    surface would require very fine-grained abrasive

    particles, which are incompatible with the high-

    energy fluvial environment of the site where OH 34

    was found (Kleindienst, 1973). Finally, Day and

    Mollesons (1976) radiographic study, together

    with the distribution of the cortical thicknessrevealed by the mid-shaft break, provides no evi-

    dence for a substantial asymmetrical loss of super-

    ficial bone. It is, however, difficult to imagine a

    simple process by which the entire circumference

    was affected evenly by abrasion.

    It seems more likely that the tibia fragment and

    the whole femoral shaft up to the inferior margin

    of the neck were embedded in sand or matrix when

    water currents eroded the head and upper margin

    of the neck, whereas the trochantera major and

    minor and the distal femur could have been lostdue to carnivore activity before fossilization, as

    suggested by the presence of several tooth marks

    (see Day and Molleson, 1976). The femoral shaft

    suffered thereby numerous scratch marks and lost

    superficial splinters of bone, but was probably

    protected from a marked reduction of its girth. On

    the other hand, Day and Mollesons (1976) failure

    to detect external basic lamella does not con-

    clusively prove substantial abrasion because the

    external circumferential lamellar bone is usually

    completely resorbed in modern humans over the

    age of approximately 40 years (Kerley, 1965;

    Schultz, 1999)and the absence of the epiphyseal

    line of the head indicates that OH 34 was fullyadult. In any case, it is unlikely that the shaft

    diameter of OH 34 was affected more by abrasion

    than that of OH 62. The rugged external appear-

    ance and the loss of surface detail leaves no doubt

    that heavy abrasion occurred in the latter specimen

    (Johanson et al., 1987).

    The length of the OH 34 femur can be fairly

    reliably estimated since a slight swelling at the

    inferior rim of the neck corresponds to the begin-

    ning of the head, and the expansion of the distal

    shaft and the supracondylar lines indicate the base

    of the condyles. Using the distal and proximal

    thirds of KNM-ER 1472, 1481, and KNM-WT

    15000 as analogues in a graphical reconstruction

    yields a minimum estimate for OH 34s femur

    length of 375 mm, and a maximum of 392 mm. A

    value close to the upper end of this range is

    supported by various wax reconstructions of the

    proximal and distal portions of a cast of OH 34,

    which yielded femoral lengths of 385395 mm.

    As the length of the OH 34 femur is critical to

    the estimation of OH 62s limb proportions, the

    accuracy of the graphical reconstruction methodwas tested on 13 modern human femora that

    were between 365 and 415 mm long. Their distal

    extremities were cut off graphically at about the

    same level as that of OH 34 and then reconstructed

    by fitting the superimposed distal third of another

    femur to the remaining shaft stumps. The thus

    reconstructed femora were slightly too long

    (mean deviation +2 mm, 8.7 mm, range 12 to

    +13 mm).

    An alternative estimate for the femur length of

    OH 34 of 411432 mm was provided by Day andMolleson (1976). This is markedly longer than our

    graphical and wax reconstructions, even if the

    possible error in these reconstructions is taken into

    account. Day and Mollesons femur length is hard

    to fit to the expanding mediolateral shaft diameter

    of OH 34 (see dotted outline in Fig. 2), as it would

    lead to an abnormally large bicondylar breadth.

    Day and Molleson based their length estimate on a

    comparison with a 432 mm long modern human

    femur that matched OH 34 in terms of the

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465438

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    7/33

    position along its length of identifiable features

    (Day and Molleson, 1976: 459). It was not indi-

    cated whether this femur matched OH 34 in other

    features such as shaft thickness, but the headdiameter of this modern femur was indicated to be

    40 mm, which seems to be considerably larger than

    that which can be inferred from OH 34s abraded

    head remnant (if KNM-ER 1472 or 1481 are

    graphically scaled down to fit the proximal femur

    of OH 34, a head size of about 33 mm is predicted,

    and 35 mm when KNM-WT 15000 is used as a

    guideline).

    Day and Mollesons second length estimate of

    411 mm was based on a regression formula of

    femur length on medullary cavity diameter that

    was derived from a British archaeological sample,

    but they cautioned: perhaps not too much weight

    should be placed on these calculations (Day and

    Molleson, 1976: 461).

    The OH 34 femur and tibia were found in situ

    by M. Kleindienst during the 1962 excavation at

    JK2 West. Their exact stratigraphic provenience is

    unknown, but most fossils from JK2 West are

    considered to come from Bed III, except for a few

    finds in Bed II (Kleindienst, 1973). However, she

    cautioned (p. 189) that the complexity of the

    stratigraphy in upper Bed II, Bed III, and Bed IVindicate[s] that it would have been, and still is,

    difficult to precisely correlate horizons between

    sites which are similar in field appearance.

    The dating of Bed III (1.150.8 Ma; Hay, 1976)

    suggests that OH 34 should be allocated to either

    H. erectus or P. boisei. Yet, among other features,

    the well-developed pilaster probably and the

    slenderness of the shaft clearly oppose attribution

    to the latter taxon. The only femur tentatively

    assigned to P. boisei, the fragmentary KNM-ER

    1500, is said to be similar to the stout femur of AL288-1 with its roundish shaft (Grausz et al., 1988).

    A difference as great as that between OH 34 and

    KNM-ER 1500 with respect to the proportion of

    the femur shaft cross-sectional area to the recon-

    structed length occurs in less than 1 of all

    possible pair-wise comparisons of the modern

    human sample and exceeds the range of variation

    in all extant great apes (see Fig. 7). The difference

    is even larger when OH 34 is compared to

    KNM-ER 1463, another femur that possibly

    belongs to P. boisei. In fact, its slender shaft

    aligns OH 34 exclusively with modern and fossil

    representatives of the genus Homo.

    On the other hand, the anteroposterior flatten-ing and mediolateral broadening of the proximal

    shaft in OH 34 is not as extreme as in some other

    early Homo femora. It is rare in extant hominoids

    to find differences in cross-sectional shape of the

    proximal femur as great as those between OH 34

    and the KNM-ER 737 or OH 28 femora, which

    belong to H. erectus. Such extremes occur in 5%

    and 3%, respectively, of all possible pair-wise

    comparisons in a world-wide sample of modern

    humans encompassing a large range of body-size

    (cf. Table 5 and Fig. 9). Similar degrees of differ-

    ence are observed in less than 3% of the gorilla

    sample and 1% of the chimpanzee sample and

    exceed the range of variation of orang-utans.

    Moreover, an attribution of OH 34 to H. erectus,

    as tentatively proposed by Howell (1978), would

    lead to an extreme body size range within that

    species (cf. Figs. 7 and 10), although the recent

    discovery of a surprisingly small young adult H.

    erectus calvaria (Leakey et al., 2003) could be in

    agreement with this notion.

    Furthermore, early Homo femora are said to

    possess an extremely thick femoral cortex com-pared to both Australopithecus and modern

    humans (Weidenreich, 1941; Day, 1971, 1978;

    Kennedy, 1983a,b; Trinkaus, 1984; Ruff et al.,

    1993). A radiological analysis revealed that OH 34

    falls inside the range of modern humans except for

    the sub-trochanteric posterior and lateral regions,

    which are thinner (Day and Molleson, 1976). Yet,

    this result may be different if adjusted for the small

    size of OH 34. Moreover, medullar stenosis seems

    to be typical mainly for later Eurasian H. erectus,

    whereas the only two analysed femora of theearlier African H. erectus have a thinner cortex

    and also fall in the modern human range of

    variation (see Ruff et al., 1993).

    In conclusion, there is no support for marked

    post-mortem changes in the morphology of

    OH 34. Although the current comparative sample

    of early hominid femora is larger than that which

    Day and Molleson had at their disposal, its mor-

    phology remains unique. It differs statistically sig-

    nificantly from P. boisei and H. erectus specimens,

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 439

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    8/33

    and other pene-contemporaneous early hominid

    species are unknown. The closest match is with

    OH 62. Homo habilis does, however, not occur as

    late as Olduvai Bed III. It is worthy of mention inthis context that the pattern of abrasion of the

    fossils from the JK2 site and their recovery from

    ancient stream channel sediments indicate that

    they were reworked from older deposits, although

    it is not known how old (Alan Walker and Michael

    Day, personal communication). Nevertheless,

    based on the available evidence, it seems best to

    follow Day and Molleson (1976), who preferred to

    refer OH 34 to Homo sp. indet.

    In any case, the use of OH 34 for a comparative

    analysis of the OH 62 femur is hardly less appro-

    priate than the traditional reconstruction of OH 62

    based on the stout AL 288-1 femur.

    Measurements and analyses

    All measurements were taken as described by

    Martin and Saller (1957) and Knussmann (1988),

    except (1) the anteroposterior diameter of the

    distal humerus, which was taken at a level equal to

    three capitulum heights from the distal end; (2) the

    radius shaft diameters, which were taken as maxi-

    mum and minimum diameters at the foramennutricium; (3) the anteroposterior diameter of the

    incisura trochlearis ulnae, which was measured at

    its deepest point; (4) the anteroposterior diameter

    of the distal femur shaft, which was taken at a level

    equal to two condylar heights from the distal

    extremity; and (5) the anteroposterior diameter of

    the distal tibia, which is the maximum antero-

    posterior diameter at the distal extremity of the

    tibia.

    As the KNM-ER 3735 skeleton is very fragmen-

    tary, femoral mid-shaft dimensions are probablynot measurable exactly at the original mid-shaft. A

    comparison to modern human femora of similar

    shaft size suggested that the proximal extremity of

    the left femoral fragment of KNM-ER 3735 is

    perhaps 13 cm below this level. Similarly, the

    distal extremity of the OH 62 femur fragment

    might be 0.5 cm above (compared to the maximum

    length estimate of the OH 34 femur) or up to

    4.5 cm below the original mid-shaft (compared

    to the femur length of AL 288-1). Due to damage

    to the OH 62 femur, however, mid-shaft

    dimensions can only be taken at about 3.5 cm

    proximal to the distal extremity, which means that

    they could have been measured either about 4 cmtoo proximally or 1 cm too distally. The deviations

    relative to mid-shaft circumference and antero-

    posterior and mediolateral diameters were there-

    fore recorded for a cast of OH 34 and 20 modern

    human femora measured up to 6 cm above or

    below mid-shaft (Fig. 3). The modern human

    femora were chosen based on a mid-shaft circum-

    ference of 6979 mm (KNM-ER 3735: 75 mm) and

    a length of 376410 mm. In addition, as the

    position of the distal end of the KNM-ER 3735

    femur fragment cannot exactly be determined, the

    deviation of the anteroposterior diameter of the

    distal femur shaft was recorded at a point 4 or

    3 times instead of 2 times the height of the

    condyles from the distal extremity.

    As can be seen in Fig. 3, on average the

    circumference and the product of the antero-

    posterior and mediolateral femur shaft diameter

    change only slightly up to 4.5 cm above or below

    mid-shaft. But based on the extremes, it is possible

    that, for example, the original mid-shaft circumfer-

    ence of KNM-ER 3735 was between 98% and

    105%, or that of OH 62 between 98% and 103%, ofthe measurable circumference. No relationship of

    these deviations was found with the slenderness,

    development of the pilaster, or length of the

    femora in the comparative sample, which is cor-

    roborated by a second sample of 10 modern

    human femora that measure between 366 and

    460 mm in length. The pattern of the OH 34 femur

    is encompassed by both modern human com-

    parative samples: its shaft circumference and the

    product of anteroposterior and mediolateral

    diameters increase continually from 6 cm abovemid-shaft to mid-shaft, following the lower

    extremes of the comparative samples. There-

    after, its shaft size remains constant until 6 cm

    below mid-shaft, which is similar to the average

    pattern of the modern human femora.

    Limb length and shaft proportions were

    analysed using bivariate regressions. This has the

    advantage over indices used in other studies (e.g.,

    Johanson et al., 1987; Kimbel et al., 1994;

    Richmond et al., 2002) in that the effect of body

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465440

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    9/33

    size can be separated when the variable under

    study does not change proportionally (isometri-

    cally) to body size (Aiello, 1981). In fact, limbproportions change nearly isometrically only in

    chimpanzees (i.e., slope=1), but not in other homi-

    noid species (see Table 1). Brachial and humero-

    femoral indices are therefore inappropriate means

    to study these proportions, especially if the body

    size covers such a large range as in the comparison

    of OH 62 and average-sized modern humans.

    There are different methods to analyse the trend

    of bivariate plots. It is often agreed that least

    squares regression should be used particularly for

    prediction purposes (Jungers, 1982, 1988; Martin

    and Barbour, 1989; Aiello, 1992; Dagosto and

    Terranova, 1992; Smith, 1994; Hens et al., 1998;Konigsberg et al., 1998), but it may produce biased

    results when a case does not derive from the same

    distribution as the reference sample and when

    there are no clear dependent and independent

    variables. Alternatives are Model II regressions. It

    is, however, controversial whether the major axis

    (e.g., Teissier, 1948; Olivier, 1976; Martin, 1982;

    Feldesman and Lundy, 1988; McHenry, 1988;

    Martin and Barbour, 1989; Formicola, 1993;

    Konigsberg et al., 1998) or the reduced major axis

    Fig. 3. Deviation of the (a) circumference and (b) anteroposterior and mediolateral diameters of the femur relative to mid-shaftdimensions for 20 modern human femora when measured up to 6 cm above or below mid-shaft, and (c) divergence of theanteroposterior diameter of the distal femur shaft measured at 4 and 3 times the height of the condyles compared to a point measuredat 2 times the height of the condyles. The thick black lines show the average relative deviation. (=diameter).

    Table 1

    Slopes of regressions of brachial and humero-femoral length proportions

    Proportion/species Slope of the LSR Slope of the MA Slope of the RMA

    Ulna to humerus length

    Homo sapiens 0.75 (0.700.80) 0.81 (0.750.87) 0.82 (0.770.87)

    Pan troglodytes 0.97 (0.901.04) 1.08 (1.011.16) 1.07 (1.001.14)

    Gorilla gorilla 0.80 (0.760.84) 0.83 (0.790.87) 0.83 (0.790.87)

    Humerus to femur length

    Homo sapiens 0.66 (0.630.69) 0.68 (0.650.71) 0.69 (0.660.72)

    Pan troglodytes 0.94 (0.861.02) 1.08 (1.001.17) 1.07 (0.991.15)

    Gorilla gorilla 1.11 (1.031.18) 1.19 (1.121.28) 1.18 (1.111.25)

    LSR=least squares regression, MA=major axis, RMA=reduced major axis, 95% confidence intervals in brackets.

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 441

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    10/33

    (e.g., Rayner, 1985; Aiello, 1992; Sjvold, 1992;

    Jungers et al., 1998) is preferable in determining

    functional relationships between variables. In thepresent study, the major axis was favoured in most

    cases, following the argument of Konigsberg et al.

    (1998) and others, but all analyses were done using

    all three regression techniques. To reduce the prob-

    lem of extrapolating beyond the characteristic

    body size for the reference sample, the present

    study used a modern human sample with an em-

    phasis on small-sized individuals. The major axis

    produces oval prediction intervals, which cannot

    be used for extrapolation. However, as in most

    analyses, the major axes are almost identical to the

    least squares regressions, and thus it is feasible to

    use the 95% prediction limits of the latter regres-sion technique for comparing the fossils to the

    ranges of variation of the modern reference

    samples. In those cases where the major axis devi-

    ates markedly from the least squares regression,

    the slope is mostly similar to the reduced major

    axis, which allows calculation of the prediction

    limits accordingly (see Sokal and Rohlf, 1995).

    The 95% prediction limits of the bivariate

    regressions are not only useful for comparing a

    specific metrical proportion of a fossil to the range

    Fig. 4 (a).

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465442

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    11/33

    of variation of the reference sample. As the fossil

    specimens do not belong to any of the extant

    hominoid groups, the average morphology and

    their variability is unknown. More important is the

    distance between a pair of fossils relative to the

    slope of the regression line of an extant hominoid

    group compared to the corresponding width of the

    95% prediction interval. If the distance between

    two fossils exceeds the width of the 95% prediction

    interval, they are deemed to differ significantly

    compared to the range of variation of the reference

    sample. On the other hand, if the distance between

    two fossils can be encompassed by the prediction

    interval of a reference sample, there is no reason to

    conclude that they had a different functional

    anatomy.

    A similar approach to examine the dissimilarity

    between two fossils is the permutation test, which

    Fig. 4 (b).

    Fig. 4. (a) Four examples demonstrating the position of OH 62 ( Homo habilis), AL 288-1 (Australopithecus afarensis), KNM-ER 1500(P. boisei), and KNM-WT 15000 (Homo erectus) relative to modern humans and African apes. The 95% prediction limits for humanand chimpanzee data points are indicated based on least squares regressions (continuous lines); their area of overlap is shaded darker.Major axes are represented by dashed lines. Symbols for fossils with dashed lines reflect the possible effect of not measuring femurmid-shaft dimensions at true mid-shaft. The larger sized individuals of the gorilla sample lie outside the graphs frames. (b) Four otherexamples of upper-to-lower limb shaft proportions, demonstrating also the position of KNM-ER 3735 (Homo habilis).

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 443

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    12/33

    calculates the probability of sampling proportions

    as divergent as those observed between a fossil

    pair within extant hominoid groups. To this end,

    all possible pair-wise diff

    erences between theresiduals of the reference species regression were

    computed. The distance between the data points

    of the fossils was calculated based on the slope of

    the reference samples regression line. Finally, the

    probability that the corresponding difference is

    either compatible with or exceeds intra-specific

    variation was inferred from a comparison with the

    distribution of the permutated differences between

    two points of the reference sample (see also

    Grine et al., 1993; Richmond and Jungers, 1995;

    Richmond et al., 2002).

    In contrast to common practice, logarithmic

    transformations were not employed in this present

    study because an examination of the residuals

    revealed no improvement of the linear regression.

    In addition, it might have been illegitimate to use

    95% prediction limits after logarithmic transfor-

    mation (cf. Sprugel, 1983; Dagosto and Terranova,

    1992).

    Results

    Limb shaft proportions

    A comparison of the shaft diameters and cir-

    cumferences of the OH 62 skeleton with those of

    AL 288-1 yields a mixed picture: OH 62 has in

    many respects a more massive upper limb than AL

    288-1. This is true, for example, for the antero-

    posterior diameter of the distal humerus shaft

    and the proximal ulna and radius, whereas the

    humerus is slightly more slender at the mid-shaft

    and the ulna is slimmer distally (Johanson et al.,

    1987). On the other hand, all diameters of OH 62sproximal femur fragment are thinner than in the

    Hadar skeletons, except for the anteroposterior

    sub-trochanteric diameter.

    Compared to the intra-specific variation of

    hominoids, however, these differences turn out not

    to be significant. Upper-to-lower limb shaft pro-

    portions have been analysed in modern humans,

    chimpanzees, gorillas, and orang-utans by means

    of regressions of humerus mid-shaft circumference,

    anteroposterior diameter of the distal humerus

    shaft, articular width of the distal humerus, radius

    head, radius shaft, and anteroposterior diameter of

    the incisura ulnae to sub-trochanteric and mid-

    shaft femur diameters, femur mid-shaft circumfer-ence, and anteroposterior diameter of the distal

    femur shaft. Some of these 24 pair-wise compari-

    sons are represented in Fig. 4. In most cases, the

    major axes are virtually identical to the least

    squares regressions, except for the plots of

    humerus mid-shaft circumference, radius shaft,

    and radius head to femur mid-shaft circumference,

    and humerus articular width, radius shaft, and

    radius head to distal femur shaft, where the major

    axes are similar to the reduced major axes.

    Fig. 4 shows that both OH 62 and AL 288-1 fall

    inside the 95% prediction interval of our sample of

    modern humans, which is true for almost all

    pair-wise comparisons between the upper-to-lower

    limb shaft dimensions of these fossils. The only

    exception is the proportion of humerus mid-shaft

    circumference to sub-trochanteric femur size

    (=APML diameter), in which OH 62 and, to a

    slightly greater degree, AL 288-1 fall just above the

    95% prediction limits for modern humans. This

    departure from average modern proportions is

    slightly more marked in a reduced major axis

    regression. However, both fossils, and especiallyOH 62, are in the lower body size range of the

    comparative modern human sample. At the edges

    of the comparative samples size range, the choice

    of the proper regression technique becomes more

    critical and the conclusions less confident (cf.

    Jungers, 1985). Nevertheless, even in a reduced

    major axis regression, OH 62 still falls inside the

    99% prediction limits of modern humans. If the

    femur mid-shaft dimensions are corrected for

    the fact that the measurement may have been

    taken too far proximally, OH 62 could even fallinside the 95% prediction limits of modern

    humans. In addition, in almost all comparisons,

    a Swiss Neolithic female of about the same

    inferred body size as OH 62 had very similar or

    even slightly larger upper limb to femur shaft

    proportions.

    On the other hand, OH 62 falls inside the range

    of chimpanzees not only for humerus mid-shaft

    circumference vs. femur sub-trochanteric size, but

    also for humerus distal shaft and proximal ulna

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465444

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    13/33

    size vs. both femur sub-trochanteric and mid-shaft

    size. This is because at small body sizes the ranges

    for chimpanzees and modern humans overlap.

    However, due to the large intra-specific variability

    of upper-to-lower limb shaft proportions in all

    comparative samples, it cannot be concluded that

    OH 62s limb shaft proportions are more

    chimpanzee-like than those of AL 288-1. Likewise,

    it would be incorrect to characterize, for example,

    the Neolithic female that in many of these com-

    parisons lies close to OH 62 as chimpanzee-

    like due to its large upper limb shaft dimensions

    relative to femur shaft size.

    In fact, independent of the regression technique

    used, the points of OH 62 and AL 288-1 lie

    relatively close together compared to the range of

    variation of extant hominoid species, with the

    exception of AL 288-1s proximal ulna relative tofemur size. Consequently, there is a high prob-

    ability that the differences in upper-to-lower limb

    shaft proportions between OH 62 and AL 288-1

    could be drawn from within any extant hominoid

    group. Table 2 summarizes these probabilities. The

    results for three proportions are shown, represent-

    ing the smallest (humerus mid-shaft circumference

    vs. femur shaft size), average (distal humerus and

    radius shaft vs. femur shaft size), and largest

    proportional difference within this fossil pair

    (proximal ulna vs. femur shaft size). In the lattercase, the two fossils lie at opposite edges of the

    modern human 95% prediction interval (cf. Fig.

    4a). Consequently, differences in proximal ulna-to-

    femur shaft proportions as great as those between

    OH 62 and AL 288-1 are found in only 5% of

    the modern human sample. In chimpanzees and

    gorillas, they are, however, much more common

    (20% and 28%, respectively). Although great, the

    differences between the two fossils are therefore

    not statistically significant.

    The second H. habilis partial skeleton,

    KNM-ER 3735 is considerably larger than OH 62.

    For all possible comparisons of upper-to-lower

    limb shaft size, it falls close to the mean of modern

    humans, i.e., for anteroposterior diameter of the

    distal humerus shaft, distal humerus articular

    width, radius neck, and radius shaft size relative to

    femur mid-shaft dimensions and anteroposterior

    diameter of the distal femur (cf. Fig. 4b). Although

    the original femur length of KNM-ER 3735 is

    unknown, the result is not different even if mid-

    shaft dimensions are maximally corrected, i.e., for

    an up to 4.5 cm too distal measurement; upper-to-

    lower limb proportions are still outside the range

    of chimpanzees.

    Similar conclusions can be drawn from upper

    limb to distal femur shaft proportions. Because the

    anteroposterior diameter of the femur shaft gener-ally increases when approaching the condyles

    (Fig. 3c), the measurable thickness of the distal

    extremity of the KNM-ER 3735 femur shaft frag-

    ment is a minimum estimate, and any correction

    would push upper limb to distal femur shaft pro-

    portion further away from great ape proportions

    (Fig. 4b).

    Apart from the effect of the fragmentary nature

    of the femur, it might also be possible that the

    human-like upper-to-lower limb proportions of

    KNM-ER 3735 partly result from the erodedcondition of its arm skeleton. An attempt was

    made to take this into account, but it may be that

    additional superficial bone is missing. The radius

    shaft to femur mid-shaft proportion of KNM-ER

    3735 would, however, only fall beyond the human

    95% prediction limits and inside those of chimpan-

    zees if one assumes that 3 mm are missing from the

    anteroposterior and an additional 3 mm from the

    mediolateral shaft diameter, and simultaneously

    the femoral mid-shaft diameters are corrected

    Table 2

    Probabilities of sampling differences in upper-to-lower limb shift proportions as great as those observed between OH 62 and

    AL 288-1 from extant hominoid groups based on the slopes of the major axis (=diameter)

    Proportion Homo Pan Gorilla PongoHumerus mid-shaft circumference vs. femur sub-trochanteric size 0.89 0.98 0.92 0.70

    Humerus distal shaft AP vs. femur sub-trochanteric size 0.38 0.25 0.39 0.24

    Incisura trochlearis ulnae AP vs. femur sub-trochanteric size 0.05 0.20 0.28 0.14

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 445

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    14/33

    according to the most negative deviation in

    Fig. 3b. Similarly, 1.5 mm must be missing from

    the anteroposterior diameter of the distal humerus

    shaft if femoral mid-shaft dimensions are maxi-mally corrected, or 2 mm without correction for

    the femur mid-shaft. Finally, with correction to the

    femur mid-shaft, 4.6 mm must be missing from the

    articular width of the distal humerus (5.8 mm

    without correction). Such an extreme loss of

    superficial bone can be excluded.

    For comparison, Fig. 4a and b also shows the

    positions of two other early hominid skeletons.

    The adolescent H. erectus KNM-WT 15000 falls

    close to the mean of adult modern humans for

    most limb shaft proportions. Noteworthy is the

    size of its distal humerus shaft relative to femur

    mid-shaft size, which lies in the upper range of the

    modern human variation and thus resembles the

    proportion of OH 62 and AL 288-1.

    Another early hominid for which limb shaft

    proportions can be analysed is KNM-ER 1500,

    which might represent P. boisei (Grausz et al.,

    1988). Relative to sub-trochanteric femur size, its

    radius shaft and proximal ulna dimensions are

    much larger than those of OH 62 and AL 288-1,

    although the differences between these fossils are

    still within the ranges of intra-specific variation forextant hominoid groups. Compared to the distal

    femur shaft, the size difference between KNM-ER

    1500s and AL 288-1s radius head, however, just

    exceeds the width of the 95% prediction interval of

    modern humans: KNM-ER 1500 lies at its upper

    limit, AL 288-1 falls just beyond the lower limit

    (data not shown). The discrepancy between the

    proportion of radius shaft to distal femur shaft of

    KNM-ER 1500, on the one hand, and KNM-ER

    3735 and AL 288-1, on the other hand, surpasses

    markedly intra-specific variation in modern homi-noids (Fig. 4b). This strongly suggests that P.

    boisei had different body proportions from both

    H. habilis and A. afarensis.

    Limb length proportions

    Brachial proportions

    The humerus of OH 62 is relatively well pre-

    served, lacking only the proximal and distal

    extremities. Its length can be estimated with only

    minor error. A graphical reconstruction using

    diverse modern and fossil humeri as analogues for

    the missing parts suggests a maximum length of270 mm and a minimum length of 258 mm. This

    compares well with the estimate of 264 mm given

    by Johanson et al. (1987), but the range of the

    estimates is smaller than predicted by Koreys

    (1990) study. By visually aligning different com-

    plete modern human humeri with damaged

    humeri, where the two fragmented ends have been

    masked according to the damage in OH 62, he

    suggested that the standard deviation for the re-

    construction is 10.5 mm. This equals about 8 mm if

    scaled down to the length of OH 62. The goodness

    of the match between the contours of two humeri

    is, however, probably better in our graphical

    reconstruction method than with Koreys tech-

    nique, which might explain his larger confidence

    interval. A minor factor is that he was familiar

    with the OH 62 specimen only from photographs.

    Nevertheless, all these values for OH 62s humerus

    are considerably longer than the humerus length

    of AL 288-1 (236.8 mm, Johanson et al., 1982; or

    246 mm if corrected for the crushed head, see

    Hausler, 2001).

    Of the OH 62 radius, only a 143 mm long shaftfragment is preserved, making it impossible to

    estimate its total length.

    The ulna consists of four fragments. Joining

    them as closely as possible yields a value of

    225 mm as the minimum ulna length if the lacking

    olecranon and caput ulnae are taken into

    account. However, a length of approximately 245

    to 255 mm is more realistic considering the differ-

    ences of the individual fragments cross-sectional

    diameters. This is almost the length of the large

    and very robust AL 438-1a (A. afarensis) ulna,and far longer than AL 288-1, as well as the

    smallest known A. africanus ulna, Stw 326 (Fig. 5).

    Since the ulna in hominoids is generally roughly

    5% longer than the radius, our estimate for

    OH 62s ulna length also closely agrees with

    Hartwig-Scherer and Martins (1991) estimate of

    its radius length, for which they give a range of

    210 mm (in their view a less likely value) to

    246 mm. In conclusion, given the considerably

    greater length, but only slightly greater shaft

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465446

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    15/33

    thickness compared to AL 288-1, both OH 62s

    humerus and ulna were probably remarkably

    slender and less robust than in the Hadarspecimen.

    If the value of 245 mm for OH 62s ulna length

    is plotted against the mean reconstructed humerus

    length, it falls close to the average brachial pro-

    portion of chimpanzees (Fig. 6). Even combined

    with the upper limit of the reconstructed humerus

    length, the resulting brachial proportion lies out-

    side the 95% prediction limit of modern humans.

    The ulna might, however, also have been longer,

    but the corresponding upper-to-forearm pro-

    portion does not exceed the probable range of thelarge-sized A. afarensis specimens (Kimbel et al.,

    1994), or that of A. africanus (Hausler, 2001) or

    BOU-VP-12/1 (Hominidae gen. et sp. indet; Asfaw

    et al., 1999). Rather, it is indistinguishable from

    them. On the other hand, although the minimum

    estimated ulna length is less likely, with a slightly

    shorter ulna of 236245 mm, according to the

    corresponding humerus length, OH 62s brachial

    proportion would fall inside the 95% prediction

    limits of modern humans.

    Relative femur length

    Upper limb proportions of OH 62 may there-

    fore not differ from those of the australopithecines.Yet, its femur is in all measurable dimensions

    considerably thinner than the AL 288-1 femur and

    possesses a different morphology (Table 3 and

    Fig. 2). In fact, none of the Plio-Pleistocene femora

    closely matches the morphology and metrics of

    that of OH 62, save for OH 34. This enigmatic

    femur has been described to be surprisingly

    slender, but a regression of femur shaft diameter

    against femur length demonstrates that OH 34 is

    not unusually slender: its reconstructed length falls

    within the modern human range of variation basedon the least squares regression (Fig. 7). The major

    axis is nearly identical. With the reduced major

    axis regression, it approaches but is still below, the

    upper 95% prediction limits for modern humans.

    Only on the basis of the maximum length estimate

    of Day and Molleson (1976), which has already

    been shown to be unrealistic, would it fall outside

    the human range of variation. Thus, its apparent

    slenderness seems to be an optical illusion due to

    its small size rather than a puzzle, as Day and

    Fig. 5. The ulna of OH 62 compared with the contours of the Stw 326 (completed with the proximal fragment 398; A. africanus), AL288-1 (A. afarensis), and AL 438-1 (A. afarensis). All comparative specimens are from the left side and have been mirrored for thisillustration.

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 447

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    16/33

    Molleson (1976) called it. Therefore, there is no

    reason to suspect a substantial degree of abrasion

    of its surface bone, of which Day and Molleson

    (1976) themselves were not convinced.

    Nevertheless, OH 34 is remarkably slender

    when compared to the few contemporaneous

    fossils. Only the subadult femur of KNM-WT

    15000 has a similar robusticity, although this

    Fig. 6. Plot of ulna length vs. humerus length (mm). The 95% prediction limits for human and chimpanzee data points are indicatedbased on reduced major axes (thin, long-dashed lines). Major axes (thick, short-dashed lines) are almost identical, least squaresregressions (not shown) slightly flatter. OH 62, AL 288-1, Stw 431, and BOU-VP-12/1 are shown with the most likely proportions andthe ranges of the estimates (shaded rectangles). The minimum ulna length of AL 288-1 (191 mm) results from directly joining the two

    fragments, the maximum length (219 mm) is Kimbel et al.s ( 1994) estimate; the minimum humerus length was estimated by Johansonet al. (1982), whereas correcting for crushing suggests a greater length (Hausler, 2001). For the range of Stw 431 see Hausler (2001).The data points for AL 137-50 and MAK-VP-1/3 were obtained by combining these reconstructed humeri with the AL 438-1a ulna.Minimum and maximum humerus lengths of BOU-VP-12/1 correspond to the estimate of Asfaw et al. (1999) based on regressions andan anatomical reconstruction, respectively; the range of its ulna length results from the lower and upper 95% prediction limits forAfrican apes and modern humans, respectively, of the regression on Asfaw et al.s (1999) radius length.

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465448

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    17/33

    Table 3

    Comparison of some measurable dimensions of the OH 62 femur fragment with those of OH 34 and AL 288-1 (=diameter, mm)

    OH 62 OH 34 AL 288-1

    Femur neck ap 13.2 (12.513.5) 15.4Femur shaft sub-trochanteric ap 19.4 17.4 17.5

    ml 20.1 21.2 25.3

    Femur mid-shaft ap 18.7a 19.3 21

    ml 19.4a 18.9 21.9

    minimum 17.5a 15.6 19.9

    Femur mid-shaft circumference 62a 59.5 67.5

    Estimated femur length 375392 280286b

    aMeasured about 3.5 cm proximal to the distal extremity.bLength estimates: 280 (Johanson et al., 1982), 281 (Jungers, 1982); after reconstruction of the crushed distal end: 283 (Schmid,

    1983), 286 (McHenry and Berger, 1998b).

    Fig. 7. Plot of femur length (mm) vs. sub-trochanteric femur shaft size (anteroposteriormediolateral diameter, mm2). The 95%prediction limits for human data points are indicated based on least squares regressions (continuous lines). Major axes (thick,short-dotted lines) are nearly identical, reduced major axes (thin, long-dashed lines) are steeper. The range of the estimated femurlength is given for OH 34, OH 53, AL 288-1, and the A. africanus specimens Stw 99 and Stw 121 (Hausler 2001). Length estimates forKNM-ER 1500 (P. boisei), OH 53, KNM-ER 1809 (P. boisei or H. habilis), KNM-ER 3728 (possibly H. rudolfensis), KNM-ER 993and 1463 (P. boisei or H. erectus), KNM-ER 737, 1808, and OH 28 (H. erectus) are derived from McHenry (1991). KNM-ER 1808is also corrected for pathological hyperostosis (Walker et al., 1982). KNM-ER 1472 and 1481 (possibly H. rudolfensis) and KNM-WT15000 (H. erectus) were measured on the original. The length of KNM-WT 16002 (Australopithecus sp.) was estimated byreconstruction and comparison with other fossils.

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 449

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    18/33

    might partly be due to its young age (cf. Ruffet al.,

    1994). On the other hand, the femur of KNM-ER

    1808 might have been even more slender if cor-

    rected for its pathological hyperostosis (Walkeret al., 1982; Rothschild et al., 1995). In any case,

    the number of early Homo femora for which femur

    robusticity can be inferred is very small. There are,

    indeed, only two specimens that do not need a

    considerable amount of estimation for determi-

    nation of their length (KNM-ER 1472 and 1481).

    The difference between OH 34 and KNM-ER 1481

    are exceeded by 31% of the comparisons within the

    modern human sample, while the difference be-

    tween OH 34 and KNM-ER 1472 are exceeded by

    25% of those comparisons. Owing to the paucity of

    contemporaneous femora, it therefore cannot be

    concluded at present that the OH 34 femur is

    abnormally slim in comparison to the available

    specimens. The question of whether OH 34 is an

    unusual specimen remains open.

    As can be seen from Fig. 7, all early Homo

    specimens fall within the range of modern humans

    with regard to femur length vs. shaft diameter,

    whereas all australopithecines, including AL 288-1,

    fall in the range of modern great apes. This indi-

    cates that a general feature of australopithecines

    and paranthropines was an apelike short femur,possibly not only relative to shaft dimensions, but

    also relative to body size (see also Jungers, 1982,

    1991). The body size-femur length relationship of

    AL 288-1 and its relatives certainly deserves

    another look. Yet, the present data suggest that

    the shortness of the femur is hardly an effect of

    small body size and the convergent regression lines

    of modern humans and chimpanzees, as has

    been repeatedly suggested (e.g., Franciscus and

    Holliday, 1992; Vancata, 1996; Hens et al., 2000;

    Holliday and Franciscus, 2001). Even if the steeperslope of the reduced major axis is preferred,

    AL 288-1 falls below the range of modern humans;

    only the smaller-sized KNM-WT 16002 (Australo-

    pithecus sp.) would lie within the lower range of

    humans.

    The femoral proportion makes it likely that

    KNM-ER 1809, which on geochronological

    grounds alone could either belong to H. habilis,

    H. erectus, or P. boisei, can be assigned to the

    latter taxon. With the same logic, the slightly

    younger KNM-ER 993 and KNM-ER 1463 might

    belong to P. boisei rather than to H. erectus, and

    OH 53 to H. habilis.

    Although Fig. 7 illustrates the diffi

    culty ofestimating femur length based merely on shaft

    diameters, OH 62s morphological similarity to

    the proximal segment of the OH 34 femur sug-

    gests that a Homo model might be more appro-

    priate for femur length reconstruction than an

    Australopithecus model.

    Another similarity between OH 62 and OH 34 is

    the presence of a well-developed pilaster (a linea

    aspera is absent). Interestingly, this parallels the

    morphology of the left femur shaft fragment of

    KNM-ER 3735. Although a pilaster is said to be

    found rarely in early hominids (see references in

    Ruff, 1995), it can also be found in AL 333-61,

    KNM-ER 736, 737, 738, and others. It starts in

    OH 62 at the same level as in OH 34 and is

    identical in its degree of development to the corre-

    sponding part of the latter specimen (see Fig. 2).

    The pilaster usually expands over the middle third

    of those femora in which it is present. Its origin

    near the distal end of the preserved OH 62 femur

    fragment suggests that almost two thirds of the

    entire femur length are missing, and not one-half,

    as it would be expected if it had the length ofAL 288-1. In the latter case, the pilaster must have

    ended just after it had started. This seems unlikely,

    as the biomechanical function of this buttress is to

    increase bending rigidity of the femur (Pauwels,

    1965; Kummer, 1978). It might therefore be nor-

    mal to find exceptionally well-developed pilasters

    in long and slender bones (see also Day and

    Molleson, 1976).

    Thus, if the OH 62 femur is reconstructed with

    the length of OH 34, it results in an essentially

    humanlike humero-femoral proportion (Figs. 8and 11). In fact, assuming human-like proportions,

    the 95% prediction limits for the femur length

    based on a humerus length of 270 mm correspond

    to 354 mm and 404 mm (least squares regression),

    as compared to OH 34s reconstructed length of

    between 375 and 392 mm.

    The associated upper and lower limb bones

    from Bouri, BOU-VP-12/1 (Hominidae gen. et sp.

    indet.), might also have possessed humanlike pro-

    portions if the anatomical reconstruction of the

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465450

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    19/33

    humerus and femur length by Asfaw et al. (1999) is

    correct. As with the brachial proportion, limb

    lengths of BOU-VP-12/1 predicted by regressions

    seem to be less trustworthy because the resulting

    humero-femoral proportion falls outside the range

    of all extant hominoids. Thus, both intermembral

    proportions might have been similar to those of

    OH 62.

    Discussion

    Limb shaft proportions

    Johanson and co-workers stated in their

    original description: Even allowing for slight

    exfoliation of the OH 62 femur, visual comparison

    makes it obvious that this individuals femur was

    Fig. 8. Plot of femur length vs. humerus length (mm). The 95% prediction limits for human and chimpanzee data points are indicatedbased on reduced major axes (thin, long-dashed lines). Major axes (thick, short-dashed lines) are almost identical, least squaresregressions (not shown) barely steeper. Shaded rectangles symbolise the ranges of the estimates for OH 62 (femur length correspondingto that of AL 288-1 and OH 34, respectively), AL 288-1 (for the humerus length cf. Fig. 6), Stw 431 (A. africanus; see Hausler, 2001),and BOU-VP-12/1 (see Asfaw et al., 1999).

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 451

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    20/33

    smaller and less robust than the A.L. 288-1 femur

    (Johanson et al., 1987: 208). Combined with the

    evidence for a considerably longer humerus in

    OH 62, they speculated that body proportions inH. habilis fell in the range ofPan. Although Korey

    (1990) showed that their humero-femoral index

    had such a large confidence interval that it can be

    placed anywhere between gorillas and humans,

    Hartwig-Scherer and Martin (1991) inferred from

    upper-to-lower limb shaft and length proportions

    that OH 62 displays closer similarities to African

    apes than does AL 288-1. The present study, in

    contrast, using a much largercomparative data set,

    found that all examined upper-to-lower limb shaft

    proportions of OH 62 and AL 288-1 are inside the

    range of modern humans. Only in humerus mid-

    shaft circumference relative to sub-trochanteric

    femur size does OH 62 and AL 288-1 fall just

    above the 95%, but still within the 99%, prediction

    limits, for the modern human sample. This obser-

    vation only illustrates that the difference between

    the two fossil hominids can be encompassed within

    the range of variation of a single species and is

    therefore not statistically significantit does not

    mean that limb shaft proportions of A. afarensis,

    H. habilis, and H. sapiens cannot be distinguished

    at the species level. It is therefore unimportant toknow that a certain specimen falls slightly closer to

    the modern human or the chimpanzee regression

    line. Indeed, for all upper-to-lower limb shaft

    proportions, the average individual variation

    within extant hominoid groups is greater than

    between these two early hominids (see also

    Richmond et al., 2002). Accordingly, it cannot be

    inferred, as did Hartwig-Scherer and Martin

    (1991), that one of them has more primitive limb

    shaft proportions than the other.

    Other reasons for the divergent conclusions ofHartwig-Scherer and Martin (1991) are their much

    smaller comparative sample size and the lack of

    small-bodied modern humans in their data set.

    Moreover, their consolidation of chimpanzee and

    gorilla data in a single African ape sample is

    problematic because gorillas are not simply scaled

    up chimpanzees (Shea, 1981), and they often have

    different slopes and intercepts of the regression

    lines (cf. Fig. 4). In addition, especially with a

    sample size that covers a limited range of size, the

    correlation coefficient of the regression can be low,

    and the exclusive use of reduced major axis often

    produces steep slopes, which obviously had a

    dramatic eff

    ect on the position of OH 62.However, as can be seen in Fig. 4, modern

    humans and chimpanzees overlap slightly for these

    proportions in the small body-size range. It there-

    fore cannot be determined whether H. habilis more

    closely follows human or ape proportions, unless a

    larger-sized specimen is analysed. An unexpected

    support for rather modern upper-to-lower limb

    proportions comes from the fragmentary skeleton

    KNM-ER 3735 from East Turkana, Kenya

    (Leakey and Walker, 1985; Leakey et al., 1989).

    Being considerably bigger than OH 62, it was

    tentatively identified as a male individual (Leakey

    et al., 1989). The size of the fragments suggested

    that the skeletons limb proportions possibly

    approximate those of chimpanzees, although the

    sacrum is much smaller and shaped differently.

    Nonetheless, whereas Leakey et al. (1989) stated

    that the distal humerus of KNM-ER 3735 is as

    large as that of their chimpanzee, they also wrote

    that hindlimb dimensions are larger in the fossil.

    In accordance with this observation, the few

    measurements that can be taken on this partial

    skeleton do not verify the claim of having limbshaft proportions that are, compared to AL 288-1,

    more apelike than what can be explained by

    individual variation. Indeed, even with the most

    generous corrections for its incompleteness, none

    of our chimpanzees approaches KNM-ER 3735 in

    any of these proportions. The present re-analysis

    of its humerus, radius, and femur fragments

    demonstrates that its shaft proportions fall close to

    the modern human mean and outside the range of

    variation of chimpanzees. This has, of course,

    implications for OH 62 as well. Although therelatively large humerus shaft cross-section of

    OH 62 could indicate relatively higher arboreal

    activity levels, the proportions of KNM-ER 3735

    suggest that any behavioural differences between

    H. habilis and later Homo were not so extensive

    as to have had a major effect on limb shaft

    proportions.

    Upper-to-lower limb shaft proportions of the

    KNM-WT 15000 skeleton from Nariokotome

    (H. erectus) lie entirely inside the range of modern

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465452

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    21/33

    humans. It is conspicuous only in that its relative

    distal humerus shaft size resembles that of OH 62

    and AL 288-1 in falling in the upper half of the

    modern human distribution. As the specimen isthat of a juvenile with a skeletal age of about

    1313.5 years (range 1115 years) and a dental age

    of 11 (Smith, 1993; see also Ruffand Walker, 1993;

    Clegg and Aiello, 1999), it is possible that some of

    its limb shaft proportions are not yet those of an

    adult H. erectus. At the age of 13, modern human

    males have on average 12% lower humerus to

    femur strength proportions at midshaft than at

    17 years of age (Sumner and Andriacchi, 1996;

    Ruff, 2003; Ruff, personal communication). Thus,

    speculations about the functional implications of

    the relatively large distal humerus shaft size must

    wait until a complete enough adult H. erectus

    skeleton is recovered.

    Another remarkable pattern of upper-to-lower

    limb shaft proportions is found in KNM-ER 1500,

    which is tentatively attributed to P. boisei (Grausz

    et al., 1988). Compared to the H. habilis skeleton

    KNM-ER 3735 and to AL 288-1 (A. afarensis), its

    radius shaft relative to femur shaft size signifi-

    cantly surpasses intra-specific variation in modern

    hominoids (Fig. 4b). In contrast to Grausz et al.

    (1988), who proposed nearly identical body pro-portions to those of AL 288-1, this strongly

    suggests that P. boisei at least possessed a con-

    siderably larger forearm than either AL 288-1 or

    OH 62.

    Limb length proportions

    Although OH 62, KNM-ER 3735, AL 288-1,

    and KNM-WT 15000 do not differ significantly in

    their upper-to-lower limb shaft proportions com-pared to the intra-specific variation in extant homi-

    noid species, it is quite possible that they diverged

    in their limb length proportions. Richmond et al.

    (2002) suggested that diaphyseal dimensions are

    fairly variable in hominoids and therefore do not

    have great power in uncovering species-level differ-

    ences in skeletal form. Indeed, in contrast to

    relative shaft dimensions, the length of OH 62s

    upper limb skeleton considerably exceeds that of

    AL 288-1, as is well demonstrated in Fig. 1.

    Brachial proportions

    Length estimates for the forearm of OH 62 are

    less secure. Combined with the length of the

    humerus, the fragmentary ulna suggests that H.habilis had a high brachial proportion. Although

    modern human proportions cannot be excluded, a

    great forearm length is supported by the relatively

    long radius neck of KNM-ER 3735, which is in

    the range of chimpanzees (Leakey et al., 1989;

    Haeusler and McHenry, 2004). Similar body pro-

    portions with a relatively long forearm and an

    elongated lower limb compared to humerus length

    have been reported for the w2.5 million year old

    associated limb bones BOU-VP-12/1 from Bouri

    (Hominidae gen. et sp. indet.; Asfaw et al., 1999),

    if the anatomical reconstruction is used. The

    other method used by Asfaw and colleagues for

    estimating its limb lengths, which is based on

    regression formulas, seems to be less reliable. It

    would suggest a femur length relative to humerus

    length that is much greater than in modern

    humans, combined with a relative forearm

    length that exceeds the range of extant African

    apes.

    BOU-VP-12-1s brachial proportion is the

    highest of all fossil hominids. It was, however,

    probably not as highly derived among hominids asclaimed by Richmond et al. (2002). If the humerus

    length is anatomically reconstructed, and accord-

    ing to the exact length of the forearm, it does not

    seem to be markedly different from those of OH

    62, large-sized A. afarensis specimens, or A. africa-

    nus (see Table 4). The discrepancy with Richmond

    and colleagues conclusions might be attributable

    to their inappropriate use of indices. Indeed, only

    when the humerus length is estimated from a

    regression (Asfaw et al., 1999) does the brachial

    proportion of BOU-VP-12/1 lie beyond the 95%prediction limits of chimpanzees and the corre-

    sponding difference to OH 62 and the australo-

    pithecines is less often observed in the extant

    hominoid samples (see also Fig. 6). Consequently,

    the humerus length estimated by the regression

    method seems much less reliable. What is more,

    Asfaw et al. (1999) did not provide 95% prediction

    limits for their estimate. There is therefore no

    indication for unique brachial proportions of

    BOU-VP-12/1 among hominids. Based on the

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 453

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    22/33

    above evidence, it is conceivable that BOU-VP-

    12/1 belonged to H. habilis, but full description

    of this specimen must be awaited before further

    conclusions can be drawn. So far, the Bouri post-

    cranials have not been assigned to a particular

    hominid genus or species. Asfaw et al. (1999)hesitated to attribute it to A. garhi, which they

    discovered in the same strata (see also DeGusta,

    2003). Indeed, the megadontia of A. garhi would

    not easily fit to a close phylogenetic relationship

    with H. habilis.

    Relative femur length

    Whereas Jungers (1988) estimated OH 62s

    length to be slightly shorter than that of AL 288-1,

    Johanson and Shreeve (1989: 286) wrote: But

    based on the distances calculated between anumber of different anatomical landmarks, we

    could estimate the femur was slightly longer than

    Lucys, somewhere between 300 and 330 milli-

    meters, or roughly a foot long. This makes the

    OH 62 femur 20 to 50 mm longer than that of

    AL 288-1, yielding equivalent body proportions

    in these two hominids. There are, however, no

    landmarks preserved on the distal end of OH 62s

    femur fragment on which such a calculation

    could be based. Thus, the only way to estimate

    OH 62s femur length is by a comparison with

    other hominid femora.

    If a modern human femur lacking its distal half

    is reconstructed by comparison with a another

    femur, Korey (1990) found that its length can be

    estimated with an accuracy (i.e., a standard devia-tion) of3.05 cm if the mean diameters of the two

    femora differed by no more than 2 mm. Scaled to

    the length of the AL 288-1 femur, the accuracy of

    OH 62s length estimate would probably be within

    about 2 cm. However, as demonstrated in Fig. 7,

    the femur of australopithecines has, like that of

    great apes, a different length to girth ratio than

    that of modern humans. This suggests that the

    actual range for the estimate of OH 62s femur

    length might be much greater than assumed by

    Korey.In fact, not only do the humerus, ulna, radius,

    and craniodental morphology and proportions dif-

    fer considerably between OH 62 and AL 288-1

    (Johanson et al., 1987), but the morphology and

    proportions of the proximal femur diverge quite

    markedly as well, suggesting that there might be

    better analogues for reconstructing the fragmen-

    tary femur of OH 62. Both the humerus and ulna

    of OH 62especially if adjusted for their greater

    lengthare much more slender than those of

    Table 4

    Probabilities of sampling differences in ulna-to-humerus length proportions as great as those observed between pairs of fossils

    from extant hominid groups based on the slope of the major axis

    Fossil pair Homo Pan Gorilla PongoOH 62AL 288-1 min. 0.02 0.03 0.02 0.18

    OH 62AL 288-1 max. 0.92 0.43 0.90 0.62

    OH 62AL 137-50 0.23 0.47 0.36 0.66

    OH 62MAK-VP-1/3 0.63 0.73 0.82 0.84

    OH 62BOU-VP-12/1 anatomical rec. 0.09 (0.010.28) 0.01 (0.000.06) 0.09 (0.010.28) 0.10 (0.010.38)

    OH 62 max.BOU-VP-12/1 rec. by regression 0.03 (0.000.10) N/A (N/A 0.00) 0.02 (0.000.10) 0.04 (0.010.09)

    AL 137-50BOU-VP-12/1 anatomical rec. 0.42 (0.080.89) 0.05 (0.000.23) 0.41 (0.070.87) 0.23 (0.060.46)

    AL 137-50BOU-VP-12/1 rec. by regression 0.18 (0.020.47) 0.00 (N/A 0.04) 0.17 (0.020.45) 0.09 (0.020.20)

    MAK-VP-1/3BOU-VP-12/1 anatomical rec. 0.14 (0.020.40) 0.00 (N/A 0.03) 0.13 (0.010.38) 0.07 (0.010.17)

    MAK-VP-1/3BOU-VP-12/1 rec. by regression 0.04 (0.000.16) N/A (N/A 0.00) 0 (0.000.15) 0.02 (0.000.06)

    N/A signifies that an equally large difference was not observed in the comparative sample. OH 62: brachial proportions are based

    on a humerus length of 270 mm and an ulna length of 245 mm. BOU-VP-12/1: humerus length according to the anatomical

    reconstruction or the regression-based estimate (Asfaw et al., 1999); ulna length corresponds to the mean estimate of anall-hominoid-regression of ulna to radius length; upper and lower 95% prediction limits in brackets. AL 288-1: directly joining the

    two ulna fragments combined with a humerus corrected for crushing of the head yields the minimum brachial proportion

    (Hausler, 2001); the maximum proportion is based on Kimbels et al. (1994) estimate. AL 137-50 and MAK-VP-1/3: the brachial

    proportions were obtained by combining the reconstructed humeri with the AL 438-1a ulna.

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465454

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    23/33

    AL 288-1. It is therefore conceivable that the

    remarkably thin femur was originally relatively

    long as well.

    The only known Plio-Pleistocene femur that

    matches the OH 62 fragment in morphology and

    metrics is the OH 34 femur. Although its geologi-

    cal age might be more than 650,000 years younger

    than OH 62 (not taking into account reworking of

    the deposits at JK2), this is much less of a discrep-

    ancy than the ca. 1.4 million years between OH 62

    and AL 288-1. Critical for the use of this specimen

    as an analogue for estimating femur length of

    OH 62 is the issue of its abnormal slenderness

    raised by Day and Molleson (1976). Considering

    the entire body of evidence, there is no unambigu-

    ous support for substantial abrasion of the femoralshaft. Apart from numerous superficial scratches,

    the idea that abrasion affected the morphology of

    OH 34 is contradicted by the retention of fine

    muscle markings and sharp edges of the foramen

    nutricium and pilaster. Only the head and neck are

    clearly abraded. With respect to the ratio of vari-

    ous shaft diameters to femur length, it falls within

    the range of modern humans. Moreover, due to

    the dearth of contemporaneous early Homo

    femora, it cannot be concluded that its robusticity

    exceeds their range of variation. There is thereforeno conclusive objection against using this femur

    instead of AL 288-1s to estimate femur length of

    OH 62. With this approach, an essentially human-

    like humero-femoral proportion is indicated for

    Homo habilis.

    Alternatively, presuming the same femur length

    as that of AL 288-1, the humero-femoral propor-

    tions of OH 62 would fall close to the lower

    95% prediction limits for chimpanzees (Fig. 8).

    However, only if the upper limit of its humerus

    length estimate is simultaneously assumed does the

    resulting difference between the humero-femoral

    proportions of the two specimens become rare in

    extant hominoids, i.e., it is observed in less than

    5% of the sample (see Table 5). This contrasts with

    the results of Richmond et al. (2002), who con-

    cluded that the difference between OH 62 and AL

    288-1 significantly exceeds the variation in extant

    hominoid species. This discrepancy is partly due to

    their use of indicessince humero-femoral pro-

    portions do not change isometrically with body

    size (cf. Table 1), the use of indices can yield

    misleading results (Aiello, 1981). Another factor

    that explains a minor part of this discrepancy is

    their use of estimates for AL 288-1s humerus and

    femur length that do not make allowance forcrushing (237 mm and 280.5 mm, vs. 246 mm and

    286 mm, which we prefer, see Table 5).

    Further support for a femur that is humanlike

    in being long relative to its girth in H. habilis

    comes from the KNM-ER 1472, 1481, and 3728

    femora. Their robusticity is close to the mean of

    modern humans (Fig. 7). If they are correctly

    assigned to H. rudolfensis, and if H. habilis

    sensu stricto had a femur as short as claimed by

    Johanson et al. (1987) and Hartwig-Scherer and

    Martin (1991), it would follow that these closelyrelated species, which some scholars do not

    recognize as different species at all, showed major

    differences in relative hindlimb proportions.

    On the other hand, Ruff (1995) states that the

    OH 62 proximal femur differs from other early

    Homo specimens but might resemble P. boisei

    (KNM-ER 1500) and other australopithecines

    in having a nearly circular cross-sectional shaft

    morphology. Cross-sectional shape is, however,

    slightly positively allometricmajor axes and

    Table 5

    Probabilities of sampling differences in humero-femoral length proportions as great as thoses observed between AL 288-1

    (corrected for crushing) and various short-legged variants of OH 62 from extant hominid groups

    Femur and humerus length estimates for OH 62 Homo Pan Gorilla PongoFemur length=AL 288-1, lower limit for humerus length (258 mm) 0.17 0.21 0.50 0.61

    Femur length=AL 288-1, mean humerus length (264 mm) 0.08 0.10 0.32 0.44

    Femur length=AL 288-1, upper limit for humerus length (270 mm) 0.03 0.04 0.21 0.30

    Femur length=300 mm, upper limit for humerus length (270 mm) 0.12 0.29 0.65 0.72

    Femur length=AL 288-1 (uncorrected), mean humerus length (264 mm) vs. AL 288-1 (uncorrected) 0.02 0.04 0.20 0.28

    M. Haeusler, H.M. McHenry / Journal of Human Evolution 46 (2004) 433465 455

  • 8/14/2019 Journal of Human Evolution 46 (2004) 433465

    24/33

    reduced major axis are nearly identical [slopes for

    modern humans 1.13 (95% confidence limits 1.051.23) and 1.09 (1.001.18)], indicating that smaller

    specimens such as OH 62 and OH 34 have a more

    circular proximal shaft cross-section than larger

    individuals (Fig. 9). Figure 9 also shows that,

    compared to the modern human regression line,

    the 11 A. afarensis proximal femora possess,

    on average, a slightly wider mediolateral shaft

    diameter relative to the anteroposterior shaft

    diameter (but see Ruff, 1995, 1998; Ruff et al.,

    1999). An even wider proximal shaft diameter is

    exhibited by specimens attributed to H. rudolfensis

    and H. erectus (Ruff

    , 1995). They all cluster closeto the upper limit of the range of variation for

    modern humans. Yet, based on the relatively

    restricted range of variation compared with

    other fossil and extant hominoid groups, this

    could in part be an artefact due to failure of

    taxonomic recognition of specimens that have a

    proportionally smaller mediolateral diameter.

    On the other hand, the P. boisei proximal

    femur KNM-ER 1500 lies almost on the regression

    line of modern humans with respect to its

    Fig. 9. Bivariate plot with reduced major axes (long dashed lines) and major axes (short dashed lined) of sub-trochanteric mediolateralvs. anteroposterior diameter (mm) for modern humans, chimpanzees, gorillas, orang-utans, and early hominid fossils. Least squaresregressions (not shown) are flatter. The inner and outer dotted lines repres