Morphometrics of the Neandertal Talus JOHK G. RHOADS' AND ERIK TRINKAUS Lkportment of Anthropology, Peabody Mccseum, Harvard University, Cambridge, Massachmetts 021 38 KEY WORDS Talus . Morphometrics . Neandertals . Fossil hominids. ABSTRACT A number of morphometric analyses of Neandertal tali since the turn of the century have failed to reach a consensus on the functional affinities of these fossil foot bones. 'To clarify the problem a univariate and multivariate analysis of the available Neandertal and Skhd tali in relation to those of modern humans was performed using nine linear dimensions and four angles. Our analysis indicates that Neandertal tali are indistinguishable from modern human tali in the implied locomotor capabilities and similar in overall size and proportions. The primary differences between the fossil and modern tali involve the greater articular robustness of the fossils, probably to compensate for higher levels of biomechanical stress. The evolution of an efficient bipedal gait among the hominids involved the development of a unique pedal morphology, one which enables the support, propulsion and balance of the body during a single legged stance. This pedal structure is reflected primarily in the hominid pedal arches and hallucial robusticity. It is evident as well in the hominid talus. As the central element in the universal joint between the leg and the subtalar skeleton, the talus must form a stable structure with the tibia, fibula and tarsals while conforming to the mechanical requirements of the talocrural and subtalar articulations. The hominid talus has, therefore, evolved as a short, high bone which concentrates the joint reaction forces close to the midline of the calcaneus. Appreciation of the importance of the talus for hominid locomotor variation and evolution has led to the creation of numerous methods for quantifying talar morphology (e.g.,Volkov, '03-04;Martin and Saller, '57; Day and Wood, '68; Lisowski et al., '74; Trinkaus, '75a) and to the detailed description of fossil hominid tali, in particular those of the European and Near Eastern Neandertals. The Neandertal talus was initially characterized as distinct from modern human tali in ways which implied a less AM. J. PHYS. ANTHROP., 46: 29-44. efficient bipedal gait (e.g,, Fraipont, '12; Boule, '11-13). Yet, recent studies of Neandertal tali (e.g., Endo and Kimura, '70; Heim, '72; Trinkaus, '75a) have indicated that Neandertal tali are indistinguishable from those of modern humans in the implied locomotor capabilities. All of the differences appear to be related t o the robustness of the tali. This robustness is seen in the other Neandertal foot bones (Trinkaus, '75a) and in the rest of their postcranial anatomy (Taylor, '68; Musgrave, '70; Heim, '72; Trinkaus, '76b). Yet, several recent multivariate analyses of fossil hominid tali (Oxnard, '72, '73a,b, '75a; Lisowski et al., '743 have separated Neandertal tali from samples of modern human tali ("Neandertal tali (Spy, Skhd) are less like modern man than previously thought" (Oxnard, '72: p. 12! 1, placing them about four "standard deviation units" from modem humans (Oxnard, '72, '75a; Lisowski et al., '74).2 These statistical difI Current address: Department of Anthropology, Yale University, New Haven, Connecticut 06.520. 2 Recently Oxnard U 5 h , personal communication), referring to the same rnriltivariate analyses, has denicd the cxistence of a significant difference hehveen Neandertal and modern human tali ("Thrse studirs show unequivocally that several different Neanderthalrrs are similar to man and quite different from non-human primattes" (Oxnard, '7%:p. 393) 1. He does not specify the precise functional implications of his statistical Ftatemmts. 29 vi 0 0 0 N m N W ID U U 0 U U 0 W U Ln 0 c n r In U 01 h zr e1 5 In I- cn cn v) h 5 5 m cn m 3- Y m N c ?J W F. a m a m c 3 r a n M IT M c W ?I M 0 I- c F L) t- 3 r W c a 5 W !3 s el P .-t 1 m U W a, W N N N w m c W c a E c T B 0- N U W c I- N W N w I- 0 W w I I I 0 0 0 0 W N c N w Y Ln W .A P P m U W c” n M P n c in u) Ln w m N N W 0 0 w 0 I- m N m N u W N U U I I N 3 W w N rg Ic I m N w 0 0 0 N I I N - 0 N m u3 P 0 v 0 N 0 rb 0 N 0 W V P w 0 m N m w m 0 L- in 0 Ln m m n F Ln 0 01 n c c ID n II I W r( n U W Ln P- n ’P m I- n I c- + n W Ln 0 h U m c W 0 v I N Ln e w N ID 0 I N 0 0 Ln I II Ln 0 N u N) W I I N -4 W W m P N m ? 0 W .@ N P c m w N IN D c w 0 N U m N W in W N W W 0 0 W 0 W P w U Ln N P- II N N W m N m P W 0 P v 0 4 N I- I- m N P w w 0 VI 0 W 0 u N W I- 0 u N) 0 Ln W 0 L” N 0 0 Y In I- W W w II -4 P m 0 0 I W I- P v( 0 N Ln 0 m 0 I- W m m 1 I I I w w 0 U m 0 Y Ln w m 0) P P w N m U u) w , I 0 N N 0 P u3 m w 0 P m W c W 0 rb N U W 01 VI \D U w m U W v 0 w N 0 U W c N P W N c 0 N ca 0 P W 0 in N 0 0 6 s F 1 50 -¶ m M M m m !- m 0 U c P W 0 P M m 01 01 I- N L? E. ID ? 0 0 m Ln m Q) N m W W N W w w 0 W 0 u W) u 0 0 0 w 0 0 N I- 0 v1 c P rb N 0 W in W N 0 P- O N P UI P 0 0 c c 0 P z n. c V Ic m c c n c VI U N m u W m Ln cn 0 u3 N w I?- v1 N U w m N PI -4 w 0 4 0 0 N c W U m 0 c in 0 N f U c 0 U N m W -4 w N in sN !- n M I- n Ln P N w Ln W w m II D - c P h P f W v N 01 U h U J. Y 0 I- 0 4 N m W 0 0 ?- Pu) U 0 n 0 v N N F c N c W IN N m W P Ln P N Y N W I u) 0 N c 0 0 N v I- I m + 0 U N ‘0 v1 P 0 N N u P I c N .A U W < N m P Ln U P m W - Ln 0 N Y N 0 0 U N n I I I e R Ln c Length w N m N w m 0 c 0 m I- W 0 P m Physiol. Height Trochlear Length Trochlear Height Head-Neck Length Articular Breadth u) N m U c m P N in w N 0 N 0 w m :m P Trochlear Breadth Lat era1 Malleolar Breadth La t er a 1 Ma l l e o l a r Height c E 4 fir 5g. 2 3‘ Trochlear Angle Neck Angle Torsion Angle Subtalar Angle NEANDERTAL TALAR MORPIIO.\lETRICS 31 and body inertia. In response to these habitual stresses of locomotion, the trabecular bone, through a process of microfractures and secondary formation of trabeculae, will remodel so as to concentrate bone in the regions of maximum stress (Pugh et al., '73). The remodeling is reflected in the reorientation of the trabeculae and the relative increase in the MATERIALS, MEASUREMENTS AND METHODS volume of trabecular bone, and therefore, This analysis is based on the study by one in the shape of the bone (Ascenzi and Bell, of us (E, T.) of the original European and '72; Lanyon, '74). Additionally the articular surfaces will Near Eastern Neandertal and Skhd hominid postcranial remains (Trinkaus, '75a; ta- change in orientation and size to best ble 1). The "Neandertals" are defined as transmit joint reaction forces while conthe Upper Pleistocene hominids from forming to the geometric requirements of Europe and the Near East included in the the articulations. Articular cartilage is taxon Homo sapiens neanderthalensis highly sensitive to excess pressure and (Campbell, '65). The remains from Mug- under high impact compression may unharet es-Skhd and Djebel Kafzeh, which dergo fibrillation and degeneration (Trias, are morphologically separate from this '61; Radin and Paul, '71). Since the presgroup (Vandermeersch, '72; Howells, '75; sure on a unit area of cartilage is inversely Trinkaus, '76a), are considered early Homo proportional to the surface area bearing sapiens of uncertain phylogenetic affinities the force, an increase in the total articular surface transmitting the joint reaction to the Neandertals. All of the measurements were taken on forces will proportionately reduce the the original fossil specimens except those compressive stress on the articular caron Kiik-Koba 1 (BM(NH1cast No. EM 2081, tilage. Therefore, under conditions of high Shanidar 1 (cast No. 381,217 of T. D. joint reaction forces, an expansion of the Stewart), and Skhd 4 (McCown and Keith, articular surface would be adaptive. TO '39 data and PUM cast No. 603). For com- reflect the results of these processes on the parison, samples of modern human adult talus, we have used length, physiological tali from the Late Woodland Amerind site height and articular breadth to indicate of Libben (collections of the Department the overall proportions of the talus, and of Anthropology, Kent State University), trochlear length, trochlear breadth, trochthe predynastic Egyptian site of Keneh, lear height, head-neck length, articular and the medieval Yugoslav site of Mistihalj breadth, lateral malleolar breadth and (collections of the Peabody Museum, Har- lateral malleolar height to quantify the relative sizes of the trochlear and malleolar vard University) were measured. The measurements were chosen to quan- surfaces (APPENDIX). The articulations of the talus, proximally tify the relative proportions, articular dimensions, and articular geometry of the with the tibia and fibula (the talocrural hominid talus. Since the talus occupies a joint) and distally with the calcaneus and central position between the leg and the navicular (the subtalar and mid-tarsal subtalar skeleton, its morphology is closely joints), permit the plantarflexion-dorsiflexcorrelated with the requirements of the ion of the foot and the pronation-supinaankle and subtalar joints. Additionally, the tion of the subtalar skeleton (Hicks, '53; talus must transmit most of the body Elftman, '60). The angles employed were weight and reaction forces generated by chosen to quantify the articular mechanics the tibia1 musculature, body momentum of the talus. The trochlear angle is a meaferences do not appear to reflect accurately the anatomical differences between Neandertal and modern human tali. We have therefore collected metrical data from all of the available Neandertal and Skhul tali sufficiently intact for analysis (table 1) and from samples of modern human tali and have compared them statistically. 32 JOHN C RHOADS 4ND ERIK TRINKAUS sure of the anteroposterior wedging characteristic of recent hominid talar trochleae (Day and Napier, '64). The wedging is necessary to accommodate the trochlea between the malleoli as the talus is abducted and adducted on the tibia during dorsiflexion and plantarflexion (Barnett and Napier, '52).This movement contributes to the medial displacement of force trajectories through the foot at push-off, thereby concentrating the stress on the stronger medial pedal arch. The subtalar angle is an estimate of the orientation of the subtalar joint, the primary articulation responsible for subtalar movement and pronation-supination of the foot (Manter, '41; Hicks, '53); therefore limitation of movement at this joint strengthens the foot as a lever for locomotion. The subtalar articular axis among the hominids is oriented more anteroposteriorly than among the pongids so as to permit more efficient resistance to the bending stresses on the pedal arch (Elftman and Manter, '35;Preuschoft, '70). On the dorsal aspect of the talus this articular orientation, as well as the mediolateral splaying of the talus, is indicated by the neck angle. The pedal arches are also maintained by the limitation of movement at the mid-tarsal joint (Hicks, '53;Elftman, '60). Its articular stability is reinforced by the nun-parallel orientation of the talonavicular and calcaneocuboid articular axes, which is accentuated by the torsion of the talar head (Elftman, '60). This feature is measured by the talar torsion angle. Other measurements possible on the talus are either redundant with those included here or contribute little to our understanding of talar variation. Non-metric or discrete variations of the talus, although they provide valuable information for the functional interpretation of Neandertal tali (Trinkaus, '7%), are not relevant to this analysis. The linear measurements were taken with standard sliding, spreading and coordinate calipers, while the angles were measured with a protractor with a moveable indicator. The linear measurements are considered reliable to the nearest 0.5 mm; the angles are reliable to the nearest whole degree. The computer programs used were MULTIVARIANCE (Finn, '74) for the multivariate analysis of variance and covariance and the single degree of freedom contrasts with their associated discriminant functions, SPSS (Nie et al., '75) for one-way analysis of variance and discriminant functions, D/DA (Rhoads, '73) for distance analysis and canonical variates, and BMD (Dixon, '75) for crosschecking analysis of variance and discriminant functions. Where possible all computations were cross-checked with at least two, and usually three, different programs. Numerical errors are common, even in the widely used programs, so cross-checking is necessary. In the statistical analyses the linear dimensions were used directly, rather than as indices. This simplifies interpretation of the results, and avoids the potentially nonGaussian distributions of indices (Simpson et al., '60).Angles have their own statistical problems, but were included to represent features of joint orientation otherwise difficult to quantify. They fit a normal model satisfactorily over the restricted range of variation in the present samples. Relations among the measurements The joint variation of the dimensions of a structure within a population deserves careful study, both as a necessary preliminary to detailed interpretation of group differences and for its interest as a reflection of the influences determining the form of the structure. The talar measurements from the modern samples were fist tested for heterogeneity of covariance matrices. Since there were no significant differences, a pooled covariance matrix was computed which provided the basis for further study of the intrapopulational measurement relationships. Table 2 shows that positive correlations exist among nearly all linear measurements except lateral malleolar breadth. The rela- 33 NEANUERTAL TALAR bIORPHOMETRICS m 4 3 u7 0 3 r. m N . I m 4 N 0 4 a U 0 .yl, n 3 m g, 0 9 0 0 i 9 m m N In m 0 Q .3 r m 0 N 3 x r I n . 0 a N 3 d ul OI 9I 0 3 * N 4 I N N N m i d 3 ul 0 0 I I I I N U 3 In 0 m m N m m r. 0 4 0 9 I m m 9 4 in N 0 9 r. i yl, d Is U 9 N 0 0 yl, 9I 0 .Y ? .r 3 3 -? m I m m rl m 0 0 N I N 9 0 9 N U x N In d 2 0 0 0 m m m C-l 0 Y .I . 4 I I 0 .rin 0 9 9 9I I U y1 I 0 2 II N m VI ul 4 2 0 m 1 d N m I 9 I N m vl m I I I U 8 I II 4 m .n In m m I I I I 1 I I I N I1 9I * : 4 YI 2 m 2 a h 3 N 03 0 I I i 0 0 m 0 3 “7 0 u3 0 3 N .J 0 In 0 U 0 0 01 N 0 I I .4 4 m .r I 0 U N m 9 I r. 3 N 0 I I I N 9 N 3 m r. 0 I I I In N 0 CD U 3 1 I 4 3 a r. N r- d 1 u3 .r N 0 N O ul m 0 I m m 9 N s 0 m m I I I 0 U h 0 m 3 yl, m 0 I I I I I ul 1 ID ul N 1 N * I I 2 m u7 ? ul m ? ID N m 0 .n cl U IA m m .n ? N “1 \D m o u M W SM R“ 62 rD m m \D ul 0 m .rm 03 ID 0 r- ? ul r. 0 . I m w 4 ul QI m m 0 ul N 9I 9 4 rl In m h 9 m .n 0 *m 0 0h 9 U u 0) 0 J. N In .3 r 0 rl 0 I \O 0 A 3 N m N .n A C\ -3 9 0 3 0 ul 0 2 I 4 2 N rc 4 . 9 0 m, B 9I m 4 u U 3 I 8 3 0 0 u m N 0 m I ? d 4 00 0 cl N m 0 .4 I 4 UJ !-I m u3 m In I 9I N 0 N rD 7YI4 0 u3 9 m 3 4 N 0 0 I w i M &I m 34 JOHN C;. RHOADS AND ERIK THINKAUS tionships indicated by the correlation lesser degrees of displacement (Anthony, coefficients are moderately strong ( r = '23; Trinkaus, '7%). As this would also 0.4-0.75)but nowhere near deterministic. affect trochlear length, it is consistent with This configuration suggests that much of the finding that the highest correlation bethe variance in the linear measurements is tween trochlear height and another talar associated simply with the general size of dimension is with trochlear length. the bone. Length best reflects general size, Further details of the covariation among since it has the highest and most generally measurements are obscured in the simple distributed positive correlations with the correlation matrix by the association of other linear measurements, as well as the nearly all linear dimensions with the highest loading on the first principal com- general size of the bone. New details ponent, which accoimts for 40% of the emerge when general size effects are restandardized variance in the measure- moved from the correlation matrix by comments and clearly represents general size. puting partial correlations holding length Lateral malleolar breadth is conspicuous constant (table 2). in its high variability and its low level of When length is held constant, most of correlation with other measurements. the other relations between variates disapSome of this is no doubt due to the pear. Trochlear dimensions are a sigdifficulties of measuring such a sinall nificant exception. Although the breadth of dimension consistently, but variance of ob- the trochlea loses its sizable correlations servation is not sufficient to account for when length is controlled for, suggesting this entire pattern. It must be principally that it is largely determined by the size of due to the relative absence of direct rela- the bone, there is a moderate residual tionships between lateral malleolar association (partial r = 0.43) between the breadth and the other talar measurements, height and the length of the trochlea. This since lateral malleolar breadth is deter- supports the interpretation of the trochlear mined by a complex biomechanical and sagittal dimensions as jointly dependent architectural interaction of the trochlear, upon the sagittal extent of the trochlea lateral malleolar and posterior calcaneal rather than merely upon overall size. articular surfaces. The highest correlation The associations of the head-neck length of lateral malleolar breadth with another show another aspect of talar measurement dimension, articular breadth, is tautologi- relationships. Its simple correlations are of cal, since it is included in articular breadth a straightforward general size configuraas defined here. Lateral malleolar breadth tion (correlation with length = 0.69).The is also weakly associated with the height of most telling of its partial correlations with the lateral malleolar facet, which has a length removed is a slight negative one closer but still weak pattern of association (partial r = - 0.13) with trochlear length. with other talar dimensions. Since head-neck length is measured from The dimensions of the trochlea have an the anterior limit of the trochlea (APPENinteresting pattern of correlations. Both DIX), the inverse variation of head-neck the length and breadth of the trochlea length and trochlear length indicates that show strong general size associations, but some of the variation in head-neck length trochlear height has a lower set of correla- as it is measured is actually variation in the tions with other linear dimensions of the distal extent of the trochlea. talus ( r = 0.4-0.6).These correlations may The angular measurements show only reflect the dependence of trochlear height the weakest associations with each other on the degree of habitual angular displace- and with the linear measurements. The ment at the ankle, which promotes the ex- highest correlation between two angles, tension or retreat of the articular surface 0.28 between the neck angle and the subanteriorly and posteriorly with greater or talar angle, reflects their similar anatomical 35 XEiLl UERTAL TALAR MORPHOMETRICS bases (see above). With linear measurements the correlation of 0.23 between the subtalar angle and trochlear height, and correlations of neck angle with length ( - 0.20)and trochlear breadth ( - 0.21) are high enough to mention; most of the remaining correlations involving angles are very low. As these are too low to explain much of the variation within populations, talar angles clearly bear no substantial allometric relations within populations of modern adults. Either they poorly reflect real structural variation (technical inadequacy), they are largely determined by separate genetic or biomechanical influences, or the range of morphologies present in our modern samples is insufficient to show allometry. significance levels of the analysis of variance contrast: X, - (1/3X , + 113 XL + 1/3 X,) (11 where XN is the Neandertal mcan, and XK, &, and X,, are the modern group means. Program MULTIVARIANCE (Finn, ’74) was used to compute both univariate and multivariate versions of this “Neandertal effect,” taking into account the differences in group size. The Neandertal-modern differences in the actual measurements are physically small and statistically non-significant at the 0.05 level for 8 of the 13 measurements (table 3). While the Neandertal sample is small, the confidence intervals are narrow enough to assure us that if marked differences existed they should be evident. It is Metric comparison of Neandertal and the functiona2 magnitude of group differmodern Homo sapiens tali ences that is important, not whether the The measurement means of the Nean- groups are statistically distinguishable. dertals and the three modern groups (table Biologists often confuse the distinct issues 3) show no major differences. For most in- of biological significance and of statistical dividual measurements the Neandertal significance; the latter is merely a measure mean falls within the spread of means for of the state of our knowledge, a measure of the modern groups. The means of the the completeness of our sampling. Given Skhul talar measurements are also close to sufficiently large samples, statistically “sigthe modern means for most measurements. nificant” group differences are a foregone When the equality of variances in the conclusion. Depending upon their genetic groups was tested for each variate using or functional basis, to the biologist these Bartlett’s test (Snedecor and Cochran, ’671, may signify anything, or nothing at all. One of the statistically reliable Neanderthere was no significant heterogeneity except in trochlear length, for which the tal-modern differences is the greater laterNeandertals had a slightly elevated vari- al extension of the lateral malleolar facet in ance due to one anomalous value (an ex- the Neandertals (increased lateral malleolar breadth), a difference in means of ceptionally low value for Tabm C l ) . The spread among the modern groups almost 2 mm in a dimension of about 8 mm. throws open to question the precise mean- The overall articular breadth, measured ing of any overall mean vector of talar across the lateral malleolar facet, the dimensions for modern H . sapiens, except trochlea and the medial malleolar facet, is perhaps as a rough functional Gestalt for greater by 3.4 mm in the Neandertal samstriding bipeds. Even so, it is instructive to ple. The difference in lateral malleolar examine the univariate contrasts between breadth accounts for some of this, and the eleven Neandertal specimens on which since the trochlear breadths of the Neanall measurements could be taken and a dertal and modern samples appear to be common modern mean obtained b y substantially the same (observed differweighting the Keneh, Libben and Mistihalj ence of less than 0.1 mm), presumably samples equally. Table 3 presents least there is on the order of 1 mm greater squares estimates with standard errors and average medial extension of the medial 36 .s m 4 C 0 a 9 9 f 03 N 3 c1 0 9 z4 r. v) m N Y) N G 0 0 0 - m m -.3 n m ti N 3 3 c . * 0 3 0 c 0 N N ; $ g " N c m "l 3 0 .c a 0 +' :: h +1 g t: N 3 c 0 N $1 Y N 4 tI x.Y) 01 N 4 +! ti Y 0 0. z d h N rn OI U 4 0 0 rn Y 0 '7 N N N U U 9 c 3 U + N pl 0 " h ' I I "7 N 0 Lo m +, 0 0 3 .4 n ". m 9 a ?! . 0 +I 0 m 4 "7 c) "7 .? " N 3 4 0 c) N 3 9 0. 3 N +I sN . U N U 2 m 3 c Y) d 0 P# 5 Y .G - +I 0 1 : " 4 e N N N w .o 3 m rn N 3 c) r. N n . N "l El 9 9 i i c -4 N h N rn c1 0 N N U n ? . 3 N ? N 0 j! ? v7 c 01 N U h CI m 3 Y) 3 0 yi 81 0: N 3 h 1 0 +I f' U N N ' I I d CI N Y) m 0: " rn N U N .o YI 0 m N +I ? 3 m h Q h N 'PI If L- N * N 0 N 0 fi n 0 m 3 0 0 2 " "l 01 h ID I n . m . l YI rn T.'l 0 N N N If 8 m h m * 2 v) Y) 4 'D ;t g1 m N Y) 3 0 I? m " c 3 m 3 Y) " 2 4 Ln 0 TI $1 e 3 N 21 4 2 Zd h m 0 +I N N N Y) m h 0 2 5 5 0 9 c L-) m N 3 +I 0 (Y -f r c 4 d 4 9 y1 .-I s. 0 e N $1 m. 3 9 +I 0 c In rn Y c) ul +I m ". 5 3 Y) N m 9 0: 0 3 m x +I "3 " Q ? m N . I 0 0 +I ? 0 d 4 9 $1 h 01 A, .r ul 9 m 3 +I 0 0 In -4 m 0 -f m 3 U N 0 m m +I r- N h G $1 9 n 0 if $1 0 3 +I .m n m N -f. 2 +I U v7 3 9 n * N 0 +I +I 9 3 c 4 M c c 37 NEANDERTAL TALAH M O R P H O ~ ~ E T H l ( 5 malleolar facet (which was not included as an individual measurement to avoid redundancy). This supports the observations of Gorjanovic-Kramberger (’061, Martin (’101, Fraipont (’12) and Boule (’11-131 that the Neandertals had enlarged malleolar surfaces. Relative to the modern samples, the mean length of the trochlea in the Neandertal sample is greater by 2.2 mm, and the head-neck length is reduced by the same amount. Since the mean overall length of the Neandertal tali falls between the modern means and since head-neck length is measured from the distal edge of the trochlea, it is likely that the apparent difference in head-neck length is merely a reflection of the difference in trochlear size (see above). This entire suite of differences, both from n priori considerations and from the study of the covariation of measurements within modern groups, would be expected to reflect use-connected modeling of articular surfaces. Neandertals are seen to differ primarily in the increased robusticity of the talocrural joint. The only remaining statistically reliable difference, a greater physiological height of the talus among the Neandertals, complements the articular robustness by concentrating trabecular bone along the primary stress paths in the talus (see above). MULTIVARIATE RESULTS TAB1.F 4 0 2 c;alzre~ brtween modPrn samples and hetween modern satnples and the Neandertalr Keneh Lihben Mistihalj 3.19 3.04 2.82 2.06 2.12 2.30 ble 4) shows that the Neandertal mean vector is not much more remote from modern group means than the modern groups are remote from one another. The distinctiveness of the Neandertals is even less impressive when individual variation is taken into account, some individual modern specimens are closer to the Neandertal mean vector than to their own population mean, and vice versa. (The Skhul tali were not included in the multivariate analyses due to the small sample size and the incompleteness of the specimens.) This scale of individual variation must be kept in mind when interpreting multivariate analyses. There is a further trap for the unwary in the practice of expressing Mahalanobis distances between samples or individuals in “standard deviation units.” A distance of three or four “standard deviation units” will sound quite remote to anyone accustomed to thinking of two standard deviations as an approximate 95% point for the “two-tailed” comparison with a univariate normal distribution. This apparent analogy is seriously deceptive. The multivariate “standard deviation units” do not correspond to the standard deviation units of the familiar univariate case. The univariate “distance,” (x-d)/s, between a single specimen with measurement value r and a sample mean %, standardized b y the sample standard deviation s, has a null distribution (probability distribution under the null hypothesis that the single specimen is drawn from the same normally distributed population as the sample) related to Student’s t. Specifically, the quantity Our analysis of individual measurements (table 3) and our multivariate distance analysis based on all measurements (table 4)show marked variations among modern populations. In many individual measurements differences between modern groups are greater than between moderns and Neandertals. This does not imply there are functional distinctions between modern groups. It does indicate, however, that the comparison of Neandertals and modern H. sapiens must be made with a full appreciation of the amount of variation existing both within and between modern popula(21 tions. Inspection of multivariate distances (ta- has the Student’s t distribution with N-1 38 JOHN C. RHOADS AND ERIK TRINKAUS degrees of freedom, where N is the size of the sample to which the single specimen is k i n g compared. This distribution approaches the normal as N becomes large. This result, however, does not apply where more than one variate is involved. In the general p-variate multinormal case, the Mahalanobis L I Z distance can he represented as: uz = (TC-x)‘S-’(ir-x) 13) which is the squared Euclidean distance between the sample mean vector x and the individual specimen’s measurement vector x in the metric defined by the sample variance-covariance matrix S . It has a null distribution related to Hotelling’s T2, so that (4) is distributed as Hotelling’s P w i t h param- eters p and N-p-1. This distribution may be transformed to the more commonly tabulated F distribution by using the expression, 15) (All of the above equations are special cases of well known results in distribution theory, for which proofs can be found in Rao, ’65).As N becomes large in relation to the number of variates, the null distribution of DZ between an individual and a sample approaches a chi-square with p degrees of freedom. So the Euclidean distance between the individual specimen and the sample does not have a simple relation to the normal distribution, but rather its square is asymptotically chi-square distributed with a number of degrees of freedom equal to the number of measurements on which it is based. It is the matter of degrees of freedom which invalidates the analogy to the normal deviate. For example, the Euclidean distance between the Spy 2 talus and the pooled modern sample in Day and Wood’s (’68) multivariate analysis of primate tali is 3.36 in “standard deviation units.” A normal deviate of 3.36 would have a very low probability. However, a proper approxi- mate comparison is of L I Z = (3.3612 = 11.3 to a chi-square variate with seven degrees offreedom. A distance from the population mean vector this great or greater will occur in about 12% of the individuals in a multivariate normal population. (This figure is obtained using the chi-square approximation; using the exact distribution will give an even higher probability.) So these multivariate “standard deviation units,” when referred to the proper scale, indicate that the Spy 2 talus is closer to the modern mean in that study than are many modern tali. Similarly, the multivariate distances obtained in the present study, based on a larger measurement set, indicate extensive population overlap. Standard coefficients of the individual measurements on the canonical variates proved disappointing in the evaluation of the morphological bases of the multivariate distinctions. In general, unless the group differences have an especially simple mathematical form, the composition of the canonical variates is some occult function of the particular groups one chooses to include. The usual analysis yields canonical variates which discriminate maximally among all the groups, with a democratic lack of regard for which differences may be particularly interesting under the investigator’s hypotheses. Thus in the present study, a traditional canonical variates analysis gave a typically unintelligible combination of measurements differentiating the modern groups mixed in with the Neandertal-modern distinction. We wished to separate the overall multivariate among-group variation into portions for (1) the Neandertal-modern distinction, and (21 modern interpopulation variation. This was done by setting up a multivariate analysis of variance with a single degree of freedom contrast opposing the Neandertals to all the modern groups (the multivariate analogue of the “Neandertal effect” used in the univariate analyses), and two orthogonal contrasts reflecting the variation between modern groups. A canonical variate associated with just the Neandertal-modern contrast was YEANDER I i L -14 L \I{ 39 \IC)RPHOhlFTHIC~ TALILb 5 Standardized camniccil tjariute coefficient5 Nrandertal-modern conNedndert,ll-rn~Jdc.rI, trast. with covariance djustrrient for length contrast Length Physiological height Trochlrar length Trochlear height Head-Neck length Articular breadth Trochlear breadth Lat. rnalleolar breadth Lat. malleolar height - 0.08 0.59 0.57 - 0.43 - 0.76 0.46 0.36 0.2: - 0.14 Trochlear angle Neck angle Torsion angle Subtalar angle - 0.29 0.06 - 0.20 - 0.14 ~ ~ computed using the MULTIVARIANCE program (Finn, '74). It showed differentiation of the Neandertals on the basis of increased malleolar widths, greater trochlear length and corresponding lesser headneck length, and greater physiological height (table 5 ) . Thus a canonical variate was obtained having an interpretation consistent with the univariate results, even though the traditional overall canonical variates had no clear anatomical interpretation. The Neandertal mean for talar length is quite close to a central value of modern man (table 31, so it appears that the multivariate distinction between the Neandertals and the modern samples is not due in any large measure to simple relations of the linear measurements to overall size (table 5). This notion was further tested by performing a multivariate analysis of covariance, controlling for length. With this model, that part of the between-group variation in other measurements accounted for by their linear correlation with the controlled variable is removed from the multivariate comparison. The distinctions among the modern groups were substantially reduced b y this adjustment. On 13 measurements without covariance, the Fvalue for the multivariate 0.40 0.38 0.38 - 0.55 0.32 0.23 0.27 - 0.12 First canonicd n r iatr hetween modern gi-onps attrr controlling for length 0.41 - 0.63 0.16 ~ ~ - 0.28 - 0.06 - 0.20 - 0.14 0.24 0 4-3 0.62 0.11 0.10 - 0.11 - 0.31 - 0.17 0.28 ~ heterogeneity among modern groups was 7.64 (on 26 and 240 degrees of freedom, p < 0.0001).On the other 12 measurements with length as a covariatc, this was reduced to 4.88 (on 24 and 270 degrees of freedom, p < 0.0001). So overall size was responsible for a considerable part of the between-group variation in modern tali. Of the remaining between-group variation, 78% was associated with a canonical variate clearly interpretable as reflecting trochlear size (table 5 ) . This recalls a similar pattern in the within-group variation and is further evidence for trochlear size being in some measure determined separately from overall size of the talus. The covariance adjustment for length had a quite different effect on t h e multivariate distinction between Neandertal and modern tali. The Neandertal-modern distinctness was greater for the other twelve measurements with length as a covariate (I;= 5.90 on 12 and 135 degrees of freedom, p < 0.0001)than it was for all thirteen measurements without covariance ( F = 5.46 on 13 and 135 degrees of freedom, p < 0.0001). Thus, with a covariance adjustment to simulate equal overall size of Neandertal and modern tali, they were less alike than without this adjustment. Therefore, the difference may 40 JOHN G. HHOADS AYD ERIK TRINKAUS be attributed to a true distinction of shape rather than merely size. The covariance adjustment did not change the pattern of the canonical variate associated with the Neandertal-modern distinction (table 5 ) . Implications of the Neandertal talar metrics The morphometric analyses of the Neandertal tali presented, univariate and multivariate, show the Neandertal tali to be largely overlapping the ranges of variation of the modern human samples. The only differences involve a relative enlargement of the talocrural articular surfaces, especially the trochlea and lateral malleolar surface, and a slightly greater physiological height. These differences can all be related to the postcranial robustness characteristic of the Neandertals. The relative increase in talocrural articular surface area represents an adaptation to exaggerated joint reaction forces associated with high activity levels. Since an increase in articular dimensions would reduce the stress per unit area of the joint surface, the large dimensions of the Neandertal trochleae would be an adaptive response to the conditions of habitually high joint reaction force undoubtedly present on the Neandertal lower limbs (as inferred from their general postcranial robustness). The relative increase in the physiological height of Neandertal tali is a related phenomenon; it concentrates bone capable of absorbing and transmitting large joint reaction forces below the trochlea and between it and the calcaneal surfaces of the talus. The enlargement of the malleolar surfaces of Neandertal tali appears to be correlated with the increase in joint reaction force across their trochleae. With high activity levels the lateral shear stress at the ankle would increase with trochlear joint reaction force, thereby placing greater stress on the medial and lateral malleoli. This is especially true at heel-strike, when the downward movement of the fibula stabilizes the talocrural joint but also increases joint reaction forces (Weinert et al., '73). An adaptive response would be the circumferential enlargement of the malleoh, via trabecular remodeling (Piigh et al., '731. It is not possible to assess directly the robustness of the Neandertal fragmentary malleoli. The dimensions of the talar malleolar surfaces, however, demonstrate that they were indeed accentuated. This does not support the contention of Bode ('11-13) and others that the Neandertal fibulae supported a greater proportion of the body weight than among modern humans. Due to their architectural positions relative to the talar neck and the posterior calcaneal surface respectively, the medial and lateral surfaces are limited in the extent to which they can extend distally by the dorsoplantar dimensions of the talus. They must also provide appropriately positioned grooves for the long flexor tendons. The enlargement of the malleoli and the corresponding malleolar surfaces of the talus to compensate for large joint reaction forces must therefore be primarily medial and latcral. The greater articular breadth of the Neandertal tali therefore correlates with the expansion of their trochlear surfaces, since both are related to higher levels of stress at the talocrural joint. None of the angles employed showed any significant differences between the Neandertal and modern samples (table 3). The talar neck angle, the one considered by Boule ('11-131 to differentiate the Neandertals most clearly, is virtually identical in the modern and fossil samples. The differences in the trochlear angle and torsion angle means, which are lower among the Neandertals, are overshadowed by the great variation in both modern and fossil samples. The measurements of the limited sample of Skhd tali (tables 1, 31 are aligned with those of the Neandertals rather than of modern humans for most of the cases where a statistical difference exists between the fossil and modern tali. In some cases, such as physiological height and trochlear length, the Skhd mean is situated further from the modern means than KEAhDERThL TAL.AR MORPHOMETRICS the Neandertal one, but the relative shortening of the head-neck length of the Neandertals is not present. Although these measurements do not support the supposed phylogenetic alignment of the Skhul hominids with modern H. sapiens rather than with the Neandertals (Stewart, ’60; Howells, ’701, they are nonetheless in agreement with the robustness of portions of the Skhul postcranial skeletons bctween that of the Neandertals and of modern humans (McCown and Keith, ‘391. CONCLUSIONS The Neandertal and Skhul hominid tali appear to be largely indistinguishable from those of modern Homo sapiens. The only consistent differences can be referred to the articular and structural reinforcement of the bone to absorb habitually high levels of joint reaction force at the talocrural and subtalar articulations. Furthermore, this analysis demonstrates the need in all such studies to evaluate carefully the suitability of statistical techniques, the relative contributions of the various measurements to the statistical differences between the samples, the true statistical significance of the distance statistics, and most importantly, the underlying biological significance of the quantitative differences. ACKNOWLEDGMENTS We would like to thank the many individuals who made original specimens available for study. This research was supported by Wenner-Gren Grant 2979 and N. I. H. Grant 5 TO1 GM01938. Drs. S. J. Gould, W. W. Howells, D. Pilbeam, C. E. Oxnard and A. C. Walker provided helpful comments, and to them we are most grateful. LITERATUHE CITED Anthony, R. 1923 Le dkterminisme et l’adaptation morphologiquos en biologie animale. Arch. morph. gen. cxpkrim., 14. Ascenzi, A,, and G . H. Bell 1972 Bone as a mechanical engineering problem. In: The Biochemistry and Physiology of Bone. Second ed. Academic Press, New York, 1: 311-352. 41 Barnett, C. H., and J. R. Napier 1952 The axis of rotation at the ankle joint in man. Its influence upon thc: form of the talus and the mobility of the fibula. J. h a t . , 86: 1-9. Bode, M. 1911-13 L’homme fasille de La Chapelleaux-Saints. Ann. Palhnt., 6111-172; 7: 21-56, 85-192; 8: 1-70. Campbell, B. G. 1965 The Nomenclature of the Hominidae. Royal Anth. Inst. Occ. Papers, 22. Day, M. H., and J. R. Napier 1964 Horninid fossils from Bed I. Olduvai Gorge, Tanganyika: Fossil foot bones. Nature, 201: 969-970. Day, M. H., and B. A. Wood 1968 Functional affinities of the Olduvai Horninid 8 talus. Wan., .3: 440-455. Dixon, W. J. 1975 BMDP: Biomedical Computor Programs. University of California Press, Berkeley. Elftman, H. 1960 The transverse tarsal joint and its control. Clin. Orthop.. 16: 41-46. Elftman H., and J. T. Manter 1935 The evolution o f tho human foot, with espc:cial refertmcc to tht: joints. J. h a t . , 70: 56-67. Endo, B., and T. Kirnura 1970 Postcranial skeleton of the Amud Man. In: Thc: Amlid Man and his Cavc: Site. H. Suzuki and F. Takai, eds. Academic Press of Japan, Tokyo, pp. 231-406. Finn, J. D. 1974 A Genc:ral Modcl for Multivariattl Analysis. Holt, Hinehart and Winston, New York. 1912 L’astragale chez I’hommc: Fraipont, C. moustkrien dc Spy. S e s affinittis. Bull. Soc. h t h r o p . Bruxelles, 31: 195-221. GorjanoviC-Kramhrger, D. 1906 Der diluvialc Mensch von Krapina in Kroatien. Ein Beitrag zur Palaoanthropologie. C. U’. Krridel’s Verlag, Wiesbaden. Heim, J.-L. 1972 Les Nkpdertaliens adultes de 1 3 Ferrassie (Dordogne). Etudes anthfopologiquc et comparative. These de Doctorat d’Etat, UniversitP de Paris \’I. Hicks, J. H. 1953 ’I’he mechanics of the foot. I: The joints. J. h a t . , 87: 345357. Howells, w. W. 1970 Mount Carrnel Man: Morphological relationships. Proc. VIIIth Intl. Cong. Anthrop. Ethnol. Sci., 1 : 269-272. -1975 Neanderthal man: Facts and figures. In: Paleoanthropology: Morphology and Paleoecology. R. H. Tiittle, ed. Mouton Put)., Paris, pp. 389-407. Lanyon, L. E. 1974 Experimental support for the trajectorial theory of bone structure. J. Bone Joint Siirg., S6R: 160-166. Lisowski, F. P., G. H. Alhrecht and C. E. Oxnard 1974 The form of the talus in some higher primates: A multivariate study. Am. 1 . Phys. Anthrop., 41: 191-216. Manter, J. T. 1941 Movements of the subtalar and transverse tarsal joints. h a t . Rec., 80: 397-410. Martin, H. 1910 Astragale hnmairi du Mousterien moyen de La Quina. Ses afhitks. Bull. Soc. prkh. franpise, 7; 391-397. Martin, R., and K. Saller 1957 Lehrbuch der Anthropologie. Verlag von Gustav Fischer, Stuttgart. McCown, T. D,, and A. Keith 1939 The Stone Age of 42 JOHN G. HHOADS AND ERIK TRINKAUS Mount Carmel 11: The Fossil Human Heniains from the Levalloiso-Moustcirian. Clarendon Press, Oxford. Musgrave, J. H. 1970 An anatomical study of the hands of Pleistocene and Recent man. Ph.D. Thesis, University of Cambridge. Nie, K.H., C. H. Hull, J . G. Jenkins, K. Steinbrenner and D. H. Rent 1975 SPSS: Statistical Package for the Social Sciences. Second ed. Mc(;raw-Hi!l, New York. Oxnard, C. E. 1972 Some African fossil foot hones: A note on the interpolation of fossils into a matrix of extant species. Am. J. Phys. Anthrop., 37: 3-12. l973a Form and Pattern in Human Evolution. LJniversity of Chicago Press, Chicago. 197311 Functional inftrrr:nces from morphometrics; problems posed by uniqueness and diversity among the primates. Syst. Zool., 22: 409-424. 1975a Uniqueness and Diversity in Human Evolution. University of Chicago Press, Chicago. 197513 The place of the australopitheoines in human evolution: grounds for doubt? Nature, 258: 389-395. Preuschoft, H. 1970 Functional anatomy of the lower extremity. In: The Chimpanzee. G. €1. &)urne, ed. S. Korger, New York, 3: 221-294. Pugh, J. W., H. M. Hose and E. L. Radin 1973 A possible mechanism of Wolffs Law: Trabecular niicrofractures. Arch. internat. Phy-siol. Bioch., 81: 27-40. Radin, E. L., and I. L. Paul 1971 Rcsponsr ofjoints to impact loading. I. In ciitro wear. Arthritis Rheum., 14: 356-362. Rao, C. H. 1965 Linear Statistical Inferencc and its Applications. John Wiley and Sons, New York. Rhoads, J. G. 1973 D/DA: Dspersion/discriminant analysis program. Unpub. program report, Dept. of Anthropology, Harvard University. Simpson, G. G., A. Roe and R. C. Lewontin 1960 Quantitative Zoology. Harcourt. Brace & World, Inc., New York. Snedecor, G. W., and W. G. Cochran 1967 Statistical Methods. Iowa State University Press, Ames. Stewart, T. D. 1960 Form of the pubic bone in Neanderthal man. Science, 131: 1437-1438. Taylor, J. V. 1968 The Neanderthal Tibia. Ph.D. Thesis, Columbia University. Trias, A. 1961 Effect of persistent pressure on the articular cartilage. An experimental study. J. Bone Joint Surg., 43B: 376-386. Trinkaus, E. 1975a A functional analysis of thr Neondertal foot. Ph.D. Thesis, University of Pennsylvania, in press. 1975b Squatting among tht: Neandertals: A proldcm in the behavioral interpretation of skeletal morphology. J. Arch. Sci., 2: 327-351. 1976 The morphology of European and Southwest Asian Neandertal pubic bones. Am. J. Phys. Anthrop., 44: 95-104. 1976h The evolution of the hominid femoral diaphysis during the Upper Pleistocene in Europe and the Near East. Z. Morph. hithrop., 67: 291-319. 1976 Les metatarsiens et phalanges du pied des Neandertaliens cle Spy. Bull. Inst. roy. Sci. nat. Belgique, in press. Vandermeersch, B. 1972 Rkcentes dkcouvertes de squelettes humains ii Qafieh (Israel): essai d’interpretation. In: The Origin of Homo sapiens. F. Bordes, ed. UNESCO, Paris, pp. 49-54. Volkov, T. 1903-04 L e s variations squelettiques du pied chez les Primates et dans les races humaines. Bull. Mkm. Sac. Anthrop., Paris, SBrie 5, 4: 632-708; 5: 1-50, 201-331. Wcinert, C. R., Jr.,J. M.McMaster and R. J. E’erguson 1973 Dynamic function of the human fibula. Am. J. Anat., 138: 145-150. NEANDERTAL TAIAR VORPHOMETRICS APPENDlX Talur measurements Length (M-1) 1 From the M. flexor hallucis longus groove to the most anterior point 011 the head measured parallel to the sagittal axis of the trochlea. Physiological height (M-31) From the most dorsal point on the trochlea to the most dorsal point on the posterior calcanoal surface measured perpendicular to the sagittal axis of the trochlea in a dorsoplantar direction. Trochlear length (M-4) Maximum length of the trochlear articular surface on the midline mcasured parallel to the sagittal axis of the trochlea. Trochlear height (M-6) From the chord defined by the Trochlear Length to the highest point on the midline sagittal axis of the troclllea measured in the sagittal plane of the trochlea. Head-neck length (M-8) From the anterior median edge ofthe trochlea to the furthest point on the head measured parallel to the long axis of the head and neck. Articular breadth (M-2b) From the lateral edge of the lateral inalleolar surface to the medial edge of the mcdial malleolar surface measured perpendicular to the sagittal plane of the trochlea. Trochlear breadth (M-5) From the mid-lateral edge of the trochlea to its mid-medial edge measured perpendicular to the sagittal plane ofthe trochlea. Lateral malleolar breadth (M-7a) From the lateral edge of the trochlea to the lateral edge of the lateral malleolar surface measured perpendicular to the sagittal plane of the trochlea. Lateral malleolar height From the dorsolateral edge of the trochlea to the lateral edge of the lateral inalleolar surface measured parallel to the sagittal plane of the trochlea in a dorsoplantar direction. Trochlear angle The angle between the medial edge of the trochlea and the lateral edge of the trochlea measiired in thc horizontal plane. Neck angle (M-16) The angle between the sagittal axis of the trochlea and the long axis (sagittofrontal)ofthe head and neck measured in the horizontal plane. Torsion angle (M-17) The angle between the long axis of the navicular articulation (medioplantar to laterodorsal) and the horizontal plane of the trochlea measured in thr: frontal plane. Subtalar angle The angle between the sagittal axis of the trochlea and the midline axis across the mcdial and posterior calcaneal surfaces measured in the horizontal plane. I (M - no.) refers to the number of the equivalent measurement in Martin and Saller 157). 43

1/--страниц