Journal of Human Evolution 46 (2004) 299–315 A re-evaluation of the metric diversity within Homo erectus James H. Kidder*, Arthur C. Durband University of Tennessee, Department of Anthropology, 250 S. Stadium Hall, Knoxville, TN 37996, USA Received 29 July 2003; accepted 12 December 2003 Abstract Previous work by several researchers has suggested that the cranial sample from Zhoukoudian possesses a unique metric pattern relative to the African and Asian specimens assigned to Homo erectus. The current study readdresses this issue with an expanded fossil sample and a larger and more comprehensive set of cranial measurements. To test the patterns present in the assemblage, canonical variates analysis was performed using a covariance matrix generated from the Howells data set. From this, interindividual Mahalanobis distances were computed for the fossils. Random expectation statistics were then used to measure statistical significance of the Mahalanobis distances. The results show that the Zhoukoudian hominids exhibit a unique metric pattern not shared by the African and Indonesian crania sampled. In these tests the Hexian calvaria resembled the African and Indonesian specimens and differed significantly from the craniometric pattern seen in the Zhoukoudian fossils. The Zhoukoudian specimens are characterized by a wide midvault and relatively narrow occipital and frontal bones, while the African and Indonesian crania (including Hexian) have relatively broad frontal and occipital dimensions compared to their midvaults. These results do not suggest that a multiple-species scenario is necessary to encompass the variation present in the sample. Based on the current evidence it is more probable that this variation reflects polytypism influenced by environmental adaptation and/or genetic drift. 2003 Elsevier Ltd. All rights reserved. Keywords: Zhoukoudian; Ngandong; Sangiran; Sambungmacan; Homo erectus; Species; Speciation Introduction The number of human species that existed during the Pleistocene has been the subject of considerable debate during the past 20 years. The origins of the current discussion can be traced back to the 1983 Senckenberg conference, when Andrews (1984), Stringer (1984), and Wood (1984) proposed that the fossil material subsumed under * Corresponding author. Tel.: +1-865-974-4408; fax: +1-865-974-2686 E-mail addresses: jimpithecus@utk.edu (J.H. Kidder), adurband@utk.edu (A.C. Durband). Homo erectus should be split into African and Asian species. According to this scheme, a number of autapomorphic features, including the angular torus, sagittal keel, and the tympanomastoid fissure, are limited to specimens from Asia. Thus, the Asian fossils, by virtue of including the type specimen from Trinil, would retain the species name H. erectus (Andrews, 1984; Stringer, 1984; Wood, 1984). The African material, which lacked these derived features, would be assigned to H. ergaster, with the KNM-ER 992 mandible as its type specimen (Groves and Mazák, 1975). Critics of this view argue that the traits in question have a 0047-2484/04/$ - see front matter 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.jhevol.2003.12.003 300 J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 continuous distribution within the H. erectus sample, as well as other hominid species (e.g., Kennedy, 1991; Bräuer and Mbua, 1992). Multivariate analysis has also been brought to bear on this issue. Work by Kramer (1993) suggested that the H. erectus sample from Africa and Asia is metrically homogenous and cannot be separated into regional groups. Kramer (1993) examined a relatively large fossil sample, but due to poor preservation of that sample only three cranial variables were utilized in his multivariate study. A number of recent projects by Antón (2001a, 2002; Antón et al., 2002) have highlighted some geographic differences in cranial morphology among the Asian fossils attributed to H. erectus. However, these studies all utilized a maximum of four linear cranial measurements (and included cranial capacity as a fifth variable). This small number of variables leaves a great deal of potential variation unexplored. In addition, Antón (2001a, 2002; Antón et al., 2002) did not provide information on the statistical significance of plots generated by her analyses. In this paper we reexamine the possibility that multiple taxa are currently subsumed within H. erectus. We attempt to quantify the level of metric variation contained within this temporally and geographically diverse assemblage using both an expanded fossil sample and improved cranial coverage. Kidder (1998) suggested that the Chinese specimens from Zhoukoudian exhibit a metric pattern that is not found in the crania from other sites in Asia and Africa. While he was careful to note that this seemingly unique metric pattern in the Chinese specimens probably did not warrant a separate taxonomic designation for those fossils, Kidder (1998) did urge further tests of those results. The current project expands on this earlier work and attempts to explicate any patterns of variation that might exist in the fossil sample currently subsumed under H. erectus sensu lato. Materials and methods The sample used for this project is shown in Table 1. It consists of twenty of the most complete hominid crania commonly described in the Table 1 Fossil sample used in this project KNM-ER 3733 KNM-ER 3883 OH 9 Ngandong 1 Ngandong 6 Ngandong 7 Ngandong 11 Ngandong 12 Sangiran 2 Sangiran 17 Pithecanthropus Skull IX (Tjg-1993.05) Sambungmacan 1 Sambungmacan 3 Sambungmacan 4 Zhoukoudian III Zhoukoudian V Zhoukoudian X Zhoukoudian XI Zhoukoudian XII Hexian literature as belonging to H. erectus (Weidenreich, 1943; Jacob, 1975; Santa Luca, 1980; Rightmire, 1990; Wu and Poirier, 1995; Arif et al., 2001, 2002; Antón et al., 2002; Baba et al., 2003). This is an extremely diachronic sample, ranging from approximately 1.8 Ma for the earliest African material (Wood, 1991) to possibly as young as 50 Ka for the Ngandong material (Swisher et al., 1996). A reference sample of known covariance structure is provided by W.W. Howells’ (Howells, 1973, 1989) modern human data set, which includes 2354 individuals from a number of different geographic populations. For the purposes of this study, Zhoukoudian Skull III is considered a young adult specimen. While Black (1929) initially identified this skull as an adolescent, he later (1930) modified this diagnosis to that of an early adolescent male approximately equivalent to a modern human at twelve years of age. Weidenreich (1943) essentially agreed with Black, but gave the specimen a slightly younger age of eight to nine years. Mann (1981) disagreed with these assessments, and instead took the position that this specimen is a young adult based on the development of muscle markings and the overall size of the cranium. Antón (2001b) concurred, and placed Zhoukoudian III at an J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 Table 2 Cranial variables analyzed in this project; definitions for each measurement can be found in Howells (1973, 1989) Maximum Cranial Length (GOL) Maximum Frontal Breadth (XFB) Maximum Cranial Breadth (XCB) Biauricular Breadth (AUB) Biasterionic Breadth (ASB) Frontal Chord (FRC) Parietal Chord (PAC) Occipital Chord (OCC) older adolescent/young adult developmental age. Due to the work of Mann (1981) and Antón (2001b), as well as other results placing this specimen near the Zhoukoudian adults in overall cranial shape (Durband et al., 2002), the decision was made to include Zhoukoudian III in the sample. The recently described “Pithecanthropus Skull IX” from Sangiran (Tjg-1993.05) is also included in this analysis, though it should be noted that sizable inconsistencies exist in the published measurement sets that are available for this fossil (e.g., Arif et al., 2001, 2002; Tyler and Sartono, 2001), and additionally the specimen requires a more concerted effort at reconstruction (Arif et al., 2002). For this project we will utilize the more complete set of measurements provided by Arif and colleagues (Arif et al., 2001) for their most recent reconstruction of Skull IX. The variables used in the study are shown in Table 2, and consist of eight linear cranial measurements. Due to the fragmentary nature of the fossil material, variables were chosen based on their availability on the fossils as well as their presence in the Howells data set. Measurements for the fossil sample were obtained from the literature (Weidenreich, 1943; Santa Luca, 1980; Rightmire, 1990; Wood, 1991; Wu and Poirier, 1995; Arif et al., 2001; Delson et al., 2001; Baba et al., 2003). To test the null hypothesis that the fossil sample encompasses a single species, the Howells data set was used to create a pooled covariance matrix from 28 samples of modern human crania. Canonical variates and interindividual Mahalanobis distances were then computed for the fossil samples after Jantz and Owsley (2001). Four dif- 301 ferent tests were run using six, six, seven, and eight variables, respectively, in order to maximize coverage of the cranial sample and include less complete specimens. The variables were transformed using the Darroch and Mosimann (1985) shape adjustment technique. While Hawks and colleagues (Hawks et al., 2000) have recently criticized the use of size adjustment techniques in craniometric studies of this nature, other workers have found ample justification for their use (e.g., Mosimann and James, 1979; Darroch and Mosimann, 1985; Simmons et al., 1991; Corruccini, 1992). When employing this methodology, each variable for each individual (row in the data matrix), is divided by the geometric mean for all variables. This results in a dimensionless shape variable. The effect of this technique is to standardize all crania to one size. Once the Mahalanobis distances have been calculated, the fossil crania can be compared to one another by the use of random expectation statistics. Through this method, the distance between pairs of individuals randomly selected from a population will be distributed as 公(2pⳮ1) with a variance of 1, where p is the number of dimensions (Defrise-Gussenhoven, 1967). This random expectation statistic can be used to determine whether the distance between the fossils exceeds that expected between individuals drawn from the Howells data set (Jantz and Owsley, 2001). A distance greater than 1.65 standard deviations above the random expectation value reflects statistical significance by a one-tailed test (Jantz and Owsley, 2001). While a modern human reference sample might not be ideal for testing distances between fossil crania, due to the constraints imposed by the relatively small fossil sample available this is currently the only alternative. Indeed, the Howells data set thus provides one of the best data sets that can be applied to this problem. Results The first analysis consists of six variables and includes sixteen fossil crania (Table 3). This set of measurements was chosen in an effort to include Ngandong 6 and OH 9 in the analysis. The results of this test are plotted in Fig. 1, and the distances 302 Analysis 1 Fossils Sangiran 17 Pithecanthropus IX Ngandong 1 Ngandong 6 Ngandong 7 Ngandong 11 Ngandong 12 Sambungmacan 3 Sambungmacan 4 KNM-ER 3733 KNM-ER 3883 OH 9 Zhoukoudian III Zhoukoudian XI Zhoukoudian XII Hexian Analysis 2 Measurements GOL XCB AUB ASB FRC OCC Fossils Sangiran 2 Sangiran 17 Pithecanthropus IX Ngandong 1 Ngandong 7 Ngandong 11 Ngandong 12 Sambungmacan 3 Sambungmacan 4 KNM-ER 3733 KNM-ER 3883 Zhoukoudian III Zhoukoudian V Zhoukoudian XI Zhoukoudian XII Hexian Analysis 3 Measurements GOL XCB XFB AUB ASB OCC Fossils Sangiran 17 Pithecanthropus IX Ngandong 1 Ngandong 7 Ngandong 11 Ngandong 12 Sambungmacan 1 Sambungmacan 3 Sambungmacan 4 KNM-ER 3733 KNM-ER 3883 Zhoukoudian III Zhoukoudian X Zhoukoudian XI Zhoukoudian XII Hexian Analysis 4 Measurements GOL XCB XFB AUB ASB FRC PAC Fossils Sangiran 17 Pithecanthropus IX Ngandong 1 Ngandong 7 Ngandong 11 Ngandong 12 Sambungmacan 3 Sambungmacan 4 KNM-ER 3733 KNM-ER 3883 Zhoukoudian III Zhoukoudian XI Zhoukoudian XII Hexian Measurements GOL XCB XFB AUB ASB FRC PAC OCC J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 Table 3 Fossil samples and cranial measurements used in multivariate tests J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 303 Fig. 1. Canonical variate plot of subset 1. obtained are shown in Table 4. In this test, clear separation exists between the Zhoukoudian crania and the rest of the sample along the first canonical variate. The only non-Chinese specimen grouping with the Zhoukoudian fossils in this analysis is Sambungmacan 3. The first variate accounts for 38.3% of the variance in this test, and is primarily influenced by a high negative loading on biauricular breadth and a positive loading on maximum cranial breadth. The second canonical variate accounts for a further 33.6% of the variation present in this sample and is primarily driven by a high positive loading on biasterionic breadth and a high negative loading on frontal chord. Thus, the analysis separates crania with relatively broad midvaults (the Zhoukoudian crania) from those with relatively narrower midvaults (African and Indonesian crania) along variate 1, while variate 2 separates skulls with relatively broad occipitals and relatively short frontals from those with narrower occipital breadths and longer frontal chords. The Mahalanobis distances generated by this test, provided in Table 4, show two clear tendencies. First, the distances among the Zhoukoudian crania are not statistically significant, indicating similar cranial shape despite the potential time depth at this site (Grün et al., 1997; Shen et al., 2001). Secondly, the African and Indonesian samples do not show a consistent pattern of significant distances from one another, though nearly all of these crania are significantly different from all of the Zhoukoudian fossils. The only exceptions to this pattern are Pithecanthropus Skull IX and Hexian, both of which show significant distances to all but three other crania in the analysis, and Ngandong 12 and KNM-ER 3883, which are significantly different from only three other crania in the analysis. Sambungmacan 3 is also an 304 Sang17 PithIX Ngan1 Ngan6 Ngan7 Ngan11 Ngan12 Sam3 Sam4 ER3733 ER3883 OH9 ZHIII ZHXI ZHXII Hexian Sang17 PithIX Ngan1 Ngan6 Ngan7 Ngan11 Ngan12 Sam3 Sam4 ER3733 ER3883 OH9 ZHIII ZHXI ZHXII Hexian 0.000 6.404 4.243 3.928 2.589 3.155 2.823 4.246 3.836 5.125 2.456 3.868 4.742 6.092 6.822 5.324 0.000 4.625 4.336 5.127 3.656 5.449 7.389 5.768 7.563 5.891 5.950 6.826 6.191 6.793 9.821 0.000 4.817 2.610 2.202 4.396 6.344 3.795 6.178 4.197 5.998 7.277 7.660 8.326 7.437 0.000 3.815 3.178 2.382 5.665 5.325 4.992 4.117 3.268 4.907 5.203 6.388 7.554 0.000 2.418 2.308 4.068 2.658 4.089 2.028 4.610 4.974 5.642 6.445 5.407 0.000 3.385 5.763 3.501 5.867 3.304 4.161 5.799 6.340 7.115 7.181 0.000 3.818 4.181 3.157 2.617 3.701 3.989 4.634 5.783 5.625 0.000 5.498 4.959 4.222 6.397 3.954 4.088 4.200 5.982 0.000 4.879 1.937 4.514 5.063 6.250 7.090 4.657 0.000 3.769 5.246 4.833 5.635 6.979 4.334 0.000 3.389 3.719 5.175 6.168 4.067 0.000 4.042 5.646 6.951 6.080 0.000 2.486 3.613 5.663 0.000 1.701 7.587 0.000 8.550 0.000 J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 Table 4 Mahalanobis distances obtained for subset 1. Distances >4.960 are statistically significant at the P=0.05 level and are shown in bold type J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 305 Fig. 2. Canonical variate plot of subset 2. exception to this pattern, exhibiting similarities to the Zhoukoudian fossils based on these six cranial measurements. The second analysis consists of six variables and includes sixteen fossil crania (Table 3). This analysis was designed to include Zhoukoudian V and Sangiran 2. The results of the test are plotted in Fig. 2, and the Mahalanobis distances obtained are shown in Table 5. As in the previous six-variable test, the Zhoukoudian crania and the remainder of the sample separate along the first canonical variate. Here, this variate accounts for 44.3% of the variation and is driven by high negative loadings on maximum frontal breadth and biasterionic breadth and a high positive loading on biauricular breadth. Variate 2 accounts for 32.5% of the remaining variation, and is influenced by a high negative loading on maximum cranial breadth. Thus, breadths of the frontal and the occipital sort the crania along variate 1, while maximum cranial breadth drives the distribution seen on variate 2. The Mahalanobis distances obtained from this analysis, displayed in Table 5, show a clear separation between the Chinese grouping and the African and Indonesian specimens. Once again, the Zhoukoudian crania show no statistically significant distances from one another, though each shows several significant distances from the nonChinese fossils. The African and Indonesian samples group together to the exclusion of the Chinese fossils, with the notable exception of Pithecanthropus Skull IX, which is statistically significantly distant from every other fossil in the analysis. As with the previous six-variable test, Ngandong 12 and KNM-ER 3883 show the greatest similarity to the overall sample, differing significantly from only two and five other crania, respectively. Finally, Hexian continues to group 306 Sang2 Sang17 PithIX Ngan1 Ngan7 Ngan11 Ngan12 Sam3 Sam4 ER3733 ER3883 ZHIII ZHV ZHXI ZHXII Hexian Sang2 Sang17 PithIX Ngan1 Ngan7 Ngan11 Ngan12 Sam3 Sam4 ER3733 ER3883 ZHIII ZHV ZHXI ZHXII Hexian 0.000 5.550 5.959 3.406 3.394 4.206 4.537 7.131 1.616 6.774 2.186 5.283 5.495 6.751 7.548 4.531 0.000 8.131 4.529 3.074 3.511 3.525 4.116 4.858 4.598 3.898 6.189 5.761 7.095 8.014 4.646 0.000 5.439 6.145 5.011 6.056 9.243 5.768 8.924 5.871 6.276 5.300 6.150 6.925 9.504 0.000 2.423 2.033 3.918 6.863 2.904 5.570 3.395 6.771 6.351 7.476 8.550 5.738 0.000 2.425 2.172 4.550 2.616 3.809 1.989 5.146 5.202 5.872 6.951 3.991 0.000 3.291 6.092 3.476 5.607 3.283 5.954 5.051 6.561 7.604 6.137 0.000 4.357 4.236 3.832 2.735 4.342 4.413 4.891 6.113 4.884 0.000 6.203 3.534 5.141 5.506 6.219 5.524 6.224 4.750 0.000 6.048 1.928 5.141 5.277 6.272 7.060 4.347 0.000 5.293 7.053 7.697 6.986 8.075 4.855 0.000 3.829 4.068 5.207 6.115 3.766 0.000 2.233 2.334 2.810 6.097 0.000 3.353 3.867 7.105 0.000 1.327 7.417 0.000 8.136 0.000 J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 Table 5 Mahalanobis distances obtained for subset 2. Distances >4.960 are statistically significant at the P=0.05 level and are shown in bold type J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 307 Fig. 3. Canonical variate plot of subset 3. consistently with the African and Indonesian specimens and shows statistically significant distances from the Zhoukoudian fossils, including Zhoukoudian V. The third analysis examines seven variables on a sample of sixteen fossil crania (Table 3) and was designed to include Sambungmacan 1 and Zhoukoudian X. Results are plotted in Fig. 3, and the Mahalanobis distances generated are shown in Table 6. As with the prior two analyses, the African and Indonesian crania sampled are widely separated from the Chinese specimens along the first canonical variate. Variate 1 accounts for 43.2% of the variation with high negative loadings on biauricular breadth and maximum cranial length and high positive loadings on maximum frontal breadth and biasterionic breadth. Variate 2 provides an additional 29.4% of the variation and is driven by a high negative loading on biauricular breadth, a high positive loading on frontal chord, and a moderate positive loading on maximum cranial breadth. Thus, as in the first two analyses, the crania are separated by the breadths of the frontal, occipital, and midvault. The Mahalanobis distances for this analysis, shown in Table 6, are broadly similar to those obtained in the earlier tests. The Zhoukoudian crania show no significant distances among themselves, while each is significantly separated from the majority, or even all, of the remaining fossils tested. The African and Indonesian samples show a general overall homogeneity with the notable exceptions of Sambungmacan 3 and Pithecanthropus Skull IX, each of which is significantly distant from virtually all of the specimens examined from every region. The fourth analysis examines eight variables on a sample of fourteen fossil crania (Table 3). 308 Sang17 PithIX Ngan1 Ngan7 Ngan11 Ngan12 Sam1 Sam3 Sam4 ER3733 ER3883 ZHIII ZHX ZHXI ZHXII Hexian Sang17 PithIX Ngan1 Ngan7 Ngan11 Ngan12 Sam1 Sam3 Sam4 ER3733 ER3883 ZHIII ZHX ZHXI ZHXII Hexian 0.000 8.112 4.729 2.869 3.604 3.024 3.883 3.483 4.959 4.333 4.038 6.330 6.393 7.148 7.847 5.515 0.000 5.413 6.255 4.939 6.133 5.126 9.203 6.016 9.500 6.220 6.942 6.818 6.513 6.877 10.492 0.000 2.669 2.320 4.275 3.050 6.642 3.862 6.685 4.344 7.568 7.694 8.017 8.581 7.466 0.000 2.542 3.079 2.425 4.390 3.008 5.457 2.803 5.755 6.010 6.819 7.326 5.372 0.000 2.644 2.119 5.836 3.761 5.451 3.607 6.331 6.463 6.460 7.291 7.253 0.000 1.870 5.320 3.524 4.354 2.225 4.627 5.806 5.146 6.281 5.879 0.000 5.640 2.192 5.059 1.796 4.896 5.926 5.276 6.181 6.051 0.000 6.175 6.331 5.453 6.241 5.071 7.039 6.953 6.068 0.000 5.116 1.812 5.065 6.767 5.917 6.792 4.984 0.000 4.763 7.133 8.727 7.361 8.783 5.303 0.000 3.890 5.849 5.133 6.096 4.555 0.000 3.910 3.153 3.741 6.346 0.000 4.266 3.331 8.487 0.000 2.022 8.418 0.000 9.180 0.000 J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 Table 6 Mahalanobis distances obtained for subset 3. Distances >5.276 are statistically significant at the P=0.05 level and are shown in bold type J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 309 Fig. 4. Canonical variate plot of subset 4. Results are plotted in Fig. 4, and the Mahalanobis distances generated are shown in Table 7. Not surprisingly, the Zhoukoudian crania continue to cluster by themselves and are widely separated from the African and Indonesian samples along the first canonical variate (with the notable exception of Pithecanthropus Skull IX). Variate 1 accounts for 39.2% of the variation and is driven by moderate negative loadings on maximum cranial length and biauricular breadth, and high positive loadings on maximum frontal breadth and biasterionic breadth. Variate 2 accounts for another 31.6% of the variation and is influenced by a high negative loading on biauricular breadth, a high positive loading on frontal chord, and a moderate positive loading on maximum cranial breadth. This plot is particularly noteworthy because the Zhoukoudian crania separate from the majority of the African and Indonesian crania on both axes, unlike the earlier tests where separation was manifested primarily along the first canonical variate. The Mahalanobis distances generated by this test, shown in Table 7, generally mirror those of the previous three examinations. The Zhoukoudian crania group neatly by themselves and share statistically significant distances from virtually all of the African and Indonesian specimens. And, as before, no consistent pattern of significant distances separates the African and Indonesian samples from one another, aside from Pithecanthropus Skull IX and KNM-ER 3733, which are statistically significantly distant from all but two other non-Chinese crania in the analysis. Hexian is also notable in being statistically significantly distant from all but three other specimens, including all of the Zhoukoudian skulls examined. The status of Ngandong 12 and KNM-ER 3883 is 310 Sang17 PithIX Ngan1 Ngan7 Ngan11 Ngan12 Sam3 Sam4 ER3733 ER3883 ZHIII ZHXI ZHXII Hexian Sang17 PithIX Ngan1 Ngan7 Ngan11 Ngan12 Sam3 Sam4 ER3733 ER3883 ZHIII ZHXI ZHXII Hexian 0.000 8.354 4.909 3.389 3.727 3.635 4.793 5.142 5.439 4.142 6.601 7.517 8.271 5.815 0.000 5.531 6.432 5.152 6.367 9.687 6.314 9.944 6.399 7.125 6.742 7.147 10.771 0.000 2.965 2.503 4.574 7.318 4.113 7.290 4.459 7.781 8.266 8.891 7.715 0.000 3.215 3.079 4.754 3.778 5.686 3.162 5.857 6.868 7.446 5.520 0.000 3.480 6.895 3.893 6.538 3.738 6.671 6.935 7.804 7.626 0.000 5.486 4.397 4.568 2.861 4.766 5.199 6.383 6.055 0.000 7.343 6.308 6.279 6.580 7.131 7.092 6.468 0.000 6.443 2.022 5.517 6.540 7.429 5.458 0.000 5.697 7.482 7.515 8.950 5.741 0.000 4.126 5.504 6.491 4.803 0.000 3.268 3.889 6.496 0.000 2.042 8.591 0.000 9.401 0.000 J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 Table 7 Mahalanobis distances obtained for subset 4. Distances >5.565 are statistically significant at the P=0.05 level and are shown in bold type J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 Table 8 Measure of biasterionic breadth relative to biauricular breadth, determined by the formula ASB/AUB Sangiran 2 Sangiran 4 Sangiran 10 Sangiran 12 Sangiran 17 Ngandong 1 Ngandong 6 Ngandong 7 Ngandong 11 Ngandong 12 Sambungmacan 1 Sambungmacan 3 Sambungmacan 4 Average= 0.960 1.000 0.952 0.893 0.957 0.977 0.913 0.947 0.933 0.933 0.927 0.881 0.931 0.939 ER 3733 ER 3883 OH 9 Average= Zhoukoudian Zhoukoudian Zhoukoudian Zhoukoudian Zhoukoudian Average= Table 9 Measure of maximum frontal breadth relative to biauricular breadth, determined by the formula XFB/AUB 0.960 0.930 0.911 0.934 III V X XI XII 311 0.837 0.835 0.782 0.790 0.762 0.801 Sangiran 2 Sangiran 17 Ngandong 1 Ngandong 7 Ngandong 11 Ngandong 12 Sambungmacan 1 Sambungmacan 3 Sambungmacan 4 Average= 0.810 0.900 0.923 0.879 0.910 0.844 0.847 0.851 0.840 0.867 ER 3733 ER 3883 Zhoukoudian Zhoukoudian Zhoukoudian Zhoukoudian Zhoukoudian Average= 0.920 0.814 III V X XI XII 0.720 0.754 0.748 0.741 0.715 0.736 scores and are higher than those obtained for the Zhoukoudian fossils. Discussion consistent with the previous tests and show the fewest number of statistically significant distances from other fossils in the analysis. To further examine and explain these significant differences in cranial shape, a simple formula of biasterionic breadth/biauricular breadth was applied to a sample of 13 Indonesian, three African, and five Zhoukoudian fossils listed in Table 8. The ratios indicate that the cranial vault does not narrow appreciably at the occipital in the African or Indonesian samples, indeed the samples have virtually identical means, but the Zhoukoudian occipitals are considerably narrower. It is also noteworthy that there is no overlap between the range of the Chinese crania and those from Africa and Indonesia. The Chinese and Indonesian means are significantly different at the P=0.005 level using a two-tailed t-test. When frontal breadth is evaluated relative to biauricular breadth, the Zhoukoudian crania are again found to show a different pattern than that obtained for the African and Indonesian samples (Table 9). Again, there is no overlap between the Chinese and the African and Indonesian samples and the means are significantly different at the P=0.005 level using a two-tailed t-test. While the Hexian cranium is not included in those tables, that specimen shows ratios of 0.986 for ASB/AUB and 0.819 for XFB/AUB. Both of those scores much more closely resemble the African and Indonesian The results of this study concur with those of Kidder (1998), Kidder and Durband (2000), and with recent work by Antón (2001a, 2002; Antón et al., 2002) suggesting that the crania from Zhoukoudian are characterized by a unique metric pattern not found in the crania from Africa or Indonesia. The addition of a test for statistical significance, which was lacking in most of the previous projects, allows additional weight to be placed on these findings as well. The Zhoukoudian specimens are distinguished by very large biauricular breadths and relatively small frontal and biasterionic breadths, while specimens from Africa and Indonesia have frontal, midvault, and occipital dimensions that are more similar in breadth. This latter pattern of consistently broad cranial dimensions apparently persisted throughout the considerable time depth of this fossil sample, except at the site of Zhoukoudian. We also found that the Zhoukoudian pattern does not extend to the specimen from Hexian, which instead exhibits a craniometric pattern that is much more similar to that seen in Africa and Indonesia. Thus, it would appear that the craniometric pattern seen throughout much of the Pleistocene was essentially static. Beginning with the Koobi Fora crania at approximately 1.8 Ma, a consistent pattern of relatively broad frontal and occipital dimensions as compared to maximum cranial breadth (or, 312 J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 conversely, relatively narrow maximum breadth as compared to frontal and occipital dimensions) emerged and remained the norm in both Africa and much of Asia, including the late Ngandong crania. We cannot state with any certainty whether this craniometric pattern is primitive or derived in comparison to the immediate ancestors of H. erectus, whomever they may be, though Antón (2002: 318) posited that a relatively broad posterior vault is “liable to be a primitive trait.” However, at the site of Zhoukoudian a derived pattern compared to that seen in the remainder of the sample is apparent. This Zhoukoudian pattern consists of frontal and occipital dimensions that are narrow relative to maximum cranial breadth (or a relatively broad maximum breadth compared to frontal and occipital dimensions) and it consistently separates these particular fossils from the rest of the H. erectus sample. The relatively compressed degree of variation seen in the Chinese sample is noteworthy. None of the Zhoukoudian fossils was significantly separated from another specimen from that site in any of the multivariate analyses, and in most cases the Mahalanobis distances did not exceed the expected value for a single biological population. Considering the potential time depth present in the Zhoukoudian assemblage (Grün et al., 1997; Shen et al., 2001) the consistency of the metric pattern is unexpected. A greater degree of variation was observed in the African and Indonesian samples, with a number of significant distances generated between those specimens (approximately one-third of the distances generated between African and Indonesian fossils were statistically significant in the eight variable analysis, for example). However, this variation is not patterned geographically. Indeed, in each of the four multivariate analyses crania from the same site (Koobi Fora and Sambungmacan) were significantly separated from one another. The underlying causes of the pronounced variation in the African and Indonesian samples are not yet clear. They could be attributable to temporal variation or perhaps even reflect some component of sexual dimorphism. These explanations are unlikely to suffice for the clear pattern of differences from the Zhoukoudian sample, as the Chinese specimens are likely inter- mediate in age between the earlier African and Indonesian material and the later Ngandong sample (Swisher et al., 1996; Grün et al., 1997; Shen et al., 2001; Antón, 2002). In addition, both sexes are almost certainly represented in the Zhoukoudian cranial sample utilized in this study (Weidenreich, 1943; Mann, 1981). The profound differences seen between the Zhoukoudian sample and the Hexian cranium are also intriguing. This separation has been previously noted by Antón (2001a, 2002; Antón et al., 2002) and Durband and colleagues (Durband et al., 2002). Etler (1994: 112a) suggested the possibility that H. erectus from Zhoukoudian represents a “cold-adapted racial variation within the population structure of middle Pleistocene east Asia.” Hexian, on the other hand, is not associated with faunal elements that would indicate a cold climate (Etler, 1994). The hypothesis that the Zhoukoudian hominids potentially reflect a more cold-adapted morphology is worthy of future study (see Antón [2002] for a more extensive discussion of this issue). The position of Zhoukoudian III in these examinations is worthy of mention. As noted earlier, some workers have treated it as a subadult (Black, 1929, 1930; Weidenreich, 1943). Our study places it very close to the other specimens from this site and would suggest that Skull III had approached adulthood at least in terms of its craniometric pattern. These results would support the notion that Zhoukoudian III is likely a young adult nearing the completion of growth, following the conclusions of Mann (1981), Antón (2001b), and Durband et al. (2002). The results obtained for the recently described Pithecanthropus Skull IX were also striking. This specimen was statistically significantly distant from the vast majority of specimens tested in each of the four analyses. Additionally, the fossil was often placed far from the clusters of specimens in the distance plots. However, due to the preliminary nature of the descriptions so far afforded this specimen (Arif et al., 2001, 2002; Tyler and Sartono, 2001) and the ongoing nature of its reconstruction (Arif et al., 2002) we feel that these results should be approached with extreme caution. J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 Kidder (1998) concluded that insufficient evidence existed for multiple species within H. erectus sensu lato. However, this study indicates that the Zhoukoudian cranial vaults possess metric characteristics that appear to be derived relative to the condition present in the African and Indonesian crania. These findings counter results from many non-metric examinations that assert that H. erectus cannot be subdivided (e.g. Kennedy, 1991; Bräuer and Mbua, 1992). The variation found in the current study is not patterned geographically, however. The specimen from Hexian does not exhibit metric characteristics that resemble those found at Zhoukoudian. In addition, the nonmetric traits used by many authors (e.g., Andrews, 1984; Stringer, 1984; Wood, 1984) to differentiate specimens within H. erectus do not correlate with this craniometric variation (Durband and Kidder, 2000). Thus, there is no consistent pattern of metric and non-metric variation that would indicate that a multiple-species scenario is necessarily appropriate for the fossils currently subsumed in H. erectus sensu lato, with the possible exception of autapomorphic cranial base features of the Ngandong and Sambungmacan fossils (Durband and Kidder, 2000; Durband, 2002a,b; Baba et al., 2003). The differences in craniometric pattern explored here are more likely attributable to two factors. Clinal variation could have influenced the shape of the skull at Zhoukoudian through adaptation to a colder climate (e.g. Etler, 1994). Work by Beals (1972; Beals et al., 1983) shows a correlation between skull shape and climate and indicates a trend towards brachycephaly in colder areas. Data provided for H. erectus supports this conclusion (Beals et al., 1983). Wolpoff (1999: 499) suggested that the Chinese assemblage provides evidence for a “gradient in morphology running from the north to the south of East Asia.” Antón (2002) is also inclined to accept this view. This morphological gradient could be due, at least in part, to these climatically induced changes in skull shape. Genetic drift might also explain the differences in craniometric pattern. Currently the variant skull form is limited to northern China at Zhoukoudian, and it may also be present at Nanjing (Antón, 2002). Because of the more fragmentary condition of the Nanjing fossils we were unable to include them in 313 this analysis, but they are said to resemble the Zhoukoudian crania (Etler, 1996; Wolpoff, 1999). Due to its location on the edge of the known geographic range of H. erectus, limited opportunities for gene flow would have been available in northern China. This relative isolation could have contributed substantially to the manifestation of this new craniometric pattern. Thorne (1980:36) has noted that such isolation may lead to “pronounced divergences in the characteristics of different geographically peripheral populations, especially if these are distinguished environmentally.” Conclusion The results of this project indicate that the Zhoukoudian hominids are characterized by a cranial vault shape unlike that seen in other representatives of Homo erectus. Each of the four multivariate analyses performed indicates that a pattern of statistically significant distances separates the Zhoukoudian specimens from other African and Indonesian fossils as well as the Hexian calvaria. The Zhoukoudian hominids are characterized by a wide midvault and relatively narrow frontal and occipital bones, while the African and Indonesian hominids and the Hexian specimen have similarly broad frontal, midvault, and occipital dimensions. This separation between Zhoukoudian and the other crania sampled is unlikely to be the result of sexual dimorphism or temporal variation, as the Chinese fossils are intermediate in age and it is probable that both sexes are represented in that sample. It is more likely that this variation is caused by either adaptation to a colder environment or genetic drift. These results do not suggest that a multiple-species scenario is necessary to encompass the variation present in the sample. However, further examination is needed in order to understand the complex morphologies currently subsumed in H. erectus. Acknowledgements The authors would like to thank Richard Jantz for use of his DISPOP computer program for 314 J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 this analysis as well as his considerable help and patience with our many questions. Fachroel Aziz graciously provided an unpublished biauricular breadth measurement for the Sambungmacan 4 specimen. Bill Kimbel and two anonymous reviewers provided helpful comments that improved the paper. References Andrews, P., 1984. An alternative interpretation of the characters used to define Homo erectus. Cour. Forsch.-Inst. Senckenberg 69, 167–175. Antón, S.C., 2001a. Cranial evolution in Asian Homo erectus: the Ngandong hominids. Geological Research Development Centre Special Publication No. 27, 39–46. Antón, S.C., 2001b. Cranial growth in Homo erectus. In: Minugh-Purvis, N., McNamara, K.J. (Eds.), Human Evolution Through Developmental Change. Johns Hopkins University Press, Baltimore, pp. 349–380. Antón, S.C., 2002. Evolutionary significance of cranial variation in Asian Homo erectus. Am. J. Phys. Anthrop. 118, 301–323. Antón, S.C., Marquez, S., Mowbray, K., 2002. Sambungmacan 3 and cranial variation in Asian Homo erectus. J. Hum. Evol. 43, 555–562. Arif, J., Baba, H., Suparka, M.E., Zaim, Y., Setoguchi, T., 2001. Preliminary study of Homo erectus skull IX (Tjg-1993.05) from Sangiran, Central Java, Indonesia. Bull. Natn. Sci. Mus., Tokyo Series D 27, 1–17. Arif, J., Kaifu, Y., Baba, H., Suparka, M.E., Zaim, Y., Setoguchi, T., 2002. Preliminary observation of a new cranium of Homo erectus (Tjg-1993.05) from Sangiran, central Java. Anthrop. Sci. 110, 165–177. Baba, H., Aziz, F., Kaifu, Y., Suwa, G., Kono, R.T., Jacob, T., 2003. Homo erectus calvarium from the Pleistocene of Java. Science 299, 1384–1388. Beals, K.L., 1972. Head form and climatic stress. Am. J. Phys. Anthrop. 37, 85–92. Beals, K.L., Smith, C.L., Dodd, S.M., 1983. Climate and the evolution of brachycephalization. Am. J. Phys. Anthrop. 62, 425–437. Black, D., 1929. Preliminary notice of the discovery of an adult Sinanthropus skull at Chou Kou Tien. Bulletin of the Geological Society of China 8, 15–32. Black, D., 1930. On an adolescent skull of Sinanthropus pekinensis in comparison with an adult skull of the same species and with other hominid skulls, recent and fossil. Palaeontol. Sin. Ser. D 7, 1–144. Bräuer, G., Mbua, E., 1992. Homo erectus features used in cladistics and their variability in Asian and African hominids. J. Hum. Evol. 22, 79–108. Corruccini, R.S., 1992. Metric reconsideration of the Skhul IV and IX and Border Cave I crania in the context of modern human origins. Am. J. Phys. Anthrop. 87, 433–445. Darroch, J.H., Mosimann, J.E., 1985. Canonical and principal components of shape. Biometrika 72, 241–252. Defrise-Gussenhoven, E., 1967. Generalized distance in genetic studies. Acta Genet. Med. Gemellol (Roma) 17, 275–288. Delson, E., Harvati, K., Reddy, D., Marcus, L.F., Mowbray, K., Sawyer, G.J., Jacob, T., Márquez, S., 2001. The Sambungmacan 3 Homo erectus calvaria: a comparative morphometric and morphological analysis. Anat. Rec. 262, 380–397. Durband, A.C., 2002a. The squamotympanic fissure in the Ngandong and Sambungmacan hominids: a reply to Delson et al. Anat. Rec. 266, 138–141. Durband, A.C., 2002b. The view from down under: a test of the multiregional hypothesis of modern human origins using the basicranial evidence from Australasia. Invited paper presented at the 17th Congress of the Indo-Pacific Prehistory Association, Taipei, Taiwan, September 9–15. Durband, A.C., Kidder, J.H., 2000. The question of speciation in Homo erectus revisited II: the non-metric evidence. Am. J. Phys. Anthrop. 30(Suppl.), 144. Durband, A.C., Kidder, J.H., Jantz, R., 2002. A multivariate analysis of the Hexian Homo erectus calvarium. Am. J. Phys. Anthrop. 34(Suppl.), 66. Etler, D.A., 1994. The Chinese Hominidae: new finds, new interpretations. Ph.D. Dissertation, University of California, Berkeley. Etler, D.A., 1996. The fossil evidence for human evolution in China. A. Rev. Anthrop. 25, 275–301. Groves, C.P., Mazák, V., 1975. An approach to the taxonomy of the Hominidae: gracile Villafranchian hominids of Africa. Casopis Min. Geol. 20, 225–247. Grün, R., Huang, P.H., Wu, X., Stringer, C.B., Thorne, A., McCulloch, M., 1997. ESR analysis of teeth from the palaeoanthropological site of Zhoukoudian, China. J. Hum. Evol. 32, 83–91. Hawks, J., Oh, S., Hunley, K., Dobson, S., Cabana, G., Dayalu, P., Wolpoff, M.H., 2000. An Australasian test of the recent African origin theory using the WLH-50 calvarium. J. Hum. Evol. 39, 1–22. Howells, W.W., 1973. Cranial variation in man: a study by multivariate analysis of patterns of difference among recent populations. Pap. Peabody Mus. Arch. Ethnol, Harvard University Press, Cambridge. Howells, W.W., 1989. Skull shapes and the map: craniometric analyses in the dispersion of modern Homo. Pap. Peabody Mus. Arch. Ethnol., Harvard University Press, Cambridge. Jacob, T., Morphology and paleoecology of early man in Java. In: Tuttle, R. (Ed.) Paleoanthropology, Morphology, and Paleoecology. Mouton, The Hague, pp. 311–25. Jantz, R.L., Owsley, D.W., 2001. Variation among early North American crania. Am. J. Phys. Anthrop. 114, 146–155. Kennedy, G.E., 1991. On the autapomorphic traits of Homo erectus. J. Hum. Evol. 20, 375–412. Kidder, J.H., 1998. Morphometric variability in Homo erectus. Am. J. Phys. Anthrop. 24(Suppl.), 138. J.H. Kidder, A.C. Durband / Journal of Human Evolution 46 (2004) 299–315 Kidder, J.H., Durband, A.C., 2000. The question of speciation in Homo erectus revisited I: the metric evidence. Am. J. Phys. Anthrop. 30(Suppl.), 195. Kramer, A., 1993. Human taxonomic diversity in the Pleistocene: does Homo erectus represent multiple hominid species? Am. J. Phys. Anthrop. 91, 161–171. Mann, A., 1981. The significance of the Sinanthropus casts, and some paleodemographic notes. In: Sigmon, B.A., Cybulski, J.S. (Eds.), Homo erectus: Papers in Honor of Davidson Black. University of Toronto Press, Toronto, pp. 41–61. Mosimann, J.E., James, F.C., 1979. New statistical methods for allometry with application to Florida red-winged blackbirds. Evolution 33, 444–459. Rightmire, G.P., 1990. The Evolution of Homo erectus: Comparative Anatomical Studies of an Extinct Human Species. Cambridge University Press, Cambridge. Santa Luca, A.P., 1980. The Ngandong fossil hominids: a comparative study of a Far Eastern Homo erectus group. Yale Univ. Publ. Anthrop, New Haven. Shen, G., Ku, T.-L., Cheng, H., Edwards, R.L., Yuan, Z., Wang, Q., 2001. High-precision U-series dating of Locality 1 and Zhoukoudian, China. J. Hum. Evol. 41, 679–688. Simmons, T., Falsetti, A.B., Smith, F.H., 1991. Frontal bone morphometrics of southwest Asian Pleistocene hominids. J. Hum. Evol. 20, 249–269. 315 Stringer, C.B., 1984. The definition of Homo erectus and the existence of the species in African and Europe. Cour. Forsch.-Inst. Senckenberg 69, 131–143. Swisher III, C.C., Rink, W.J., Antón, S.C., Schwarcz, H.P., Curtis, G.H., Suprijo, A., Widiasmoro, 1996. Latest Homo erectus of Java: potential contemporaneity with Homo sapiens in Southeast Asia. Science 274, 1870–1874. Thorne, A.G., 1980. The longest link: human evolution in Southeast Asia and the settlement of Australia. In: Fox, J.J., Garnaut, R., McCawley, P., Mackie, J.A.C. (Eds.), Indonesia: Australian Perspectives. Australian National University, Canberra, pp. 35–43. Tyler, D.E., Sartono, S., 2001. A new Homo erectus cranium from Sangiran, Java. Hum. Evol. 16, 13–25. Weidenreich, F., 1943. The skull of Sinanthropus pekinensis: a comparative study on a primitive hominid skull. Palaeontol. Sin. New Ser. D 10, 1–298. Wolpoff, M.H., 1999. Paleoanthropology, second ed. McGraw Hill, Boston. Wood, B.A., 1984. The origin of Homo erectus. Cour. Forsch.Inst. Senckenberg 69, 99–111. Wood, B.A., 1991. Koobi Fora Research Project Vol. 4: Hominid Cranial Remains. Clarendon Press, Oxford. Wu, X., Poirier, F.E., 1995. Human Evolution in China: A Metric Description of the Fossils and a Review of the Sites. Oxford University Press, Oxford.