Congruence of distance matrices based on cranial discrete traits cranial measurements and linguistic-geographic criteria in five Alaskan populations.код для вставкиСкачать
BRIEF COMMUNICATION Congruence of Distance Matrices Based on Cranial Discrete T r a its, C r ania I Meas urement s, and Linguist i c-G eog r a phic Criteria in Five Alaskan Populations N. S. OSSENBERG Department ofAnatomy, Queen's University, Kingston, Ontario, Canada, K7L 3N6 KEY WORDS Discrete traits . Skull . Eskimo . Aleut - Biological distance - Correlation ABSTRACT Biological distances (C.A. B. Smith's Measure of Divergence) based on 25 nonmetric skull variants have been compared with distances (Mahalanobis' D2) based on cranial measurements of four Eskimo populations representing the Yupik subdivision of the Eskaleut linguistic stock, and one Aleut population. The ranking of Measures of Divergence (pooled-sex samples) for ten pairwise comparisons is significantly correlated (Spearman's r,) with both the male and female rankings of the F-values of D'. In addition, the nonmetric distances showed stronger concordance than the metric distances with a hierarchy based on linguistic and geographical affinities. These findings indicate that, depending upon the particular battery of variants used, discrete traits provide valid taxonomical information in the study of extinct human populations. A battery of nonmetric cranial variants demonstrated to be taxonomically valid in terms of its power to discriminate within- and between-race distances (Ossenberg, '76) is being used in a current project to trace prehistoric relationships in Western North America. Five of the approximately 35 Amerind groups represented in the project are among those analyzed craniometrically by Zegura ('71, '75), permitting a further test of the validity of this particular trait battery. The purpose of the present paper is t o examine the congruence of distance matrices based on cranial discrete traits, cranial measurements, and linguistic-geographic attributes in these five populations. MATERIALS AND METHODS Samples One Aleut and four Eskimo samples housed in the U.S. National Museum furnished the data for this study. Table 1lists for each sample the geographical derivation, approximate time level, putative linguistic classification, and size of that portion of the total retrieved sample employed for nonmetric observations AM. J. PHYS. ANTHROP.. 47: 93-98 and measurements, respectively. Owing to the inclusion of male and female, subadult and fragmentary specimens, more individuals are represented in the portions of the samples used for nonmetric analysis. Pooled-sex nommetric data are used since it has been demonstrated that distance measures based on this particular trait battery are not greatly distorted by a component due to sex. Separate male and female distance analysis for t h e three largest (Kodiak, Nunivak and Dakota Sioux) of 16 samples included in the earlier study (Ossenberg, '76) revealed that for each of the three samples: the male-female distance is not significantly greater than zero, males of the population are closer to their own females than to any one of the other 15 pooled-sex samples, and the rank correlation between male and female distance measures vs 15 pooled-sex samples is significant a t the 0.001 level. The rationale for inclusion of subadult data for certain features has also been discussed previously (Ossenberg, '69, '76). Many of the Kodiak skulls exhibit slight occipital flattening; the other series are undeformed. 93 94 N. S. OSSENBERG TABLE 1 The samples Number of individuals in sample Group and lncation Approximate time depth Aleut, Kagamil Island Eskimo, Kodiak Island (upper level, Uyak Bay) Eskimo, South Mainland Alaska Eskimo, Nunivak Island Eskimo, St. Lawrence Island ' A.D. 1,500-1,700 (Turner, '67) A.U. 1,000-1,500 (Clark, '66) A.D. 1,700-1,900 (Collins, '63) A.D. 1,700-1,900 (Collins, '63) A.D. 1,700-1,900 (Collins, '63) Linguistic classification IDumond. ' 6 5 ) Nonmetric study male and femalr Metric study (Zegura. '71, '763 Male Female Aleut 111 28 32 Yupik: Suk 100 24 14 Yupik: Yuk 173 18 12 Yupik: Cux 102 39 32 76 71 38 Yupik: Yupak ' Includes sites frnm the regions of Rristol Bay, Nushagak River, and Kushokwim River below Old Bethel. The portioii of the sample used for the metric study dprwei from the latter reDon only. L in& istic-geographic ranking of the samples A nonbiological ranking of these populations is derived fundamentally from a scheme of internal Eskaleut linguistic relationships, refined according to geographical criteria. Figure 1 is based on Dumond's ('65) hypothetical reconstruction of evolutionary relationships within t h e Eskaleut linguistic stock. Krauss ('73) along with other authorities would oppose, on theoretical grounds, a Stammbaum scheme such as the one represented in figure 1. Nevertheless, in the particular case of Eskaleut the affinities defined by Krauss appear to correspond in general to those implicit in Dumond's scheme. Both agree that the first order dichotomy is between Eskimo and Aleut, the second between Yupik and Inupik; and that Yupak is more distinct from Suk and Yuk than the latter two are from each other. One important difference is that, whereas Dumond considers Yuk and Cux to be separate languages, Krauss places them together, regarding Nunivak Island speech (Cux) as a n aberrant dialect within a single Central Alaskan Yupik language community. The latter definition would alter the diagram by shifting Cux from the status of equidistant branching with Yuk from the Yupik stem up to a subbranch of Yuk. Owing to lack of data on the relevant cognate percentages, quantitative ordering of all ten pairwise comparisons in linguistic terms is not possible. However, taking into account the analyses of both Dumond and Krauss, one may construct a crude hierarchy in which the closest linguistic affinity is between Yuk and Cux, and the most distant between Aleut and each of the other four groups, as follows: 1. Yuk: Cux 2. Suk: Yuk and Cux 3. Yupak: Yuk, Cux and Suk 4. Aleut: Suk,Yuk, Cux and Yupak I have drawn upon geographical and, to some extent, archaeological and ethnohistorical considerations (Oswalt, '67) to refine the hierarchy, as follows: 1. Yuk: Cux 2. Suk: Yuk 3. Suk: Cux 4. Yupak: Yuk 5. Yupak: Cux 6. Yupak: Suk 7. Aleut: Suk 8. Aleut: Yuk 9. Aleut: Cux 10. Aleut: Yupak It should be emphasized that this scheme is a linguistic-geographic composite. Thus, for example, the clustering of all Aleut comparisons a t the bottom is linguistic; but within that cluster the order is geographical and does not imply linguistic affinity. Traits To the 24 features analyzed previously (Ossenberg, '76) I have added odonto-occipital articulation, a n anomaly represented by a 95 CRANIAL DISCRETE TRAITS IN POPULATION STUDIES Aleutian Islands Pacific Region Kodiak Island f South Mainland Alaska Siberia and St.Lawrence Island Nunivak Island I I 1 Seward Peninsula to Greenland I INUPIK YUPIK ALEUT ESKIMO I ESKALEUT Fig. 1 Diagram representing internal Eskaleut linguistic relationships (after Dumond, '65) facet for articulation with the dens of the axis on the anterior margin of the foramen magnum (Ossenberg, '691, giving a total of 25 traits. StatisticaE methods The distance measure used for discrete traits is the Measure of Divergence (MD) devised by C. A. B. Smith (Grewal, '62; Berry, '631, modified according to the recommendation of Green and Suchey ('76) using the Freeman-Tukey inverse sine transformation of trait frequencies. The standard deviation of MD is calculated according t o the formula recommended by Sj~vold('73). MD is statistically significant when equal to or greater than twice its standard deviation. The distance statistic employed by Zegura ('71, '75) for the cranial measurements is Mahalanobis' D2. A battery of 59 linear and 15 angular measurements were reduced to 11 canonical variates used in the computation of D2. An important caveat pertaining t o the comparisons in this report is that the alignment of D' pertains to the dimensionality of a 12-sample analysis, heavily weighted towards Inupik Eskimo not represented by any of the five samples in this study. Rigorous methodology would require computation of a new variancecovariance matrix. Though Zegura found that a 10-sample analysis (excluding 2 deformed series) produced almost identical inter-group dispersion within 2-dimensional canonical variate plots as the 12-sample analysis, it is possible that a 5-sample analysis might produce a different dispersion and alter the rank order of D2. The F-statistic was employed by Zegura ('71) to test the significance of D2.As the Fvalues proved to be more highly correlated than D2with the linguistic-geographic ranking described above, I have used F-values for the comparisons in this study. Spearman's coefficient, r, (Siegel, '561, was used to test rank correlation between orderings of ten pairwise comparisons based on: linguistic-geographic criteria, F-values of D2 for measurements, and MD values for discrete traits. RESULTS The geographic-linguistic ranking, F-values and MD's for the ten pairwise comparisons are listed in table 2. The Spearman rank correlation coefficients are in table 3. DISCUSSION As expected, the highest correlation obtains 96 N. S . OSSENBERG TABLE 2 Rank order of lingulstzc-geographicrelationships, and biological distances based on cranlal measurements and cranial discrete traits in one Aleut and four Eskimo samples. Standard deuiation o f M D an parenthesis Measurement distance Male Group comparison Linguisticgeographic rank SMA ': Nunivak 1 Female ____ F-value of D' 4.1 Discrete trait distance F value Rank 1 of D' 2.1 Rank MD Rank 1 0.051 3 (0.003) Kodiak: S M A ' 2 5.2 3 3.0 2 Kodiak: Nunivak 3 11.0 8 6.6 7 4 4.4 2 3.3 3 St. Lawrence: SMA ' 0.016 (0.003) 0.060 (0.004) 0.034 1 4 2 (0.003) St. Lawrence: Nunivak 5 13.3 10 7.1 8 S t . Lawrence: Kodiak 6 6.8 6 4.1 5 Kagamil: Kodiak 7 5.7 4 3.5 4 Kagamil: SMA 8 5.8 5 5.4 6 Kagamil: Nunivak 9 12.0 9 13.4 10 10 10.0 7 10.3 9 Kagamil: St. Lawrence 0.120 (0.004) 0.066 (0.004) 0.088 (0.004) 0.067 (0.003) 0.138 (0.0041 0.114 (0.0041 9 5 7 6 10 8 ' South Mainland Alaska TABLE 3 Spearman 5 rank correlation coefficients,r,, for rankings in tablet? Probability leuels in parentheses Linguisticgeographic F-value male F-value female MD F-value male F value female 0.92 (p<0.011 0.77 (p<O.Ol) 0.84 (p<O.Ol) 0.50 (nsJ 0.75 (p<O.Ol! 0.78 (pi0.01i for the male-female hierarchy of F-values. The r, (0.92) for 5 groups is identical to that reported by Zegura for male-female ranking of D2 for 12 groups. This implies similarity of inter-group metric distances in a multidimensional framework regardless of sex. The nonmetric distances are significantly correlated with the metric distances for both males and females; more strongly with the latter. The correlation of linguistic-geographic with biological distances is equally good for nonmetric and female metric data (r, 0.78 and 0.75 respectively), but not significant for male metric data (rs 0.50). The better overall concordance of nonmetric variables could be due partly to larger sample size. The slight occipital flattening of skulls in the Kodiak series may account for the discrepancy between the metric and linguisticgeographic rank of Kodiak:Nunivak and Kagami1:Kodiak (table 2). However, there are no discordant ranks for Kodiak that might be attributed t o the effect of this type of artificial cranial deformation on discrete traits. The most marked deviation from the linguistic-geographic hierarchy is St. Lawrence:Nunivak, ranked fifth. In contrast, three sets of cranial data agree closely, ranking this comparison tenth (male metric), eighth (female metric) and ninth (nonmetric). It may be that linguistic-geographic criteria in this case underestimate the genetic consequences of isolation of one island from another. In contrast to the findings in this study, Zegura ('75) reported rather low taxonomic congruence between distances based on measurements and distances based on 15 discrete traits in 12 Eskimoid populations. Moreover, CRANIAL DISCRETE TRAITS IN POPULATION STUDIES when concordance between biological and linguistic relationships was examined, cranial measurements performed better than discrete traits. Factors to be considered in comparing my findings with those of Zegura for nonmetric traits include the different distance measures (MD vs Balakrishnan and Sanghvi's B2),composition and size of samples, and trait batteries. Other cases of poor concordance have been reported: for North American Plains (Jantz, '701, North American Southwest (El-Najjar, '74) and Africa (Rightmire, '72). I t is possible that unsatisfactory performance of discrete traits in some instances could be attributable-not to this type of variable per se-but to the particular battery of traits used. More than 200 variants have been described on the human skull. Several factors could account for certain traits being possibly of less value than others (Sjavold, '73; Corruccini, '74; Zegura, '75; Suchey, '75). That different batteries produce different group alignments is shown by the present 25trait analysis contrasted with Zegura's 15trait analysis. Only five features are common to both traitlists. Absence of correlation between biological and nonbiological data in any particular instance would not, in itself, justify dismissing the biological measures as untrustworthy. This is particularly true for skeletal studies where, as pointed out by Zegura, there is no way of verifying the language spoken by the people. Furthermore, there are well-known cases among living peoples of lack of correspondence between cultural attributes and racial identity, owing to factors such as migration, cultural diffusion and assimilation of one group by another; for example, in Siberia (Dolgikh, '65). The use of skeletal analysis for historical reconstruction is limited also by the fact that our samples represent lineages, not biological populations in the strict sense (Cadien et al., '76). Nevertheless, where cultural, geographical and biological data do concur, the evidence for group relationships is considerably reinforced. CONCLUSION Significant correlations between distances based on 25 discrete cranial attributes, cranial measurements, and linguistic-geographic criteria indicate that this battery of non- 97 metric variants is useful for the study of extinct North American populations. ACKNOWLEDGMENTS The cranial series were studied a t the U.S. National Museum with the aid of a National Research Council of Canada doctoral fellowship, 1963-64; and a grant from the Boreal Institute of the University of Alberta, 1970. The help and advice of Doctor J. L. Angel, Doctor H. B. Collins, Doctor D. Ortner, Doctor T. D. Stewart and Doctor L. St. Hoyme of that Museum are gratefully acknowledged. Data analysis is funded by Canada Council Grant S75-0074. I am grateful to Doctor S. Zegura for permission to use his thesis data, and for his helpful criticism of the manuscript. LITERATURE CITED Berry, R. J . 1963 Epigenetic polymorphism in wild populations of Mus rnusculus. Genet. Res., 4: 193-220. Cadien, J. D., E. F. Harris, W. P. Jones and L. J. Mandarin0 1976 Biological lineages, skeletal populations and microevolution. Yearbook of Phys. Anthrop., 18: 194-201. Clark, D. W. 1966 Perspectives in the prehistory of Kodiak Island, Alaska. Am. Antiq., 31: 358-371. Collins, H. B. 1963 Personal communication. Corruccini, R. S. 1974 An examination of the meaning of cranial discrete traits for human skeletal biological studies. Am. J. Phys. Anthrop., 40: 425-445. Dolgikh, B. 0. 1965 Problems in the ethnography and physical anthropology of the Arctic. Arctic Anthrop., 3: 1-9. Dumond, D. E. 1965 On Eskaleutian linguistics, archaeology and prehistory. Am. Anthrop., 67: 1231-1257. El-Najjar, M. Y. 1974 People of Canyon de Chelly: A Study of their Biolom .," and Culture. Ph.D. dissertation. Arizona State Univ., Tempe. Green, R. F., and J. M. Suchey 19'76 The use of inverse sine transformations in the analysis of non-metric cranial data. Am. J. Phys. Anthrop., 45: 61-68. Grewal, M. S. 1962 The rate of genetic divergence of sublines in the C57 BL strain of mice. Genet. Res., 3: 226237. Jantz, R. L. 1970 Change and Variation in Skeletal Populations of Arikara Indians. Ph.D. dissertation, Univ. Kansas, Lawrence. Krauss, M. E. 1973 Eskimo-Aleut. Current Trends in Linguistics, 10: 796-902. Ossenberg, N. S.1969 Discontinuous Morphological Variation in the Human Cranium. Ph.D. dissertation, Univ. Toronto. 1976 Within and between race distances in population studies based on discrete traits of the human skull. Am. J. Phys. Anthrop., 45: 701-716. Oswalt, N. H. 1967 Alaskan Eskimos. Chandler Publishing Co., San Francisco. Rightmire, G. P. 1972 Cranial measurements and discrete traits compared in distance studies of African Negro skulls. Hum. Biol., 44: 263-276. Siegel, S. 1956 Nonparametric Statistics for the Behavioral Sciences. McGraw-Hill, New York. 98 N. S. OSSENBERG Sjevold, T. 1973 The occurrence of' minor non-metrical variants in the skeleton and their quantitative treatment for population comparisons. Homo, 24: 204-233. Suchey, J. M. 1975 Biological Distance of Prehistoric Central California Populations Derived from Non-metric Traits of the Cranium. Ph.D. dissertation, Univ. California, Riverside. Turner, C. G. 1967 The Dentition of Arctic Peoples. Ph.D. dissertation, Univ. Wisconsin, Madison. Zegura, S. L. 1971 A Multivariate Analysis of the Interand Intra-population Variation Exhibited by Eskimo Crania. Ph.D. dissertation, Univ. Wisconsin, Madison. _ _ 1975 Taxonomic congruence in Eskimoid populations. Am. J. Phys. Anthrop., 43: 271-284.