AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 52139-144 (1980) Components of Racial Variation in Finger Ridge-Counts RICHARD L. JANTZ AND CLEONE H. HAWKINSON Department of Anthropology, University of Tennessee, Knoxville, Tennessee KEY WORDS Dermatoglyphics, Finger ridgecounts, Race ABSTRACT Principal components analysis was used to evaluate finger ridge-count variability as a n indicator of genetic relationships between populations. The analysis was carried out on American White, American Black and African Black samples, each including both sexes. Each individual is represented as a vector of 20 counts, a radial and a n ulnar count for each digit. No assumptions were made prior to analysis concerning the number of meaningful components, and all were examined sequentially. The first five eigenvectors extracted from the within-groups correlation matrix have loadings very similar to those previously described by Roberts and Coope ('75). However, it is the component scores derived from the sixth eigenvector which show the most marked variation, accounting for 45% or more of the D' in all Black-White comparisons. A number of other components also show significant intergroup heterogeneity, but they often do not accord with what is known of the genetic relationships between the populations. Apparently a large amount of ridge-count variation is not genetically meaningful, a t least as far as these populations are concerned. Dermatoglyphics have been used in population studies in one way or another since it was recognized over 70 years ago that they show racial and population variation. No consensus has developed as to what analytical approaches might prove most useful in population studies, and the problem has become progressively worse as anthropologists ask more sophisticated questions of their data. An early approach, which continues to have adherents, is simply to tabulate frequencies of arches, loops and whorls. As a result of Holt's ('56 ; '57) genetic studies of quantitative traits, total ridgecount seemed useful with its high heritibility estimates and statistical qualities. As is now well known, TRC is the sum of the larger counts for each digit, a procedure which greatly simplifies matters, reducing twenty variables to one. It may also, however, obscure a great deal of information relevant to genetical studies of human populations. Holt's parent-child ('56) and sib-sib ('57) correlations are in almost perfect agreement with theoretical expectations under the assumption of additivity, suggesting a heritibility of nearly lOWo. Since heritibility estimates are population specific, we might expect them to vary somewhat. Loesch's ('71) Polish data and Froelich's ('76) southwest Pacific data both yield 0002-948318015201-0139 $01.40 0 1980 ALAN R. LISS, INC heritibility estimates for TRC of around 70%. These are substantially below Holt's estimates, but even so indicate that the majority of variation is genetic. Unfortunately little is known about heritibility of components of ridge-count variability apart from TRC. There may be some aspects of ridge-count variability which are primarily environmental. It is apparent that dermatoglyphics, when incorporated into population studies along with other types of data, often produce results difficult to interpret. The most extensive analysis of this type is that of Friedlaender ('75) on Bougainville. He calculated intervillage distances based on anthropometric, serological, odontometric, kinship, geographical, and dermatoglyphic data. Good agreement was found between anthropometric and serological data, which in turn agreed well with geography, anthropometry being the better of the two. Dermatoglyphics had low agreement with all three, producing a unique population arrangement which was not easily interpretable in genetic terms. Moreover, this poor agreement between dermatoglyphics and other biological data is the rule rather than the exception. Nee1 et al. ('74) found poor agreement between dermatoglyphics and serology or anthropometry, Chai ('72) between derma- 139 140 RICHARD L. JANTZ AND CLEONE H. HAWKINSON toglyphics and anthropometry, and Jantz and Chopra ('76) between dermatoglyphics and anthropometry or serology. Jantz and Chopra ('76) found further that population relationships based on dermatoglyphics alone change, depending upon the method used. Relationships based on ridge-counts vary with those based on pattern classification. The latter further varies with the type of system used. There is clearly a need for research directed toward the problem of how, or even whether, dermatoglyphics provides useful information in anthropological contexts. Steps in this direction were taken by Roberts and Coope ('751, who carried out a principal components analysis on the correlation matrix of the twenty finger ridge-count variables. They identified five components and advanced field theory as a n explanatory model. Jantz and Owsley ('77) carried out a similar analysis, rotating five extracted principal components to attempt to simplify the interpretation. Both studies obtained results unattainable by previous investigations. It turns out that radial and ulnar sides of the digits are relatively independent, indicating that ignoring orientation in choosing the larger count is not appropriate. Moreover, certain digits covary to the extent that they may be considered meaningful entities. If t h e components emerging from such analyses have genetic meaning, they should prove useful in population studies, and may reveal unsuspected aspects of population variation. To date the components of variation as elucidated by principal components analysis have not been analyzed from the perspective of interpopulation variation. That such a n approach may yield more meaningful and unsuspected information has been suggested by Jantz and Hawkinson's ('79)analysis of African ridge-counts. The object of this paper is to carry out an analysis using component scores of three samples to evaluate the genetic significance of components extracted from a correlation matrix. MATERIALS AND METHODS The three samples employed in the analysis are an American White sample (185males, 184 females),a n American Black sample (96 males, 119 females), and an African Black Yoruba sample (120 males, 55 females). These are the three samples previously used i n a factor analytic study (Jantz and Owsley, '77) and the Yoruba sample has recently been described (Jantz and Brehme, '78). Standard procedures were followed for ridge-counting (Holt, '68). Each individual is represented as a vector of twenty counts, a radial and an ulnar count for each digit. The 20 x 20 covariance matrix was formed for each group. The matrices were pooled over groups and sexes and converted to the withingroups correlation matrix, yielding a correlation matrix based on 741 degrees of freedom. All 20 eigenvalues and eigenvectors were extracted using the SPSS subroutine FACTOR, which also calculates a factor score coefficient matrix. If A is the matrix of factor loadings, the factor score matrix is F = (A'A)-'A. The principal component scores are then calculated as Y = F Z where Z is the data matrix expressed as standard scores. Principal components scores obtained in this way are then standard scores, having a grand mean of zero and a standard deviation of one. These scores have the further advantage that they may be used to '72) as: obtain Mahalanobis' - D' - (Goodman, - D' = (Y,-Y,) (Yl-YJ)' where Y ,and YJare the mean vectors of component scores for groups i and j. The pooled correlation matrix was used to assure that all distances would have the same meaning to facilitate comparison. There is some evidence that the correlation matrices differ between Blacks and Whites (Jantz, '77; Jantz and Owsley, '77) in that Blacks have higher average correlations. Factor analysis failed to reveal any systematic differences in structure so we are probably justified in pooling the matrices over groups. We have computed component scores using the unrotated factor loadings. It has been shown that rotation often renders factors easier to interpret, but it is also the case that the loadings change with the number of factors included in the rotational solution. Since we wish to m a k e no assumptions concerning t h e number of meaningful components, we have examined all of them sequentially. RESULTS It is not our purpose to attempt any thorough examination of the component loadings from the standpoint of structure. It may be useful to look briefly a t the most important eigenvalues and associated eigenvectors, which a r e presented in Table 1. The eigenvalues behave in a manner similar to that found previously (Robertsand Coope, '75). The first eigenvalue is large, accounting for almost half of the variance. Subsequent eigenvalues decrease i n value very gradually, so that 12 are required to account for 91% of the variation. The first five 141 RACIAL VARIATION IN RIDGE-COUNTS TABLE I , First sir eigenuectors and eigenvalues of the finger ridge-count correlation matrix Eigenvectors Digits 1 2 3 4 5 6 0.74 0.44 0.80 0.71 0.81 0.68 0.69 0.64 0.63 0.56 0.72 0.46 0.76 0.75 0.76 0.69 0.63 0.67 0.61 0.58 -0.25 0.55 -0.22 0.39 -0.21 0.40 -0.14 0.11 -0.36 0.10 -0.25 0.55 -0.24 0.28 -0.24 0.34 -0.10 0.09 -0.42 0.07 -0.08 0.38 -0.07 0.14 0.13 -0.16 0.32 -0.43 0.13 -0.32 -0.13 0.40 -0.06 0.06 0.25 -0.25 0.44 -0.45 0.15 -0.33 -0.04 0.20 -0.19 -0.08 -0.18 -0.21 -0.07 -0.10 0.33 0.58 -0.04 0.21 -0.15 -0.10 -0.12 -0.23 -0.05 -0.16 0.31 0.58 -0.43 -0.25 -0.25 -0.01 0.07 0.22 0.20 0.15 0.31 -0.01 -0.45 -0.20 -0.28 0.07 0.11 0.21 0.17 0.15 0.25 -0.03 0.12 0.24 -0.03 -0.08 -0.10 0.07 -0.27 0.17 0.35 -0.33 0.12 0.17 -0.10 -0.10 -0.10 0.05 -0.29 0.15 0.37 -0.26 9.03 1.81 1.47 1.23 1.01 0.82 45.10 9.10 7.30 6.20 5.10 4.10 -~ L5 R U L4 R L3 U R U L!! R U L1 R IJ R5 R4 R3 R U R U R lJ R2 R U Rl R U Eigenvalues Percent of variation components are very similar to those described by Roberts and Coope ('75).The first component is clearly size, all loadings being positive, reflecting the positive correlation between all variables. The second component contrasts radial and ulnar sides of the digits, except for digit 11. This suggests some sort of negative interaction after removing the size effect. Component 3 contrasts the ulnar count on digit V with the ulnar counts on digits I1 and I. In addition, the radial and ulnar sides of digit I1 are contrasted. Component 4 is a thumb component, both radial and ulnar counts loading positively but emphasizing the ulnar count more than the radial. Component 5 takes the form of an ulno-radial gradient, starting with strong negative loadings on digit V and ending in strong positive loadings on the radial count of the thumb. The ulnar thumb count does not participate in the gradient. Roberts and Coope attempt to interpret only five components, which generally correspond to those with eigenvalues of one or greater. However, our component 6 is easily interpreted and turns out to be very important. It contrasts radial and ulnar counts on digits I and 11,and in effect, the digits themselves since the direction of the contrasts on the two digits is reversed. Even later components are interpretable to some extent. All components after 11 reflect asymmetry, homologous variables having opposite loadings. While these collectively account for a very small fraction of the total variation, it is likely that even these contain some biologically meaningful information. We will return to some of these later. The component scores for each sample for all twenty components are presented in Table 2, along with the F ratios derived from a one way analysis of variance between groups. Several points are noteworthy from these results. The general finding is that there are several components which show significant heterogeneity, the majority of which are in males. In males, components 2,3,6,7,8,12, and 15show significant heterogeneity, while in females only 6,12, and 17 are significantly heterogeneous. Only components 6 and 12 show sex concordance for statistical significance. Component 12 is principally asymmetry of the ulnar counts of digit V. Component 15 is a rather complex reflection of the asymmetry of radial ulnar contrasts on digit 111, and also involves the ulnar count of digit IV. Component 17 is a rather more straightforward whole hand asymmetry centered on the radial count of digit 111. It is perhaps somewhat surprising that asymmetry components show significant interpopulation heterogeneity. However, the concordance of sexes on component 12, and the RICHARD L. JANTZ AND CLEONE H. HAWKINSON 142 TABLE 2. Principal component scores ofAmerican Whites (AWI, American Blacks (ABj und Yoruba (Y), and F ratios for between-group heterogeneity Males Component number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 F ratio Females AW AB Y AW AB Y d 0.058 -0.203 0.051 -0.003 -0.061 0.553 -0.191 -0.205 0.048 -0.080 -0.069 0.222 0.110 0.012 -0.171 0.067 0.025 0.006 -0.069 -0.169 0.033 0.146 0.201 0.066 0.111 -0.293 0.179 0.164 0.128 0.094 -0.003 0.003 -0.045 0.050 0.328 -0.012 0.017 0.014 -0.021 0.064 0.110 0.103 -0.148 0.008 -0.119 -0.319 -0.203 0.138 0.015 0.028 -0.006 -0.088 -0.036 -0.001 -0.002 0.012 -0.073 -0.081 0.138 0.045 -0.095 -0.062 -0.040 0.024 0.035 0.338 0.123 -0.048 -0.052 -0.108 0.006 0.081 0.014 0.051 0.140 -0.612 -0.188 0.234 -0.097 -0.309 -0.146 -0.457 0.354 0.076 0.048 0.142 0.100 -0.185 -0.144 -0,122 0.069 0.019 0.064 -0.164 0.169 0.064 0.17 10.5W 3.34* 0.17 1.51 20.9P* 11.51"" 6.27"* 0.35 1.05 0.21 3.84* 1.13 0.08 7.87*' 0.23 0.38 0.34 1.61 2.47 -0.032 0.040 0.026 -0.001 0.024 -0.014 0.118 0.033 -0.036 0.019 0.003 0.085 -0.136 0.120 0.118 -0.236 -0.073 0.000 0.008 -0.095 -0.192 0.085 -0.036 0. I07 0 0.78 2.08 0.25 2.77 1.55 36.73*" 2.33 0.76 0.66 2.48 0.93 3.07" 0.75 0.34 0.19 0.33 3.584 1.20 0.98 0.28 **P- 005 -*P. 0 0 1 fact that both sexes show variation on a second asymmetry factor, suggests that there may be a n overlooked biological reality here. The specific finding of greatest interest is that component 6 shows remarkable variation between groups. Its F ratio exceeds the next highest ratio by a factor of two in males and ten in females. Quite clearly, this factor is the single most important source of interpopulation variation. It accounts for 45% or more of the D2, shown in Table 3, in all Black-White comparisons. In the case of the White-Yoruba TABLE 3 . LY values for race and sex American Whites American Blacks Yoruba DISCUSSION AW AB Y 0.36* 1.5P* 1.29% 1.26** 0.49 0.51 1.22*" 0.68 0.64 Sex differences within race are shown in the diagonal. Race differences for males are In the lower left triangle, far females in the upper right triangle. D' values were tested for significance by the variance ratio n , * n 2 ( n , + n,-m- 1 ) F A __ ___ D* n , + n, ( n , + n,-2) m where n , and n, are the sample sizes and m is the number of variables. The F ratio has m and ( n , + n, - 1 - m) degrees of freedom. *P<0.05. * * P < 0.01. . female comparison, component 6 is responsible for 7PkJof the D' between them! The D2 values themselves seem to reflect a general genetic reality. American Whites show relatively large differences in relation to American Blacks and Yorubas. The latter two show lower distances between each other. However, in both sexes Yoruba are actually slightly closer to American Whites than American Blacks are, which would not accord with our expectations based on what is known of White gene flow into the American Black population. The sex D' is low for all groups and statistically significant in Whites only. We now turn to the several implications of these results for the anthropologic utilization of finger ridge-count data. An interesting finding is t h a t t h e most important component extracted from the correlation matrix shows no statistically significant population variation. The first component, being principally a size factor, is nearly equivalent to the sum of the twenty counts, a commonly used summary character known as absolute ridge-count. This same result has been found for total ridge-count as well (Jantz, '74; Jantz and Hawkinson, '791, so t h e conclusion t h a t summing counts RACIAL VARIATION IN RIDGE-COUNTS obscures a great deal of information now seems well established. A second finding, perhaps the most important of our findings, pertains to the extraordinary Black-White difference seen on component 6. This component has an eigenvalue of only 0.82 and accounts for only 4.1% of the total variation in within-group terms. Yet it is by far the most important source of racial variation. From a methodological point of view, this finding seems to offer strong support for the idea of using twenty counts, rather than summing counts to simplify matters. This component is exclusively a radial-ulnar contrast so there is no way the information could be obtained using only the larger count from each digit. Some of this has been alluded to in pattern type analysis, as in the lower frequency of radial loops on digit I1 (Jantz, '74) and relatively higher frequency of arches on the thumb (Glanville and Huizinga, '66) found in Black populations. The developmental significance of component 6 also requires preliminary discussion a t this time. Roberts and Coope ('75) applied field theory as an explanatory model for their components, and Jantz and Owsley ('77) also felt field theory to be appropriate. Thus component 6 may represent some sort of developmental field affecting the radial digits only. This is quite in line with the results from embryological data (Babler, '77; Okijima, ' 7 5 ) which shows a radial-ulnar gradient in maturation of the ridged skin system. But the most important feature of this component would seem to be the radial-ulnar contrasts drawn between sides of digits I and 11. This type of contrast is quite apparent elsewhere, as in components 2 and 3 , so it represents a common phenomenon of the finger ridge system. One way to interpret it, although at this stage it is purely speculative, is that there is a racial difference in the nature or strength of a growth gradient during morphogenesis of the ridged skin system. The relationship of pattern types to ridge maturation, as well as Black-White differences in ridge maturation, have been documented (Babler, '77). The form this maturational difference might take is, of course, not clear. That it relates to differences in timing of pad regression in relation to ridge development seems likely. It is noteworthy here that the sum of the radial and ulnar counts on each digit in Blacks and Whites is very similar, so differences pertain almost exclusively to radial-ulnar distribution. Additional inferences might arise from our expectations from admixture studies. To the extent that a component variable represents 143 genetic variation, we might expect that Whites and Africans would be most dissimilar, with American Blacks displaced towards Whites somewhat, due t o admixture. This expectation is borne out in the male scores for component 6, but in females the American Blacks are more extreme than the Yoruba. Given the small sample size of the Yoruba females, this result does not invalidate a genetic interpretation. Even though component 6 produces somewhat ambiguous results in terms of our genetic expectations, they are much better than results of the other components. Component 7 is especially noteworthy in this regard. It shows highly significant heterogeneity in males, but it is not patterned in any intelligable way. Yoruba and American Whites have similar scores, American Blacks being the extreme group. American Black males are also extreme on component 15. Perhaps these represent some sort of unidentified gene-environment interaction. ACKNOWLEDGMENTS The data for this research were analyzed on the IBM 360165 at the University of Tennessee Computer Center. LITERATURE CITED Babler, W.J. (1977) The prenatal origins of populational differences in human dermatoglyphics. Diss. Abst. Int., 38:(6)A:3591. Chai, C.K. (1972) Biological distances between indigenous populations of Taiwan. In: The Assessment of Population Affinities in Man. J.S. Weiner and J. Huizinga, eds., Clarendon Press, Oxford, pp. 182-210. Friedlaender, J.S. (1975) Patterns of Human Variation. Harvard University Press, Cambridge. Froehlich, J.W. (1976) The quantitative genetics of fingerprints. In: The Measures of Man. Methodologies in Biological Anthropology. E. Giles and J.S. Friedlaender, eds., Peabody Museum Press, Cambridge, pp. 26&320. Glanville, E.V., and J. Huizinga (1966) Digital dermatoglyphics of the Dogon, Peul, and Kurumba of Mali and Upper Volta. Proc. Koninkl. Nederl. Akad. Wetensch., Series C, 69t664-674. Goodman, M.M. (1972) Distance analysis in biology. Syst. Zool., 21:174-186. Holt, S.B. (1956) Genetics of dermal ridges: parent-child correlations for total finger ridge-count. Ann. Hum. Genet., 20:270-281. Holt, S.B. (1957) Genetics of dermal ridges: sib-pair correlations for total finger ridge-count. Ann. Hum. Genet., 21t352-362. Holt, S.B. (1968)The Genetics of Dermal Ridges. Charles C. Thomas, Springfield, Ill. Jantz, R.L. (1974) Finger ridge-counts and inter-finger variability in Negroes and Whites. Hum. Biol., 46:66%675. Jantz. R.L. (1977) Sex and race differences in finger ridgecount correlations. Am. J. Phys. Anthropol., 461171-176, Jantz, R.L.. and H. Brehme (1978)Finger and Dalmar dermatoglyphics of a Yoruba (Nigeria)sample. Ann. Hum. Biol., 5t539-546. 144 RICHARD L. JANTZ AND CLEONE H. HAWKINSON Jantz, R.L., and V.P. Chopra (1976) A comparison of dermatoglyphic methodologies i n population studies. Paper presented to the 11th International Congress of Human Genetics, Mexico City. Jantz, R.L., and C.H. Hawkinson (1979) Finger ridge-count variability in sub-Saharan Africa. Ann. Hum. Biol., 6:41-53. Jantz, R.L., and D.W. Owsley (1977)Factor analysis of finger ridge-counts in Blacks and Whites. Ann. Hum. Biol., 4.357-366. Loesch, D. (1971) Genetics of dermatoglyphic patterns on palms. Ann. Hum. Genet., 34:277-293. Neel, J.V., F. Rothhammer, and J.C. Lingoes (1974) The genetic structure of a tribal population, the Yanomama Indians. X. Agreement between representation of village distances based on different sets of characteristics. Am. J. Hum. Genet., 26.281-303. Okijima, M. (1975) Development of dermal ridges in the fetus. J. Med. Genet., 12:24%250. Roberts, D.F., and E. Coope (1975) Components of variation in a multifactorial character: a dermatoglyphic analysis. Hum. Biol., 47:16%188.