Quantitative Hair Form Variation in Seven Populations

Quantitative Hair Form Variation in Seven Populations DANIEL HRDY Department of Anthropology, Harvard University, Cambridge, Massachusetts 021 38 KEY WORDS Hair form. Principal components analysis Melanesia Human variation. ABSTRACT Although hair form has received much attention in the past, it has rarely been studied systematically, and never using direct curling variables. In the present study, seven groups were scored on eight variables, including four newly-devised curling variables. These data were analyzed using univariate and multivariate techniques to give information about the population relations and mechanisms of hair form. Racial groups were separated using a principal components analysis. African and Melanesian populations were shown to have significantly different quantitative hair form traits, especially in regard to their regularity of curvature. The physiological, environmental, and genetic factors contributing to hair form variation are discussed. Cranial hair form has been one of the major racial criteria since the beginning of anthropological thought, but it is surprising how little systematic research has been done to delimit these population differences. For centuries such imprecise expressions as frizzy or curly have been in popular use. Such terms do not connote exact meanings to their users, and their vagueness has increased with continued use. Typologies using these terms have had an influence on current anthropological thought, not due so much to their perpetuation as to the failure of any adequate substitute to appear. There have been many classifications of hair form ranging from the classic formulation of Huxley (Leiotrichi, Ulotrichi) (1865), to Deniker (Straight, Wavy, Frizzy, Woolly) ( 00), to Martin (Glatthaar, Wellighaar, Kraushaar) ( 28), to Hooton (Straight, Wavy, Kinky, Woolly) ( 46), to Garn (Straight, Wavy, Helical) ( 48). As Trotter, one of the great quantifiers of hair said ( 38), There has been a slow and gradual change in the point of view of anthropologists... The tendency has been away from the two or three large groupings and toward smaller and more clearly defined categories. The next obvious step is the quantification of hair form. Metric studies of hair have been done since the beginning of the century, but AM. J. PHYS. ANTHROP., 39 7-18. with little emphasis on population differences. Hausman ( 25) and Wynkoop ( 29), for instance, were anthropologists interested in microscopic description and structural relations of hair. Later studies, such as those of Steggerda and Seibert ( 41) and Trotter et al. ( 56) have been more quantitative, but have generally been restricted by the lack of curling variables and the limitations of univariate statistics. The selection of variables is an important problem in the description of hair form. The variables must as accurately and completely as possible describe the hair. The most popular traditional hair form variables have been diameter of the hair shaft, cuticular scale count along the hair shaft, and cross-sectional index of the hair shaft. Cross-sectional index was usually measured by preparing thin, transverse sections of the hair shaft. However, spurious data can result from oblique sectioning, a technical difficulty which is increased in highly curled hairs. Cross-section had a great vogue in connection with Pruner-Bey s (1864) theory associating cross-sectional index with curling. As Garn ( 48) and others have pointed out, there is no consistent functional correlation between the two. The variables are associated in that a population with curly hair is likely to have elliptical crosssections, but this does not hold within the 7

8 DANIEL HRDY population; the curliest hairs do not necessarily have the most elliptical crosssection. This paper is an attempt to apply quantitative and statistical methods to the description of cranial hair form of several different populations in order to understand the relations among the populations and the nature of hair form itself. In this paper only the term curling is used to describe nonlinear hair form; more specific terms cannot be used because of the quite complex variation of hair form. Only hair form variables will be discussed. Eight variables were measured or each hair, including several devised to measure curling accurately: (1) Average diameter The average diameter was measured by placing the hair between glass slides and measuring with a micrometerequipped microscope. The length of the hair was rapidly scanned and measured at many different places along the shaft, and the average value (in p) recorded. (2) Medullation The presence or absence of medulla was scored with zero for no medullation, one for discontinuous medullation and two for continuous medullation. It was necessary to bleach some pigmented samples slightly to see the microscopic structure. Commercial ammonia powder bleach was used, which left the microscopic structure intact. (3) Scale count The number of cuticular scales on a straight line per given distance - in this case 0.52 mm - was counted. As values of this measurement vary widely from worker to worker, one should use only one s own data for comparative purposes (Garn, 48). (4) Kinking A kink is a sudden constriction and twisting of the hair shaft, producing an obvious discontinuity in curvature. Each hair was scored zero or one, or no kink or kink. Vernacular usage has the term kinky hair describing any highly curly and twisting hair, but a true kink is a rare and impressive thing. (5) Average curvature Each hair was placed between two glass slides, allowing measurement of the curvature of hairs that vary in three dimensions. The radius of curvature of each curve of the hair was determined by placing a transparent template with circles of known radius over the sample and matching an arc of the appropriate circle with the curve. Average curvature itself is the inverse of the average radius of curvature; a high average curvature is represented by a high number. (6) Ratio of maximum to minimum curvature The maximum and minimum curvature are those curls of the hair that have the highest and lowest curvatures, as measured above. The ratio was computed by dividing the minimum into the maximum curvature; the ratio is hence a measure of the regularity of curling. A hair that has both small and large curves (an irregular hair) will have a high ratio, while a hair that has curves of similar size (a regular hair) will have a low ratio - approaching one. A hair that has only a single curve will have the same maximum and minimum curvature and hence a ratio of one. (7) Crimp The crimp (a term used originally to describe wool) is the number of times the direction of curvature changes per unit length. Crimp is not only related to turning of the hair on its axis of propagation, as has been erroneously assumed by Danforth ( 25), but is also related to differential bilateral cellular alignments of the cortex (Chapman, 65). A spiralized hair, although highly curled, would presumably have very little crimp. (8) Ratio of natural to straight length This is a measure of the functional result of curling-a measure of the effect of curling on length. First the maximum length of the curled hair (i.e., natural length) was measured, then the straight length was determined by straightening the hair with two forceps. The ratio is the actual, uncurled, straight length of the hair divided by its maximum curled length. Once the hair is removed from the

QUANTITATIVE HAIR FORM VARIATION 9 influence of scalp oils and other nearby hairs, its configuration is remarkably constant; hence, the curled length is a reproducible physical trait of the hair. The analysis was carried out on hair samples from seven groups. Both population and individual samples were drawn from each group studied and used for between population, within population and within individual comparisons. The population sample for each group consisted of single hairs from a number of people; the individual sample for the group consisted of a number of hairs from a single individual selected at random from the group. Strictly speaking, an average of all individuals in the population should be used for within individual statistics, but this shorter method is adequate for general comparative purposes. Hair was taken from the vertex in all populations except the Sioux and Ifugao, which are museum collections, and contain hair from unspecified locations on the head. None of the hairs had been chemically treated. The whole existing length of each hair was used without any special preparation (except as mentioned under medullation). The hairs were generally about 6 cm long, with the Northwest European and Japanese samples usually longer and the Ifugao somewhat shorter. The groups were: (1) Bougainville. Collected by the 1970 Harvard Solomon Islands Project from the Aita, a non-austronesian-speaking tribe on Bougainville Island. The population sample contained single hairs from 30 individuals, and the individual sample of 40 hairs from a single individual. (2) Malaita. Collected by the 1968 Harvard Solomon Islands Project from the Baegu, a Melanesian speaking coastal tribe. Population sample - 30; individual sample - 40. (3) East Africa. Collected by the author in Kampala, Nairobi, Beira (Mozambique), and Lusaka (Zambia) in 1970. The group was fairly homogeneous even though spread out geographically. Population sample, 20; individual sample, 40. (4) Northwest European. Collected from American students of Northwest European descent at Harvard University in 1971. Population sample, 30; individual sample, 40. (5) Sioux. From a large collection in the Peabody Museum, ca. 1930. Population sample, 30; individual sample, 40. (6) Ifugao (Philippines). From the Beyer collection in the Peabody Museum. This is an old collection ( 06), but the hair is impervious to dry storage of this length due to the high stability of the many cross-chain disulfide bonds of the keratin proteins which constitute the bulk of hair (Mercer, 61). Population sample, 30. Neither Ifugao or Japanese have individual samples, due to insufficient sample size. (7) Japanese. From workers at the Massachusetts General Hospital and students at Harvard, collected in 1971. Population, 30. The univariate and multivariate analyses were carried out on Harvard s IBM 360 computer using the Statistical Package for the Social Sciences (Nie et al., 70) and Data Text (Armor and Couch, 71). RESULTS (1) Diameter As expected the Mongoloid groups were significantly higher than the other populations (see tables 1 and 2 for basic univariate statistics and significance tests). The Ifugao had especially thick hair, being significantly higher than the Sioux and Japanese. African and European hair showed no significant difference in diameter, contrary to several reports, (e.g., Steggarda and Seibert, 41; Garn, 48). The Bougainville group was higher than the Malaita group, but not as high as the Mongoloids. The within individual variance was generally less than the within population variance, but not nearly significantly so. The high individual variance relative to the population variance has been noted by Trotter and Duggins ( 48). (2) Medullation Medullation is closely correlated with diameter. Once again the Mongoloid population means were separated from the other group means by a significant margin. However, the thinner-haired populations clustered, with only the extremes - African and Eurpean - being significantly different (p < 0.05). It is interesting that medullation is not completely related to diameter; for instance the Afri-

c 0 TABLE 1 Means and standard deviations of eight hair form variables for seven populations Bougainville P:n=30 I:n = 40 Malaita East A*ica p:n=30 I:n=40 P:n=20 I:n=40 N.W. Europe P:n=30 I:n = 30 Sioux Japan Ifugao Total ~ _ P:=30 I:n=40 P:n=30 P:n=30 N=390 Diameter Mean 84.06 SD 13.15 81.25 10.27 76.80 12.42 82.25 11.09 76.10 15.83 77.32 12.03 79.17 11.27 73.50 11.89 89.97 13.50 93.32 11.36 95.53 10.35 104.20 12.77 84.58 14.73 Medullation Mean 0.467 SD 0.571 0.600 0.496 0.500 0.509 0.475 0.554 0.700 0.657 0.500 0.599 0.333 0.547 0.267 0.450 1.133 0.629 1.635 1.353 1.233 0.626 1.800 0.484 0.805 0.838 Scale count Mean 18.33 SD 2.00 Kinking Mean SD Average curvature Mean 4.719 SD 0.849 Ratio of curvature Mean 2.114 SD 0.813 19.05 2.15 0.050 0.221 4.518 0.498 2.373 0.805 18.33 2.28 0.033 0.183 3.221 1.077 1.859 0.565 19.17 1.72 3.292 0.645 2.058 0.633 17.95 2.04 0.100 0.308 7.553 1.780 1.456 0.429 18.30 1.65 0.075 0.267 6.607 1.291 1.484 0.362 15.07 1.80 0.186 0.132 1.103 0.281 15.20 1.16 0.308 0.158 1.594 1.091 15.83 2.07 0.079 0.067 1.000 15.55 1.52 0.073 0.032 1.025 0.158 15.47 1.46 0.069 0.040 1.ooo 17.47 2.16 0.103 0.097 1.033 0.183 17.21 2.40 tl 9 0.021 0.142 5 r 3: w 2.542 2.687 ' 1.533 0.735 Crimp Mean 1.117 SD 0.806 0.714 0.291 0.530 0.266 0.688 0.207 0.949 0.569 0.874 0.448 0.028 0,056 0.093 0.090 0.004 0.028 0.008 0.043 0.419 0.521 Ratio of natural to straight length Mean 2.669 SD 0.750 3.083 0.797 2.008 0.634 1.865 0.337 2.971 1.387 2.945 0.866 1.212 0.308 1.186 0.138 1.019 0.038 1.003 0.01 1 1.013 0.035 1.038 0.02 7 1.845 1.010 P, population sample statistics (drawn from n individuals in a population). I, individual sample statistics (n hairs drawn from a single individual).

QUANTITATIVE HAIR FORM VARIATION 11 TABLE 2 Variables signi$cantly different at p < 0.01; two-way comparisons of seven populations East N. W. Malaita Africa Europe Sioux Japan Ifuago Bougainville AC,CR SC,AC RC,CR Malaita East Africa N.W. Europe Sioux Japan SC,AC RC,CR SC,AC RC,CR D, Diameter. AC, Average curvature. M, Medullation. RC, Ratio of curvature. SC, Scale count. CR, Crimp. K, Kinking (none)., Ratio of natural to straight length. M,SC,SC D,SC,SC,SC,SC none D,AC RC,CR SC,AC,SC SC, can population is higher than the thickerhaired Bougainville sample. The individual variances were again comparable to the population variances, with no significant differences between within individual and within population variances. (3) Scale count Two significant clusters were formed - the curly haired populations plus the Ifugao on one hand, and the straight haired groups and Europeans on the other. (4) Kinking The African group was higher than any of the other groups; besides the African, only the Malaitan population sample and the Bougainville individual sample exhibited kinking. Both African and Bougainville samples were highly curled, but the Bougainville samples do not show as much kinking as the African. As shown in tables 4 and 5, kinking is largely uncorrelated with curvature; it is an independent characteristic of some populations. (5) Average curvature This is quite an important variable in the description of hair form. The African population had by far the highest curvature, the Solomon Islands groups next (Bougainville significantly higher than Malaita), followed by the European, with the Mongoloid populations grouped with low values. Each of the populations (except for the Mongoloid cluster) was significantly isolated from other groups. Even the African and Bougainville groups were distinct from each other at the 0.001 level, a fact that is not evident to gross observation. The within individual and within population variances were again comparable. (6) Ratio of curvature Another important variable in regard to population differences, ratio of curvature, also exhibited a wide range of variation. The Solomon Islands populations, especially Bougainville, showed a high degree of irregularity in the curling of a single hair. The Solomon Islands groups were more irregular than the African group, which may account for the early description of Melanesian hair as frizzy rather than woolly. The Mongoloid and European samples were much more regular. Many of the problems of hair description come from the treatment of the two variables - Average curvature and Ratio of curvature-by the casual observer. The tendency is for any curly

12 DANIEL HRDY hair to be described as being merely to compare within and between popula- curly, etc., without realizing the type tion variances (table 3). In all variables or amount of curling. There are quite except kinking, the between group varisignificant differences that exist between ance was higher than the within group African and Melanesian hair that give them a completely different metric charvariance with a significance below the p = 0.001 level, while kinking was sigacter. nificant at the p = 0.05 level. Average (7) Crimp curvature had by far the largest F ratio, while kinking was the smallest. In marked The with highest were contrast to the between and within popu- BougainviUe and East Africa, with Mala- lation F-ratios, the variances within popita significantly lower. The European sam- ulations are quite to the var- PIe was significantly higher than the Man- iances of different hairs from a single goloid cluster, but quite distant from the individual, the within individual variance curly haired groups* Although the African being occasionally larger (tables 1, 3). group had a higher curvature than the Bougainville sample, the Bougainville Correlations group had greater crimp. The correlation matrix for the eight (8) Ratio of natural to straight length variables is shown in table 4. For the p%rposes of this study, the entire sample (N = This variable is quite similar to average 390) was used to compute this matrix. curvature and crimp, with the African The basic matrix is assumed to be a fair and Melanesian groups high and Mongo- sample of the hair form universe. A single loid low. Even though the Bougainville hair is a member of the universe beand African hairs are different in the sides a single population. Later, populamethod of curling - note the average tion inferences can be drawn from the curvature and ratio of curvature variables relation of a specific population to the - the functional result is similar; the universe. Because of the similarity of hair is shortened by an almost equal amount. F-tests F-tests were done on the eight variables TABLE 3 variances. both the notmlation and individual samples were *uskd to compute the universe matrix. As expected, there is a large positive correlation between diame- Between populations and among population F-ratios for eight variables and seven populations Mean squares Within single Within Among individual population population Diameter 128.760 160.900 3153.180 Medullation 0.589 0.329 8.352 Scale count 2.791 3.944 58.558 Kinking 0.024 0.014 0.031 Average 0.472 0.599 218.719 curvature within pop. within indiv. 1.25 0.56 1.41 0.58 1.27 F-ratios Among population within pop. 19.60 3 25.40 3 14.85 3 2.19 1 364.84 3 Ratio of 0.479 0.182 6.275 0.38 34.463 curvature Crimp 0.068 0.141 6.626 2.071 47.083 Ratio of 0.304 0.349 18.49 1.14 53.01 3 natural to straight length d.f. = 39,161 and 6,193 1 p<.05. 2 p<.ol. 3 p <.001.

~~ QUANTITATIVE HAIR FORM VARIATION 13 TABLE 4 Total correlation matrix for eight variables and seven populations; N = 390 1 2 3 4 5 6 7 1 Diameter 2 Medullation 3 Scalecount 4 Kinking 5 Average curvature 6 Ratio of curvature 7 Crimp 8 Ratio of natural to straight length 1 p < 0.05. 2 p<o.ol. 0.678 2-0.057-0.080-0.014-0.031-0.085-0.360 2-0.302 2 0.477 2 0.176 2-0.181 2-0.2162 0.3702 0.032 0.3752-0.217 2-0.229 2 0.422 2 0.123 1 0.733 2 0.402 2-0.304 2-0.288 2 0.439 2 0.162 2 0.798 2 0.436 2 0.489 2 ter and medullation. There is a slight correlation of scale count with the curling variables. The four curling variables have a strong intercorrelation with each other and a negative correlation with diameter. Kinking is correlated slightly with curling. The average within-population correlations and the between-population correlations were computed using a program of Prof. W. W. Howells (table 5). Only the population samples were used (N = 200). It can be seen that absolute values of the between-population correlations are much greater than the within-group correlations. This indicates that most of the total correlations are due to between group associations. For instance, the correlation between diameter and crimp may not be significant in a population, but when this population is compared to other populations, it happens that crimp is negatively correlated with diameter. In other words, in the hair form universe a thick hair is likely to be an uncrimped one; this is not due to any intrinsic relation of shaft size and crimping, but to the nature of the groups studied. Multivariate descriptions based on the total correlation matrix will be largely Iimited to describing the rela- TABLE 5 Between population and average within population correlation matrices for six variables and eight populations; population samples, N =200 1 2 3 4 5 Between population correlations 1 Diameter 2 Scalecount 3 Average curvature 4 Ratio of curavature 5 Crimp 6 Ratio of natural to straight length - 0.270 2-0.576 2 0.791 2-0.551 2 0.799 2 0.635 2-0.581 2 0.875 2 0.963 2 0.816 2-0.563 2 0.798 2 0.965 2 0.755 2 0.981 2 Average within population correlations 1 Diameter 2 Scalecount 3 Average curvature 4 Ratio of curvature 5 Crimp 6 Ratio of natural to straight length 0.070-0.1 72 2-0.048 0.041 0.066-0.092 0.079-0.016 0.135 1-0.006-0.078 0.044 0.237 2 0.081-0.350 2 1 p<0.05. 2 pio.01.

14 DANIEL HRDY tion of the populations to one another and the universe rather than any functional relations between the variables. The within group correlations have more meaning for understanding the variables themselves. Principal components analysis Principal components analysis (PCA) is an efficient method of reducing the variables to more manageable numbers by reducing the basic matrix to a few orthogonal components, accounting for both unique and non-unique variance. The PCA was carried out on a six variable correlation matrix. The variables of medullation and kinking were discarded because they do not have continuous distributions and they have low F-ratios. The components were extracted and simplified by varimax rotation, with three components accounting for 80% of the variance (see table 6 for rotated factor loadings). This three component model has a specialized curling component, a size component, and a regularity of curling component. The third factor is of interest because it is highly loaded on a variable (ratio of curvature) which is significantly different in African and Melanesian populations. The actual scoring of the populations was done using a linear function of the standardized scores of the conceptual variables and factor score coefficients. The scaled factor scores are given in table 7. The populations broke into separate racial groups in most cases (see table 8 for significant tests of populations for individual factor scores). The Solomon Islands groups were in all cases quite distinct (p < 0.05) from the Africans, the Mongoloids forming a cluster (Sioux and Japanese very similar, with the Ifugao less so), and the European group isolated. These data are graphed in figure 1, with each component given equal weight. There is a question whether or not to weigh each component according to its percentage of the total variation. Generally, equal weightings are used for description (as in this case), and percentage weightings for computing distances. See Howells ( 70), for a discussion of the problem. The first component, high correlation with curling and scale count, separated Africans and Solomon Islanders from the other groups but with a significant difference between Africans and Melanesians. The second, mostly correlated with size, separated the Mongoloids from the other groups; the Ifugao had by far the largest score, relatively isolating it from the other Mongoloid populations. In the third component (highly correlated with regularity ) the Solomon Islanders were separated from the more regularly curled Africans and noncurled groups. The principal components method is satisfying conceptually, and gives clearer results than a discriminant functions analysis, which is rather ambiguous because of the high F-ratio for average curvature. DISCUSSION Because average curvature is so variable between groups (F-ratios, table 3), it will play the largest part in their differentiation; other curling variables such as regularity of curling and crimp are also important. Diameter, medullation, and scale count, which have been used to describe populations in the past (e.g., Trotter et al., 56) have lower population F-ratios than the curling variables, and are therefore more difficult variables with which to obtain significant comparative results. The use of scale count (e.g., Anderson, 69), or indeed any variable, as an individualization technique is largely wishful thinking; the within-individual variance of hair form is prohibitively large. Hair form, like most metric traits, is undoubtedly polygenic (see, for example, Day, 32, studies in Negro-white hybridization). If an individual can exhibit as much phenotypic variance as a population, as is the case for most variables (table 3), the implication is that the range of genes in the population pool is limited. Interaction with the environment undoubtedly plays an important part, giving a wide range of expression to a single genotype. It is more likely that there are relatively few genes in the pool and many expressions thereof, rather than the trait being controlled by a large number of loci acting differentially on individual hair follicles. = A.R-I.Z; factor scorecoefficients = A.R-I (see Harmon, 70) where- is factor score matrix R is correlation matrix A is component loadings matrix Z is standardized score matrix

QUANTITATIVE HAIR FORM VARIATION 15 TABLE 6 Loadings offirst three principal components on six variables after varimax rotation Component I Component I1 Component Ill Diameter -0.178 0.929-0.087 Scale count 0.633 0.288 0.399 Average curvature 0.915-0.253 0.107 Ratio of curvature 0.230-0.122 0.944 Crimp 0.804-0.078 0.163 Ratio of natural to straight length 0.777-0.244 0.238 eigenvalue 3.141 0.968 0.706 per cent of variance 52.4 16.1 11.8 Bougainvine 1,002 Malaita 0.258 East Africa 1.640 N. W. Europe 0.761 Sioux 0.651 Japan 0.654 Ifugao - 0.380 TABLE 7 Scaled components scores for seven populations Component I Component I1 Component 111 mean S.D. mean S.D. mean S.D. 0.704-0.075 0.953 1.015 1.429 0.558-0.423 0.745 0.848 0.975 0.697-0.636 1.064-0.329 0.708 0.274-0.674 0.756-0.349 0.497 0.226 0.095 0.801-0.460 0.186 0.173 0.345 0.635-0.499 0.129 0.249 1.172 0.813-0.280 0.360 IF~GAO JAPAU.SIOUX N. W: EUROPE I. MALAITA EAST AFRICA Fig. 1 Centroids of seven populations graphed on three hair form principal components. Of course the gene frequencies vary from group to group; however, within each population there is much less genetic variation than between populations. The between population correlations (table 5) are largely due to the genetic differences between the groups. However, the within population correlations represent correlations of the variables themselves (table 5). Thus, most variables are either uncorrelated with other variables within a group or correlated with others due to geometric relations as are some of the curling variables. However, the negative correlation of average curvature and diameter may have some significance in the explanation of the origin of curling. The relation of diameter (and size) to curvature supports models based on mechanical as well as chemical forces, such as Chapman s ( 65) hypothesis of the action of the arrector pili muscles on crimp formation in wool. The separation produced by principal component analysis of metric traits (fig. 1) has obvious relation to the genetic separation of the groups, and hence to some of the evolutionary history of the groups. Although data of this type are best analyzed using clustering or distance methods, as will be demonstrated in a future paper, they nonetheless give qualitative relationships. Using three components, four clusters are isolated in different sec-

16 DANIEL HRDY TABLE 8 Component scores significantly different at p <0.01; two-way comparisons of seven populations Malaita East N. W. Afkica Europe Sioux Japan Ifugao 1, component I. 2, component 11. 3, component 111. tors of the three-dimensional space (table 7). These clusters are the Melanesian (+, component I; -, component 11; +, component 111), African ( +, -, - ), European(-,-,-,)andMongoloid(-,+,-). The differentiation of the European, African, and the Mongoloid groups on the basis of metric traits is not surprising, but the relation of the Melanesian groups to other groups had not been clearly established in such a non-anecdotal way. In all the components, the Malaita group was nearer the non-curled groups relative to the Bougainville sample (significantly in the first and second components, testing using the dispersion of individual scaled factor scores in each population sample). It has been postulated (A. Damon, personal communication) that there exists a significant proportion of Polynesian genes in the gene pool of the coastal peoples of Malaita; this theory is supported by the apparent influence of noncurling genes in the Malaitan hair form pool. Although the Bougainville and African groups were found to be quite distinct statistically, it is apparent that functionally their hair form is similar (i.e., ratio of length variable). Is this likeness indicative of a genetic relation or the result of environmental pressures? It is becoming apparent that the Melanesian and African populations have been separate for a very long time. Blood groups and other serological traits, of course, are different (e.g., Boyd, '50; Gajdusek et al., '67), as are metric characteristics of the skull (Howells, '68), and melanosome packaging (Szabo et al., '69, and personal communication). The similarity of func- tional hair form is most probably due to parallel evolution under similar environmental pressures. However, without a more precise knowledge of the physiological mechanisms producing curling, such discussion can be only speculative. ACKNOWLEDGMENTS I would like to acknowledge the help of Drs. A. Damon, W. W. Howells, L. A. Goldsmith and H. P. Baden of Harvard University. This research was supported by grant GM13482 and PHS Training grant TO1 GM01938 from the National Institute of General Medical Sciences, and United States Public Health Service grant AM0 6838. This research is part of the Human Adaptability section of the International Biological Program, and was conducted with the permission and kind assistance of the British Solomon Islands Protectorate and the Territory of Papua and New Guinea, Commonwealth of Australia. LITERATURE CITED Anderson, H. P. 1969 A simple scheme for the individualisation of human hair. Microscope, 17: 221-227. Armor, D. J., and A. S. Couch 1971 The Data Text Primer (M.S.). Boyd, W. C. 1950 Genetics and the Races of Man. Little, Brown, Boston. Chapman, R. E. 1965 Ovine arrector pili musculature and crimp formation in wool. In: The Biology of Skin and Hair Growth. A. G. Lyne and B. F. Short, eds. American Elsevier Publishing Co., New York, pp. 201-232. Danforth, C. H. 1925 Hair, with Special Reference to Hypertrichosis. American Medical Association, Chicago. Day, C. B. 1932 A Study of Some Negro-White Families in the United States. Peabody Museum of Harvard University, Cambridge. Deniker, J. 1900 The Races of Man. Schriber, London.

QUANTITATIVE HAIR FORM VARIATION 17 Gajdusek, D. C., J. Gwart, R. L. Kirk, R. W. Carrell, D. Irvine, P. A. M. Kynoch and H. Lehmann 1967 Haemoglobin J Tongariki (a 115 Alanine + Aspartic Acid): The first new haemoglobin variant found in a Pacific (Melanesian) population. J. Med. Genet., 4: 14. Garn, S. M. 1948 Human Hair: Its Anatomy, Growth, and Distribution. Ph.D Thesis. Harvard University, Cambridge. Harmon, H. H. 1970 Modern Factor Analysis. University of Chicago Press, Chicago. Hausman, L. A. 1925 The relationships of the microscopic structural characters of human head-hair. Am. J. Phys. Anthrop., 8. 17S177. Hooton, E. A. 1946 Up from the Ape. Mac- Millan, New York. Howells, W. W. 1968 Multivariate Analysis of Human Crania. Proceedings, VIIIth International Congress of Anthropological and Ethological Sciences, Tokyo and Kyoto; I (published 1970). 1970 Anthropometric grouping analysis of Pacific peoples. Arch. and Phys. Anthro. in Oceania, 5: 192-217. Huxley, T. H. 1865 On the Methods and Results of Ethnology. Man s Place in Nature and Other Essays, New York (1908). Martin, R. 1928 Lehrbuch der Anthropologie. Jena, I: 213. Mercer, E. H. 1961 Keratin and Keratinization; an Essay in Molecular Biology. Pergamon Press, Oxford. Nie, N. H., D. H. Bent and C. H. Hull 1970 SPSS (Statistical Package for the Social Sciences). McGraw-Hill and Co., New York. Pruner-Bey 1864 On the human hair as a race character. Anthropological Reviews, London, 2: 1. Steggerda, M., and H. C. Seibert 1941 Size and shape of head hair from six racial groups. J. Hered., 32: 315-318. Szabo, G., A. B. Gerald, M. A. Pathak and T. B. Fitzpatrick 1969 Racial differences in the fate of melanosomes in human epidermis. Nature, 222: 1081-1082. Trotter, M. 1938 A review of the classifications of hair. Am. J. Phys. Anthrop., 24: 105-126. Trotter, M., and 0. Duggins 1948 Age changes in head hair from birth to maturity. I. Index and size of hair of children. Am. J. Phys. Anthrop., 6: 489-506. Trotter, M., 0. Duggins and F. M. Setzler 1956 Hair of Australian Aborigines (Arnhem Land). Am. J. Phys. Anthrop., 14: 649459. Wynkoop, E. M. 1929 A study of the age correlations of the cuticular scales, medulla and shaft diameters of human head hair. Am. J. Phys. Anthrop., 23: 177-188.