AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 127:342–350 (2005) Dispersion Dimorphism in Human Populations Elisabetta Marini,1 Esther Rebato,2 Walter Racugno,3 Roberto Buffa,1 Itziar Salces,2 and Silvana M. Borgognini Tarli4* 1 Sezione di Scienze Antropologiche, Dipartimento Biologia Sperimentale, Università di Cagliari, 09042 Monserrato (Cagliari), Italy 2 Dipartamento de Genética, Antropologı́a Fı́sica y Fisiologı́a Animal, Facultad de Ciencias, Universidad del Paı́s Vasco-Euskal Herriko Unibertsitatea, 48080 Bilbao, Spain 3 Dipartimento di Matematica, Università di Cagliari, 09124 Cagliari, Italy 4 Unità di Antropologia, Dipartimento di Etologia, Ecologia, e Evoluzione, Università di Pisa, 56126 Pisa, Italy KEY WORDS sexual dimorphism; living populations; bootstrap test; female canalization ABSTRACT The aim of this paper is to verify the existence of dispersion (or variability) dimorphism in several anthropometric traits, i.e., some skeletal, muscular, and adipose dimensions, and to provide an evaluation of this scarcely considered aspect of sex differences. This research focuses on two human samples from two different European regions, Sardinia (Italy) and the Basque Country (Spain). Twenty-three anthropometric measurements were taken. A ﬁrst qualitative analysis was carried out comparing the proﬁles of the coefﬁcients of variation of each variable in both sexes. Secondly, the equality of variability was veriﬁed with different tests. In the normal case, Student’s t-test, as proposed by Sokal and Braumann ( Syst. Zool. 29:50 – 63), was applied. In the more general case, it was necessary to resort to resampling techniques. A suitable bootstrap test was constructed and the results were compared, when possible, with those obtained by other methods. The measurements showed parallel trends in the two populations. In particular, subcutaneous skinfolds showed signiﬁcant dispersion dimorphism, with the greatest coefﬁcients of variation in men. We suggest that this result was linked to stronger canalization in female dimensions related to the reproductive function and inﬂuenced by sociocultural factors. We also suggest deﬁning dispersion dimorphism, whose existence is conﬁrmed by the results presented in this paper, as a descriptive pattern of phenotype variability in both sexes that can be speciﬁc of a given class of anthropometric traits. Am J Phys Anthropol 127:342–350, 2005. Intrasexual variability is a distinct level within the hierarchical classiﬁcation of variation, starting with the ontogeny of individuals and going through intrasexual, intersexual, interdemic, subspeciﬁc, and interspeciﬁc levels (Albrecht and Miller, 1997). As concerns the study of sexual dimorphism in primates, intrasexual variability has received sporadic consideration (Cope and Lacy, 1992; Ipiña and Durand, 2000; Kelley and Plavcan, 1998; LaVelle, 1995; Leutenegger and Cheverud, 1982, 1985; Leutenegger and Larson, 1985; Marini et al., 1999; Meindl et al., 1985; Oxnard, 1987; Plavcan, 2000, 2001; Plavcan and Cope, 2001; Plavcan and Kay, 1988; Tague, 1989, 1991, 1992, 1995; Wood, 1976). However, intrasexual variability plays a double role in the expression of sexual dimorphism: in case of equal distance between mean values, a decreased dispersion of measurements in both sexes produces a decrease in the overlap of corresponding distributions and hence a higher degree of dimorphism; moreover, intrasexual variability can be in itself a component of dimorphism when sex differences in the extent of variability of a given trait are signiﬁcant (Marini et al., 1999). We propose here the new label “dispersion dimorphism” which is more appropriate than the usual term “variance dimorphism” (Marini et al., 1999; Plavcan, 2000), since variance gives an absolute measure of variability that depends on the order of magnitude of the variables considered. The literature aimed at verifying the existence of dispersion (or variability) dimorphism in primates is characterized by divergent results. Application of inadequate statistical methods, or violation of the assumptions of the statistical model used, may be possible explanations for such discrepancies. A series of papers published during the 1980s, building on research by Schultz (1926), pointed to a © 2005 WILEY-LISS, INC. © 2005 Wiley-Liss, Inc. Grant sponsors: Ministero Istrucione Universitá Ricria (Italy) and Univirsidad Paı́s Vasco-Euskal Herriko Universitatea (Spain). Grant number: 1/UPV 00154.310-E-13972/2001. *Correspondence to: Silvana M. Borgognini Tarli, Unità di Antropologia, Dipartimento di Etologia, Ecologia, e Evoluzione, Università di Pisa, via S. Maria 55, 56126 Pisa, Italy. E-mail: firstname.lastname@example.org Received 3 December 2003; accepted 25 June 2004. DOI 10.1002/ajpa.20134 Published online 6 December 2004 in Wiley InterScience (www. interscience.wiley.com). 343 DISPERSION DIMORPHISM IN HUMANS greater male than female variance in body weight, skeletal, and dental dimensions in several primate species (Leutenegger and Cheverud, 1985; Leutenegger and Larson, 1985; Wood, 1976) including humans, where a smaller variability of female pelvic traits was also noticed (Meindl et al., 1985). On the other hand, Plavcan and Kay (1988), analyzing dental dimensions in several platyrrhine species, found that trait variance in males was less or greater than that in females in an equal number of cases. Tague (1989, 1995), studying intrasexual variability in pelvic traits of humans and nine species of nonhuman anthropoids, did not ﬁnd signiﬁcant variability dimorphism. Plavcan (2000) reported a greater male within-sex coefﬁcient of variation only in 54% of all comparisons performed with craniometric data in 35 primate species. On the whole, investigations aimed at verifying the signiﬁcance of sex differences refer to a ﬁnite set of variables, mostly represented by body weight and dental or cranial dimensions. As yet, a systematic analysis of dispersion dimorphism of different metric traits that can be referred to the various body compartments (e.g., skeletal, muscular, or adipose) is not available. These components vary in their ecosensitivity, and in their expression in the two sexes (Waddington, 1942). Moreover, it is likely that various dimensions are differently affected by possible causal mechanisms of variability dimorphism, such as natural and sexual selection, canalization, and sociocultural factors. The aim of the present study was to evaluate the expression of dispersion dimorphism in anthropometric traits. The investigation was carried out on a range of variable categories and on two different human populations, characterized by diverse environments and history, and devoid of appreciable reciprocal gene ﬂow. The study was directed toward young adults exclusively, in order to avoid possible confounding sex-by-age interaction effects. The relevance of the study of dispersion dimorphism lies in the fact that it expands our knowledge of human variability and sex differences. Moreover, since most proximate and ultimate factors affect phenotype variability, the evaluation of dispersion dimorphism is a necessary step toward the identiﬁcation of mechanisms working differently on the two sexes and, therefore, toward the understanding of causes and correlates of sexual dimorphism. In particular, the study of dispersion dimorphism is a direct way of testing the better female canalization hypothesis (Hamilton, 1982; Stini, 1972, 1982; Stinson, 1985), which postulates that males are more sensitive to the environment than females and hence generally more variable. MATERIALS Basque sample The sample from the Basque Country consisted of 200 individuals of both sexes (100 males and 100 females) aged 20 –25 years (21.48 ⫾ 1.20 years in males, and 21.93 ⫾ 1.44 years in females). This sample came from a broad cross-sectional study on university students from the Basque Autonomous Community started in 2001. Most individuals (89.6% males and 87.5% females) came from urban areas in one of the three Basque provinces. On the basis of a preliminary interview used to deﬁne socioeconomic status, the sample can be considered middle class. Sardinian sample The Sardinian sample consisted of 280 individuals of both sexes (140 males and 140 females) aged 20 –25 years (22.01 ⫾ 1.51 years in males, and 22.05 ⫾ 1.38 years in females). The sample came from a large database collected between 1995–1998 containing personal, social, and anthropometric data of the Sardinian population. Individuals were randomly selected from the population born in the town of Cagliari. On the basis of socioeconomic characteristics, the sample can be considered middle class. METHODS Body measurements In both samples, 20 measurements were taken: weight (kg); stature, sitting height, and iliospinal height (cm); biacromial, biiliac, elbow, and knee breadths (cm); relaxed and tensed midarm circumferences, and waist, thigh, and calf circumferences (cm); and biceps, triceps, subscapular, suprailiac, abdominal, midthigh, and calf skinfolds (mm). In addition, in the Sardinian sample alone, three cephalic/facial measurements were taken: cephalic length and breadth (cm), and bizygomatic breadth (cm). In each sample, measurements were taken by the same experienced observer (E.M. or E.R.), following standard international criteria (Weiner and Lourie, 1981; Lohman et al., 1988). Statistical analysis The tests more widely used for the comparison of variability in two populations require assumptions of the variables under study. Such assumptions are not always justiﬁable. In particular, F statistics for comparison between variances assume the normality and stochastic independence of the two variables. Moreover, when the variables have different scales, the signiﬁcance of the difference between the variances may not be indicative of the different dispersion of the distributions. The use of an index of relative variability, such as the coefﬁcient of variation CV⫽/ removes this last difﬁculty, but it does not permit reference to the F-Fisher probability law, because the ratio of variation coefﬁcients does not follow that distribution (except in the obvious case of equality between the two means). 344 E. MARINI ET AL. TABLE 1. Descriptive statistics and t-test results of sample from Basque Country1 Men Weight Stature Sitting height Iliospinal height Breadths Biacromial Bi-iliac Elbow Knee Circumferences Midarm relaxed Midarm tensed Waist Thigh Calf Skinfolds Biceps Triceps Subscapular Suprailiac Abdominal Midthigh Calf Women Mean SD N Normality Mean SD N 73.15 175.72 91.46 103.65 10.66 6.63 3.72 5.98 39.20 29.03 6.59 9.51 100 100 100 100 n.s. n.s. n.s. n.s. 58.96 162.10 85.86 95.72 9.50 6.89 3.36 5.25 100 100 100 100 P ⬍ 0.01 n.s. n.s. n.s. 3.70 2.44 0.41 0.47 100 100 100 100 P ⬍ 0.01 P ⬍ 0.01 P ⬍ 0.01 n.s. 35.60 27.05 5.79 8.89 2.03 2.15 0.36 0.49 100 100 100 100 n.s. P ⬍ 0.01 P ⬍ 0.01 P ⬍ 0.01 28.48 30.71 79.65 52.39 36.95 2.80 2.95 8.62 4.48 2.81 100 100 100 100 100 n.s. n.s. P ⬍ 0.01 n.s. n.s. 25.79 26.60 70.71 49.30 35.28 2.70 2.75 8.14 4.41 2.70 100 100 100 100 100 n.s. P ⬍ 0.01 P ⬍ 0.01 n.s. P ⬍ 0.01 6.13 9.59 12.21 13.56 16.30 15.21 10.08 3.33 4.99 6.15 7.43 8.55 6.18 5.03 85 85 85 85 85 85 85 P⬍ P⬍ P⬍ P⬍ P⬍ n.s. P⬍ 8.87 16.44 13.32 14.89 16.37 23.99 17.75 3.33 4.99 5.52 6.20 5.88 6.05 5.15 98 98 98 98 98 98 98 n.s. n.s. P ⬍ 0.01 n.s. n.s. P ⬍ 0.01 n.s. 0.01 0.01 0.01 0.01 0.01 0.01 Normality t-test P ⬍ 0.01 P ⬍ 0.01 P ⬍ 0.01 P ⬍ 0.01 P ⬍ 0.01 1 Weight is expressed in kg; skinfolds in mm; all other measurements in cm. Under “Normality” results of Lilliefors test are reported. Variables above signiﬁcant level cannot be considered as normally distributed. Results of t-test are shown only when variables were normally distributed in both sexes. n.s., not signiﬁcant. Neither is it possible to bypass the obstacle assuming any functional link between variables, such as their proportionality for example (Lewontin, 1966), because it would contradict the hypothesis of variable independence. Finally, for some anthropometric variables, the normality assumption is not veriﬁed, and therefore tests of the Student kind cannot be applied. All things considered, the more general case refers to variables whose probabilistic model is unknown, so that the only available information is contained in the data. In such circumstances, the use of empirical methods, generally based on resampling techniques, is suggested. For this purpose, a bootstrap test for the null hypothesis H0: CVX1 ⫽ CVX2 vs. the alternative H1: CVX1 ⫽ CVX2 was designed by using the following test statistic 冉 兩CV X 1⫺CV X 2兩 冊 CV X2 1 CV X2 2 ⫹ 2n 1 2n 2 1/2 , which is asymptotically normal (Sachs, 1984), where CVXi ⫽ SXi/x i, SXi denotes the standard deviation and SXi and ni is the size of sample from Xi, I ⫽ 1,2. In order to evaluate the signiﬁcance of the test conditional on Ho, i.e., in order to obtain the “bootstrap distribution” on which the test is based, the following transformation of the original variables was used: T ⫹ n 2X 2T nX iT ⫹ 1 1 X⬘i ⫽ X iT ⫺X , i ⫽ 1,2, n 1⫹n 2 冉 冊 n1S12 ⫹ n2S22 1/2 iT is the mean and X niSi2 value of XiT (more detailed information on the mathematical/statistical properties of the test can be found in Cabras et al., 2004). It is easy to show that CVX⬘1⫽CVX⬘2. The behavior of the test in various anthropological situations and in comparison with other tests was considered by Borgognini Tarli, Marini, and Racugno (unpublished results). Here, as an example, we will show some comparisons permitted by the assumptions on the variables: in particular, in the case of normally distributed variables, with Student’s t-test as suggested by Sokal and Braumann (1980), and with the usual F-test, when the two means are not signiﬁcantly different. The inferential analysis of the coefﬁcients of variation will be preceded by a short descriptive study of the variability proﬁles (Sokal and Braumann, 1980) for the comparison of the two sexes in each population. We employed the test of Lilliefors (1967), a modiﬁed Kolmogorov-Smirnov test that permits using the sample mean and the sample standard deviation in place of mu and sigma, in order to identify variables departing from normality. where XiT ⫽ Xi RESULTS Tables 1 (Basque Country) and 2 (Sardinia) show the results of the test of Lilliefors (1967) on the normality of intrasexual distribution and of the comparison between male and female means (t-test). The ﬁrst two columns also show means 345 DISPERSION DIMORPHISM IN HUMANS TABLE 2. Descriptive statistics and t-test results of sample from Sardinia1 Men Weight Cephalic length Cephalic breadth Bizigomatic breadth Stature Sitting height Iliospinal height Breadths Biacromial Bi-iliac Elbow Knee Circumferences Midarm relaxed Midarm tensed Waist Thigh Calf Skinfolds Biceps Triceps Subscapular Suprailiac Abdominal Midthigh Calf 1 Women Mean SD N Normality Mean SD N 67.22 19.26 15.10 13.26 169.89 87.96 101.45 8.46 0.70 0.64 0.75 6.36 3.82 4.68 39.42 27.31 7.02 9.73 140 141 141 137 140 139 93 n.s. n.s. n.s. P ⬍ 0.01 n.s. n.s. n.s. 53.71 18.33 14.53 12.46 156.70 83.17 93.50 7.59 0.62 0.58 0.61 6.14 3.63 4.65 140 172 172 159 140 140 101 P ⬍ 0.01 n.s. P ⬍ 0.01 n.s. n.s. n.s. n.s. 2.42 1.54 0.39 0.55 138 138 103 101 n.s. n.s. n.s. n.s. 35.14 26.37 6.01 8.88 2.01 2.04 0.35 0.51 139 139 117 110 n.s. n.s. P ⬍ 0.01 n.s. P ⬍ 0.01 26.91 30.83 74.90 50.21 34.08 2.60 2.74 6.17 4.70 3.11 139 139 139 97 93 n.s. n.s. n.s. n.s. n.s. 23.47 25.28 64.41 50.20 32.55 2.01 2.15 5.54 4.47 2.68 134 133 140 99 98 P ⬍ 0.01 P ⬍ 0.01 n.s. n.s. n.s. P ⬍ 0.01 n.s. P ⬍ 0.01 4.10 7.76 10.34 9.23 12.15 12.35 11.50 1.79 3.76 3.78 4.34 6.52 6.09 5.07 137 140 137 140 97 98 93 P P P P P P P 6.08 13.33 12.71 10.21 12.77 24.65 18.63 2.51 4.78 4.53 4.48 5.74 7.33 6.02 133 140 140 140 111 118 100 P⬍ P⬍ P⬍ P⬍ P⬍ n.s. n.s. ⬍ ⬍ ⬍ ⬍ ⬍ ⬍ ⬍ 0.01 0.01 0.01 0.01 0.01 0.01 0.01 Normality t-test P ⬍ 0.01 P ⬍ 0.01 P ⬍ 0.01 P ⬍ 0.01 P ⬍ 0.01 P ⬍ 0.01 0.01 0.01 0.01 0.01 0.01 Legend as in Table 1. and standard deviations of all metric traits under study. At ﬁrst glance, in both samples, men showed greater mean values than women in variables related to the skeletal component (heights, breadths, and cephalic dimensions) and to the muscular component (circumferences). On the other hand, women had a subcutaneous adipose component more developed than that of men. A similar picture ﬁts well with general knowledge on the patterns of human sexual dimorphism in metric traits. The results of the test of Lilliefors (1967) indicate that in the Basque Country, many variables do not meet the normality condition (10 out of 20 in men, and 9 out of 20 in women). In general, longitudinal dimensions were more often normally distributed, whereas skinfolds showed an opposite tendency. The only ﬁve variables with normal distribution in both sexes (stature, sitting height, iliospinal height, midarm relaxed circumference, and thigh circumference) were examined by t-test to verify the hypothesis of equality between means. As shown in Table 1, such a hypothesis must be rejected in all cases at the 1% level. Thus, in spite of the assumption of normality, it will be not legitimate to apply the F-test, and it will be necessary to resort to coefﬁcients of variation for the comparison of dispersion in males and females. In the Sardinian sample, variables not normally distributed were 8 out of 23 in men and 10 out of 23 in women. The t-test was applied to 10 variables. The difference between male and female means was always signiﬁcant, with thigh circumference the only exception. In this last case only was it possible to use the F-test to verify the equality of intrasexual variances, and to adopt this variable as a case for the comparison by the bootstrap method. Intrasexual variability profiles The amount of variation in a range of anthropometric dimensions can be analyzed by means of variability proﬁles, which are line graphs coefﬁcient of variation (of CV) values (Sokal and Braumann, 1980). Figures 1 and 2 show the variability proﬁles relative to all variables examined in the two samples. It can be noted that variability proﬁles show very similar trends in both samples and in the two sexes. In the sample from the Basque Country, the variables with a smaller CV were those of longitudinal dimensions (in the two sexes, coefﬁcients ranged between 0.038 – 0.058), followed in order of increasing value by transverse breadths (coefﬁcients ranging between 0.050 – 0.094), circumferences (coefﬁcients between 0.076 – 0.115), weight (men CV ⫽ 0.146; women CV ⫽ 0.161), and skinfolds, which showed the greatest CV values (ranging between 0.252– 0.548). In the Sardinian sample, the CV proﬁles almost overlap with those of the Basque sample. The variables with the lowest CV were dimensions of height, followed in order of increasing value by transverse measurements, circumferences, body weight, and ﬁnally skinfolds. It must be noted that CV values of cephalic measurements, taken only in this sample, were close to those of the other skeletal dimensions. 346 E. MARINI ET AL. Fig. 1. Variability proﬁles in sample from Basque Country. Fig. 2. Variability proﬁles in Sardinian sample. Dispersion dimorphism Tables 3 and 4 show the intrasexual CV values, and the results of tests for comparisons between coefﬁcients, for the populations of the Basque Country and Sardinia, respectively. In both samples, trait CVs had a slight tendency toward being greater in males than in females overall. This was due to the consistently higher trait CVs in males in the skinfold measurements. As shown in Table 3, containing the data of the sample from the Basque Country, the bootstrap test highlighted signiﬁcant differences, due to greater coefﬁcients of variation in men, in seven characters: all subcutaneous skinfolds (with the exception of the subscapular), and biacromial breadth. Also in the Sardinian sample (Table 4), the bootstrap test found signiﬁcantly dimorphic variables, with greater male CVs: the triceps, midthigh, and calf skinfolds, the abdominal skinfold excepted, which is close to the limit of statistical signiﬁcance. The biiliac breadth had a greater CV in women than in men. DISPERSION DIMORPHISM IN HUMANS TABLE 3. Coefficients of variation (CV) in two sexes and relative statistical comparisons in sample from Basque Country1 Men Women (CVW) (CVM) Weight Stature Sitting height Iliospinal height Biacromial breadth Bi-iliac breadth Elbow breadth Knee breadth Midarm relaxed circumference Midarm tensed circumference Waist circumference Thigh circumference Calf circumference Biceps skinfold Triceps skinfold Subscapular skinfold Suprailiac skinfold Abdominal skinfold Midthigh skinfold Calf skinfold 0.146 0.038 0.041 0.058 0.094 0.084 0.062 0.050 0.098 0.096 0.108 0.085 0.076 0.543 0.521 0.504 0.548 0.525 0.406 0.499 0.161 0.042 0.039 0.055 0.057 0.079 0.062 0.055 0.105 0.103 0.115 0.089 0.077 0.375 0.304 0.415 0.416 0.359 0.252 0.290 B tSB n.s. n.s. n.s. n.s. n.s. B, bootstrap test; tSB, t-test proposed by Sokal and Braumann (1980), only for normally distributed variables. n.s., not signiﬁcant. * P ⬍ 0.05. ** P ⬍ 0.001. TABLE 4. Coefficients of variation (CV) in two sexes and relative statistical comparisons in sample from Sardinia1 Weight Cephalic length Cephalic breadth Bizigomatic breadth Stature Sitting height Iliospinal height Biacromial breadth Bi-iliac breadth Elbow breadth Knee breadth Midarm relaxed circumference Midarm tensed circumference Waist circumference Thigh circumference Calf circumference Biceps skinfold Triceps skinfold Subscapular skinfold Suprailiac skinfold Abdominal skinfold Midthigh skinfold Calf skinfold B 0.126 0.036 0.042 0.057 0.037 0.043 0.046 0.061 0.056 0.056 0.057 0.097 0.141 0.034 0.040 0.049 0.039 0.044 0.050 0.057 0.077 0.058 0.058 0.086 n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. * n.s. n.s. n.s. 0.089 0.085 n.s. 0.082 0.094 0.091 0.435 0.484 0.366 0.470 0.536 0.493 0.441 0.086 0.089 0.082 0.413 0.358 0.357 0.439 0.449 0.297 0.323 n.s. n.s. n.s. n.s. ** n.s. n.s. n.s. *** ** tSB F n.s. n.s. n.s. *** n.s. n.s. n.s. n.s. and 4. They are always in accordance with those obtained by the bootstrap test. DISCUSSION Intrasexual variability profiles n.s. n.s. n.s. n.s. ** n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. * ** n.s. * ** ** ** 1 Men Women (CVM) (CVW) 347 Intrasexual variability proﬁles of the anthropometric phenotypes observed in the two population samples (Basque Country and Sardinia) showed consistent parallelism (Figs. 1, 2). The more evident phenomenon was an appreciable divergence in the variability levels of skinfolds as compared to all other dimensions, especially skeletal measurements. This pattern of “skinfold phenotype” could be due to the hereditary and/or environmental component, but it might also be inﬂuenced by intra- and interobserver and instrumental error. Indirect information on the proportion of phenotype variability that can be attributed to the genetic or environmental component comes from research on heritability coefﬁcients. Although heritability estimates are affected by some theoretical limitations (e.g., they are speciﬁc for the population and environment analyzed), there is a general consensus on the assumption that skeletal measurements have greater heritability than muscular or adipose dimensions (Mueller, 1977; Salces, 2002; Susanne, 1975). In the case of skinfolds, a substantial inﬂuence of the environment (ecosensitivity) is also indicated by their appreciable individual variability during life. As to residual variability due to technical error, the reliability values of different anthropometric dimensions are available (Lohman et al., 1988; Marks et al., 1989; Ulijaszek et al., 1998). Since the reliability of skinfolds is on the same order of magnitude as that of skeletal diameters (R ⫽ 0.81– 0.95), technical error cannot be the only reason for the observed differences in variability. Therefore, the greater phenotype variability of skinfolds as compared to other measurements might be mainly ascribed to a stronger environmental inﬂuence. n.s. 1 B, bootstrap test; tSB, t-test proposed by Sokal and Braumann (1980), only for normally distributed variables; F, F-test, for normally distributed variables with equal means. n.s., not signiﬁcant. * P ⬍ 0.05. ** P ⬍ 0.01. *** P ⬍ 0.001. In order to conﬁrm the adequacy of the bootstrap test used, t- and F-tests were also applied whenever possible. The relative results are shown in Tables 3 Dispersion dimorphism The range of anthropometric traits analyzed in the present paper allowed us to show the different manifestations of dispersion dimorphism in soft-tissue vs. skeletal traits. As already demonstrated (Plavcan and Kay, 1988; Plavcan, 2000; Tague, 1989, 1995), cranial and postcranial skeletal dimensions did not show signiﬁcant dispersion dimorphism or a noticeably higher frequency of CV values greater in men than in women. Body circumferences and breadths also did not show signiﬁcant differences in variability between the sexes. The only exceptions were biacromial breadth in the sample from the Basque Country (CV value greater in men) and bi-iliac breadth in the sample from Sardinia (CV greater in women). Since the last two results had opposite directions in the 348 E. MARINI ET AL. two samples, it is difﬁcult to assign a speciﬁc significance to these results. More focused analysis might help in better assessing the role of these differences. The main results of this research clearly indicate that in both samples, the dimensions of subcutaneous fat showed signiﬁcant dispersion dimorphism, with greater variability in men. Although variations of the phenomenon during the life-course may be hypothesized, a preliminary analysis of a larger database including individuals aged 6 –75 years from both countries showed that the trend of dispersion dimorphism was similar in all age classes (Marini, Rebato, Buffa, Salces, Floris, and Borgognini Tarli, unpublished results). As already discussed in the case of variability proﬁles, the different degree of dispersion observed in the two sexes could be related to the environmental component, to the genetic component, and/or to a combination between observer and instrumental error. However, we think it reasonable to exclude the last factor, because it is unlikely that instrument or observer error would be systematically different between the sexes. The environmental component could decrease female phenotype variability in case of a reduced sensitivity of the character to macroenvironmental factors (environmental canalization; Wagner et al., 1997), or in case of lower environmental variation for women. This last circumstance could be due to cultural factors, in cases of a more homogenous female lifestyle. In this connection, it was hypothesized that women are more concerned in coping with body mass and fatness than men (Dornbusch et al., 1984; Rocandio et al., 2003). Sociocultural factors might also affect intrasexual variability by inﬂuencing mate choice. As to the genetic component, a lower variability in females could be due to greater intensity of stabilizing selection, a mechanism that acts predominantly on quantitative traits (Johnson, 1976; Phillips and Arnold, 1989; Via and Lande, 1985), or to epistatic interactions among genes (canalizing selection; Waddington, 1957). Traits related to fertility can also be subjected to the action of sexual selection (Darwin, 1871), and the combination of the two types of selection would again reduce variability (Via and Lande, 1985). In fact, Gaulin and Sailer (1985) showed that female variability in body weight is lower in cases of sexual selection, asymmetric parental investment, and nutritional constraints. Sexual selection, with male choice preferentially directed toward women with particular amounts and distributions of subcutaneous fat (Morris, 1967), which is an important determinant of body shape, might therefore also reduce female variability. From a theoretical standpoint, each of these potential correlates of dispersion dimorphism might be inﬂuential. At present it is difﬁcult to select the most effective one(s). The measurements of soft tissues showing dispersion dimorphism are more phenotypically ﬂexible and ecosensitive, and therefore more susceptible to the inﬂuence of cultural factors. However, they are also very relevant in reproduction and lactation, and for this reason they might be subjected to environmental canalization and maybe stabilizing or canalizing selection. The hypothesis that females express greater canalization in growth and body composition traits than do males was previously advanced (Hamilton, 1982; Stini, 1972, 1982; Stinson, 1985). It predicts that women are less variable in traits related to reproduction. Experimental evidence supporting the above theory is based on comparisons of the degree of sexual dimorphism in populations experiencing different levels of environmental stress (Hamilton, 1975; Stini, 1969, 1972; Stinson, 1985; Pucciarelli et al., 1993). In this case, weaker male canalization would cause a greater effect on male average dimensions, while females would remain relatively stable. Other less direct information comes from comparisons of sex-speciﬁc morbility and mortality levels, and from studies on sex differences in secular trends, or in growth under conditions of environmental stress. Of particular interest are the results obtained in Guatemalan children, showing greater environmental inﬂuence on boys (Bogin and McVean, 1982) and on the adipose tissue (Bogin and McVean, 1981). The adipose tissue is an active endocrine organ that inﬂuences many aspects of human metabolism and contributes to adaptation to physiological and environmental challenges (Mohamed-Ali et al., 1998). Humans are unique among primates in having major sex-speciﬁc fat stores and patterns of fat distribution (Norgan, 1997; Pond, 1997). Humans also show sex differences in the mechanisms that regulate nutrient utilization and energy homeostasis, showing a stronger propensity toward the retention of fat mass during times of energy surfeit in females than in males (Cortright and Koves, 2000). Adipose reserves guarantee women the energetic surplus required during pregnancy and lactation, and are crucial in case of starvation (McFarland, 1997; Pond, 1997). In this connection, in morphologically complex organisms such as humans, the power of stabilizing selection may be strong enough to include environmental canalization in characters with protective effects for early developmental stages (Wagner et al., 1997). CONCLUSIONS The literature on dispersion dimorphism (Cope and Lacy, 1992; Ipiña and Durand, 2000; Kelley and Plavcan, 1998; LaVelle, 1995; Leutenegger and Cheverud, 1982, 1985; Leutenegger and Larson, 1985; Marini et al., 1999; Meindl et al., 1985; Oxnard, 1987; Plavcan, 2000, 2001; Plavcan and Cope, 2001; Plavcan and Kay, 1988; Tague, 1989, 1991, 1992, 1995; Wood, 1976), directed toward the study of skeletal and dental traits in nonhuman primates and in humans, is characterized by discor- DISPERSION DIMORPHISM IN HUMANS dant results sometimes obtained by inadequate statistical methods. The present paper tackles this problem using a range of metric traits sampled in two different human populations: one from the Basque Country and one from Sardinia. The results were very similar in the two samples examined. In agreement with results in the literature (Plavcan and Kay, 1988; Plavcan, 2000; Tague, 1989, 1995), dispersion dimorphism was absent in some anthropometric variable categories (such as breadths and circumferences) that are substantially related to the skeletal and muscular compartments. On the other hand, dispersion dimorphism was signiﬁcantly present, with greater male variability, in the case of skinfolds. We ﬁrst excluded instrumental and/or observer error as a cause. The possible inﬂuential factors (environmental canalization, sociocultural pressure, stabilizing selection, and sexual selection) were discussed. The adipose body compartment would be especially subject to environmental canalization and stabilizing selection because of its role in pregnancy and lactation. The role of cultural factors that affect most human behavioral traits might be relevant in reducing female environmental variation by means of direct effects and by indirect inﬂuence on male choice. ACKNOWLEDGMENTS We thank the anonymous referees for their comments and observations, which contributed to substantial improvement of this paper. LITERATURE CITED Albrecht GH, Miller JMA. 1997. Variability proﬁles in fossil studies: the case of Homo habilis. Am J Phys Anthropol [Suppl] 24:65. Bogin B, McVean RB. 1981. Body composition and nutritional status of urban Guatemalan children of high and low socioeconomic status. Am J Phys Anthropol 55:543–551. Bogin B, McVean RB. 1982. Ethnic and secular inﬂuences on the size and maturity of seven year old children living in Guatemala City. Am J Phys Anthropol 59:393–398. Cabras S, Mostallino G, Racugno W. 2004. A non-parametric bootstrap test for the equality of coefﬁcients of variation. Tech. Report, Dipartimento di Matematica, Universitá di Cagliari 5:1–12. Cope DA, Lacy MG. 1992. Falsiﬁcation of a single species hypothesis using the coefﬁcient of variation: a simulation approach. Am J Phys Anthropol 89:359 –387. Cortright RN, Koves TR. 2000. Sex differences in substrate metabolism and energy homeostasis. Can J Appl Physiol 25:288 – 311. Darwin C. 1871. The descent of man and selection in relation to sex. London: Murray. Dornbusch SM, Carlsmith JM, Duncan PD, Gross RT, Martin JA, Ritter PL, Siegal-Gorelick B. 1984. Sexual maturation, social class and the desire to be thin among adolescent females. J Dev Behav Pediatr 5:308 –314. Gaulin SJC, Sailer LD. 1985. Are females the ecological sex? Am Anthropol 87:111–119. Hamilton ME. 1975. Variation among ﬁve groups of Amerindians in the magnitude of sexual dimorphism of skeletal size. Ph.D. dissertation, Ann Arbor, University of Michigan. Hamilton ME. 1982. Sexual dimorphism in skeletal samples. In: Hall RL, editor. Sexual dimorphism in Homo sapiens. New York: Praeger. p 107–163. 349 Ipiña SL, Durand AI. 2000. A measure of sexual dimorphism in populations which are univariate normal mixtures. Bull Math Biol 62:925–941. Johnson C. 1976. Introduction to natural selection. Baltimore: University Park Press. Kelley J, Plavcan JM. 1998. A simulation test of hominoid species number at Lufeng, China: implications for the use of the coefﬁcient of variation in paleotaxonomy. J Hum Evol 35:577–596. LaVelle M. 1995. Natural selection and developmental variation in the human pelvis. Am J Phys Anthropol 98:59 –72. Leutenegger W, Cheverud J. 1982. Correlates of sexual dimorphism in primates: ecological and size variables. Int J Primatol 3:387– 402. Leutenegger W, Cheverud J. 1985. Sexual dimorphism in primates: the effect of size. In: Jungers WL, editor. Size and scaling in primate biology. New York: Plenum Press. p 33–50. Leutenegger W, Larson S. 1985. Sexual dimorphism in the postcranial skeleton of New World primates. Folia Primatol (Basel) 44:82–95. Lewontin RC. 1966. On the measurements of relative variability. Syst Zool 15:141–142. Lilliefors HW. 1967. On the Kolmogorov-Smirnov test for normality with mean and variance unknown. J Am Stat Assoc 62:399 – 402. Lohman TG, Roche AF, Martorell R. 1988. Anthropometric standardization reference manual. Champaign, IL: Human Kinetics. Marini E, Racugno W, Borgognini Tarli SM. 1999. Univariate estimates of sexual dimorphism: the effects of intrasexual variability. Am J Phys Anthropol 109:501–508. Marks GC, Habicht JP, Mueller WH. 1989. Reliability, dependability, and precision of anthropometric measurements. The Second National Health and Nutrition Examination Survey 1976 –1980. Am J Epidemiol 130:578 –587. McFarland R. 1997. Female primates: fat or ﬁt? In: Morbeck ME, Galloway A, Zihlman AL, editors. The evolving female. A life history perspective. Princeton: Princeton University Press. p 163–175. Meindl RS, Lovejoy CO, Mensforth RP, Don Carlos L. 1985. Accuracy and direction of error in the sexing of the skeleton: implications for paleodemography. Am J Phys Anthropol 68: 79 – 85. Mohamed-Ali V, Pinkney JH, Coppack SW. 1998. Adipose tissue as an endocrine and paracrine organ. Int J Obes Relat Metab Disord 22:1145–1158. Morris D. 1967. The naked ape. London: Jonathan Cape. Mueller WH. 1977. Sibling correlations in growth and adult morphology in a rural Colombian population. Ann Hum Biol 4:133– 142. Norgan NG. 1997. The beneﬁcial effects of body fat and adipose tissue in humans. Int J Obes Relat Metab Disord 21:738 –746. Oxnard CE. 1987. New perspectives on human evolution. Seattle: University of Washington Press. Phillips PC, Arnold SJ. 1989. Visualizing multivariate selection. Evolution 43:1209 –1222. Plavcan JM. 2000. Variance dimorphism in primates. Am J Phys Anthropol [Suppl] 30:251. Plavcan JM. 2001. Sexual dimorphism in primate evolution. Yrbk Phys Anthropol 44:25–53. Plavcan JM, Cope DA. 2001. Metric variation and species recognition in the fossil record. Evol Anthropol 10:204 –222. Plavcan JM, Kay RF. 1988. Sexual dimorphism and dental variability in platyrrhine primates. Int J Primatol 9:169 –178. Pond CM. 1997. The biological origins of adipose tissue in humans. In: Morbeck ME, Galloway A, Zihlman AL, editors. The evolving female. A life history perspective. Princeton: Princeton University Press. p 147–162. Pucciarelli HM, Carnese FR, Pinotti LV, Guimarey LM, Goicoechea AS. 1993. Sexual dimorphism in schoolchidren of the Villa IAPI neighborhood (Quilmes, Buenos Aires, Argentina). Am J Phys Anthropol 92:165–172. Rocandio AM, Arroyo M, Ansotegui L, Herrera H, Salces I, Rebato E. 2003. Objective and subjective body mass index in a group of university students of Spain. 12th European Congress on 350 E. MARINI ET AL. Obesity. Helsinski, Finland, 29 May–1 June. Int J Obes Relat Metab Disord [Suppl] 27:4 –92. Sachs L. 1984. Applied statistics (second edition). New York: Springer-Verlag. Salces I. 2002. Determinantes genéticos y ambientales del crecimiento y composición corporal en la provincia de Vizcaya. Investigación sobre la transmisión familiar y variaciones con la edad. Ph.D. dissertation, Universidad del Paı́s Vasco-Euskal Herriko Unibertsitatea, Bilbao, Spain. Schultz AH. 1926. Studies on the variability of platyrrhine monkeys. J Mammal 7:286 –305. Sokal RR, Braumann CA. 1980. Signiﬁcance tests for coefﬁcient of variation and variability proﬁles. Syst Zool 29:50 – 63. Stini WA. 1969. Nutritional stress and growth: sex difference in adaptive response. Am J Phys Anthropol 31:417– 426. Stini WA. 1972. Reduced sexual dimorphism in upper arm muscle circumference associated with protein-deﬁcient diet in a South American population. Am J Phys Anthropol 36:341–352. Stini WA. 1982. Sexual dimorphism and nutrient reserves. In: Hall RL, editor. Sexual dimorphism in Homo sapiens. New York: Praeger. p 391– 419. Stinson S. 1985. Sex differences in environmental sensitivity during growth and development. Yrbk Phys Anthropol 28:123–147. Susanne C. 1975. Genetic and environmental inﬂuences on morphological characteristics. Ann Hum Biol 2:279 –287. Tague RG. 1989. Variation in pelvic size between males and females. Am J Phys Anthropol 80:59 –71. Tague RG. 1991. Commonalities in dimorphism and variability in the anthropoid pelvis, with implications for the fossil record. J Hum Evol 21:153–176. Tague RG. 1992. Sexual dimorphism in the human bony pelvis, with a consideration of the Neandertal pelvis from Kebara Cave, Israel. Am J Phys Anthropol 88:1–21. Tague RG. 1995. Variation in pelvic size between males and females in nonhuman anthropoids. Am J Phys Anthropol 97: 213–233. Ulijaszek SJ, Johnston FE, Preece MA. 1998. The Cambridge encyclopedia of human growth and development. Cambridge: Cambridge University Press. Via S, Lande R. 1985. Genotype-environment interaction and the evolution of phenotypic plasticity. Evolution 39:505– 522. Waddington CH. 1942. Canalization of development and the inheritance of acquired characters. Nature 14:563–565. Waddington CH. 1957. The strategy of the genes. New York: MacMillan Co. Wagner GP, Booth G, Bagheri-Chaichian H. 1997. A population genetic theory of canalization. Evolution 51:329 – 347. Weiner JS, Lourie JA. 1981. Practical human biology. London: Academic Press. Wood BA. 1976. The nature and basis of sexual dimorphism in the primate skeleton. J Zool Lond 180:15–34.