2.1. Samples

2.2. Measures

3.1. Relationship between resource scarcity and the valuation of fatness in women

In Sample 1, the results were contrary to Hypothesis 1 and to those of Anderson et al. (1992). Correlations between the valuation of fat and each resource scarcity measure were negative, and one was significantly so: for threat of natural disasters, ρ=−.51 (two tailed P<.05, n=20); threat of famine, ρ=−.41 (P<.10, n=17); and chronic scarcity, ρ=−.36 (P<.20, n=14). Henceforth, we focus on the threat of natural disasters as our best measure of resource scarcity because it was both most strongly related to the valuation of fatness in women and most frequently scoreable.

In Sample 2, we included only societies that Anderson et al. (1992) had also rated. The result matched neither theirs nor our Sample 1 result: Threat of natural disasters was not correlated with the valuation of fat at all (ρ=.07, n=17).

One reason why our results differ from those of Anderson et al. (1992) could be the difference in the dependent variables. Following Brown and Konner (1987), Anderson et al. (p. 208) coded ideals of female bodily “beauty” as (1) plump or fat, (2) not too fat or too thin, or (3) slim or slender, whereas we sought to assess whether “fatness” or “thinness” is valued in the society, for whatever reason: beauty, sexuality, health, or reproductive considerations. We assessed the relationship between these two measures on all cases that were included in both studies and found that ρ=−.62 (P<.001); the coefficient is negative because the valuation of fat is the high end of our scale. Thus, although the two dependent variables are not perfectly related, they tap something in common, and the opposing results of the two studies remain to be explained.

Another possible source of differences is the definitions of resource scarcity. Anderson et al. (1992, p. 208) created a four-point “nutritional stress” scale (see our Appendix) from the “Preservation and Storage of Food” codes of Murdock and Morrow (1970), which are meant to reflect food storage adequacy within the context of resource fluctuation, whereas our measures of resource scarcity concern adequacy of food supplies, not storage. Unlike the dependent variables, the two studies' independent variables are surprisingly unrelated: the rhos between the nutritional stress scale of Anderson et al. and our resource scarcity measures are −.02 (n=37), −.02 (n=34), and −.13 (n=31) for threat of natural disasters, threat of famine, and chronic scarcity, respectively. We infer that the measure of Anderson et al. is not, in fact, a measure of resource scarcity, and we further checked this inference by correlating it with codings from Dirks (1993): none of the seven measures of severity of famine and food shortage of Dirks was significantly related to the Anderson et al. nutritional stress measure at P<.05 (one tailed). Instead, the Anderson et al. (1992) measure appears to tap food storage (see below), with which it is highly correlated (ρ=.56, P<.001, one tail).

Then why did not both of our samples show a relationship between valuing fatness and threat of disasters? We think that another variable may be modulating the relationship–degree of food storage. We can measure that variable by using “food surplus via storage” (Variable 21 in World Cultures, 2001, Vol. 12, No. 2,), which is recoded from the same Murdock and Morrow scale 10 used by Anderson et al. (1992). The recoding, which emphasizes storage, is (1) none or barely adequate; (2) simple or adequate; or (3) complex or more than adequate. (We changed the coding for one culture, the Ganda, whom we recoded as a 1 because the asterisked remark in Murdock and Morrow, (1970), indicates no appreciable storage.)

Food storage is not significantly related either to the valuation of fatness in women or to any of our three measures of resource scarcity, but it does modulate their relationships to one another. In societies with little or no storage (score 1), high threat of disasters significantly predicts low valuation of fatness (in our combined samples, ρ=−.56, P=.02, two tails, N=17), as we found in Sample 1, but in societies with some storage (scores 2–3), threat of disasters has no relationship to the valuation of fatness in women.

3.2. Other possible predictors of the valuation of fatness

Resource scarcity did not predict the valuation of fatness in the presence of storage. What does? Two factors that Anderson et al. suggest may be relevant are climate and male dominance. It is plausible that extra fat is favored in cold climates for the same reasons that cold selects for shorter limbs and stocky bodies, and Anderson et al. (1992) found some support for cold's relevance, that is, a significant relationship between higher latitude and the idea that fat is beautiful. We looked at two measures that should tap colder climate: mean annual temperature (from Whiting, 2001: Variable 186) and absolute latitude [Variable 833.1 in World Cultures, Vol. 12 (2)].

Are societies in colder climates more likely to value fatness in women? We find no evidence in support of this claim. As Table 1 shows, whether food storage is absent, present, or ignored (compare the three columns), we find no significant relationships between valuation of fat and either latitude or mean annual temperature. (A reviewer suggested that inconsistent results on this point may again reflect the effect of an unmeasured modulating variable, namely, the adequacy of clothing/shelter. This is a possibility for future research.)

Table 1. Other possible predictors of the valuation of fat in women proposed by Anderson et al. (1992), controlling on food storage

		Storage absent or minimal (Score 1)	Storage more than minimal (Scores 2–3)	Ignoring storage
	Climate
	Latitude	.03 (n=23)	−.14 (n=24)	−.06 (n=47)
	Mean annual temperature	.02 (n=23)	.05 (n=24)	.05 (n=47)
	Male dominance
	View that men dominate	−.33 (n=8)	−.45 (n=8)	−.38† (n=18)
	High value on males being aggressive, strong, and sexually potent	−.81⁎⁎⁎ (n=9)	−.61⁎ (n=9)	−.58⁎⁎(n=16)

Correlations shown are Spearman rhos.

⁎ P<.05, one tail. ⁎⁎ P<.05, two tails. ⁎⁎⁎ P<.01, two tails. † P<.10, one tail.

Anderson et al. (1992) also claim support for the idea that male dominance is relevant: They found that two variables in the cross-cultural study of the status of women of Whyte (1978) significantly predict the valuation of fatness in women, namely, “an explicit view that men should and do dominate their wives” [Variable 621 in World Cultures, Vol. 12 (2)] and “a high value on males being aggressive, strong and sexually potent" [Variable 625 in World Cultures, Vol. 12 (2)]. We retested the conclusion of Anderson et al. with respect to both variables.

Because both “male dominance’ variables are coded such that higher scores indicate less of the trait in question, the conclusions of Anderson et al. predict negative correlations with the valuation of fatness in women, and indeed, all the correlations are negative (Table 1). Whyte only coded half of the SCCS sample cases, hence, the ns with controls on storage are small. Even if we look at the relationships without controlling on storage (see the bottom cell in Column 3), an effect of Variable 625 is significantly supported. An effect of Variable 621 is not, but the results are in the same direction.

4.1. Food storage and the valuation of fatness in women

How do we interpret the fact that fat is not always valued? Could thinness be adaptive when there is little food storage and a high threat of disasters? A controversial “small but healthy” hypothesis (Seckler, 1980, Seckler, 1982, as cited by Pelto & Pelto, 1989) suggests that adjusting growth to low food availability in childhood may be adaptive in situations with frequent resource scarcity, but this hypothesis does not suggest that being thin would be adaptive, only that smaller stature might sometimes be fitter if a child can maintain normal weight for height. Most scholars reject the notion that stunting can be construed as healthy, citing many medical and social reasons why stunted children would be disadvantaged (Beaton, 1989, Martorell, 1989, Messer, 1989, Pelto & Pelto, 1989, Scrimshaw & Young, 1989). In individual selection terms, someone with stored body fat would surely be more likely to survive a serious food disaster than would a very thin person. However, if selection between groups were relevant, it may be a different matter. For example, Stini (1971, p. 1027; cf. Beaton, 1989, p. 36; Thomas, 1973, as cited by Messer, 1989) suggests that more individuals could survive on available resources with proportional reductions of body size because each individual would then have lower nutritional requirements. Given the relationship between body size and the amount of food needed to maintain that size (Beaton, 1989, p. 36; cf. Gaulin & Sailer, 1985, p. 112), a group selection argument could be applicable to thinness. Collectively, thinner individuals need fewer calories, thus, if one cannot “save up” enough to cover unpredictable scarcity (measured here by threat of disasters), it may be more adaptive to require less food. Dirks (2004) points to cases in the ethnographic record that have cultural eating patterns consistent with the society preparing for famine; it seems as if the social group anticipates the need to control eating. For example, the Cagaba of northern Colombia glorify fasting (Reichel-Dolmatoff, 1951, as cited by Dirks, 2004). And the Gurage of Ethiopia take very light meals, behaving as if food were scarce (Shack, 1971, as cited by Dirks, 2004). Paque (1980) suggests that religious fasting (e.g., Ramadan) may have served as a buffer against famine and lack of water in desert climates. In the face of unpredictable famine, it also might be better for the group to lower rates of reproduction, at least while resources are poor. A pregnant and lactating woman has high caloric needs, and if women are encouraged to be thin, they are less likely to reproduce as quickly (Frisch, 1980).

Why would food storage change the relationship between the valuation of fatness in women and the threat of natural disasters? Food storage provides an extrasomatic way to protect against unpredictable resource scarcity. Many societies have “famine” food that they tap only if necessary, and where there are cultural means of mitigating famine, it may be less important for the group to save resources via small bodies or thinness, and it may be advantageous (for both groups and individuals) to have some plumpness. Calories can be stored in two ways—on and off the body. General thinness could also protect people from unpredictable resource scarcity, especially in the absence of storage.

4.2. Male dominance or protest masculinity?

Although “a high value on males being aggressive, strong, and sexually potent” (ASP henceforth) is strongly related to valuation of fatness in women, we question whether this variable measures male dominance. This variable is ideological and may or may not reflect actual male dominance. The codes come from a larger study of the status of women by Whyte (1978, pp. 96–100), who concludes that there is no unitary phenomenon of status; at best, there are nine weak clusters of status variables. What is important to note here is that ASP did not fall into any cluster suggesting actual male dominance (e.g., property control, power in kinship contexts, domestic authority, or control of women's marital and sexual lives), but into a cluster that Whyte calls “ritualized fear of women,” apparently because the cluster also includes “belief that sexual activity is dangerous.” In sum, the fact that ASP is not correlated with various aspects of male dominance suggests to us that ASP reflects something else, namely, what has been called “machismo” or “protest masculinity.”

In the literature of psychological anthropology, machismo or protest masculinity is viewed by some as stemming from conflict and insecurity in males about their masculinity. One model of the conditions favoring protest masculinity and its consequences, referred to as “status-envy theory,” was proposed by the Whitings and their colleagues (Burton & Whiting, 1961, Whiting, 1960, Whiting, 1965), who suggested that the roots of protest masculinity begin early in life, with father absence or low father salience. If the boy perceives that the mother controls most resources, he will develop an early, unconscious feminine identification. Later, if the boy learns that men actually control most resources, he will develop a secondary masculine identification. Conflict between the early and later identifications should be greatest with low father salience early in life and actual high male dominance. To resolve their sex–identity conflict, boys will try hard to show how masculine they are by overcompensating and exaggerating what is viewed as masculine behavior. Whiting (1965) suggests that males exhibiting protest masculinity will also show antagonism and even fear of women and may avoid the company of women as much as possible. Such behaviors are assumed to be defensive responses to the unconscious fear of being feminine.

Broude and Greene (1983) rated husband–wife activities to measure aloofness versus intimacy. While some husband–wife activities such as sleeping arrangements may be governed by other considerations, such as men's houses and men having to work far from home, two activities that seem particularly reflective of psychodynamics are whether husband and wife eat together and whether they spend leisure time together. We correlated measures of these traits [World Cultures (2001), Vol. 12 (2), Variables 752 and 753; originally from Broude & Greene, 1983] with ASP and found that where ASP is high, men eat separately from women (ρ=−.35, P<.01, one tail, n=52) and spend their leisure time separately (ρ=−.42, P<.001, one tail).

Does ASP reflect sexual potency, or just the protestation thereof? If it reflects interest in a great deal of sex, it should be correlated with a high desirability of sex in marriage. But it is not. [The ρ is .03; the data on desirability of sex in marriage come from World Cultures (2001), 12 (2), Variable 160; originally from Broude & Greene, 1976].

A meta-analysis by psychologists, covering 39 studies, suggests that “hostile masculinity” and “hypermasculinity” are highly correlated with sexual aggression, including rape by males against females (Murnen, Wright, & Kaluzny, 2002). A purportedly direct measure of “rape frequency” [World Cultures (2001), 12 (2), Variable 174, originally from Broude & Greene, 1976] is highly correlated with ASP: ρ=−.77 (n=17; P<.001), which is consistent with the meta-analysis. For additional tests of the hypothesis that ASP reflects protest masculinity, it would be good to get measures of the propensity of men to behave aggressively in reaction to perceived slights to their masculinity or to the perception that their wives were interested in other men. Such measures are not currently available, unfortunately. But we do find that societies with a high value on ASP are much more likely to accept violence towards members of the local community [World Cultures (2001), 12 (2), Variable 781; originally by Ross, 1983: ρ=.62, P<.001, one tail, n=22].

Even if we are correct about high ASP reflecting protest masculinity, we are still left with the need to explain why protest masculinity should predict valuing fatness in women. If males are trying hard to show how male they are or how inferior women are, they are not likely to seek wives for companionship or for a close emotional relationship. Left to their own (not society's) choice, such males might even prefer not to marry. But if they do, it will likely be for pragmatic reasons, such as having a wife to manage a household and produce children. If fatness is a proxy for fertility, then macho males might prefer fatness in women.

What we are suggesting here seems consistent with the argument of Draper and Harpending (1982) that males with protest masculinity are following a directly adaptive reproductive strategy that they label the “cad” strategy. But alternatively, we might view protest masculinity as an inadvertent consequence of other adaptive processes. In particular, we have previously presented evidence that warfare results from people's attempts to gain resources in the face of expectable but unpredictable scarcity (Ember & Ember, 1992a). When there is internal warfare (within the society or language group), which is the most frequent kind of warfare in the ethnographic record, men often marry women from enemy communities (Ember, 1978). Husbands and wives in such situations tend to have aloof relationships, as indicated by their sleeping separately (Whiting & Whiting, 1975, Ember, 1978, p. 674), which will create relatively high father absence. In addition, with high male mortality in warfare, which also may be likely, polygyny may be an adaptive response to an excess of women because it allows all women to marry and reproduce (Ember, 1974). Hence, polygyny may also increase protest masculinity. The Draper/Harpending scenario postulates that protest masculinity is directly adaptive; our scenario suggests that protest masculinity is not itself adaptive, but results from other adaptive processes. More research is needed to decide between these two scenarios. In any case, the results presented here are consistent with the idea that valuing fatness in women may be associated with protest masculinity.

4.3. Conclusions

In sum, our study does not support the hypothesis that more resource scarcity predicts valuing fatness in women. Using our own measures of the valuation of fatness or thinness in women and of resource scarcity, we obtained results opposite to those of Anderson et al. (1992) in one sample, and no relationship in another. We suggest that the dependent variables in the Anderson et al. study and ours were measuring similar attributes, but the independent variables were not: Anderson et al. were not measuring resource scarcity, but food storage, which apparently modulates the relevance of resource scarcity to fat valuation. We also failed to support Anderson et al. with respect to the relevance of cold climate, but we did replicate their result that “a high value on males being aggressive, strong, and sexually potent” (ASP) predicts valuing fatness in women. We suggest that ASP does not really reflect male dominance, as Anderson et al. assume, but rather protest masculinity. Why protest masculinity may relate to valuing fatness in women needs further evaluation.

In the United States, shifts towards valuing thinness versus fatness seem to have coincided with the rise of women's movements in the 1920s and late 1960s. Consider that Marilyn Monroe epitomized beauty in the 1950s; she was well-rounded, not thin. “Thin” became more popular when people (particularly women) began to question early marriage and staying home to be mothers of large families. They appear to have rejected male chivalry and preferred men who were sensitive and caring. (Dennis Werner, personal communication, suggested this argument.) Was machismo behavior also less acceptable at those times? We suggest that it was, particularly among intellectuals and the elite. Whether and how these changes are causally connected require further investigation.

The cross-cultural results here are more complex than expected. We have tried to clarify the possibilities, even if we have complicated them. We hope that our work will stimulate further research on the conditions under which fatness is culturally valued.