Main

The Matisgenka marriage system is matrilocal, meaning that a man leaves his natal home to marry into the woman's family. As a result, families lose their economic investment in their sons, not daughters, and so sons are more costly than daughters. Also, elder sons traditionally marry polygamously, meaning that the expected reproductive success of younger sons is lower than that of any daughter, and thus, if anything, there should be a preference for female offspring. As it turns out, sex ratios seem to be female-biased. A genealogical survey of an even more culturally isolated group of Matsigenka, the Sotileja population, recorded a birth ratio of 73 females to 62 males, although this female bias was later reduced by higher female mortality2.

Nor does the model of Manning et al. explain why westernized indigenous populations in South America should prefer hourglass-shaped females. First, what traditional society does not value strength (or sons)? Second, Peru's male-dominated economy should increase, not decrease, the value of males in westernized populations, and, by extension, the value of high-WHR females. Finally, how could such preferences change evolutionarily in a single generation?

Tovée and Cornelissensuggest that WHR is a proxy measurement for BMI. In Yomybato village, under all three decision criteria, figures were ranked O9, O7, N9, N7, U9, U7 (sorted by median then mean ranks; notation: O, N, U are overweight, normal and underweight, respectively, and numbers refer to high (9) and low (7) WHRs), which is consistent with the hypothesis that figures were ordered by BMI alone. However, Shipetiari males ranked the figures O7, O9, N7, N9, U7, U9 under the attractiveness and spouse criteria, which might not be consistent with the BMI hypothesis, because the ranks do not vary smoothly with BMI. Moreover, Alto Madre and United States males gener-ally grouped figures first by WHR (7, 9) and then by weight, which is inconsistent with the BMI hypothesis. Also, although both the latter populations generally ranked N7 first, O7 was the second choice of Alto Madre males, and U7 was that of United States males. On the other hand, by using a different set of (headless) figures, Tovée et al. found that British university students, in contrast, strongly disliked females with low or high values of BMI but were agnostic with regard to WHR3. Finally, a recent study has revealed that the culturally isolated hunter-gatherer Hadza men in Tanzania are also indifferent to WHR but prefer pictures of high-BMI women (A. Wetsman and F. Marlowe, personal communication).

What can we make of these cross-cultural differences? First, not only might WHR preferences be affected by westernization, but the very concept of using WHR at all to assess attractiveness might be culturally variable and, in Peru, an effect of westernization. Second, these results show that BMI preferences vary markedly across cultures (as well as through time4). If we were to rely on the weak cultural invariance test of adaptation, BMI and WHR preferences would not be considered as adaptations.

We obviously need a more sophisticated evolutionary analysis to explain variation in preferred body shapes and sizes, which does not ignore the effects of environment on ontogeny. For now, the nature of female beauty must remain mysterious, as perhaps it should.

See also Manning et al. and Tovée et al.

Erratum Colour was inappropriately added by Nature to Fig. 1 of Yu and Shepard's original paper5, which should have appeared in black and white. We apologize for any confusion that may have arisen as a result.