Does women’s body attractiveness predict indices of reproductive capacity? Prior research has provided evidence that large breast size and low waist-to-hip ratio (WHR) are positively associated with women’s estrogen and progesterone concentrations, but no previous studies appear to have directly tested whether ratings of women's body attractiveness are predicted by higher concentrations of ovarian hormones measured across broad regions of the menstrual cycle. Here, we collected daily saliva samples across 1-2 menstrual cycles from a sample of young women; assayed the samples for estradiol, progesterone, and testosterone; obtained anthropometric measurements of the women’s bodies; and also obtained attractiveness ratings of the women’s bodies from photographs of them taken in standardized clothing with faces obscured. Contrary to previous research, mean hormone concentrations were uncorrelated with breast size and WHR. Body mass index (BMI) was a very strong negative predictor of body attractiveness ratings, similar to previous findings. Zero-order associations between women’s mean hormone concentrations and mean attractiveness ratings were not significant; however, after controlling for BMI, attractiveness ratings were independently and positively associated with both estradiol and testosterone concentrations. Discussion focuses on the implications of these findings for whether attractiveness assessment mechanisms are specialized for the detection of cues of differential fecundity in young women’s bodies.

The extant literature on human mate preferences demonstrates that mate preferences are difficult to measure: findings are surprisingly contradictory across methodologies. In self-report and vignette studies, traits related to cooperative partner or parental value, such as warmth, kindness, and trustworthiness, are found to be of paramount importance. These methods also tend to find the sex differences in preferences that are predicted by evolutionary theory, with men valuing physical attractiveness relatively more than women do, and women valuing social status or cues of resource acquisition ability relatively more than men do. However, methods like these where participants must consciously evaluate different traits and report their preferences for them are problematic, because of issues like social desirability bias, demand characteristics, and the fact that humans probably have little conscious insight into the majority of their cognitive programs.

In research measuring revealed preferences, such as speed-dating studies, in which choices have real consequences for mating, cues of cooperative partner or parental value do not appear to influence attractiveness, and physical attractiveness is the main driver of mating choices, for both sexes. However, in speed-dating interactions, information about other relevant traits may be largely absent, and physical attractiveness may be the only trait that can be accurately assessed during such a brief encounter. The correlational nature of speed-dating studies also precludes the ability to make causal claims. Thus, the validity of both self-reported preferences and revealed preferences from speed-dating research are questionable, and this literature is currently at an impasse.

Here, we attempted to solve many of the issues in these previous methodologies with a novel experimental paradigm for assessing revealed preferences. We resurrected the memory confusion protocol developed by Taylor et al. (1978) and adapted it for the investigation of human mate preferences. This method afforded us experimental control, a strong but implicit manipulation of traits via behavioral information, and a built-in manipulation check by virtue of the memory confusion data that are emblematic of this protocol. Across a series of seven studies, we manipulated cues of kindness and social status or provisioning ability and tested effects on mate attractiveness. Opposite-sex targets represented by faces, who varied in physical attractiveness, were randomly paired with sentences describing behaviors that were pre-rated as indicating various levels of the manipulated traits. After viewing these pairings, subjects rated targets’ mate attractiveness, and revealed preferences were assessed from these ratings.

Results for kindness manipulations indicated that manipulated kindness positively predicted mate attractiveness ratings for both sexes. Unkind behaviors towards others were detrimental to the mate attractiveness of both sexes, and highly kind behaviors towards others had particularly positive effects on the mate attractiveness of female targets. Trends suggested that women especially valued high kindness by men when it was directed towards them compared to others.

Findings from the first resource access study suggested that cues of social standing alone did not impact the mate attractiveness of either sex. However, behavioral information about provisioning ability had significant effects on the mate attractiveness of both sexes, with no clear sex difference in this effect. Physical attractiveness was consistently the strongest predictor of mate attractiveness, for both sexes.

These studies suggest that the conclusions from neither self-report nor speed-dating methodologies are wholly accurate: cues of cooperative partner or parental value are indeed important components of mate attractiveness, contrary to the null findings for these qualities in speed-dating studies. Cues of provisioning ability may be similarly consequential to mate attractiveness judgments. However, self-report studies may have underestimated the importance of physical attractiveness in the initial stages of mate selection. The present studies demonstrate the utility of this novel experimental method for the assessment of human mate preferences.