Reply to “A Preregistered Study of Competing Predictions Suggests That Men Do Overestimate Women’s Sexual Intent”

In our original work (Perilloux & Kurzban, 2015) critiqued by Murray, Murphy, von Hippel, Trivers, and Haselton (2017), our goal was to address an epistemic challenge posed by the finding that men report women’s sexual interest to be greater than women report their own interest to be (Haselton & Buss, 2000). From the reported sex difference alone, one cannot infer whether men overestimated, women underreported, or some combination of the two. Distinguishing among these possibilities is difficult because the truth of the matter—women’s actual sexual interest—is unknown.

We were pleased to work with Murray et al. in their development of a simple but clever experiment to clarify a portion of our prior work (Perilloux & Kurzban, 2015). As did Murray et al. in their new experiment, we relied on Gricean pragmatics to get at the issue in our original experiment. By allowing participants to answer both “say” and “want” questions, we signaled that we expected participants might answer them differently, licensing them to do so.

In both our work and that of Murray et al., this indeed occurred. Just as we found that women believe that women have greater sexual intentions than they say they do, so did Murray et al. We are persuaded that, contrary to what we predicted, the first question asked “anchors” responses: When the “say” question is asked first, ratings for the “want” question increase; but when the “want” question is asked first, ratings for the “say” question decrease. We take from this that women do not have specific values in mind for “say” and “want” questions. That is, women have beliefs about the ordinal relationship between actual and reported sexual intentions, but not about the cardinal values for those intentions. This conclusion, although contrary to our original view, is in line with the reasoning of Murray et al.

In this reply, we wish to make three points. First, although the results Murray et al. obtained undermine our original interpretation of our result, it bears emphasizing that the findings, taken together, still cannot answer the initial research question: Is the difference between men’s and women’s answers on these scales caused by men’s overestimation, women’s underreporting, or both? On this note, our understanding of the Gricean interpretation Murray et al. advance is that participants will “differentiate” between the two questions (p. 253). Differentiation does not imply a prediction regarding direction; we think it significant that asking the “want” question first did not, as it could in principle have done, lead to increased ratings for the “say” question. Clearly, our initial research question remains open, and we look forward to Murray et al. and other scholars devoting their methodological creativity to designing methods that put both the underreporting hypothesis and the overperception hypothesis at risk of falsification.

Second, there are data in the Supplemental Material provided by Murray et al. that speak, in a modest way to be sure, to this question. As Figure S2A shows, most women reported that they themselves do in fact at least sometimes act less interested than they really are when beginning to date someone new. Additionally, most women reported that they themselves never or rarely acted more interested than they actually were (Fig. S2B). These data appear to be consistent with the view that women, on average, both say and act as though they are less interested than they actually are; this leaves open the possibility that men pick up on this “coyness” and that men’s ratings therefore reflect women’s true intentions more accurately than women’s self-reported interest does.

Third, and finally, we were gratified to see that in their closing paragraph addressing the question of why biases in judgment and belief might exist, Murray et al. focus exclusively on people’s effects on other people (in agreement with previous scholars: Kurzban, 2011; Perilloux, 2014; Trivers, 2000). This is a welcome shift from prior putative explanations; Haselton and Buss (2000), for instance, suggested that men overperceive women’s sexual interest not to facilitate persuasion but rather to avoid missed sexual opportunities (p. 82). Our own view, echoed in Murray et al.’s focus in their closing, is that although evolution might well have selected for false-belief-generating systems in the service of persuasion (Kurzban & Aktipis, 2007; Trivers, 2000), it is unlikely to have done so in the service of other functions, which are better served by true-belief-generating systems (Kurzban, 2011; Perilloux, 2014). Our view corresponds to that of Pinker (2011), who argued that biased beliefs, such as overperceiving a woman’s interest, “are adaptive only when a deception-detecting audience is in the loop, not when an inaccurate representation is invoked as an internal motivator” (p. 36). The fact that Murray et al. highlight persuasion as the single explanation for false-belief-generating mechanisms brings our respective views closer together.

We disagree, then, with Murray et al. when they claim that data and theory are inconsistent with our view. The kinds of perceptual phenomena to which they refer at best are subject to debate (Firestone & Scholl, in press) and at worst reflect methodological errors, such as unaccounted-for demand characteristics (e.g., Durgin et al., 2009) or theoretical confusion (Firestone, 2013). For example, Murray et al. refer to Proffitt, Bhalla, Gossweiler, and Midgett’s (1995) study to support the claim that people “judge a hill as steeper when they are standing at the top rather than at the bottom” (p. 254). However, as Li and Durgin (2009) made clear, “the categorical comparison of ‘top’ and ‘bottom’ viewing is unwarranted” (p. 1). The perception of steepness depends on direction of gaze and optical slant (which varies with distance from the edge). These dependencies are easily accommodated by principles of perception and seem designed to increase sensitivity to steepness by harmlessly sacrificing some perceptual constancy for some sloped surfaces (Li & Durgin, 2009). Therefore, no recourse to perceptions of danger is needed to explain the effects reported by Proffitt et al. (1995). Indeed, Firestone and Scholl (in press) make an argument about visual perception that closely parallels our own view, suggesting that it is “better for vision to report the facts as honestly as possible and let the other systems relying on it (e.g., action, navigation, social evaluation, decision-making) use that information as they see fit” (section R2.4.1).

In closing, we applaud Murray et al. for clarifying our finding; their Commentary has persuaded us that before being asked what other women say their sexual intentions are and what other women’s sexual intentions actually are, women do not have a cardinal representation of these variables in mind. We look forward to future work that can arbitrate among the competing hypotheses that men overestimate women’s sexual intentions, women underreport their sexual intentions, or some combination of the two.