Abstract
This reply addresses the issues raised by the thoughtful commentaries on Wood, Kressel, Joshi, and Louie’s (2014) meta-analysis. We maintain that menstrual cycle influences on women’s mate preferences are obtained inconsistently in the literature and are linked to research artifacts. This pattern provides little support for the simple evolutionary psychology biology-to-behavior models that inspired this research. As illustrated by the commentaries, more promising theories of human reproduction situate biological and psychological processes within societal structures.
The five respected scientists all provided thoughtful comments on our meta-analytic review. The goal of our meta-analysis was to understand the inconsistent research findings on menstrual cycle influences on mate preferences. Even by the most generous estimate, more than a third of study findings in this literature report menstrual cycle effects trending in the direction opposing predictions (Gildersleeve, Haselton, & Fales, in press). Our analysis suggested that the few supportive effects were research artifacts. Women did not shift their preferences across the cycle in short-term relationship contexts, as predicted by evolutionary psychology theories. The few predicted findings emerged only in published reports, declined in more recent publications, and reflected researcher degrees of freedom in defining the fertile phase such that larger effects emerged in studies with less precise estimates.
Meta-analytic reviews, like any other research, require researchers to make subjective decisions (see Ferguson, 2014). Because meta-analysts explicitly report these decisions, other scientists can evaluate them for validity (Wood & Eagly, 2009). For example, the more sanguine meta-analysis aforementioned (Gildersleeve et al., in press) failed to report the null effects in unpublished studies and did not test for the decline of effects over time or for researcher degrees of freedom in fertile phase definition—despite that reanalysis of their data revealed exactly the same patterns as in our review (Wood & Carden, in press). Other recent work (Harris, Chabot, & Mickes, 2013) suggests that the published literature may be particularly prone to exploitation of researchers’ degrees of freedom. For example, even within a single group of researchers, fertility is often calculated differently across published reports, and moderators are reported inconsistently, potentially increasing Type 1 errors (Simmons, Nelson, & Simonsohn, 2011).
Our finding that menstrual cycles have little effect on behavior contributes to an established literature on women’s cycle fluctuations. As Hyde and Salk (2014) suggest, early social stereotypes and popular media depictions of women as victims of hormonal fluctuations (Chrisler & Levy, 1990) might have informed psychologists’ early theories about menstrual influences. During the 1970s, researchers tested these ideas through studies of premenstrual symptoms and mood disorders tied to cycle phase. However, self-reports of disorders proved to be influenced by research artifacts tied to women’s cultural beliefs about their cycles (e.g., Ruble, 1977). The present meta-analytic findings similarly challenge simple hormone–behavior models and highlight the need for more sophisticated theories that place hormonal processes within sociocultural contexts.
Defending Evolutionary Psychology Predictions
As Ferguson (2014) notes, empirical failures are not likely to sway scientists who already believe in a theory. Scientists evaluate data within a conceptual paradigm, and they defend their favored paradigm against empirical falsification (Kuhn, 2012). Jones’s (2014) comment is illustrative. First, he argued that evolutionary psychology predicts the obtained pattern in which mate preferences shifted primarily in studies with longer, less precise estimates of the fertile phase. He guessed that such broad definitions of the fertile phase captured low progesterone at the beginning of the cycle (forward counting). However, fertile phases of 13 to 15 days would necessarily incorporate the elevation in progesterone after ovulation (given that most studies exclude days 1–3 of menses). Thus, we make the opposite guess that, by including more days in the fertile phase, researchers failed to isolate the low progesterone days in the beginning of the cycle. However, these are all just guesses. More reliable evidence on the effects of women’s cycles is provided from studies that validated menstrual phase with hormonal or follow-up assessments. In our review, these studies failed to provide much evidence for cycle shifts in preferences.
In further defense of evolutionary psychology predictions, Jones (2014) attributed the absence of menstrual cycle effects to shortcomings in our review. Yet our review of six different male attributes validated findings across multiple data sets and interpreted only the reliable patterns. 1 The evidence that cycle shifts in mate preferences are tied to research artifacts was further secured in Wood and Carden’s (in press) reanalysis of Gildersleeve et al.’s (in press) data set.
The artifacts we identified are open to interpretation. Publication status, according to research registry studies, is influenced heavily by the statistical significance of findings (Dwan, Gamble, Williamson, & Kirkham, 2013). Significance and not study quality also might underlie the publication status effects in mate preferences literature. By our coding of this literature and by Gildersleeve et al.’s (in press), unpublished studies were not lower in quality. Also relevant to interpreting our findings is the strength of the artifactual processes in Gildersleeve et al.’s and in our review. Menstrual shifts approached zero in studies that were unpublished, were published recently, or had more precise fertile phases. Given this pattern, the most plausible conclusion is that initial published reports of significant shifts in women’s preferences for quality men were false positives that were produced in part by researchers defining the fertile phase in ways that capitalized on chance in their data.
Understanding the Relation between Hormones and Mate Preferences
We hope that our meta-analytic findings spur psychological researchers to adopt more integrative models of women’s reproduction that include sociocultural processes along with evolutionary biology. What would such a model look like? According to Brown, Cross, Street, and Brand’s (2014) thoughtful commentary, women’s reactions to novelty, threat, and reward shift across the menstrual cycle and interact with socially transmitted information. They argued that menstrual cycle shifts are more likely to reflect such domain-general evolved dispositions than the specific mate preferences favored by evolutionary psychologists. However, because of our failure to find that women’s judgments shifted reliably across the menstrual cycle, we withhold comment until a systematic review is conducted of cycle shifts in sensitivity to threats and rewards.
Hyde and Salk (2014) proposed an alternative evolutionary hypothesis in which women have evolved specific mate preferences for masculinity and dominance in a partner that are relatively constant across the menstrual cycle. Given that women’s preferences shift with changes in women’s roles across recent time periods in industrialized countries as well as across different countries (e.g., Zentner & Mitura, 2012), any such evolved dispositions appear to be highly sensitive to current social contexts. As Hyde and Salk noted, cultural influences need to be a central component of any such evolutionary theory.
The social neuroendocrinology analysis of van Anders (2014) meaningfully incorporated both social constructions and evolutionary biology into models of human reproduction. According to her steroid/peptide theory (van Anders, 2013), hormonal responses to nurturing and competitive behaviors reflect gendered social constructions and provide insights into human’s evolved biology. In line with this thinking, we reviewed cross-cultural studies demonstrating that both menstrual cycles and mate preferences are socially situated and vary with women’s roles in society. Thus, menstrual frequency and other reproductive behaviors depend on women’s child-rearing and other productive societal roles. The science of gender will progress with these kinds of evolutionary models built not only on biology but also on culture, the most human of evolutionary processes.
Footnotes
Author note:
The author thanks Priyanka Joshi and Brian Louie for their thoughtful comments on this reply.