On hits and being hit on: error management theory, signal detection theory, and the male sexual overperception bias. Jordann L. Brandner, Jadyn Pohlman, Gary L. Brase. Evolution and Human Behavior, February 18 2021. https://doi.org/10.1016/j.evolhumbehav.2021.01.002
Abstract: Although Error Management Theory (EMT) can explain male sexual overperception, more advanced Signal Detection Theory (SDT) analyses can identify sensitivity and bias separately. An SDT analysis of perceptions of relatively clear interest/disinterest signals (Study 1) found that sensitivity to sexual interest/disinterest signals drove participants' perceptions, rather than an overall bias to perceive sexual interest. Cues of interest were generally underperceived, while sensitivity and accuracy were uniformly high. EMT analysis also found overall sexual interest underperception, but with men slightly overperceiving interest relative to women. These discrepant results were due to EMT using difference scores, which obscure baseline perceptions for men and women. Individual differences in life history strategy, mating strategy, and mate value did not affect sensitivity or bias. Study 2 largely replicated these results using more ambivalent scenarios, except EMT analyses found men's misperception to be significantly larger than women's, despite being closer to pre-rated communication levels. These results show that sexual communication may be more nuanced than previously thought, and that an SDT analysis is more appropriate for such data.
Keywords: Sexual overperception biasError management theorySignal detection theoryHuman sex differencesIndividual differences
4. General discussion
Signal Detection Theory analyses show that sensitivity to the differences between signals of sexual interest and disinterest drove participants' perceptions of sexual interest, rather than an overall bias to perceive sexual interest. Men and women did not significantly differ in sensitivity, which was high, even when including more ambivalent combinations of behaviors in Study 2. Generally, cues of sexual interest were underperceived, with sex-based effects varying. In Study 1, women had a significantly more liberal bias than men, whereas in Study 2, there was no significant effect of sex on bias. Other individual differences, such as life history strategy, mating strategy, and mate value did not generally affect sensitivity or bias. Error Management Theory analyses also found an overall underperception of sexual interest, but with men showing higher misperception scores than women. Further examination found that EMT usage of difference scores did not accurately reflect the average levels of perception between men and women due to overall underperception. This results in negative difference scores which were exacerbated in Study 1 by different baselines for men and women. Baselines were adjusted in Study 2 to be more equal, however, issues with interpretation of the results persisted again due to negative difference scores. EMT analyses lead to the conclusion that men's misperception scores were greater than women's, even as men's perceptions were closer to true levels of interest communicated.
4.1. Theoretical and practical implications
Although the “male sexual overperception effect” was not found, there are numerous implications of these results which indicate that sexual communication is actually more nuanced than previously thought. SDT analysis is a valuable methodology for better exploring the male sexual overperception effect and other biases explained through EMT. EMT analyses of the evolutionary costs and benefits of a decision outcome can be integrated with SDT to evaluate optimal decisions under uncertainty from an evolutionary standpoint. This integration is particularly important, as SDT analyses provide measures of both bias and sensitivity, which is more informative than typical EMT analyses which only provide a measure of bias. The separation of bias (tendency to respond in one particular way) from sensitivity (how distinct signals and noise are from each other) allows for a better understanding of the psychological mechanisms underlying a behavior. For instance, the present findings indicate that sexual interest perceptions were significantly driven by sensitivity, rather than bias, suggesting that evolutionary optimality in sexual communication may be attained through the ability to separate signals of sexual interest from signals of sexual disinterest. Moreover, this high level of separation more closely matched the patterns of high participant accuracy than did the EMT analyses, which found general underperception of interest, with men perceiving more interest than women. Notably, the EMT analysis has no measure of overall accuracy or cue discriminability, leading to conclusions of men perceiving more interest. This contrasts with the more accurate SDT conclusions of high discrimination of interest from disinterest, resulting in high accuracy when perceiving sexual communication.
Additionally, the standardized measures of sensitivity and bias provided by SDT are an improvement from the non-standardized measure of bias from EMT. Measures of c and d’ are comparable to other research using SDT, whereas EMT measures of bias are only comparable to other measures of bias on the same scales. Although EMT measures of bias could potentially be standardized, such as through a z-score calculation, they would still be problematic due to known issues with difference scores in terms of information lost from the original data (Cronbach & Furby, 1970), particularly in relationship research (Griffin, Murray, & Gonzalez, 1999). A single difference score can come from a variety of situations that could be driven by a main effect in one of the variables or by one of the variables having different variance than the other. For EMT in particular, the use of difference scores leads to analyses that can only assess bias and neglect sensitivity. Because bias and sensitivity interact as people strive to make optimal judgments, this can lead to incomplete understandings of behavior.
As an example of this issue in using difference scores, it turns out that the women's vignettes in Study 1 communicated more sexual interest than men's vignettes and after these baselines were subtracted from average perceptions, women showed a larger absolute difference between perception and truth despite men and women having similar average levels of perceived interest. Moreover, because this difference was negative, men's misperception scores were greater (i.e., closer to zero) resulting in misperception scores that indicated that men perceived more interest than women, despite having a lower average perception of interest than women. While this difference in baseline scores could be accounted for through SDT's use of multilevel modeling, EMT analyses can only account for this through manipulation of stimuli. Study 2 manipulated the stimuli to address this difference, but issues with interpretation due to negative misperception scores persisted. Additionally, EMT analyses typically aggregate across trials and within sex, losing valuable information and statistical power, whereas SDT analyses can use a within-subjects multilevel approach that can account for individual and stimuli variation through inclusion of individuals and stimuli in the random effects structure (Wright & London, 2009).
In the particular context of the male sexual overperception effect, the assessment as to if men are overperceiving female sexual interest will depend on how overperception is operationalized. Is overperception simply men perceiving more sexual interest than women? Or is overperception perceiving more sexual interest than is communicated? Currently in this line of research, overperception is operationalized most commonly as men perceiving more sexual interest than women perceive. Overperception is typically tested as a sex difference in average misperception scores, which are often calculated as the difference between expressed interest and perceived interest (e.g., Perilloux et al., 2012). However, this analysis does not necessarily indicate if men are overperceiving, just that they have larger misperception scores than women. The issue of this operationalization was demonstrated in these studies. EMT's use of difference scores to operationalize misperception is flawed, so having “higher” misperception when underperceiving would be a good thing, as it would mean more accuracy, however EMT does not have the ability to address this, as it assumes misperception is always overperception and thus higher misperception values indicate poorer responding.
When using SDT methods however, researchers have more control over how overperception is operationalized. Researchers can compare a participant's perceptions to numerous baselines: members of the participant's sex (e.g., does this man perceive more sexual interest than other men say is communicated?), members of the target's sex (e.g., does this man perceive more sexual interest than women say is communicated?) or members of both sexes (e.g., does this man perceive more sexual interest than both men and women together say is communicated?). These baselines can be established during stimuli development and validation, similar to how it is approached in this study (see Supplemental Materials for details), which used members of the target sex as the baseline. Since sexual communication is often cross-sex communication, sex-specific behaviors and sex-specific ratings of behaviors are necessary for proper operationalization of sexual perception. This can get complicated, short of being able to peer into participants' souls for their absolute truth. Luckily, the estimations of baseline truth described above and in the supplementary materials can suffice for research and analysis purposes.
Lastly, SDT analyses also allow for more precise exploration of the effects of individual differences. Whereas previous EMT research has correlated individuals' personality measures with their average difference scores (e.g., Perilloux et al., 2012) or used regression to predict average difference scores using personality measures (e.g., Howell et al., 2012; Kohl & Robertson, 2014), SDT analysis can examine the effect of individual differences on sensitivity and/or bias, depending on the hypothesis. Although correlation and regression are valid tools, applying them on the average difference scores used by EMT can be problematic due to the issues noted above.
4.2. Limitations
It is important to note that the present research used vignettes as a proxy for communication. While the use of descriptions of actions rather than real actions is common for research in male sexual overperception (e.g., Abbey & Harnish, 1995; DeSouza et al., 1992; Edmondson & Conger, 1995; Fisher & Walters, 2003; Haselton & Buss, 2000; Kowalski, 1993), it nevertheless could have affected the results and conclusions of this study. Prior research has indicated that different formats of presentation can affect perceptions of sexual interest, with formats including less information showing more overperception (Edmondson & Conger, 1995). This could indicate that more realistic stimuli could result in even more underperception of sexual interest, as the vignettes used here simplify an interaction to three behaviors. Another possible issue with these vignettes is that they were created using an act-nomination frequency approach (Buss & Craik, 1983). This resulted in some behaviors which included a certain degree of interpretation implicitly included in the descriptions (e.g., giving a “nasty look”). These potential issues with vignettes could be addressed in future studies by creation of video stimuli which include information about whether or not participants were truly attracted to each other, and thus covertly signaling sexual interest.
Like all studies on male sexual overperception, these studies rely on indirect measures of sexual interest. No truly objective measures of sexual intent exist, resulting in methods that rely on self-report and 3rd-party observers. Self-ratings may understate true intent, especially for women, as concealment of sexual interest and disinterest can be advantageous for gathering additional information about potential mates (Trivers, 1972). Conversely, same sex 3rd-party observers might overstate a woman's sexual interest, either as a misperception or as a tactic of intrasexual competition. These present studies used an iterated act-frequency nomination and evaluation approach for developing stimuli that describe typical sexual interest and disinterest behaviors, and hopefully this reduced any influences of self-concealment and 3rd-party competition. Future research in this area, though, should examine the differences between self and 3rd-party evaluations of interest, as well as include evaluations of interest from others who are close to the individual and may know them better, such as close family or friends.
This research does not directly address other theories regarding male sexual overperception, such as the general oversexualization hypothesis (Abbey, 1982, Abbey, 1991), the media hypothesis (Abbey, 1991), or the default-model hypothesis (Shotland & Craig, 1988). The revealed structure and outcomes from the SDT analysis, however, are more consistent with EMT than with any of those alternative theories. Further study designs could employ SDT analyses of variables that would contrast at least some of these competing hypotheses. The current research does provide evidence against the general insensitivity hypothesis (Farris, Viken, & Treat, 2010; Farris et al., 2008a, Farris et al., 2008b) as sensitivity to cues of sexual interest was very high and drove responses, contrary to this hypothesis which proposes sensitivity to these cues is low.
These studies also do not address is if male sexual overperception is a biased cognitive mechanism or a biased behavioral outcome. Much discussion of the male sexual overperception effect in recent years has been about whether men truly believe women are more sexually interested in them, or if they just behave as if women are sexually interested in them (e.g., McKay & Efferson, 2010; Murray, Murphy, von Hippel, Trivers, & Haselton, 2017; Perilloux & Kurzban, 2015; Perilloux & Kurzban, 2017). These studies found no effect of bias on responses and instead found that sensitivity was driving responses, and thus it may be a case of neither biased cognitions nor biased actions, but instead sensitivity to cues. Moreover, if an effect of bias had been found, these studies would not be able to examine if the bias was caused by a belief pattern or a behavioral pattern. Future studies should consider including measures of confidence in the belief that a woman is communicating sexual interest to help clarify this issue.
These studies demonstrate that the use of difference scores in prior research may have overestimated the actual prevalence of the male sexual overperception effect. This opens up an interesting topic of when and under what situations male sexual overperception may (or may not) appear. For instance, a version of this argument could note that data collection for Study 1 occurred during the fall/winter of 2018, closely following the confirmation hearings of Brett Kavanaugh and Dr. Christine Blasey-Ford's testimony as well as the #MeToo movement. These events may have influenced participants to underestimate sexual interest in a precautionary manner. However, Study 2 occurred during the summer and fall of 2020, well past the immediate effects of these contexts and yet the same results were found. Is it possible that the societal effects of these events are both persistent and outweigh more ultimate evolutionary bias adjustments? Reanalyzing previous research (which have datasets that include repeated measures and thus are amenable to SDT analysis methods) or other research across time and varying cultures may help to partially address this question.
No comments:
Post a Comment