Male vs. female-favoring sex differences

Research on sex differences is sometimes controversial [1]. From the sociobiology wars of the 1970s to the scandals surrounding Lawrence Summers and James Damore, various culture-war flare-ups have centered on unwelcome claims about differences between men and women [2]. Some commentators point out, however, that not all sex-difference claims are equally likely to arouse concern or consternation [3]. One potentially important factor shaping people’s reactions to research on sex differences is whether the findings paint men or women in a better light. At first glance, there would appear to be three main predictions about how this would impact people’s reactions (assuming it impacts them at all).

1. Anti-Female Bias: People will react more positively to male-favoring findings.
2. Pro-Female Bias: People will react more positively to female-favoring findings.
3. Gender Tribalism: Men will react more positively to male-favoring findings; women will react more positively to female-favoring findings.

Existing theories in social psychology could in principle support each of these predictions. Traditional sexism might suggest a preference for male-favoring findings, either just in males or in females as well as a result of internalized misogyny [4]. Benevolent sexism and the chivalry hypothesis might suggest a preference for findings that favor females [4, 5]. And gender-ingroup bias might suggest gender tribalism: a male preference for male-favoring findings and a female preference for female-favoring findings [6–10].

Because existing theories can predict all possible outcomes, they give us little guidance in framing predictions in this domain. To break the theoretical gridlock, we proposed a new theory grounded in universal physical sex differences, which we dubbed the greater-protectiveness-of-females (or G-PROF) theory. The central claim of the G-PROF theory is that, as a general rule, people are more protective and/or controlling of girls and women than of boys and men. This, we argue, traces back ultimately to certain physical and reproductive differences between the sexes. First, men are typically larger and stronger than women, which means that women are typically more vulnerable. Second, women rather than men get pregnant and nurse the young, which means that, if a woman gets pregnant without an investing partner, she and her family are likely to be left “holding the baby.” In both cases, these sex differences may push individuals toward greater protectiveness and/or control of females than males—tendencies that may eventually be enshrined in the norms of a culture. A third sex difference may lead to the same destination but via a different route: cultural group selection [11]. The sex difference in question is the fact that women are more “reproductively indispensable”: If half the men in a group were removed, the group could in principle produce as many offspring in the next generation; if half the women were removed, the group could not. That being the case, groups that were more protective and controlling of women may have survived longer and grown more rapidly, and those tendencies would have slowly proliferated [12].

It is worth emphasizing that, although this explanation is grounded in evolved differences between the sexes, it is first-and-foremost a sociocultural explanation: Our primary claim is that these evolved differences influence learned behavior and social norms. The reason for this emphasis is that the extent to which girls and women are protected varies a great deal across cultures, and this is easier to reconcile with a sociocultural than an evolutionary explanation. That said, it does seem possible that the physical and reproductive sex differences that sometimes push individuals and cultures to be more protective and/or controlling of females could also have helped shape an evolved inclination in the same direction. Thus, the tendency may have both sociocultural and evolutionary roots.

In any case, applying the G-PROF theory to the question of reactions to research on sex differences led us to predict that people would typically react more positively to findings that put women rather than men in a more favorable light—that is, Option 2 on the earlier list. In a pre-registered study, we presented participants with fictitious popular-science articles describing sex differences that either favored females (females draw better or lie less) or favored males (the opposite). As expected, participants of both sexes were more positively disposed toward the female- than the male-favoring findings, judging the former to be more important, and the latter to be more harmful, offensive, and upsetting. We had predicted that, although the pro-female tendency would be found in both sexes, it would be stronger in women than men due to gender-ingroup bias [6–10]. Contrary to expectations, however, there was little evidence for such an effect: The preference for female-favoring differences was comparable in magnitude for both sexes. A follow-up study in Southeast Asia replicated both the male-favoring aversion and the absence of a sex difference, suggesting that this pattern extends beyond the boundaries of the West.

Several variables moderated participants’ reactions to the research. The first was belief in male privilege. The more privileged that participants believed men are over women, the more negatively they reacted to the male-favoring differences and the more positively to the female-favoring ones (and vice versa for the minority of participants who believed that women are privileged over men). The second moderator was political orientation: Independently of male-privilege belief, the more that participants leaned politically to the left, the more negatively they reacted to the male-favoring differences. Political orientation did not predict reactions to the female-favoring differences, perhaps because the protection of females is not a consideration when females are doing better. In our Southeast Asian replication, male-privilege belief predicted negative reactions to the male-favoring differences but did not predict reactions to the female-favoring ones, and political orientation did not predict reactions at all.

Lay predictions about gender bias

An additional issue explored in our earlier study was people’s beliefs about men and women’s gender-ingroup biases. Specifically, we explored participants’ predictions about how the average man and woman would respond to the male- vs. female-favoring sex differences. Our expectation was that both sexes would greatly overestimate the level of ingroup bias in both their own sex and the other. This was based on the fact that people tend to overestimate the magnitude of most effects in psychology [e.g., 13], coupled with the fact that people may have a heightened sensitivity to any evidence that individuals or groups are acting in self-interested ways, as the costs of overlooking such information are likely to be higher than the costs of an occasional false alarm [cf: 14]. As predicted, and consistent with past research [15], participants of both sexes predicted that the average man would strongly prefer the male-favoring findings, whereas the average woman would strongly prefer the female-favoring ones. The participants’ predictions were erroneous: First, participants greatly exaggerated the magnitude of the average woman’s own-sex preference, and second, participants assumed that the average man would exhibit a strong own-sex preference when in fact he exhibited a modest other-sex preference. These results were fully replicated in our Southeast Asian sample, suggesting again that the phenomenon is not simply a WEIRD or a Western one.

Aims and hypotheses

The aversion to male-favoring sex differences is contrary to what many would expect and could have significant theoretical and practical implications—shedding light, for instance, on the conduct and reception of research on sex differences, challenging common views about gender stereotypes and gender-ingroup bias, and helping quell conflict between the sexes. As such, it is important to assess the replicability of the finding. That was the aim of the present study. More precisely, the aim was to conduct a direct replication of our initial study on the topic [2], with a comparable Western sample and the same materials. The hypotheses below are based on the results of the earlier study, which was pre-registered with OSF (https://osf.io/6n5up/).

1. Participants will react more positively to female- than male-favoring sex differences.
2. Participant sex will have no impact on the strength of this effect.
3. The more privileged that participants think men are over women, the more negatively they will react to male-favoring differences and the more positively to female-favoring differences.
4. The more that participants lean to the left politically, the more negatively they will react to male-favoring differences. Political orientation will not predict reactions to female-favoring differences.
5. Both sexes will greatly overestimate the extent to which the average man and woman exhibit a preference for own-sex-favoring sex differences.

Note that Hypotheses 1 and 5 are the same as the original study, as the results were consistent with our hypotheses; Hypothesis 2 differs from the original study (we initially predicted that the preference for female-favoring findings would be stronger for female participants, but did not find this); and Hypotheses 3 and 4 were not part of the original study (the patterns emerged through exploratory analysis).

Participants

Participants were recruited via Prolific.ac, a website offering people a small payment for participating in online studies. We calculated the minimum sample size in advance with G*Power 3.1 [16], using an a priori power test (“ANOVA: Fixed effects, main effects, and interactions”) for three independent variables: Sex Favored (female vs. male), Participant Sex (female vs. male), and Trait Valence (positive vs. negative). Alpha was set at .05, power at .95, and the minimum effect size of interest (f) at 0.25 (a medium effect, which was the most common effect size found in our initial study). The analysis produced a recommended minimum sample size of 210. Our final sample consisted of 303 individuals: 150 men and 153 women. The age range was 18 to 72 years (M = 34.16, SD = 11.43). Most participants were from either the United States (64.4%) or the United Kingdom (32.3%).

Materials and procedure

The study was run with the online survey platform Qualtrics. After providing some general biographical information, participants were presented with a fictitious popular-science article discussing a recent study finding a sex difference (also fictitious; see S1 Fig for a sample). For half the participants, this was a sex difference in a positive trait (drawing ability); for the other half, it was a sex difference in a negative trait (lying frequency). Within each condition, half the participants were shown a female-favoring difference, and half were shown a male-favoring difference. Each version of the article included both a discussion of the sex difference and a bar graph representing it. Having read the article, participants answered four short questionnaires.

Statistical analyses

All analyses were conducted with SPSS Version 26. No outliers were identified or removed; all Cook’s Distance values were < 1. Four aggregate variables were created from the individual items in the four questionnaires: participants’ reaction to the research, participants’ predictions about the average man’s reaction, participants’ predictions about the average woman’s reaction, and belief in male privilege. For the three reaction questionnaires, we reverse-scored the items indicating a negative reaction to the research (i.e., those assessing how surprising, offensive, harmful, upsetting, and inherently sexist it was). As such, higher scores on these variables indicated a more positive reaction. All four variables had good internal consistency (Reaction to Research: α = .82; Average Man Predicted: α = .88; Average Woman Predicted: α = .9; Male-Privilege Belief Scale: α = .9).

To examine the effects of Sex Favored and Participant Sex on reactions to the research, and on predictions about the reactions of the average man and woman, we ran a series of two-way ANOVAs. As with the original study, we did not include Trait Valence as a third independent variable, as preliminary analyses showed that there was no main effect of this variable and no interactions between it and the other independent variables (main effect of Trait Valence: F_{1, 295} = 0.12, p = .729, d = 0.03; interaction with Sex Favored: F_{1, 295} = 0.21, p = .644, d = 0.06; interaction with Participant Sex: F_{1, 295} = 0.00, p = .986, d = 0.01; interaction with Sex Favored and Participant Sex: F_{1, 295} = 1.69, p = .194, d = 0.15). To determine whether male-privilege belief and political orientation moderated the effect of Sex Favored on participants’ reactions to the research, we conducted a moderation analysis using PROCESS Model 2 [18]. All variables were mean-centered before running the analysis [19, 20]. Interactions were explored using post-hoc multiple regressions. Our data met all the assumptions for multiple regression, including normality, linearity, homoscedasticity, and absence of multicollinearity. The minimal dataset is freely available from OSF at https://osf.io/ctnez/.

The threshold for statistical significance for all analyses was set at p = .05. We did not correct for multiple comparisons as we had specific predictions for each of our dependent variables and are reporting all our results. See S1-S5 Tables in S1 File for a breakdown of the descriptive and inferential statistics pertaining to this study.

Sex favored and participant sex

We first examined participants’ overall reactions to the fictitious research. Consistent with Hypothesis 1, there was a main effect of Sex Favored: Participants reacted less positively to the male-favoring sex differences than the female-favoring ones (F_{1, 299} = 24.48, p < .001, d = 0.58; see Fig 1, leftmost bars). As with our initial study, there was also a main effect of Participant Sex: Collapsing across experimental conditions, men were somewhat more positive about the research (F_{1, 299} = 7.12, p = .008, d = 0.33). Finally, and consistent with Hypothesis 2, there was no interaction between Participant Sex and Sex Favored, indicating that male and female participants preferred the female-favoring over the male-favoring sex differences to a comparable degree (F_{1, 299} = 1, p = .318, d = 0.12).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Participants’ reactions to fictitious research describing a male- vs. a female-favoring sex difference.

Higher scores on the “Overall reaction” variable indicate a more positive reaction to the research; higher scores on the individual items indicate stronger agreement that the descriptor applies to the research. * p < .05; ** p < .01; *** p < .001; n.s.: not significant. Error bars = 95% CIs.

https://doi.org/10.1371/journal.pone.0266171.g001

To get a more detailed picture of the participants’ reactions, we also examined the individual items making up the aggregate variable. Fig 1 presents the results. Mirroring our original study, participants found the female-favoring sex differences more plausible, and the male-favoring differences more surprising, harmful, upsetting, and sexist. In contrast to the original study, participants also found the female-favoring sex differences more interesting but did not rate them as more important or the research as better conducted.

Male-privilege belief and political orientation

The average participant viewed men as more privileged than women (M = 1.19, SD = 1.25, on a scale ranging from -3 to 3, with positive values indicating a belief that men are privileged over women and negative values the reverse). On average, though, female participants thought the privilege gender gap was larger than did males (M = 1.55, SD = 1.09 vs. M = 0.82, SD = 1.25; F_{1, 299} = 29.46, p < .001, d = 0.63). A minority of participants (12.9%) saw women as more privileged than men. This included 22.1% of the males and 4.8% of the females (X² [1, N = 278] = 18.55, p < .001, d = 0.53).

Political orientation was measured with a Likert scale spanning from 1 (“Left/Liberal”) to 7 (“Right/Conservative”). Participants fell across the full range of the political spectrum, but with the average falling somewhat to the left (M = 3.26, SD = 1.58, where 4 indicates the centrist midpoint).

To test Hypotheses 3 and 4, we next looked at whether male-privilege belief and political orientation moderated the effect of Sex Favored on participants’ reactions to the research. The overall model was significant (R² = .19, F₅,₂₈₅ = 13.66, p < .001). Sex Favored was the strongest predictor of participants’ overall reactions (B = 0.57, t₂₈₅ = 4.97, p < .001), followed by male-privilege belief (B = -0.13, t₂₈₅ = -2.31, p = .022). Political orientation did not uniquely predict participants’ reactions (B = 0.08, t₂₈₅ = 1.93, p = .054).

There was a significant interaction between Sex Favored and male-privilege belief (B = 0.38, t₂₈₅ = 3.49, p = .001). Consistent with Hypothesis 3, the more privileged that participants believed men are over women, the less positively they reacted to the male-favoring sex differences (R² = .18, F_{1, 148} = 32.79, p < .001, B = -0.39, t₁₄₈ = -5.73, p < .001). Contrary to Hypothesis 3, though, male-privilege belief did not predict reactions to the female-favoring differences (R² = .002, F_{1, 149} = 0.36, p = .548, B = 0.04, t₁₄₉ = 0.6, p = .548).

There was no interaction between Sex Favored and political orientation, contrary to Hypothesis 4 (B = -0.1, t₂₈₅ = -1.23, p = .221). However, a post-hoc analysis using PROCESS Model 1 revealed that, without including male-privilege belief in the analysis, there was a significant Sex Favored × Political Orientation interaction (R² = .15, F_{3, 289} = 16.41, p < .001, B = -0.26, t₂₈₉ = -3.45, p = .001): Consistent with Hypothesis 4, a left-leaning political orientation predicted more negative reactions to male-favoring differences (R² = .21, F_{2, 144} = 19.5, p < .001, B = 0.13, t₁₄₄ = 2.23, p = .028), but did not predict reactions to the female-favoring differences (R² = .006, F_{2, 141} = 0.41, p = .663, B = 0.03, t₁₄₁ = 0.49, p = .623). The fact that, after adding male-privilege to the model, the Sex Favored × Political Orientation interaction disappeared indicates that, in this dataset at least, much of the link between a left-leaning political orientation and negative reactions to male-favoring sex differences is a result of the fact that left-leaning individuals typically perceive higher levels of male privilege.

Predictions about the average man and woman’s gender biases

Finally, we looked at participants’ predictions regarding the reactions of the average man and woman to the fictitious findings. Starting with the average man, there was a large main effect of Sex Favored: Consistent with Hypothesis 5, participants predicted that the average man would react much more positively to the male- than the female-favoring sex differences (F_{1, 298} = 225.02, p < .001, d = 1.66; see Fig 2). The collective prediction was false: As discussed, the average man in fact reacted less positively to the male-favoring differences. There was no main effect of Participant Sex (F_{1, 298} = 2.11, p = .148, d = 0.05). As with our earlier study, however, there was a significant Sex Favored × Participant Sex interaction (F_{1, 298} = 24.57, p < .001, d = 0.57): Female participants predicted more own-sex favoritism from the average man than male participant did. More precisely, female participants predicted that the average man would react more positively to the male-favoring differences than male participants predicted (F_{1, 148} = 7.18, p = .008, d = 0.44), and less positively to the female-favoring differences (F_{1, 150} = 18.02, p < .001, d = 0.69).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Participants’ predictions regarding the reactions of the average man and woman to a male- vs. female-favoring sex difference, alongside the actual reactions observed in the study.

* p < .05; ** p < .01; *** p < .001; n.s.: not significant. Error bars = 95% CIs.

https://doi.org/10.1371/journal.pone.0266171.g002

Turning to the average-woman predictions, there was an extremely large effect of Sex Favored on participants’ reactions to the research (F_{1, 297} = 426.38, p < .001, d = 2.38). Consistent with Hypothesis 5, participants predicted that the average woman would react much more positively to the female- than the male-favoring sex differences (F_{1, 297} = 426.38, p < .001, d = 2.38; see Fig 2). Though accurate with respect to its direction, the predicted effect was 361% larger than the actual one (d = 2.38 vs. 0.66). There was no main effect of Participant Sex (F_{1, 297} = 0.44, p = .509, d = 0.14). However, unlike our earlier study, there was a (barely) significant interaction between Sex Favored and Participant Sex (F_{1, 297} = 3.94, p = .048, d = 0.23): Males predicted slightly more own-sex bias from the average woman than females predicted—the mirror image of the pattern for the average man. This was largely because males predicted that the average woman would react more positively to the female-favoring differences than females predicted (F_{1, 149} = 3.9, p = .05, d = 0.32). The sexes did not differ in their predictions for the male-favoring differences (F_{1, 148} = 0.79, p = .375, d = 0.15).

Limitations and future directions

The present study had a number of limitations. One is that our fictitious popular-science articles omitted a piece of information often included in such articles and likely to impact people’s reactions to sex-difference research: the sex of the researcher producing the research. If, as we suspect, the aversion to male-favoring findings stems from greater protectiveness of females than males, then the effect may be compounded when the research is produced by a male. Historically, men have been more of a threat to women than other women have, and thus the female-protective tendency may be intensified when a perceived threat originates from a man. If this reasoning is correct, then people may judge a male-favoring finding promoted by a male in a dimmer light than they would the same finding promoted by a female.

A second limitation is that we looked at the effects of male- vs. female-favoring sex differences for only two traits: drawing ability and lying frequency. To assess the generality of the male-favoring aversion, the range of traits should be extended. One trait that may be particularly worth exploring is intelligence. Intelligence is a reliable predictor of many important life outcomes, including academic achievement, career success, income, health, and longevity [23]. As such, sex differences in intelligence are more consequential and more contentious than those explored in our studies to date. Would the effect of Sex Favored be amplified for such a difference? Would men still exhibit an other-sex bias, or would they instead start exhibiting an own-sex bias given the higher psychological stakes? Based on the robustness of the male-favoring aversion in our studies so far, and the fact that the phenomenon that led us to predict it (i.e., the greater protectiveness of women) would if anything be more relevant for a consequential trait than an innocuous one, our expectation is that participants of both sexes would continue to react less positively to a male-favoring difference, but that the effect would be larger than that found in our prior research. This remains, however, to be demonstrated.

A final limitation of the present study is that we did not directly explore the psychological variables underpinning the aversion to male-favoring findings. One way to do so would be to split the Reaction-to-Research variable into three parts, namely the quality of the research, the harmfulness of its findings, and the extent to which the disfavored sex should be shielded from those findings. This would enable us to address two new questions. First, we could determine whether Sex Favored has an impact on judgments of research quality alone. Unlike judgments about harm and the need to shield the disfavored sex, judgments about quality should in principle be unaffected by the direction of the effect. As such, exploring the effect of Sex Favored on quality judgments alone would provide a stronger test of the effects of motivated reasoning on participants’ reactions. Second, if Sex Favored does indeed affect quality judgments (as we suspect it will), we could then explore whether this effect is mediated by participants’ perception of the harmfulness of the research findings and their desire to shield women from them.