Overdoing Gender*

This paper explores the empirical validity of the masculine overcompensation thesis, which asserts that men react to masculine insecurity with extreme demonstrations of masculinity. Overcompensation, an often-cited popular account for exaggerated masculine-typed behaviors, is supported by theoretical work in the masculinities and identity theory literatures. I tested the masculine overcompensation thesis in an experiment in which men and women were given feedback on a gender identity survey suggesting that they are either masculine or feminine. While women showed no change as a result of the type of feedback received on the gender identity survey, men showed a variety of exaggerated masculine-typed behaviors and attitudes. Compared to men given masculinity-confirming feedback, men given feedback threatening their masculinity, 1) expressed greater homophobia, including greater support for a ban on same-sex marriage, 2) showed greater support for the Iraq war, including President Bush’s handling of it, and 3) expressed greater interest in purchasing a Sport Utility Vehicle (SUV), relative to other vehicles. Masculinity-threatened men also reported feeling more ashamed, guilty, upset and hostile than did masculinity-confirmed men. Results of the study offer strong support for the masculine overcompensation thesis. TESTING THE MASCULINE OVERCOMPENSATION THESIS INTRODUCTION In the 1964 film Dr Strangelove, Or How I Learned to Stop Worrying and Love the Bomb, the initiation of World War III is depicted as a great act of overcompensation by a sexually impotent general. This line of reasoning is not unusual. Masculine overcompensation is invoked as an explanation for a variety of behaviors from the small (men purchasing sportscars at the onset of mid-life crises) to the world-changing (as in Franz Neumann’s analysis of the rise of the Third Reich as a reaction to post-WWI German emasculation). Even though this notion is pervasive in popular culture its empirical validity remains in question. The masculine overcompensation thesis asserts that men who feel insecure about their masculinity enact extreme masculine behaviors in an effort to achieve masculine status in their eyes and others. The masculine overcompensation thesis further suggests that extreme, caricatured demonstrations of masculinity by men may thus be rooted in masculine insecurity. Masculine overcompensation is an idea rooted in Freudian psychology but which has grown to have a life of its own. It has been cited as an explanation for a wide variety of behaviors ranging from war and homophobia to the purchase of powerful sports cars and SUVs. In this paper I review the theoretical basis for the masculine overcompensation thesis and lay out specific hypotheses regarding the role of masculine overcompensation in men’s behavior. I report the results of an experiment in which I gave men and women feedback suggesting they had either masculine or feminine gender identities and measured the effects on a variety of masculine behaviors. Finally, I discuss implications of the study for the validity of the masculine overcompensation thesis, addressing possible alternative explanations for the results. THEORY The masculine overcompensation thesis is originally derived from Freudian theory, specifically Freud’s notion of “reaction formation” (1894). A recent review of Freudian defense mechanisms by Baumeister and colleagues (1998) found substantial and diverse support for reaction formation in the contemporary social psychology literature. In their analysis, reaction formation “involves converting a socially unacceptable impulse into its opposite....People respond to the implication that they have some unacceptable trait by behaving in a way that would show them to have the opposite trait.” Masculine overcompensation could be seen as a special case of reaction formation, with the implication of femininity or deficient masculinity being the unacceptable trait motivating men’s overcompensating behavior. In perhaps the most famous empirical demonstration of reaction formation, Adams, Wright, and Lohr (1996) found that more homophobic men actually showed greater sexual arousal while watching videos of homosexual intercourse, despite selfreporting low levels of arousal. This finding suggests that homophobia may be a case of reaction formation for men with homosexual urges, but stronger concerns about the social implications of homosexuality. The masculine overcompensation thesis follows a similar line of reasoning: men who suspect themselves to have insufficient masculinity overcompensate by enacting extreme masculine behaviors and attitudes designed to create the impression that they are in fact quite masculine. In modern social science, the masculine overcompensation thesis is most clearly related to the expansive gender literature on masculinity. Connell’s conception of “hegemonic masculinity” describes the ever-changing, manifold meaning of masculinity the content of which varies substantially across cultures (1987). Femininity and other masculinities are less powerful and respected than hegemonic masculinity. Hegemonic masculinity exerts a strong conformity pressure on men, though it is not normative in the sense that total conformity is possible. Rather, true masculinity is an unattainable ideal (Connell 1987; 1995). Thus, for Connell, masculine insecurity is not an occasional event, although various events may be more emasculating than others. Instead, masculine insecurity, feelings of emasculation, and the suspicion of femininity are ubiquitous for men. These concerns and feelings of deficiency instigate the enactment of masculinity in everyday life. Through this lens, because true masculinity is idealized and unattainable, a strain always exists, and the result of that strain is overcompensation and striving for ever greater masculinity. Kimmel also emphasizes the nearly constant threat to men that they may be revealed as insufficiently masculine (2000). Kimmel refers to “homophobia” as the fear of other men detecting one’s insufficient masculinity. However, this fear is itself a source of shame and must be covered up, along with any possible unmasculine or feminine characteristics, with bold demonstrations of strength and masculinity. From the perspective of Kimmell, and Connell, masculine overcompensation is not an occasional phenomenon, but is instead the fundamental process through which masculinities are enacted, What we call masculinity is often a hedge against being revealed as a fraud, an exaggerated set of activities that keep others from seeing through us, and a frenzied effort to keep at bay those fears within ourselves...the reigning definition of masculinity is a defensive effort to prevent being emasculated. (p. 103-6, 2000) Theories of identity converge on the general claim that individuals often exhibit overcompensation-type behaviors in an effort to protect identities that are central to them, as gender identities are assumed to be for most individuals. In Identity Theory, gender identification as masculine or feminine motivates behavior related to gender, such as men who identify as masculine behaving in a more dominant or competitive fashion (Stets and Burke 2000a, Burke 1989). Research in identity theory shows that much may be at stake for individuals perceived to have gender-inappropriate identities, including insults and resulting low self-esteem (Burke and Tully 1977). Individuals who have adopted an identity enact behaviors meant to maintain that identity, even in the face of threat (Stets and Burke 2000b). Social identity theory (Tajfel and Turner 1986; Abrams and Hogg 1990) asserts that group identification relies on being simultaneously prototypical of one’s own social group and distinctive from the out-group. Thus, in the case of gender identity, men typically find it essential to project masculinity and avoid appearing feminine. Social identity theory also suggests a number of threats to one’s group identification, and behaviors individuals enact in response to these threats. Environmental stimuli suggesting that one’s group is not distinctive from the out-group, that one’s group is not as good as the out-group, that one is not a valuable member of the group, and that one is more similar to the out-group than to one’s own group all present unique threats to group identity motivating behaviors to restore group identity. In the case of gender identity, this identity defense may take the form of masculine hostility and violence against women. In a recent experiment conducted in Italy, men were given feedback on a gender identity survey suggesting that they were either typical or atypical men (Maass et al. 2003). Next, participants worked with a woman on a computer-mediated image-sharing task where sexual harassment behavior was made possible by including pornographic materials among the images that could be sent back and forth. Researchers tracked the number of participants who chose to send uninvited, explicit pornographic materials to women co-workers; the rate of men engaging in this behavior was 23% when they had been given feedback suggesting they had typical levels of masculinity, but more than doubled to 50% among men told that they were feminine. A striking result, the study suggests that men may react to the implication that they lack masculinity with extreme masculine behaviors, in this case sexual harassment of women. MEASURING OVERCOMPENSATION According to the masculine overcompensation thesis, men will become more likely to enact extremely masculine behaviors after a threat to their masculinity. In order to test this assertion I follow Maass et al.’s (2003) research strategy of experimentally manipulating feedback to study participants, suggesting that they are either typical or atypical of their gender, then assessing a wide variety of attitudes and behaviors to see if men exposed to masculinity threat behave in more extreme masculine ways. I predict no effects of femininity-threat for women participants on the masculine-typed attitudes and behaviors used in the study. I predict that men exposed to masculinity threat, compared to men whose masculinity is confirmed in the study, will: Hypothesis 1: Express more negative attitudes towards homosexuality; Hypothesis 2: Express greater support for violent military action; Hypothesis 3: Express greater interest in purchasing an SUV; and Hypothesis 4: Persist longer on a public test of strength. The masculine-typed behaviors selected for study all reflect aspects of masculinity in American culture and are therefore potential measures of masculine overcompensation. Homophobic attitudes reflect interpersonal dominance and assertions of heterosexuality. Support for war involves core masculine ideals of violent aggression, strength, and competition. Interest in SUV purchasing may reflect a desire for the power, potency, and physical size the vehicle can offer the owner. Further, marketing of SUVs in America systematically targets men’s masculinity concerns (Bradsher 2002), enhancing these associations. Finally, public displays of strength directly reflect strength and physical dominance. All of these characteristics are typical of American masculinity ideals. I also predict that negative affect will be higher for men in the masculinity-threat than in the masculinity-confirmed condition, though which specific negative emotions items will differ across conditions I leave as an open question. METHODS I designed an experimental study to test these hypotheses. In the experimental study men and women were given feedback suggesting they were either masculine or feminine. Following this experimental manipulation, all participants filled out a series of surveys designed to assess their 1) attitudes towards homosexuals, 2) support for the Iraq War, 3) interest in purchasing an SUV, and 4) positive and negative affect. Participants also performed a brief strength test using a hand-grip. Participants 111 undergraduates at Cornell University participated in the study for money or extra credit in a sociology class. One man was removed from analyses after voicing suspicion regarding the feedback he received on the gender identity survey. Procedure The study featured a 2 (Participant gender: Man/Woman) x 2 (Gender Identity Feedback: Masculine/Feminine) between-subject design creating four conditions: femininity confirmed, femininity threat, masculinity confirmed, and masculinity threat. Participants were recruited by fliers advertising payment for participation in a sociology experiment or by announcements of an extra credit opportunity in their undergraduate sociology class. After reporting to the lab participants filled out a demographic questionnaire and a “gender identity survey.” After this the participants were asked to wait several minutes as the research assistant scored the gender identity survey. The gender identity survey was in fact the Bem Sex Role Inventory and was not scored until after completion of the study (Bem 1974). Participants were given results on the gender identity survey in a sealed envelope with their name printed on it. The envelopes were in fact filled with feedback sheets completed prior to the session and sealed so that research assistants would be blind to the experimental condition. The feedback sheets inside the envelopes displayed a 0 to 50 scale of possible scores on the gender identity survey as shown in Figure 1. The range from 0 to 25 was the masculine half of the scale and 26-50 the feminine half. In the middle of each range brackets indicated the supposed “average male range” and “average female range” for scores in the study. Participants’ scores were entered on a line above the scale and with an “X” placed at the corresponding place on the scale. In actuality, the average ranges depicted on the feedback sheets, as well as the participants’ scores, were false and created purely for the purposes of manipulating gender identity feedback. Men and women in the study were randomly assigned to receive either masculine or feminine feedback. Half of men were told they scored an “11” on the gender identity survey, a score corresponding to the middle of the masculine distribution noted on the feedback sheets. The other half of the men were given a score of “32,” just inside the feminine range of the scale. Half of the women in the study were told that they scored a “39” on the gender identity survey, corresponding to the middle of the feminine distribution. The other half of female participants were told they scored an “18,” a number just within the typical masculine range of the scale. This manipulation of gender 1 Masculinity/femininity scales have been criticized for conflating masculinity and femininity with sex difference (Connell 1987). I recognize this and other criticisms of the scale but use it nonetheless to enhance experimental realism because pre-testing and prior research revealed that undergraduates found it a convincing gender identity feedback. 2 This notion that masculinity and femininity exist at two ends of the same gender identity spectrum is inconsistent with the literature on gender. For example, the Bem Sex Role Inventory produces separate scores for an individual’s masculinity and femininity. Nonetheless, we gave participants feedback on a single masculine/feminine dimension because it is consistent with a lay view of gender identity. identity feedback is similar to that used by social identity researchers as a manipulation of gender prototypicality (Schmitt and Branscombe 2001; Maass et. al. 2003). Dependent Measures Following the manipulation of gender identity feedback participants were asked to fill out two survey packets and perform a strength task with a handgrip. One survey packet was labeled a “Political Views Survey” and assessed participants’ attitudes towards homosexuals and the Iraq War. Participants rated on 7-point scales their support for a constitutional amendment banning same-sex marriage, the gay rights movement, and whether they thought homosexuality was “always” or “never wrong.” These three measures were reliable (Cronbach’s alpha = .89) and therefore were averaged (support for the gay rights movement reverse-scaled) to create a single composite measure of “Homophobia.” Participants also rated on 7-point scales their support for “President Bush’s decision to invade Iraq” and approval of “President Bush’s handling of the war in Iraq.” These two measures were also reliable (Cronbach’s alpha = .93) and therefore were averaged to create a single composite measure of “Support for Iraq War.” The political views survey also included a one-page “feelings survey” on the back page asking participants to indicate on a 5-point scale how much each of 20 emotions items described how their feelings at the time. This PANAS survey was a measure of positive and negative affect used widely in social psychology research (Watson et al. 1988). Participants also filled out a survey packet entitled “Car Purchasing Survey.” The packet included one-page descriptions of four different Ford automobiles (an SUV, a Minivan, a Sedan and a Coupe) including picture, engine specs, and fuel economy information. Participants were asked to carefully study each vehicles description before filling out a survey on the back page assessing attitudes towards each vehicle. Participants rated each vehicle on a 10-point scale of “desirability,” indicated how much they would be willing to pay for each vehicle, which was their favorite, and which was the one they would be most likely to buy. The order of the Political Views and Car Purchasing surveys was counter-balanced so that half of participants were administered the former first, and half the latter. After filling out these two survey packets participants were asked to participate in a brief strength test (all agreed to participate). In the strength test participants used their dominant hand to hold a handgrip, like that used for hand exercises, closed around a scrap of paper for as long as they could. Research assistants timed how long participants went before releasing the piece of paper. Following the strength test, participants were sensitively debriefed regarding the deception (the gender identity feedback), thanked for their participation, and paid. Debriefing carefully followed the structure suggested by Aronson et al. (1990), based on research by Ross et al. (1975), for properly restoring the original state of participants given false feedback in an experiment. RESULTS Table 1 compares means for men and women’s attitudes towards homosexuality and the Iraq war following feedback that the participant was either masculine or feminine. Ratings are generally low relative to the scale’s midpoint of “4,” suggesting that the participants reported generally low levels of homophobia and support for war. This was especially true among women in the study. As can be seen, women reported roughly the same attitudes towards homosexuality after receiving either feminine or masculine feedback, as predicted. Men, however, reported more homophobic attitudes in the treatment condition than in the control condition. Men who were told that they were feminine expressed more negative attitudes towards homosexuality as compared with men who were told that they were masculine. Statistically significant differences were observed on all three individual items and the homophobia composite measure, suggesting greater homophobia among masculinity-threatened men. These findings constitute support for Hypothesis 1. No difference in attitudes towards the Iraq War was observed among women given feminine and masculine feedback. Women again exhibited no effects as a result of the experimental manipulation, as predicted. Men given feedback that they were feminine reported significantly higher support for Bush’s handling of the war and scored significantly higher on the composite measure of support for the Iraq War. Another measure of support for the decision to invade Iraq showed an effect in the predicted direction, though the difference only approached marginal significance (p = .11). These findings support Hypothesis 2. Men given masculinity-threatened feedback supported the Iraq War more than those reassured of their masculinity. It is possible that, rather than displaying masculine traits more conspicuously, men simply become more politically conservative when insecure about their masculinity. To evaluate this alternative to the masculine overcompensation thesis I analyzed participants’ responses to some filler items from the “Political Views Survey” to determine if men also showed more conservative attitudes towards political issues that weren’t strongly linked to cultural conceptions of masculinity. Specifically, I looked at participants’ support for the environmental protections and ratings of the fairness of inequality in the U.S. I found no significant differences between masculinity-threatened and masculinity-confirmed men on these measures. I also found no effects among women in the study. These findings strongly suggest that the mechanism at work is masculine overcompensation, and not simply a general conservative shift amongst masculinitythreatened men. Table 2 gives results for the various measures of interest in purchasing an SUV. As can be seen from the table, women showed no differences in their ratings of the desirability of the SUV, the amount they would pay for one, their tendency to pick the SUV as their favorite vehicle, or their belief that the SUV was the vehicle they would be most likely to buy. Men on the other hand showed differences on all items. Men rated SUVs as significantly more desirable, and were willing to pay more for one, after receiving masculinity-threatening feedback. More men in this experimental condition also chose the SUV as their favorite and the vehicle they’d be most likely to buy, though these results were marginally significant. These findings support Hypothesis 3. Men reacted to feminine gender identity feedback by displaying more interest in SUVs. We found no significant effects of the manipulation for either men or women’s interest in purchasing any of the other cars reviewed. Table 2 also gives means for persistence on the strength test. We found no effect of the manipulation on either men or women’s persistence on the strength test, thus Hypothesis 4 is rejected. These results may be partially attributable to the extremely high variance observed in demonstrated strength resulting, in part, from our failure to control for pre-existing levels of strength. Emotions I also administered the PANAS survey of positive and negative affect to explore the emotional consequences of the treatment for men and women (Watson et. al. 1988). Means for emotion items showed no differences across conditions for women with one exception. Women given feedback that they were masculine reported feeling higher levels of nervousness than did women in the control condition (t = 2.33, p < .03), but no other differences were observed across conditions for women in the study. Men in the masculinity-threat condition reported feeling more guilty, ashamed, upset and hostile than did men in the control condition (all differences p < .05). I found no statistically significant differences across conditions for men on any other emotion items. These findings confirm our earlier prediction that negative affect would increase for men, but not women, as a result of the gender identity survey feedback. Political Orientation I asked participants to report their political orientation on a single 10-point scale ranging from “Liberal” to “Conservative” prior to the administration of the gender identity survey. To be assured that our random assignment of participants to conditions was generally successful in mitigating any fundamental differences between the participants assigned to each condition, I checked to see if women assigned to the treatment condition 3 One would expect 1 of the 20 items to be significant by pure chance using a .05 significance standard. were on average as liberal as those assigned to the control condition (Ms = 5.76 and 5.92, respectively). I found that they were (t = -.26, p = .80). I also evaluated whether treatment and control condition men reported the same levels of liberalism (Ms = 6.93 and 6.52, respectively), and found that random assignment was also successful for men (t = .87, p = .39). If anything, men assigned to the treatment condition reported higher liberalism than those assigned to the control condition, an initial difference that would make the predicted effects for support for war and expressed homophobia less likely. In a reanalysis of these effects controlling for self-reported liberalism all results remain qualitatively the same, although results for men become slightly more statistically significant. Bem Sex Role Inventory I created masculinity and femininity scores for male and female participants from responses on the Bem Sex Role Inventory (BSRI). Men’s masculinity scores were higher on average than those of female participants (M’s = 5.16 and 4.80, respectively; t = 2.49, p < .02). Likewise, women’s femininity scores were higher than those of men, though this difference was only marginally significant (M’s = 4.69 and 4.44, respectively; t =