Abstract
Manhood is a precarious state that men seek to prove through the performance of masculine behaviors—including, at times, acts of aggression. Although correlational work has demonstrated a link between chronic masculine insecurity and political aggression (i.e., support for policies and candidates that communicate toughness and strength), experimental work on the topic is sparse. Existing studies also provide little insight into which men—liberal or conservative—are most likely to display increased political aggression after threats to their masculinity. The present work thus examines the effects of masculinity threat on liberal and conservative men’s tendency toward political aggression. We exposed liberal and conservative men to various masculinity threats, providing them with feminine feedback about their personality traits (Experiment 1), having them paint their nails (Experiment 2), and leading them to believe that they were physically weak (Experiment 3). Across experiments, and contrary to our initial expectations, threat increased liberal—but not conservative—men’s preference for a wide range of aggressive political policies and behaviors (e.g., the death penalty, bombing an enemy country). Integrative data analysis (IDA) reveals significant heterogeneity in the influence of different threats on liberal men’s political aggression, the most effective of which was intimations of physical weakness. A multiverse analysis suggests that these findings are robust across a range of reasonable data-treatment and modeling choices. Possible sources of liberal men’s heightened responsiveness to manhood threats are discussed.
Supplementary Information
The online version contains supplementary material available at 10.1007/s11199-023-01349-x.
Keywords: Precarious manhood, Masculinity threat, Political aggression, Political beliefs, Political ideology, Integrative data analysis (IDA), Multiverse analysis
Political aggression is on this rise. The use of harsh and demeaning rhetoric in debates (Hinck et al., 2018), the coarsening of language in elites’ and laypeople’s political communication (Cicchirillo et al., 2015), and increasingly tolerant attitudes toward political violence (Avishai, 2020) all illustrate the extent to which aggressive attitudes and behavior have become normative in the current political climate. Although this trend has seen its most alarming manifestations on the ideological right, as exemplified by the 2021 Capitol insurrection, political aggression is not solely a right-wing phenomenon; left-wing online harassment and the self-styled “dirtbag left” have also gained ground in recent years (Bowles, 2020; Stephens, 2020).
Given that violent rhetoric and behavior are most often associated with male politicians and lay-citizens, many commentators and researchers have linked the rise of political aggression to the psychology of precarious manhood (sometimes termed “fragile masculinity”; Bruni, 2017; Cauterucci, 2017; DiMuccio & Knowles, 2021; Katz, 2016; Zirin, 2017). Indeed, recent empirical work shows that men chronically high in masculine insecurity (one manifestation of precarious manhood) are more likely than secure men to support political aggression – defined here as support for policies and politicians that signal toughness, strength, and force (DiMuccio & Knowles, 2021). Yet, despite evidence implicating precarious manhood in political aggression, two factors limit our understanding of this effect: (a) most of the relevant data are correlational in nature and (b) research to date has failed to examine which men are most vulnerable to masculinity threats and, consequently, to political aggression. To address these gaps in the literature, the present research offers an experimental assessment of precarious manhood’s impact on political aggression, while attempting to identify for whom—political conservatives or liberals—this influence is strongest.
The Precarious Manhood Thesis
Decades of scholarship reveal that American males are expected to achieve their status as “real” men, and actively maintain this high-status title, or risk losing their group membership altogether (i.e., not be considered a man; Vandello & Bosson, 2013). The precarious manhood thesis posits that manhood is widely viewed, by both men and women in the United States and other societies, as requiring repeated and public social proof and validation (Vandello et al., 2008). Empirical work demonstrates that manhood is seen as a status that is both elusive (i.e., does not occur through maturation alone) and tenuous (i.e., is not automatically retained once it is achieved)—and that, if it is lost, men will engage in various compensatory actions to restore their manhood (Glick et al., 2007; Vandello & Bosson, 2013; Willer et al., 2013).
Data from history, political science, anthropology, sociology, and psychology (e.g., Bosson et al., 2021; Ducat, 2004; Gilmore, 1990; Kimmel, 2006) suggest that precarious manhood is a nearly universal phenomenon, even in countries with greater gender equality (DiMuccio et al., 2017). For instance, anthropological (Gilmore, 1990) and psychological (Bosson et al., 2021) research from over seventy countries around the world documents the widespread construction of manhood as an achieved, rather than ascribed, status. In some cultures, men are made, not born; that is, they are expected to undergo active, public, and often painful demonstrations of their societies’ masculine archetypes—thus showing their readiness to become men (Gilmore, 1990). In Western cultures, where formalized manhood rituals are rare, societal and normative pressures nonetheless require that men prove that they are “real” men and not boys or women (Connell, 1995).
Manhood’s precarity—specifically, the ever-present risk of losing the valued social status of being a man—can trigger anxiety in males, which in turn may motivate compensatory behaviors meant to reaffirm their status as “real” men (Kosakowska-Berezecka et al., 2016; Vandello & Bosson, 2013). We now turn to research examining the specific cognitions and behaviors men enact to reaffirm their manhood.
Aggressive Consequences of Manhood Threats
“Real” men are masculine men, and in any given culture there exists a dominant or hegemonic masculinity that is widely accepted by society (Vescio & Schermerhorn, 2021). Although what constitutes hegemonic masculinity changes over time, several qualities have defined American masculinity for many decades—namely, eschewal of femininity, emotionality (other than anger), and homosexuality; seeking competition, status, and achievement; being independent and confident; and taking risks and being aggressive (Brannon, 1976; Vandello et al., 2008). In American contexts, attempts to affirm manhood are likely to play out along the dimensions of American hegemonic masculinity—including displays of aggression.
Threats to manhood can cause myriad behavioral, cognitive, and attitudinal outcomes, many of which may represent means by which men seek to regain their threatened manhood status (for a review, see Vandello & Bosson, 2013). In experimental studies, manhood threat is most often induced by insinuating that male participants are somehow feminine. Common paradigms threaten men’s masculine confidence by having them engage in stereotypically feminine tasks such as braiding hair (Bosson & Vandello, 2011) or applying scented lotion (Weaver et al., 2013), giving them false test feedback suggesting that they are higher in feminine knowledge than the average man (Berke et al., 2017; Stanaland & Gaither, 2021; Vandello et al., 2008; Willer et al., 2013), or providing them with false feedback implying that they have only as much physical strength as the average women (Cheryan et al., 2015).
These (and other) experimental paradigms have shown that threatened men react with increased anxiety (Bosson et al., 2009; Caswell et al., 2014; Vandello et al., 2008), more hostility towards women (Maass et al., 2003), more aggressive cognitions (Berke et al., 2017; Stanaland & Gaither, 2021), more public discomfort, anger, shame and guilt (Vescio et al., 2021), harder punches to a punching bag (Bosson et al., 2009), administration of more shocks to a confederate (Cohn et al., 2009), and more aggression towards gay men (Bosson et al., 2012). Taken together, this work suggests that threats to manhood are especially likely to result in behaviors and cognitions that affirm one masculine trait in particular—namely, physical aggression.
Precarious Masculinity and Politics
While research has shown that a range of manhood threats can cause physical aggression, less empirical work has examined the effects of such threats on men’s political aggression—defined here as the endorsement of political stances, policies, candidates, or strategies that show toughness, strength, and force (Carian & Sobotka, 2018; DiMuccio & Knowles, 2020, 2021; Willer et al., 2013). Nonetheless, we argue that politics is an especially appealing avenue through which men can express aggression and thus reaffirm their masculinity. First, because it impacts millions of people, the political domain is highly consequential and therefore can provide a heightened sense of power and influence. Second, politics enables male voters to signal their masculinity vicariously, through support for politicians and political positions that most evince masculine traits of toughness, risk-taking, and aggression. Third, displaying aggression through politics can be an effective way of proving manliness without violating social norms against physical violence. Though politics have long been understood as a gendered and male-dominated domain (Lizotte, 2019; McDermott, 2016), only recently has the focus turned to understanding the role of precarious masculinity in shaping political attitudes and behaviors (DiMuccio & Knowles, 2021; Vescio & Schermerhorn, 2021).
Writers since the 1970s have identified masculinity as a factor in many (often destructive) decisions made by politicians attempting to prove their masculine bona fides (Ducat, 2004; Fasteau, 1974; Katz, 2016). Correlational research has since corroborated the link between masculinity and politics. For example, men tend to be more supportive than women of forceful, aggressive, and punitive policies such as the death penalty and military intervention (Hansen et al., 2020; Lizotte, 2019; Norrander, 2008)—likely due to men’s desire to behave according to their social-role expectations (Diekman & Schneider, 2010). Moreover, Americans who endorse hegemonic masculine norms and “masculine honor beliefs” (related constructs encompassing the belief that men must behave in certain ways to be considered “real” men) are more likely to support aggressive candidates (e.g., Donald Trump) than are those who do not embrace such beliefs (Martens et al., 2018; Vescio & Schermerhorn, 2021). Finally, recent work using real-world electoral outcomes suggests that a regional indicator of men’s levels of masculine anxiety—as indexed using the rate of Google searches for “precarious” search terms in different Designated Market Areas (DMAs)—predicts both support for aggressive policies and aggressive politicians’ electoral success (DiMuccio & Knowles, 2021).
Experimental research relevant to precarious manhood and politics is relatively sparse. Willer et al. (2013) found that, following a manhood threat, men showed greater support for George W. Bush’s decision to invade Iraq, as well as for his handling of the Iraq war (Willer et al., 2013). More recently, researchers found that men whose manhood had been challenged were more supportive of Donald Trump, and that this effect was mediated by the desire for a more masculine president (Carian & Sobotka, 2018).
Unanswered Questions
While existing research points to a relationship between precarious masculinity and politics, more work is needed to understand the scope and causal structure of this link. Correlational work (e.g., DiMuccio & Knowles, 2021) shows that masculine anxiety predicts a wide range of aggressive political beliefs—yet one must be cautious about claiming that masculinity anxiety causes aggressive political views. The reverse causal pattern is possible; for instance, it may be that endorsing aggressive policies and candidates serves to highlight associated notions of manhood, thus priming men’s concerns about their own masculinity.
Conversely, relevant experimental work has examined the causal effect of manhood threat on single policies (e.g., the Iraq war; Willer et al., 2013) or candidates (e.g., Donald Trump; Carian & Sobotka, 2018). While these policies and candidates are arguably “aggressive” in nature, one must examine a broad range of aggressive and nonaggressive policies to assess whether threat induces political aggression per se. For example, without assessing an array of policies that span the ideological spectrum, it is impossible to distinguish threat-induced endorsement of political aggression from a more general “conservative shift” —the tendency for people to adopt more conservative political views when the prevailing social system is threatened (Bonanno & Jost, 2006; Nail & McGregor, 2009; Nail et al., 2009). Thus, it is necessary to assess policies that, while clearly conservative or liberal, do not have strong aggressive connotations (e.g., affirmative action or climate regulation). Evidence that manhood threats increase men’s support for aggressive policies—but not for ideologically-loaded policies unrelated to aggression—would suggest that the effect of masculinity threat on political aggression cannot be explained simply by a conservative shift.
Finally, we know of no research that has investigated which men—political conservatives or liberals—are most vulnerable to manhood threat and, consequently, to increased political aggression. There is, however, reason to believe that political ideology will moderate the link between manhood threat and political beliefs. Compared to liberals, conservatives tend to hold more traditional beliefs about gender roles (including masculine ones; Feather et al., 1979; Sharrow et al., 2016), are perceived as more masculine by others (Marx Ferree, 2020; Winter, 2010), and have more masculine personality characteristics (McDermott, 2016). Compared to Democratic politicians, Republican politicians use more masculine language and imagery in their speeches, party platforms, and political strategies (Roberts & Utych, 2020).
We suggest that, because conservative men are especially likely to endorse, value, and perform hegemonic masculinity, they are also particularly invested in maintaining their status as “real” men by successfully responding to manhood threats. Indeed, men who hold more traditional beliefs about masculinity (e.g., masculine honor beliefs) are more likely to respond to manhood threats with aggression and a drive for muscularity (Saucier et al., 2015, 2018). Furthermore, men who react aggressively to manhood threats are perceived more positively by men high in masculine honor beliefs (O’Dea et al., 2018), suggesting that conservative males garner social and interpersonal benefits by responding to such threats. Informed by these findings, we initially predicted that conservative men would be more likely than their liberal counterparts to engage in increased political aggression after experiencing threats to their masculinity.
The Present Research
Three experiments investigated the effect of manhood threats on men’s support for aggressive and nonaggressive political stances. In Experiment 1, masculinity threat was induced by providing men with personality feedback suggesting they possess feminine traits; participants then rated their support for a range of aggressive and nonaggressive political policies. In Experiment 2, masculinity threat was induced by having male participants engage in a stereotypically feminine activity (nail-painting) before responding to a fictional foreign-policy scenario with either aggressive or nonaggressive actions. Finally, in Experiment 3, men were threatened by leading them to believe their hand grip was only as strong as that of the average woman; participants then responded to a subset of the policies from Experiment 1 and the foreign policy scenario from Experiment 2.
The inclusion in all experiments of policies that do not communicate aggression helped to ensure that any effect of manhood threat was restricted to aggressive policies—that is, those that might reaffirm participants’ masculinity. Moreover, the inclusion of female participants in Experiment 1 allowed us to test whether threats to womanhood (in the form masculine personality feedback) would shift women’s political attitudes; consistent with precarious manhood theory, we expected threat to leave women’s political beliefs unaffected. In all three experiments, we examined for whom—liberal or conservative men—the effect of masculinity threat on political attitudes is the strongest, predicting that conservative men would display a larger politically-aggressive response to manhood threat than would liberal men.
Transparency and Openness
We report how we determine our sample sizes in all experiments. No available data or experimental conditions were excluded from any analysis. Surveys containing all administered measures are available at https://osf.io/t8gmh/. In addition, complete data for each experiment, as well as code to reproduce all reported analyses are posted at the same OSF link. The experiments’ design and analyses were not preregistered.
Experiment 1
Experiment 1 tested the following hypotheses: (1) gender threat will significantly increase conservative men’s support for aggressive political policies and (2) the effect of threat on conservatives’ rating of aggressive policies will be significantly larger than for any other combination of ideology (conservative or liberal), policy type (aggressive or nonaggressive), and gender (male or female). Threat was induced by leading male and female participants to believe that they scored unusually high on feminine or masculine personality traits, respectively, and political aggression was indexed via scores on a battery of policy views.
Method
Participants
Three hundred and forty-one participants were electronically tested using the Prolific Academic crowdsourcing platform. The sample size was chosen to yield an 80% chance of observing a medium effect of condition (d = .50) at an alpha of .05 (two-tailed) and was determined before data analysis. Requirements for participation included being 18 years of age or older, identifying as a man or woman, and being born and raised in the United States. No participants were excluded from analyses. The study duration was approximately 15 min, and participants were paid $1.50.
Participants were 146 men (42.8%) and 195 women (57.2%), 297 of whom were aged 18 to 44 (87.1%). Two hundred and forty-eight participants identified as White (72.7%), 23 as Black (6.7%), 17 as Latino/a (5.0%), 3 as Native American (0.9%), 16 as East Asian (4.7%), 10 as South Asian (2.9%), 2 as Middle Eastern (0.6%), 3 as another race or ethnicity (0.9%), and 19 as multiple categories (5.6%). Two hundred and fifty-nine participants described their sexual orientation as heterosexual (76.0%), 12 as gay (3.5%), 10 as Lesbian (2.9%), 49 as bisexual (14.4%), and 11 as “other” (3.2%). In terms of highest educational attainment, 59 participants reported having a postgraduate degree (17.3%), 142 an undergraduate degree (41.6%), 135 a high school diploma or equivalent (39.6%), and 5 as having not finished high school (1.5%).
Politically, the sample skewed left, with 173 participants describing themselves as Democrats (50.7%), 39 as Republicans (11.4%), 120 as Independents (35.2%), and 9 as “something else” (2.6%). Participants’ average level of political conservatism was 4.4 on a 1–11 scale (SD = 2.5), with 219 identifying as liberal (64.2%), 63 choosing the scale midpoint (18.5%), and 59 identifying as conservative (17.3%).
Materials and Measures
Gender Threat Manipulation
Gender threat was induced by providing participants with false feedback on the Bem Sex Role Inventory (BSRI; Bem, 1974). The BSRI has participants rate themselves on 60 personality traits: 20 that are stereotypically masculine (e.g., assertive, self-reliant, analytical), 20 that are stereotypically feminine (e.g., affectionate, gentle, cheerful), and 20 that are gender-neutral (e.g., reliable, sincere, conscientious). All traits are positive in valence, and their order of presentation was randomized. Participants recorded their responses on a scale from 1 (never true of you) to 7 (always true of you).
Once participants completed the BSRI, a score from 0 to 100 was calculated, such that higher scores indicated more agreement with the masculine traits and lower scores indicated more agreement with the feminine traits; neutral traits were excluded from scoring. For male participants, manhood threat was induced by subtracting 30 points from their actual score, thereby placing them closer in personality to a stereotypical woman. For female participants, womanhood threat was induced by adding 30 points to their actual score, thereby placing them closer in personality to a stereotypical man. In this way, feedback provided to participants was anchored on their actual levels of (stereotypical) masculinity and femininity, helping to ensure that no participant received scores vastly—and thus unrealistically—discrepant from their actual responses. Scores could be no lower than 3 or higher than 97.
Participants in the threat condition saw their adjusted score juxtaposed with the putative score of the average person of their gender (80 for men and 32 for women). Even the most masculine man in the threat condition would receive a score of 70 (100 minus 30), which is below the average man’s score. In light of research showing that women tend to be higher in androgyny than men (Donnelly & Twenge, 2017), we adjusted the average women’s score to be further away from 0 (i.e., 32) than the average man’s score was from 100 (i.e., 80). Participants in the no-threat condition received no BSRI feedback, but all participants received the same instructions prior to taking the test. In a pilot test, participants were asked after the manipulation whether they suspected the true purpose of the research. None correctly guessed the hypothesis or purpose of the manipulation.
Support for Aggressive and Nonaggressive Policies
Policy support was measured by dividing a list of 17 foreign and domestic political policies into aggressive and nonaggressive categories; this distinction was guided by a priori considerations and previous research (DiMuccio & Knowles, 2021; Lizotte, 2017), and confirmed using principal components analysis (PCA; see Results). Participants rated their agreement with each policy on a scale from 1 (strongly oppose) to 7 (strongly support). There were nine aggressive policies (stand-your-ground laws, build the wall, ban Muslim immigration, death penalty, presidential war powers, increase military spending, use of torture, troops to Middle East, use of military force) and eight nonaggressive policies (marriage equality, affirmative action, police reform, Obamacare, climate regulation, pay equality, gun control, social welfare programs).
Political Orientation
Participants were asked to rate their overall political orientation on a scale 1 (extremely liberal) to 11 (extremely conservative).
Demographic Information
Participants were administered a standard demographic questionnaire with questions about their age, gender identity, sexual orientation, nationality, region of the country, ethnic and racial background, educational history, employment status, socioeconomic status, religion, and religiosity.
Procedure
After consenting to take part in the study, participants were asked to indicate their gender and complete the BSRI. Participants were then randomly assigned to the threat or no-threat condition. In the threat condition, participants were informed that the BSRI measures the degree of masculinity or femininity of their personality. The meaning of higher and lower scores was explained, after which participants were presented with their adjusted score along with the supposed “average man’s” and “average woman’s” scores. No BSRI feedback was provided in the no-threat condition. Participants then completed the policy support questions, followed by the ideology and demographic items. Finally, participants were fully debriefed and dismissed.
Results
Summary statistics for, and correlations between, variables assessed in Experiment 1 are displayed in Table 1.
Table 1.
Summary Statistics For, and Correlations Between, Variables Assessed in Experiment 1 (N = 341)
| Variable | M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Gender | .57 | .50 | — | −.14 | .21 | .02 | .10 | .18 | −.06 | .07 | −.18 | .33 | .03 | −.11 | −.24 | −.20 | −.15 | −.08 | −.19 | .15 | −.16 |
| 2. Conservatism | 4.37 | 2.48 | — | −.58 | .44 | −.44 | −.30 | −.53 | −.55 | .47 | −.35 | −.49 | .48 | .47 | .29 | .46 | .61 | .57 | −.49 | .36 | |
| 3. Marriage equality | 5.78 | 1.88 | — | −.24 | .35 | .26 | .39 | .47 | −.27 | .28 | .33 | −.32 | −.34 | −.22 | −.23 | −.54 | −.46 | .30 | −.29 | ||
| 4. Death penalty | 3.57 | 1.96 | — | −.16 | −.18 | −.28 | −.32 | .39 | −.20 | −.23 | .36 | .49 | .19 | .37 | .35 | .35 | −.35 | .32 | |||
| 5. Affirmative action | 4.72 | 1.55 | — | .28 | .40 | .38 | −.18 | .20 | .34 | −.19 | −.17 | −.12 | −.26 | −.31 | −.31 | .38 | −.24 | ||||
| 6. Police reform | 5.78 | 1.56 | — | .25 | .17 | −.28 | .22 | .16 | −.28 | −.26 | −.14 | −.08 | −.24 | −.21 | .31 | −.19 | |||||
| 7. Obamacare | 5.12 | 1.85 | — | .50 | −.30 | .24 | .49 | −.21 | −.25 | −.07 | −.31 | −.48 | −.38 | .40 | −.20 | ||||||
| 8. Climate regulation | 5.69 | 1.78 | — | −.35 | .39 | .44 | −.32 | −.38 | −.12 | −.38 | −.46 | −.41 | .40 | −.17 | |||||||
| 9. Presidential war powers | 1.83 | 1.32 | — | −.27 | −.35 | .46 | .50 | .36 | .31 | .47 | .45 | −.32 | .36 | ||||||||
| 10. Pay equality | 6.08 | 1.44 | — | .30 | −.17 | −.35 | −.09 | −.29 | −.29 | −.34 | .28 | −.16 | |||||||||
| 11. Gun control | 5.31 | 2.01 | — | −.29 | −.22 | −.13 | −.41 | −.47 | −.42 | .35 | −.12 | ||||||||||
| 12. Increase military spending | 2.89 | 1.77 | — | .52 | .47 | .37 | .44 | .43 | −.31 | .48 | |||||||||||
| 13. Use of torture | 2.63 | 1.83 | — | .40 | .45 | .45 | .49 | −.34 | .44 | ||||||||||||
| 14. Troops to Middle East | 3.34 | 1.73 | — | .22 | .24 | .22 | −.18 | .47 | |||||||||||||
| 15. Stand-your-ground laws | 3.91 | 1.89 | — | .36 | .39 | −.30 | .25 | ||||||||||||||
| 16. Build the wall | 2.30 | 1.97 | — | .67 | −.39 | .36 | |||||||||||||||
| 17. Ban Muslim immigration | .54 | .98 | — | .43 | .43 | ||||||||||||||||
| 18. Social welfare programs | .38 | .67 | — | .22 | |||||||||||||||||
| 19. Use of military force | 3.66 | 1.65 | — |
Note. Correlations in bold are significant at p < .05. Conservatism ranged from 1–11 (midpoint 6) and policy attitudes (in italics) from 1–7 (midpoint 4). Gender is coded such 0 = male and 1 = female; the mean of gender is interpreted as proportion female
Effect of Threat Manipulation on Conservatism
Because our predicted moderator, participant conservatism, was measured after the administration of the threat manipulation, we sought to confirm that the manipulation itself did not affect conservatism. To test this, we regressed conservatism on feedback condition, participant gender, and the interaction between these two variables. This analysis yielded no significant main effect of threat, B = .018, SE B = .024, t = .73, p = .467, 95% CI [-.030, .066], or interaction between threat and gender, B = .022, SE B = .049, t = .45, p = .655, 95% CI [-.075, .118]. These null findings freed us to examine the manner in which conservatism moderated the impact of gender threat on participants’ policy attitudes.
Aggressive and Nonaggressive Policy Composites
To test our primary predictions, we first created aggressive and nonaggressive policy composites. We had a priori expectations (informed by DiMuccio & Knowles, 2021) as to which policies were “aggressive” (e.g., use of torture, capital punishment) and which were “nonaggressive” (e.g., climate regulation, affirmative action). In DiMuccio and Knowles (2021), both liberal and conservative men and women agreed that the aggressive and nonaggressive policies were indeed aggressive and nonaggressive. Nonetheless, we sought empirical support for this distinction. We therefore subjected all policies to a principal components analysis (PCA). Inspection of the scree plot suggested an inflection point in eigenvalues between factors 2 and 3; thus, 2 factors were extracted (see Table S1 in the online supplement). Because the unrotated solution failed to produce two distinct groups of items, we applied a varimax rotation to the PCA results. The rotated solution produced two item clusters mapping well onto the aggressive–nonaggressive distinction. However, four items—stand-your-ground laws, build the wall, ban Muslim immigration, and police reform—displayed substantial cross-loadings and were thus excluded from further analysis. (Please refer to the Multiverse Analysis section for a systematic assessment of alternative data treatments across studies.) All remaining aggressive items were averaged to form an aggressive policy composite (α = .80) and all remaining nonaggressive items were averaged to form a nonaggressive policy composite (α = .80).
Analyses
In this experiment, participant ideology (conservatism) was a continuous predictor, participant gender (male vs. female) and personality feedback (threat vs. no feedback) were between-participants factors, and policy type (aggressive vs. nonaggressive) was a within-participant factor. The personality feedback manipulation was coded such that -.5 = no feedback and .5 = threat feedback, policy type such that -.5 = aggressive policies and .5 = nonaggressive policies, and gender such that male = -.5 and female = .5. Participants’ conservatism scores were centered around the sample mean. The aggressive and nonaggressive policy composites were rescaled to range from 0 (the minimum possible score) to 1 (the maximum possible score).
Because the data were nested, with two scores per participant (an aggressive policy score and a nonaggressive policy score), a mixed-model linear regression analysis was conducted using the mixed command in Stata 17 (StataCorp, 2021). This analysis had 80% power to detect an effect size of f = .30. Participants’ composite scores were regressed on gender, policy type, feedback condition and all interactions between these terms. A random intercept and random slope of policy type were specified, and the covariance between the random effects was freely estimated. The results of this analysis are shown in Table 2. The regression equation was used to plot estimates of policy attitudes among conservative and liberal men and women in each personality feedback condition, with “conservative” and “liberal” defined as 1 SD above and below the ideology mean, respectively (Fig. 1). Visual inspection of the regression predictions suggests that gender threat may have increased liberal men’s endorsement of aggressive policies—but that threat had no significant effect on liberal men’s endorsement of nonaggressive policies, on women’s support for either aggressive or nonaggressive policies, or on conservative men’s endorsement of either policy type.
Table 2.
Results of Mixed-Model Linear Regression Analysis in Experiment 1 (N = 341)
| 95% CI | |||||||
|---|---|---|---|---|---|---|---|
| Parameter | B | SE B | z | p | LB | UB | f2 |
| Fixed Effects | |||||||
| Gender (G) | −.010 | .011 | −.97 | .333 | −.032 | .011 | .001 |
| Conservatism (C) | −.048 | .024 | −1.96 | .050 | −.096 | .000 | .005 |
| Policy Type (P) | .403 | .012 | 33.12 | .000 | .379 | .426 | 1.791 |
| BSRI Feedback (F) | .009 | .011 | .86 | .387 | −.012 | .031 | .001 |
| G × C | −.139 | .049 | −2.85 | .004 | −.234 | −.043 | .010 |
| G × P | .076 | .024 | 3.14 | .002 | .029 | .124 | .016 |
| G × F | −.006 | .022 | −.29 | .770 | −.049 | .036 | .0001 |
| C × P | −1.127 | .055 | −20.65 | .000 | −1.234 | −1.020 | .696 |
| C × F | −.061 | .049 | −1.25 | .210 | −.156 | .034 | .002 |
| P × F | −.056 | .024 | −2.30 | .021 | −.104 | −.008 | .009 |
| G × C × P | −.045 | .109 | −.42 | .677 | −.259 | .169 | .0003 |
| G × C × F | .209 | .097 | 2.15 | .032 | .018 | .400 | .006 |
| G × P × F | .060 | .049 | 1.23 | .218 | −.035 | .155 | .002 |
| C × P × F | .082 | .109 | .75 | .451 | −.132 | .296 | .001 |
| G × C × P × F | −.384 | .218 | −1.76 | .079 | −.812 | .044 | .005 |
| Intercept | .534 | .005 | 98.57 | .000 | .524 | .545 | |
| Random Effects | |||||||
| Var (P) | .043 | 1.760 | .000 | 3.57 × 1033 | |||
| Var (Intercept) | .008 | 0.440 | .000 | 1.78 × 1043 | |||
| Cov (Intercept, P) | −.004 | 0.001 | −.007 | −.002 | |||
Note. BSRI Bem Sex Role Inventory, CI confidence interval, LB lower bound, UB upper bound. f2 calculation performed using the method described by Selya et al. (2012)
Fig. 1.
Men and Women’s Policy Endorsement as a Function of BSRI Feedback, Policy Type, and Political Ideology in Experiment 1
We next ran focused tests examining the impact of threat on policy support for every combination of gender, policy type, and feedback using the margins command in Stata 17 (Table 3). Contrary to prediction, but consistent with inspection of Fig. 1, the simple effect of threat on male conservatives’ ratings of aggressive policies was not significantly different from zero (p = .717). Instead, only liberal men endorsed aggressive policies more strongly after threat (p = .003). As predicted, threat did not affect men’s or women’s endorsement of nonaggressive policies; nor did threat effect women’s ratings of either type of policy.
Table 3.
Simple Effects of Ggender Threat and Linear Hypothesis Tests in Experiment 1 (N = 341)
| 95% CI | |||||||
| B | SE B | z | p | LB | UB | Glass’s Δ | |
| Liberal Men / Aggressive Policy (LMA) | .124 | .042 | 2.94 | .003 | .041 | .206 | .582 |
| Liberal Men / Nonaggressive Policy (LMN) | −.024 | .035 | −.69 | .488 | −.092 | .044 | −.132 |
| Conservative Men / Aggressive Policy (CMA) | −.013 | .035 | −.36 | .717 | −.082 | .057 | −.060 |
| Conservative Men / Nonaggressive Policy (CMN) | −.037 | .029 | −1.27 | .206 | −.094 | .020 | −.203 |
| Liberal Women / Aggressive Policy (LWA) | −.003 | .030 | −.10 | .921 | −.063 | .057 | −.016 |
| Liberal Women / Nonaggressive Policy (LWN) | −.004 | .025 | −.17 | .865 | −.053 | .045 | −.021 |
| Conservative Women / Aggressive Policy (CWA) | .041 | .034 | 1.21 | .228 | −.026 | .109 | .218 |
| Conservative Women / Nonaggressive Policy (CWN) | −.009 | .028 | −.33 | .741 | −.065 | .046 | −.046 |
| X2 | df | p | |||||
| LMA vs. LMN | 6.60 | 1 | .010 | ||||
| LMA vs. CMA | 5.93 | 1 | .015 | ||||
| LMA vs. CMN | 9.89 | 1 | .002 | ||||
| LMA vs. LWA | 5.95 | 1 | .015 | ||||
| LMA vs. LWN | 6.84 | 1 | .009 | ||||
| LMA vs. CWA | 2.30 | 1 | .129 | ||||
| LMA vs. CWN | 6.90 | 1 | .009 | ||||
| LMA vs. Σ (LMN + CMA + CMN + LWA + LWN + CWA + CWN)/7 | 8.71 | 1 | .003 | ||||
Note. CI confidence interval, LB lower bound, UB upper bound. “Liberal” and “conservative” refer to participants 1 SD below and above the mean level of political conservatism. Glass’s Δ calculated according to the method described in Shaw (2022)
Having observed a significant simple effect of threat only for liberal men rating aggressive policies, we proceeded to run linear hypothesis (Wald) tests (using the test command in Stata 17) to examine whether this effect was significantly greater than that for other combinations of gender and policy type. As can be seen in Table 3, the threat effect for liberal men rating aggressive policies was significantly larger than for every other “cell” in the design except conservative women rating aggressive policies (p = .129). Moreover, the effect on liberal men rating aggressive policies was significantly greater than all other combinations taken averaged together (p = .003).
Discussion
The results of Experiment 1 caught us by surprise. Contrary to our hypothesis, feminine personality feedback did not increase aggressive policy endorsement among conservative men. Instead, it did so among liberal men (Fig. 1 and Table 3). As expected, however, no threat effects emerged for endorsement of nonaggressive policies or among female participants (regardless of participant ideology). These results suggest that it may be liberal—rather than conservative—men who display heightened levels of political aggression when their manhood is called into question.
We sought in the next experiment to assess whether the pattern observed in Experiment 1 is reliable. Experiment 2 conceptually replicates Experiment 1 using a different threat manipulation (a nail-painting activity), sample (New York City residents, including students and non-students), and assessment of political views (responses to a fictional foreign-policy vignette). As no feedback effects emerged for women in Experiment 1, only male participants were recruited in Experiment 2.
Experiment 2
Experiment 2 tested the following hypotheses: (1) manhood threat will significantly increase liberal men’s support for aggressive political policies and (2) the effect of threat for liberals rating aggressive policies will be significantly larger than for any other combination of ideology (conservative or liberal) and policy type (aggressive or nonaggressive). Threat was induced by having male participants engage in a stereotypically feminine task—specifically activity—painting their nails pink—and political aggression was indexed via responses to a fictional foreign-policy scenario.
Method
Participants
Two hundred and thirty-five self-identified men were recruited from a participant pool that includes students at New York University as well as members of the broader community. The sample size was chosen to yield an 80% chance of observing a medium effect of condition (d = .5) at an alpha of .05 (two-tailed) and was determined before data analysis. Requirements for participation were identifying as a man and being 18 years or older. The study duration was approximately 30 min, and participants were paid either $20 or $25 (they received a “bonus” of $5 for not canceling, rescheduling or coming late).
Participants ranged in age from 18 to 60 (M = 22.6, SD = 7.6). Eighty-four participants identified as White (35.7%), 18 as Black (7.7%), 42 as Latino/a (17.9%), 68 as Asian American (28.9%), and 23 as another race or ethnicity (9.8%). Two hundred and twenty-five participants described their sexual orientation as heterosexual (95.7%), 2 as gay (0.8%), 6 as bisexual (2.6%), and 2 as “other” (0.8%). In terms of highest educational attainment, 12 participants reported having a postgraduate degree (5.1%), 42 an undergraduate degree (17.9%), 179 a high school diploma or equivalent (76.2%), and 2 as not having finished high school (0.8%).
The sample skewed left politically, with 123 participants describing themselves as Democrats (52.3%), 21 as Republicans (8.9%), 70 as Independents (29.8%), and 21 as “other” (8.9%). Participants’ average level of political conservatism was 4.8 on a 1–11 scale (SD = 2.1), with 145 identifying as liberal (61.7%), 43 choosing the scale midpoint (18.3%), and 47 identifying as conservative (20.0%).
Materials and Measures
Gender Threat Manipulation
To induce manhood threat, participants in the threat condition were randomly assigned to paint all ten of their fingernails with pink nail polish from the original bottle—a stereotypically feminine task. Participants in the no-threat condition painted ten circles on a sheet of paper with white paint, which was provided to them in a black bottle with no markings. Given the novelty of the task, we asked participants in a pilot study several follow-up questions in which they were asked to guess the true purpose of the study and the hypothesis. No participants correctly guessed the hypothesis.
Foreign Policy Scenario
To measure political aggression, we created a “warring countries” vignette. The 260-word scenario describes an unstable political situation between two fictional countries (Bombola and Liga) on the brink of war. The participants were asked to imagine that they were the president of one of the countries (the name of which was counterbalanced). Participants learned that their country shared a border with the opposing nation, and that they were being threatened with possible military force. The vignette was written so that there appeared to be no clear best course of action. See the Appendix for the full text of the vignette. After reading the vignette, participants were then asked the following questions in fixed order:
In light of the information provided, how likely do you think it is that war is imminent? (1 = extremely unlikely to 6 = extremely likely)
You have received intelligence that Bombola [Liga] is very inclined to place more troops in the contested border regions. In light of such news, how likely would you be to hurry to deploy troops first in order to gain an advantage? (1 = extremely unlikely to 6 = extremely likely)
Dropping bombs in an unpopulated territory in Bombola [Liga] may intimidate them so as to not attack. How likely would you be to carry out such a show of capacity of military force? (1 = extremely unlikely to 6 = extremely likely)
Your forces have managed to capture a spy from Bombola [Liga]. It is very likely that he has highly valuable information that will be of critical help to your own country to prepare for a possible confrontation. However, his interrogation has failed to extract anything so far. In consideration of all that may be at stake, how likely would you be to permit the use of torture in order to obtain vital information from the spy? (1 = extremely unlikely to 6 = extremely likely)
Your forces have captured citizens of Bombola [Liga], who were caught attempting to set bombs in a large city in your country as an act of terrorism. How likely would you be to endorse their capital punishment as a demonstration of force/strength? (1 = extremely unlikely to 6 = extremely likely)
Some of your advisers say that you should deploy troops to the border region as a show of strength. How many troops would you deploy? (0 to 90,000 in increments of 20,000)
Despite the tensions with Bombola [Liga], your country still has a number of trade deals with them. In light of the current conflict, how likely would you be to impose economic sanctions on Bombola [Liga], such as ceasing imports and exports of goods, in order to deter them from war? (1 = extremely unlikely to 6 = extremely likely)
How likely would you be to engage in back-door diplomacy/negotiations in order to prevent a possible conflict? Some of you advisers argue that it might weaken the country’s position (1 = extremely unlikely to 6 = extremely likely)
How likely would you be to offer a peace deal? Some of your advisers argue that it would give the appearance of weakness, both to your own citizens as well as to Bombola [Liga]. (1 = extremely unlikely to 6 = extremely likely)
Questions 2 through 6 (deploy troops, dropping bombs, torture, capital punishment, number of troops) were meant to assess the participants’ inclination toward a politically-aggressive approach to the foreign-policy dilemma. Questions 7 through 9 (sanctions, negotiations, peace deal) were meant to assess nonaggressive approaches to the scenario. Question 1 (likelihood of war) was included for exploratory purposes and was not intended as an index of policy approach.
Political Orientation and Demographic Information
Political orientation and demographic questions were the same as those in Experiment 1.
Procedure
Upon arriving at the lab, participants were told that they would be completing two separate studies. The first study was described as a research project being conducted for a marketing class at New York University, with the aim of improving the usability of an (ostensibly) random product that participants would be testing. Participants learned that the product test would be filmed, supposedly for later viewing in the marketing class.
After providing informed consent, the experimenter retrieved a black box from the cabinet filled with folded slips of paper. The participant was then asked to pick a slip of paper indicating which product they would be testing. Participants in the threat condition were presented with a box that (unbeknownst to them) only had slips of paper with the words “nail polish” written on them, whereas the slips in the no-threat condition always had the word “paint” on them. The participant read the slip aloud and the experimenter retrieved the correct product. The experimenter then went to the camera, showed the participants the viewfinder, asked them if they would be able to see themselves in it as they used the product, and turned it on.
In the threat condition, participants were asked to paint each of their ten fingernails with the nail polish, paying special attention to the look and feel of the polish and applicator brush. After they finished painting their nails, participants answered a series of “product usability” questions meant to reinforce the experimental cover story. After turning off the camera, participants in the threat condition were told that the experimenters also wished to assess the look and feel of the nail polish once it was dry, and so that in the meantime they should complete the second—ostensibly unrelated—study on the computer. In this way, the experimental manipulation remained salient throughout the completion of the dependent measures.
In the no-threat condition, participants were given a black bottle with white paint in it and a sheet of paper with 10 circles on it; the circles were situated approximately where fingernails would be. Participants were given identical instructions to those in the threat condition, except that references to nail polish and fingernails were replaced with references to the paint and the circles. No-threat participants were asked the same usability questions after the task.
After participants completed the product-testing task, they were asked to direct their attention to a laptop placed in front of them and complete an “Attitudes and Opinions” survey. In the survey, they read and responded to the foreign-policy vignette and completed political orientation and demographic items. They were then fully debriefed, paid, and dismissed.
Results
Summary statistics for, and correlations between, variables assessed in Experiment 2 are displayed in Table 4.
Table 4.
Summary Statistics For, and Correlations Between, Variables Assessed in Eexperiment 1 (N = 235)
| Variable | M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Conservatism | 4.85 | 2.09 | — | .13 | .19 | .37 | .29 | .24 | .01 | −.08 | −.20 |
| 2. Mass troops at border | 4.48 | 1.24 | — | .28 | .22 | .15 | .34 | .24 | .00 | .05 | |
| 3. Bomb enemy territory | 2.52 | 1.53 | — | .29 | .23 | .17 | .18 | −.07 | −.04 | ||
| 4. Torture enemy spy | 3.44 | 1.70 | — | .43 | .25 | .02 | −.09 | −.23 | |||
| 5. Execute enemy spy | 3.63 | 1.78 | — | .18 | .09 | −.07 | −.10 | ||||
| 6. Number of troops deployed | 2.74 | 1.32 | — | .09 | −.12 | −.05 | |||||
| 7. Impose economic sanctions | 3.93 | 1.63 | — | .17 | .01 | ||||||
| 8. Pursue backdoor negotiations | 4.72 | 1.25 | — | .09 | |||||||
| 9. Seek peace deal | 4.52 | 1.24 | — |
Note. Correlations in bold are significant at p < .05. Conservatism ranged from 1–11 (midpoint 6) and vignette responses (in italics) from 1–6
Effect of Threat Manipulation on Conservatism
As in the previous experiment, participants’ levels of conservatism were measured after the administration of the threat manipulation. We therefore sought to confirm that the manipulation did not affect conservatism by regressing conservatism on threat condition. This analysis yielded no significant main effect of threat, B = -.009, SE B = .025, t = -.36, p = .716, 95% CI [-.058, .040], freeing us to examine conservatism as a moderator of any threat effect on policy attitudes.
Aggressive and Nonaggressive Vignette Composites
Although we had a priori expectations as to which actions were aggressive (mass troops at border, bomb enemy territory, torture enemy spy, execute enemy spy, and a large number of troops deployed) and which were nonaggressive (impose economic sanctions, pursue backdoor negotiations, and seek peace deal), we sought empirical support for this distinction. To this end, we subjected all vignette items to a principal components analysis (PCA). Inspection of the scree plot suggested an inflection point in eigenvalues between factors 2 and 3; hence, 2 factors were extracted. Lending credence to our distinction between policy types, all aggressive items loaded more strongly on the first extracted factor, whereas all nonaggressive items loaded more strongly on the second extracted factor (See Table S2 in the online supplement). We therefore averaged all aggressive items to form an aggressive policy composite (α = .63) and averaged all nonaggressive items to form a nonaggressive policy composite (α = .23).
We note that while both alphas are below the commonly cited cutoff of .70, there is no universally accepted cutoff for alpha (Lance et al., 2006). Some scholars have argued that an alpha larger than .60 is acceptable for a construct with few items (Ursachi et al., 2015). The nonaggressive composite has an unreliable alpha but, given that it was only meant as a comparison for the aggressive composite, we were not overly concerned about this. In addition, we address this limitation in our multiverse analysis, below.
Analyses
In this experiment, participant ideology (conservatism) was a continuous predictor, manhood threat (threat vs. no threat) was a between-participants factor, and response type (aggressive vs. nonaggressive) was a within-participant factor. The threat manipulation was coded such that -.5 = no threat and .5 = threat and policy type was coded such that -.5 = aggressive and .5 = nonaggressive. Participants’ conservatism scores were centered around the sample mean. The aggressive and nonaggressive policy composites were rescaled to range from 0 (the minimum possible score) to 1 (the maximum possible score).
Because the data were nested, with two scores per participant (an aggressive vignette score and a nonaggressive vignette score), a mixed-model linear regression analysis was conducted using the mixed command in Stata 17. This analysis had 80% power to detect an effect size of size f = .40. Participants’ composite scores were regressed on policy type, feedback condition, and the interaction between these terms. A random intercept and random slope of policy type were specified, and the covariance between the random effects was freely estimated. The results of this analysis are shown in Table 5. The regression equation was used to plot estimates of vignette responses among conservative and liberal men in each experimental condition, with “conservative” and “liberal” defined as 1 SD above and below the ideology mean, respectively (Fig. 2). Visual inspection of the regression predictions suggests that manhood threat may have increased liberal men’s aggressive responses to the foreign-policy vignette, whereas there is little evidence that threat affected liberals’ nonaggressive responses or conservatives’ endorsement of either policy type.
Table 5.
Results of Mmixed-Mmodel Linear Regression Aanalysis in Experiment 2 (N = 235)
| 95% CI | |||||||
|---|---|---|---|---|---|---|---|
| Parameter | B | SE B | z | p | LB | UB | f2 |
| Fixed Effects | |||||||
| Conservatism (C) | .152 | .044 | 3.44 | .001 | .066 | .239 | .025 |
| Policy Type (P) | .205 | .016 | 13.16 | < .001 | .175 | .236 | .404 |
| Threat (T) | .007 | .017 | .42 | .678 | −.026 | .039 | .003 |
| C × P | −.547 | .083 | −6.58 | < .001 | −.710 | −.384 | .102 |
| C × T | −.032 | .089 | −.36 | .720 | −.205 | .142 | .0003 |
| P × T | −.062 | .031 | −2.00 | .045 | −.123 | −.001 | .010 |
| C × P × T | .326 | .166 | 1.96 | .050 | −.000 | .652 | .009 |
| Intercept | .575 | .008 | 69.26 | < .001 | .559 | .592 | |
| Random Effects | |||||||
| Var (P) | .049 | 2.172 | .000 | 1.35 × 1036 | |||
| Var (Intercept) | .014 | 0.543 | .000 | 3.08 × 1030 | |||
| Cov (Intercept, P) | −.001 | 0.002 | −.005 | .003 | |||
Note. CI confidence interval, LB lower bound, UB upper bound. f2 calculation performed using the method described by Selya et al. (2012)
Fig. 2.
Men’s Foreign Policy Endorsement as a Function of Masculinity Threat, Policy Type, and Political Ideology in Experiment 2
We next ran focused tests examining the impact of threat on vignette responses for every combination of policy type and threat condition using the margins command in Stata 17 (Table 6). Replicating Experiment 1, and consistent with visual inspection of Fig. 2, the simple effect of gender threat on male liberals’ ratings of aggressive vignette responses was significant—whereas threat did not significantly affect liberals’ endorsement of nonaggressive responses or conservatives’ preference for either response type.
Table 6.
Simple Eeffects of Gender Tthreat and Linear Hypothesis Ttests in Experiment 2 (N = 235)
| 95% CI | |||||||
| B | SE B | z | p | LB | UB | Glass’s Δ | |
| Liberal Men / Aggressive Policy (LMA) | .075 | .033 | 2.280 | .023 | .011 | .139 | .382 |
| Liberal Men / Nonaggressive Policy (LMN) | −.049 | .032 | −1.540 | .123 | −.112 | .013 | −.283 |
| Conservative Men / Aggressive Policy (CMA) | .001 | .033 | .040 | .971 | −.063 | .065 | .006 |
| Conservative Men / Nonaggressive Policy (CMN) | .001 | .032 | .020 | .985 | −.062 | .064 | .004 |
| X2 | df | p | |||||
| LMA vs. LMN | 7.86 | 1 | .005 | ||||
| LMA vs. CMA | 2.52 | 1 | .112 | ||||
| LMA vs. CMN | 2.62 | 1 | .106 | ||||
| LMA vs. (Σ (LMN + CMA + CMN)/3 | 5.90 | 1 | .015 | ||||
Note. CI confidence interval, LB lower bound, UB upper bound. “Liberal” and “conservative” refer to participants 1 SD below and above the mean level of political conservatism. Glass’s Δ calculated according to the method described in Shaw (2022)
Having observed a significant simple effect of threat only for liberal men rating aggressive policies, we proceeded to run linear hypothesis (Wald) tests (using the test command in Stata 17) to examine whether this effect was significantly greater than that for every other combination gender and policy type. As can be seen in Table 6, the threat effect for liberal men rating aggressive policies was significantly larger than for liberal men rating nonaggressive policies. The difference in the effect of threat between liberal men rating aggressive policies and conservative men rating aggressive policies (p = .112) or nonaggressive policies (p = .106) failed to reach significance. However, the threat effect for liberal men rating aggressive policies differed significantly from the average threat effect across the remaining three “cells” of the study combined (p = .015).
Discussion
Replicating Experiment 1, the threat manipulation significantly increased liberal men’s endorsement of aggressive actions in the foreign-policy vignette (Fig. 2 and Table 6). No threat effect emerged among conservative men or among liberal men rating nonaggressive policies. In the next experiment, we attempted once more to replicate the effect of masculinity threat on liberal men’s support for aggressive policies. We did so by using a new manipulation of manhood threat (false feedback on a test of grip strength), a sample of New York City residents, and a mix of policy and vignette items from the previous studies.
Experiment 3
Our goal in Experiment 3 was to replicate the findings of Experiments 1 and 2. Thus, our hypotheses were as follows: (1) masculinity threat will significantly increase liberal men’s support for aggressive political policies and (2) the effect of threat for liberals rating aggressive policies will be significantly larger than for any other combination of ideology (conservative or liberal), and policy or response type (aggressive or nonaggressive). Threat was induced by providing men with false feedback on a test of physical strength, and political aggression was indexed via policy endorsement (as in Experiment 1) and responses to a fictional foreign-policy scenario (as in Experiment 2).
Method
Participants
The sample consisted of 64 self-identified men walking in Washington Square Park near New York University. Requirements for participation included identifying as a man, being American, and being 18 years of age or older. An a priori sample size of 200 participants was decided prior to data collection, however, as data collection commenced during Spring 2020, it was unfortunately cut short due to the Covid-19 pandemic. A sensitivity analysis nonetheless revealed that the sample yielded 90% power to detect a small-to-medium effect size (f = .20). The study duration was approximately 15 min and participants were paid $5.
Participants ranged in age from 19 to 63 (M = 26.8, SD = 9.5). Twenty-four participants identified as White (37.5%), 9 as Black (14.1%), 6 as Latino/a (9.4%), 6 as East Asian (9.4%), 5 as South Asian (7.8%), 1 as Middle Eastern (1.6%), and 13 who specified multiple categories (20.3%). Sixty participants described their sexual orientation as heterosexual (93.8%), 3 as gay (4.7%), and 1 as bisexual (1.6%). Thirty-three participants reported being students (51.6%) and 33 nonstudents (48.4%).
The sample skewed left politically, albeit to a lesser extent than in the previous experiments. Twenty-eight participants describing themselves as Democrats (43.8%), 14 as Republicans (21.9%), 21 as Independents (32.8%), and 1 as “other” (1.6%). Participants’ average level of political conservatism was 5.1 on a 1–11 scale (SD = 2.7), with 35 identifying as liberal (54.7%), 11 choosing the scale midpoint (17.2%), and 18 identifying as conservative (28.1%).
Materials and Measures
Gender Threat Manipulation
To induce masculinity threat, we used a modified version of Cheryan et al.'s (2015) hang-grip paradigm. Physical strength is an important aspect of masculinity (Frederick et al., 2017; Lee-Won et al., 2017) and handgrip strength is specifically associated with masculine qualities in male college students, such as aggression and sexual promiscuity (Gallup et al., 2007). As such, telling men they have weak hand strength was expected to be an especially powerful threat to men’s masculinity.
Participants were given a hand dynamometer and asked to hold it in front of them with their dominant hand, gripping it as hard as they could for about 3 s. In the threat condition, unbeknownst to the participant, the unit of measurement was toggled to kilograms, thus showing a lower number for a given grip pressure than if had it been set to pounds. Conversely, in the no-threat condition, the unit of measurement was switched to pounds, thus showing a higher number for a given grip pressure than if had it been set to kilograms. Participants were then asked to announce their score so that the researcher could record it on an iPad. The researcher next showed participants a chart with overlapping distributions, ostensibly depicting the scores of previous male and female participants. In the threat condition, the researcher pointed to the chart and showed the men that their hand grip was closest to the average woman (the kilogram reading). In the no-threat condition, the researcher pointed to the chart to show the men that their hand grip resembled the average man’s (the pound reading).
Support for Aggressive and Nonaggressive Policies
Political aggression was measured, in part, using the same list of policies as in Experiment 1, except three policies (presidential war powers, police reform, stand-your-ground laws) were removed in the present experiment. Note that we administered two items (e.g., build the wall, ban Muslim immigration) that were ultimately excluded from the political aggression composite in Experiment 1. These items were nonetheless omitted from analysis in Experiment 3 for the sake of consistency.
Foreign Policy Scenario
Political aggression was also measured using the foreign policy vignette from Experiment 2. While the questions were the same as in the previous experiment, the text of the scenario was abridged (by removing unnecessary detail and filler words) to reduce the duration of the study.
Political Orientation and Demographic Information
Political orientation and demographic questions were the same as those in Experiments 1 and 2.
Procedure
Once approached by the research assistant, participants were told that the study was about the relationship between physical strength and attitudes. Participants gave their informed consent before being randomly assigned to a condition and engaging in the handgrip strength task. After receiving strength feedback, participants were provided with an iPad and asked to fill out a survey containing the dependent measures and demographic questions. They were then fully and thoroughly debriefed, paid $5, and dismissed. Given the very brief nature of the study, we choose not to ask any follow-up questions as we did with the previous experiments. We deemed this omission to be acceptable given the established manipulation that was used.
Results
Summary statistics for, and correlations between, variables assessed in Experiment 3 are displayed in Table 7.
Table 7.
Summary Statistics Ffor, and Correlations Between, Variables Assessed in Experiment 3 (N = 64)
| Variable | M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Conservatism | 5.11 | 2.70 | — | −.75 | .60 | −.65 | −.68 | −.71 | −.52 | .72 | .58 | .51 | .59 | .79 | .72 | −.77 | .46 | .34 | .38 | .33 | .42 | .34 | .27 | −.35 | −.36 |
| 2. Marriage equality | 5.59 | 1.64 | — | −.58 | .54 | .56 | .65 | .67 | .65 | −.58 | −.53 | −.52 | −.67 | −.68 | .67 | −.40 | −.32 | −.38 | −.37 | −.44 | −.30 | −.26 | .24 | .35 | |
| 3. Death penalty | 3.92 | 1.78 | — | −.36 | −.53 | −.57 | −.56 | −.56 | .54 | .71 | .63 | .68 | .54 | −.62 | .55 | .31 | .45 | .54 | .60 | .47 | .41 | −.28 | −.41 | ||
| 4. Affirmative action | 5.02 | 1.93 | — | .60 | .54 | .46 | .49 | −.28 | −.29 | −.41 | −.56 | −.56 | .74 | −.36 | −.17 | .01 | .01 | −.28 | −.18 | −.05 | .26 | .22 | |||
| 5. Obamacare | 4.75 | 2.00 | — | .65 | .48 | .50 | −.40 | −.38 | −.43 | −.68 | −.57 | .71 | −.15 | −.06 | −.04 | −.10 | −.17 | −.19 | −.20 | .15 | .01 | ||||
| 6. Climate regulation | 5.19 | 2.05 | — | .51 | .63 | −.66 | −.61 | −.55 | −.74 | −.62 | .67 | −.46 | −.33 | −.45 | −.30 | −.30 | −.35 | −.24 | .16 | .26 | |||||
| 7. Pay equality | 6.17 | 1.24 | — | .47 | −.41 | −.50 | −.42 | −.60 | −.50 | .51 | −.30 | −.25 | −.21 | −.26 | −.37 | −.40 | −.22 | .27 | .39 | ||||||
| 8. Gun control | 3.94 | 2.25 | — | −.56 | −.53 | −.66 | −.64 | −.50 | .60 | −.43 | −.38 | −.35 | −.30 | −.31 | −.23 | −.06 | .34 | .29 | |||||||
| 9. Increase military spending | 3.95 | 1.88 | — | .66 | .68 | .66 | .62 | −.45 | .53 | .44 | .63 | .51 | .39 | .53 | .33 | −.28 | −.38 | ||||||||
| 10. Use of torture | 3.73 | 1.85 | — | .69 | .67 | .60 | −.52 | .56 | .44 | .55 | .58 | .60 | .47 | .39 | −.34 | −.34 | |||||||||
| 11. Troops to Middle East | 4.22 | 1.86 | — | .69 | .62 | −.47 | .64 | .33 | .45 | .42 | .49 | .40 | .27 | −.44 | −.38 | ||||||||||
| 12. Build the wall | 3.22 | 2.04 | — | .82 | −.74 | .50 | .32 | .44 | .39 | .43 | .46 | .33 | −.36 | −.30 | |||||||||||
| 13. Ban Muslim immigration | 2.55 | 1.80 | — | −.74 | .43 | .34 | .49 | .42 | .52 | .47 | .32 | −.39 | −.31 | ||||||||||||
| 14 Social welfare programs | 5.13 | 1.96 | — | −.38 | −.29 | −.29 | −.31 | −.48 | −.28 | −.23 | .28 | .22 | |||||||||||||
| 15. Use of military force | 4.63 | 1.75 | — | .47 | .58 | .51 | .57 | .49 | .33 | −.24 | −.36 | ||||||||||||||
| 16. Mass troops at border | 4.25 | 1.20 | — | .74 | .69 | .64 | .59 | .52 | −.21 | −.44 | |||||||||||||||
| 17. Bomb enemy territory | 3.48 | 1.44 | — | .81 | .70 | .57 | .59 | −.16 | −.41 | ||||||||||||||||
| 18. Torture enemy spy | 3.44 | 1.46 | — | .74 | .59 | .63 | −.20 | −.48 | |||||||||||||||||
| 19. Execute enemy spy | 3.61 | 1.43 | — | .58 | .59 | −.32 | −.45 | ||||||||||||||||||
| 20. Number of troops deployed | 3.06 | 1.27 | — | .56 | −.36 | −.57 | |||||||||||||||||||
| 21. Impose economic sanctions | 4.56 | 1.38 | — | −.01 | −.26 | ||||||||||||||||||||
| 22. Pursue backdoor negotiations | 3.94 | 1.40 | — | .59 | |||||||||||||||||||||
| 23. Seek peace deal | 4.08 | 1.36 | — |
Note. Correlations in bold are significant at p < .05. Conservatism ranged from 1–11 (midpoint 6), policy attitudes from 1–7 (midpoint 4), and vignette responses from 1–6. Gender is coded such 0 = male and 1 = female; mean gender is interpreted as proportion female.
Effect of Threat Manipulation on Conservatism
As in the previous experiments, we first sought to confirm that the manipulation did not affect conservatism by regressing conservatism on threat condition. This analysis yielded no significant main effect of threat, B = .060, SE B = .061, t = .98, p = .330, 95% CI [-.062, .183], freeing us to examine conservatism as a moderator of any threat effect on policy attitudes or vignette responses.
Aggressive and Nonaggressive Composites
Aggressive and nonaggressive policy composites were created in a manner consistent with the PCA results in Experiment 1. The aggressive policies were death penalty, increase military spending, use of torture, troops to Middle East, and use of military force (α = .89) and the nonaggressive policies were marriage equality, affirmative action, Obamacare, climate regulation, pay equality, gun control, social welfare programs (α = .90).
Aggressive and nonaggressive vignette composites were created in a manner reflecting the PCA results in Experiment 2. The aggressive items were mass troops at border, bomb enemy territory, torture enemy spy, execute enemy spy, and a large number of troops deployed (α = .91) and the nonaggressive items were impose economic sanctions, pursue backdoor negotiations, and seek peace deal (α = .27).
Analyses
In this experiment, participant ideology (conservatism) was a continuous predictor, strength feedback (strong vs. weak) was a between-participants factor, and policy/response type (aggressive vs. nonaggressive) was a within-participant factor. The strength feedback manipulation was coded such that -.5 = strong and .5 = weak and policy/response type was coded such that -.5 = aggressive and .5 = nonaggressive. Participants’ conservatism scores were centered around the sample mean. The aggressive and nonaggressive policy and vignette composites were rescaled to range from 0 (the minimum possible score) to 1 (the maximum possible score).
Analyses were run separately for the policy and vignette composites. For each outcome, a mixed-model linear regression analysis was conducted using the mixed command in Stata 17. Participants’ composite scores were regressed on policy/response type, feedback condition, and the interaction between these terms. A random intercept and random slope of policy/response type were specified, and the covariance between the random effects was freely estimated. The results of these analysis are shown in Table 8 (policy endorsement) and Table 9 (vignette responses). The regression equations were used to plot estimates of policy and vignette responses among conservative and liberal men in each experimental condition, with “conservative” and “liberal” defined as 1 SD above and below the ideology mean, respectively (Fig. 3). Visual inspection of the regression predictions clearly suggests that manhood threat increased liberal men’s endorsement of aggressive policies and vignette responses, but provides little indication that threat affected liberals’ ratings of nonaggressive policy or responses, or conservatives’ ratings of either policy/response type.
Table 8.
Results of Mixed-Model Linear Regression Aanalysis in Experiment 3: Policy Endorsement (N = 64)
| 95% CI | |||||||
|---|---|---|---|---|---|---|---|
| Parameter | B | SE B | z | p | LB | UB | f2 |
| Fixed Effects | |||||||
| Conservatism (C) | −.118 | .044 | −2.67 | .008 | −.205 | −.031 | .035 |
| Policy Type (P) | .164 | .032 | 5.20 | .000 | .102 | .226 | .288 |
| Threat (T) | .067 | .022 | 3.12 | .002 | .025 | .109 | .048 |
| C × P | −1.519 | .129 | −11.74 | .000 | −1.773 | −1.265 | 1.468 |
| C × T | −.139 | .088 | −1.57 | .116 | −.312 | .034 | .012 |
| P × T | −.151 | .063 | −2.40 | .017 | −.275 | −.028 | .061 |
| C × P × T | .241 | .259 | .93 | .353 | −.267 | .748 | .009 |
| Intercept | .603 | .011 | 55.95 | .000 | .582 | .624 | |
| Random Effects | |||||||
| Var (P) | .057 | 1.857 | .000 | 2.01 × 1026 | |||
| Var (Intercept) | .006 | .464 | .000 | 4.93 × 1063 | |||
| Cov (Intercept, P) | −.007 | .003 | −.012 | −.001 | |||
Note. CI confidence interval, LB lower bound, UB upper bound. f2 calculation performed using the method described by Selya et al. (2012)
Table 9.
Results of Mixed-Model Linear Rregression Analysis in Experiment 3: Vignette Responses (N = 64)
| 95% CI | |||||||
|---|---|---|---|---|---|---|---|
| Parameter | B | SE B | z | p | LB | UB | f2 |
| Fixed Effects | |||||||
| Conservatism (C) | .091 | .064 | 1.42 | .155 | −.035 | .217 | .016 |
| Policy Type (P) | .116 | .030 | 3.83 | .000 | .057 | .176 | .121 |
| Threat (T) | .101 | .031 | 3.22 | .001 | .039 | .162 | .081 |
| C × P | −.509 | .124 | −4.09 | .000 | −.753 | −.265 | .138 |
| C × T | −.210 | .128 | −1.63 | .102 | −.461 | .042 | .021 |
| P × T | −.222 | .061 | −3.67 | .000 | −.341 | −.104 | .111 |
| C × P × T | .340 | .249 | 1.37 | .172 | −.148 | .827 | .015 |
| Intercept | .581 | .016 | 37.16 | .000 | .550 | .612 | |
| Random Effects | |||||||
| Var (P) | .051 | 5.726 | .000 | 1.30 × 1095 | |||
| Var (Intercept) | .014 | 1.431 | .000 | 1.19 × 1088 | |||
| Cov (Intercept, P) | −.001 | .004 | −.008 | .006 | |||
Note. CI confidence interval, LB lower bound, UB upper bound. f2 calculation performed using the method described by Selya et al. (2012)
Fig. 3.
Men’s Foreign-Policy Endorsement as a Function of Masculinity Threat, Policy Type, and Political Ideology in Experiment 3
We next ran focused tests examining the impact of threat on policy support and vignette responses for every combination of policy or response type and threat condition using the margins command in Stata 17 (Tables 10 and 11). As predicted, and consistent with inspection of Fig. 3, the simple effects of manhood threat on male liberals’ ratings of aggressive policies (Table 10) and aggressive vignette responses (Table 11) were significant. Also as expected, threat did not significantly affect men’s endorsement of nonaggressive policies or vignette responses.
Table 10.
Simple Effects of Gender Tthreat and Llinear Hypothesis Ttests in Experiment 3: Policy Endorsement (N = 64)
| 95% CI | |||||||
| B | SE B | z | p | LB | UB | Glass’s Δ | |
| Liberal Men / Aggressive Policy (LMA) | .206 | .062 | 3.33 | .001 | .085 | .328 | .813 |
| Liberal Men / Nonaggressive Policy (LMN) | −.004 | .045 | −.09 | .932 | −.093 | .085 | −.015 |
| Conservative Men / Aggressive Policy (CMA) | .079 | .062 | 1.29 | .197 | −.041 | .200 | .313 |
| Conservative Men / Nonaggressive Policy (CMN) | −.013 | .045 | −.29 | .774 | −.101 | .076 | −.052 |
| X2 | df | p | |||||
| LMA vs. LMN | 5.49 | 1 | .019 | ||||
| LMA vs. CMA | 2.10 | 1 | .147 | ||||
| LMA vs. CMN | 8.19 | 1 | .004 | ||||
| LMA vs. Σ (LMN + CMA + CMN) | 6.61 | 1 | .010 | ||||
Note. CI confidence interval, LB lower bound, UB upper bound. “Liberal” and “conservative” refer to participants 1 SD below and above the mean level of political conservatism. Glass’s Δ calculated according to the method described in Shaw (2022)
Table 11.
Simple Effects of Gender Threat and Linear Hypothesis Tests in Experiment 3: Vignette Responses (N = 64)
| 95% CI | |||||||
| B | SE B | z | p | LB | UB | Glass’s Δ | |
| Liberal Men / Aggressive Policy (LMA) | .305 | .063 | 4.84 | .000 | .181 | .428 | 1.310 |
| Liberal Men / Nonaggressive Policy (LMN) | −.001 | .061 | −.02 | .988 | −.120 | .118 | −.005 |
| Conservative Men / Aggressive Policy (CMA) | .119 | .063 | 1.90 | .057 | −.004 | .242 | .510 |
| Conservative Men / Nonaggressive Policy (CMN) | −.020 | .060 | −.34 | .737 | −.139 | .098 | −.116 |
| X2 | df | p | |||||
| LMA vs. LMN | 12.58 | 1 | .0004 | ||||
| LMA vs. CMA | 4.36 | 1 | .037 | ||||
| LMA vs. CMN | 13.87 | 1 | .0002 | ||||
| LMA vs. Σ (LMN + CMA + CMN) | 14.33 | 1 | .0002 | ||||
CI confidence interval, LB lower bound, UB upper bound. “Liberal” and “conservative” refer to participants 1 SD below and above the mean level of political conservatism. Glass’s Δ calculated according to the method described in Shaw (2022)
Having observed significant simple effects of threat only for liberal men rating aggressive policies and vignette responses, we proceeded to run linear hypothesis (Wald) tests (using the test command in Stata 17) to examine whether these effects were significantly greater than those for every other combination gender and policy type. As can be seen in Table 10, the threat effect for liberal men rating aggressive policies was significantly larger than for any other combination of ideology and policy type except conservative men rating aggressive policies (p = .147). The threat effect for liberal men rating aggressive policies differed significantly from the average of threat effects for the remaining three combinations of ideology and policy type (p = .010). As shown in Table 11, the threat effect for liberal men rating aggressive vignette responses was significantly stronger that for every other combination of ideology and response type, as well as the average of threat effects for these combinations (p = .0002).
Discussion
As predicted, and replicating Experiments 1 and 2, threatening men’s physical strength significantly increased liberal men’s endorsement of aggressive real-world policies and hypothetical actions in the foreign-policy vignette. At the same time, threat left liberal men’s endorsement of nonaggressive policies and vignette responses unaffected; likewise, threat did not alter conservative men’s responses to either aggressive or nonaggressive policies or vignette responses.
Integrative Data Analysis
Using Integrative Data Analysis (IDA), we sought to synthesize our findings across the three experiments, while quantifying between-experiment heterogeneity in the observed effects (Curran & Hussong, 2009). IDA is a useful technique for integrating results across studies when all original data are available. In IDA, hypotheses are tested using the combined datasets. Between-study heterogeneity is modeled either as a random effect or, when few individual datasets are available, as a fixed effect. Because we are integrating only three experiments, a fixed-effects IDA was used (Curran & Hussong, 2009).
As a first step in the IDA, we combined our three datasets. Because female participants were only recruited in Experiment 1, we omitted them from the combined dataset. Experiment was coded using weighted effect (Sweeney) coding (te Grotenhuis et al., 2017). Owing to this choice of coding schemes, model parameters constitute effects for the average participant in the combined dataset. Moreover, because Experiment 3 used two pairs of composites—one for policy attitudes and the other for vignette responses—we averaged the two aggressive composites and the two nonaggressive composites to form overall aggressive and nonaggressive composites.
We next re-tested our hypotheses in the combined dataset using the same mixed-model regression approach as in Experiment 1–3. With the mixed command in Stata 17, we regressed participants’ composite scores on experiment, conservatism, threat condition, policy type, and all interactions between these predictors. A random intercept and random slope of policy type were included in the model, and the covariance between these random effects was freely estimated. The results of this analysis (Table 12) were used to plot predicted scores for liberal and conservative men rating aggressive and nonaggressive policies and vignette responses (Fig. 4).
Table 12.
Results of Mixed-Model Linear Regression Analysis in Combined Dataset (N = 445)
| 95% CI | |||||||
|---|---|---|---|---|---|---|---|
| Parameter | B | SE B | z | p | LB | UB | f2 |
| Fixed Effects | |||||||
| Experiment 2 (E2) | .009 | .005 | 1.86 | .063 | −.001 | .019 | .004 |
| Experiment 3 (E3) | .026 | .013 | 2.01 | .044 | .001 | .051 | .004 |
| Conservatism (C) | .085 | .026 | 3.25 | .001 | .034 | .136 | .011 |
| Policy Type (P) | .248 | .011 | 22.08 | .000 | .226 | .270 | .580 |
| Threat (T) | .020 | .011 | 1.88 | .060 | −.001 | .041 | .004 |
| C × P | −.043 | .011 | −4.05 | .000 | −.064 | −.022 | .246 |
| C × T | −.095 | .052 | −1.81 | .070 | −.198 | .008 | .004 |
| P × T | −.088 | .022 | −3.91 | .000 | −.132 | −.044 | .018 |
| C × P × T | .306 | .111 | 2.76 | .006 | .089 | .523 | .009 |
| E2 × C | .066 | .025 | 2.70 | .007 | .018 | .114 | .008 |
| E2 × P | −.043 | .011 | −4.05 | .000 | −.064 | −.022 | .019 |
| E2 × T | −.013 | .010 | −1.28 | .200 | −.033 | .007 | .002 |
| E2 × C × P | .249 | .052 | 4.80 | .000 | .147 | .351 | .027 |
| E2 × C × T | .066 | .049 | 1.34 | .181 | −.031 | .162 | .002 |
| E2 × P × T | .306 | .111 | 2.76 | .006 | .089 | .523 | .002 |
| E2 × C × P × T | .024 | .104 | .23 | .815 | −.179 | .228 | .000 |
| E3 × C | −.098 | .055 | −1.79 | .074 | −.206 | .009 | .003 |
| E3 × P | −.108 | .027 | −3.93 | .000 | −.162 | −.054 | .018 |
| E3 × T | .064 | .026 | 2.47 | .013 | .013 | .115 | .006 |
| E3 × C × P | −.216 | .117 | −1.85 | .064 | −.444 | .013 | .004 |
| E3 × C × T | −.079 | .110 | −.72 | .472 | −.295 | .137 | .000 |
| E3 × P × T | −.099 | .055 | −1.81 | .071 | −.207 | .008 | .004 |
| E3 × C × P × T | −.016 | .233 | −.07 | .945 | −.473 | .441 | .000 |
| Intercept | .566 | .005 | 106.79 | .000 | .556 | .576 | |
| Random Effects | |||||||
| Var (P) | .049 | 1.104 | .000 | 8.06 × 1017 | |||
| Var (Intercept) | .011 | .276 | .000 | 8.77 × 1019 | |||
| Cov (Intercept, P) | −.003 | .001 | −.005 | −.0002 | |||
Note. CI confidence interval, LB lower bound, UB upper bound. f2 calculation performed using the method described by Selya et al. (2012)
Fig. 4.
Men’s Foreign Policy Endorsement as a Function of Masculinity Threat, Policy Type, and Political Ideology in the Combined Dataset
We next ran focused tests examining the impact of threat on policy support for every combination of policy type and threat condition using the margins command in Stata 17 (Table 13). The simple effect of manhood threat on male liberals’ ratings of aggressive vignette responses was significant. Also as expected, threat did not affect liberal or conservative men’s endorsement of nonaggressive policies.
Table 13.
Simple Eeffects of Ggender Threat and Linear Hypothesis Tests in Combined Dataset (N = 445)
| 95% CI | |||||||
| B | SE B | z | p | LB | UB | Glass’s Δ | |
| Liberal Men / Aggressive Response (LMA) | .115 | .024 | 4.85 | .000 | .069 | .162 | .566 |
| Liberal Men / Nonaggressive Response (LMN) | −.036 | .022 | −1.68 | .094 | −.078 | .006 | −.251 |
| Conservative Men / Aggressive Response (CMA) | .012 | .022 | 0.55 | .581 | −.032 | .056 | .050 |
| Conservative Men / Nonaggressive Response (CMN) | −.012 | .020 | −0.58 | .559 | −.052 | .028 | −.071 |
| X2 | df | p | |||||
| LMA vs. LMN | 21.08 | 1 | .000 | ||||
| LMA vs. CMA | 9.71 | 1 | .002 | ||||
| LMA vs. CMN | 16.47 | 1 | .000 | ||||
| LMA vs. Σ (LMN + CMA + CMN) | 21.95 | 1 | .000 | ||||
Note. CI confidence interval, LB lower bound, UB upper bound. “Liberal” and “conservative” refer to participants 1 SD below and above the mean level of political conservatism. Glass’s Δ calculated according to the method described in Shaw (2022)
Having observed a significant simple effect of threat only for liberal men rating aggressive policies, we proceeded to run linear hypothesis (Wald) tests (using the test command in Stata 17) to examine whether this effect was significantly greater than that for every other combination gender and policy type. As can be seen in Table 13, the threat effect for liberal men rating aggressive policies was significantly larger than for every other combination of ideology and policy type.
The IDA allowed us to assess the overall weight of evidence for the effect of threats to masculinity on men’s political attitudes. Contrary to our initial predictions, we found strong and consistent evidence that liberal men become more politically aggressive when their masculinity is called into question. In contrast, our experiments collectively yielded no evidence that manhood threats affect conservative men’s levels of political aggression.
IDA also allows us to directly assess between-experiment (and thus between-manipulation) heterogeneity in the effect of manhood threat on liberal men’s political aggression. Using the margins command in Stata 17, we calculated regression predictions for the threat effect in each of the three experiments. The largest effect of threat was seen in Experiment 3, which utilized the strength-threat manipulation (See Table S3 in the online supplement). Linear hypothesis (Wald) tests, conducted using the test command, revealed that impugning men’s physical strength led to reliably larger increases in liberal men’s political aggression than having them paint their nails (Experiment 2; p = .011) or ascribing to them feminine personality traits (Experiment 1; p = .076). However, the threat inductions used in Experiments 1 and 2 (personality feedback and nail-painting) did not differ significantly from one another (p = .460).
Multiverse Analysis
While we believe we made sound judgments in processing our data and selecting our modeling approach, other researchers may have preferred different—but equally justifiable—approaches. Given the “researcher degrees of freedom” (Simmons et al., 2011) that we exercised in data processing and analysis, we conducted a multiverse analysis (Steegen et al., 2016) to assess the effect of alternative choices on our conclusions. In a multiverse analysis, the researcher identifies all seemingly relevant choices in data processing and modeling—and their alternatives—and reruns the most important statistical tests within each of these separate “universes.” The robustness of the researcher’s conclusions can then be assessed by tabulating the proportion of choice combinations under which a statistically significant result is obtained, or by calculating the average p-value of a result across all choice combinations (Steegen et al., 2016).
Choice Points
We identified several important choice points in the present research, some of which pertain to data processing and others to statistical modeling; we thus conducted an analysis of the “data multiverse” and the “model multiverse” in the combined IDA dataset (Steegen et al., 2016). In terms of data processing, the following choices were identified:
In Experiment 1, three policies (stand-your-ground laws, build the wall, and ban Muslim immigration) were excluded from analysis because they loaded relatively strongly on both factors extracted in the PCA. However, these policies loaded negatively, and most strongly, on the nonaggressive policy factor. Thus, other researchers may have elected to reverse-score these items and include them in the nonaggressive policy composite. Note that our choice in this regard also affected the formation of the policy composite in Experiment 3, which used the same set of items as Experiment 1 (except for stand-your-ground laws).
In Experiment 2, a nonaggressive foreign-policy composite was created by averaging impose economic sanctions, pursue backdoor negotiations, and seek peace deal. However, these items formed an unreliable composite (α = .22). While we were not overly concerned about this—given the fact that these items functioned only as a comparison against which to judge the effect of masculinity threat on aggressive policies—other researchers may have elected not to average them to create a composite. Another approach would be to use each of the nonaggressive policies individually as exemplars of nonaggressive policies. Note that our choice in this regard also affected the formation of the policy composite in Experiment 3, which used the same set of vignette-related items as Experiment 1.
In terms of statistical modeling, the following choices were made:
In our mixed-model regression analyses, parameters’ standard errors were estimated based on the observed information matrix—the default in Stata for models using maximum likelihood estimation. We might have instead chosen to apply the robust option, which uses the Huber-White sandwich estimator for standard errors (Huber & Ronchetti, 2009). This option corrects for certain types of model misspecification, including heteroskedastic residuals.
In our mixed-model analyses, a random effect of policy type was estimated, and this random slope was allowed to covary with the random intercept. However, these random slopes were always near zero and accompanied by extremely wide confidence intervals. Thus, other researchers might have chosen simply to omit the random slope of policy type from the models.
Along with their alternatives, our data-processing and modeling choices imply a multiverse of two choices of policy-attitude composite in Experiments 1 and 3, four choices of vignette composite in Experiments 2 and 3, two choices of random-effect specification in Experiments 1–3, and two choices of standard-error estimator in Experiments 1–3. This yields 2 × 4 × 2 × 2 = 32 choice combinations to be examined in the combined dataset.
Analysis
In our multiverse analysis, we focus on the tests most relevant to the idea that masculinity threat increases aggressive policy endorsement or vignette responses among liberal men—and that this effect is unique to liberal men’s ratings of aggressive policies. The predicted threat effect is assessed by examining estimates of the simple effect of threat on aggressive policy endorsement among liberal men (see Table 13). The uniqueness of this effect can be assessed in two ways. Specifically, we can tabulate choice combinations in which: (a) the simple effect of threat is nonsignificant for conservative men and nonaggressive policies, and (b) the threat effect among liberal men rating aggressive policies is significantly greater than the effect for every other conjunction of ideology and policy type—as well as the average effect across these conjunctions.
The results of our multiverse analysis are summarized in Table S4 in the online supplement. The simple effect of threat for liberal men rating aggressive policies was statistically significant in 100% of the runs, with a mean p = .0001. The simple effect of threat for liberal men rating nonaggressive policies was significant under 25% of the choice scenarios, with a mean p = .35; these significant effects, however, are negative, implying that threat may reduce support for nonaggressive policies among liberal men. Finally, the threat effects for conservative men rating aggressive and nonaggressive policies were never significant, with mean ps of .74 and .60, respectively. This pattern of significance across choice combinations suggests that the effect of masculinity threat on liberal men’s endorsement of aggressive policies is highly robust to the choice of data-processing and modeling strategies—and does not extend to ratings of nonaggressive policies or to conservative men’s ratings of either policy type.
Direct tests strongly suggest that the threat effect for liberal men rating aggressive policies is unique. Indeed, under 100% of choice scenarios, this simple effect was greater than that for liberal men rating nonaggressive policies (mean p = .002), conservative men rating aggressive policies (mean p = .006), conservative men rating nonaggressive policies (mean p = .005), and the average of these three effects (mean p = .0003). We therefore take our multiverse results to indicate that the observed pattern of results, in which the effect of masculinity threat is localized to liberal men’s endorsement of aggressive policy, is robust.
General Discussion
The present findings provide support for our hypothesis that threats to men’s (but not women’s) gender status leads to an increase in political aggression, defined as attitudes or behavior that communicate toughness, strength, or force. At the same time, our findings run directly counter to our initial prediction as to which men—liberals or conservatives—would be most affected by masculinity threat. Although we hypothesized that conservative men would increase in political aggression after masculinity threat, they did not; instead, across our three experiments, it was liberal men who exhibited increased political aggression after a manhood threat.
In Experiment 1, we found that a personality-based false feedback manipulation, in which men learned they possessed traits resembling that of the average woman, significantly increased liberal—but not conservative—men’s endorsement of aggressive political policies (e.g., the death penalty). Consistent with our hypotheses, gender threat had no effect on endorsement of nonaggressive policies, or on women’s endorsement of either policy type. In Experiment 2, we found that a behavioral manipulation, in which men engaged in a stereotypically feminine behavior (applying pink nail polish) increased liberal—but not conservative—men’s support for an aggressive approach to a foreign-policy dilemma. We again observed no effect of threat on men’s endorsement of nonaggressive approaches. Finally, in Experiment 3, we found that a strength-based false feedback manipulation, in which men learned their handgrips were only as strong as the average woman’s, caused liberal—but not conservative—men to become more supportive of both aggressive policies and foreign-policy strategies. Once again, manhood threat affected support for aggressive, but not nonaggressive, political policies and strategies.
While liberal men increased significantly in political aggression in all three experiments, and conservative men in none of the experiments, the difference between the threat effect for liberals and conservatives was not always significant. Specifically, ideology did not significantly moderate men’s threat-induced political aggression in Experiment 2, and did so in Experiment 3 only for vignette responses. Nonetheless, our Integrative Data Analysis (IDA; Curran & Hussong, 2009) pooling the evidence across experiments suggests a significant difference in liberals’ and conservatives’ threat responses overall. A multiverse analysis (Steegen et al., 2016) of the combined data indicated that our results are highly robust to different data-treatment and modeling choices. Taken together, then, our results tell a consistent story in which liberal men show a greater tendency to reaffirm their masculinity after manhood threats by embracing more aggressive political views.
Manhood Threats
The present research allows us to compare the relative impact of three different kinds of manhood threats on liberal men. Predictions derived from the IDA revealed that the strength-based threat manipulation in Experiment 3 produced the strongest effect on aggressive policy endorsement. This suggests that intimations of physical weakness represent an especially powerful threat to liberal men’s masculinity—and, more broadly, that the onus on men to display physical strength is highly salient in American society (Frederick et al., 2017). Our personality-based threat manipulation in Experiment 1, in which men were led to believe they possessed stereotypically feminine personalities, yielded the second-strongest effect on political aggression among liberal men. Interestingly, Experiment 2’s behavioral threat manipulation, in which men were compelled to paint their nails pink, produced the weakest threat effect. This may imply that engaging in a feminine behavior at the clear behest of an experimenter does not impugn men’s masculinity as drastically as does information that one’s personal qualities, whether physical (Experiment 3) or psychological (Experiment 1), are truly feminine. This pattern of results suggests that, in the political domain, advertising that impugns men’s physical strength, or suggests that men possess feminine personality traits, may be highly effective in shifting liberal men’s electoral and policy preferences in an aggressive direction. These results are also consistent with previous research showing that public (vs. private) threats tend to be most problematic for men (Weaver et al., 2013). In our research, the experiment that induced threat in a public space (Experiment 3) produced stronger results than threats experienced in private (Experiments 1 and 2). It is worth noting, however, that both public and private threats “worked”— suggesting that, to some extent, participants wish to demonstrate to themselves that they are adequately masculine.
The Role of Ideology
The link between masculinity and conservative political ideology is well-established. Past work has found that chronic masculine insecurity predicts voting for Republican presidential and congressional candidates (DiMuccio & Knowles, 2021). In other research, threats to masculinity increased men’s support for Donald Trump—an effect mediated by the desire for a highly masculine president (Carian & Sobotka, 2018). Other studies have revealed a strong link between masculinity and conservatism including robust cultural associations between “Republican” and “masculine” (Katz, 2016; McDermott, 2016; Winter, 2010) and a tendency for political conservatives to endorse traditional gender and sex-role beliefs (Feather et al., 1979; Sharrow et al., 2016). Given this link, we were surprised to find that it was liberal—not conservative—men who reacted with increased political aggression to manhood threat. We propose four potential explanations for this unexpected finding.
First, it may be that our dependent measures of political aggression (e.g., support for military intervention and the death penalty) failed to allow sufficient room for movement among conservative participants, who already strongly endorsed such positions. Indeed, we observed a ceiling effect in which 17% of our conservative male participants scored at or near the scale maximum across studies and measures (Terwee et al., 2007). This was not the case for liberal participants, who either opposed aggressive policies less (Experiments 1 and 2) or became supportive them (Experiment 3) after a threat to their manhood. Manhood threat may nonetheless cause conservative men to venture outside the range of socially-sanctioned political aggression (e.g., military intervention) into the realm of violent extremism (as exemplified the 2021 Capitol insurrection). If this is correct, then more extreme measures of political aggression would allow such an effect to emerge. By increasing the extremity of aggressive political options, researchers can allow for effects of masculinity threat to emerge among conservative men, while also shedding light on the recent rise of right-wing extremism in the U.S. (Kapur, 2021).
Second, it is possible that, compared to liberals, conservatives are higher in chronic concern for masculinity—and that this blunts the impact of transient threat inductions on their political attitudes. If this is the case, then the effects of masculinity threat may already be “baked in” to conservative men’s political attitudes. Future research should carefully parse out the effects of, and interactions between, trait vs. state levels of masculine insecurity.
Third, it may be that liberal men are genuinely more vulnerable to masculinity threats in political contexts. In light of the fact that people stereotype liberals as feminine and conservatives as masculine (Katz, 2016; Rudman et al., 2013; Winter, 2010), it stands to reason that many liberal men are especially eager not to exhibit feminine traits in the political realm. In other words, perhaps liberals experience stereotype threat (Spencer et al., 1999) with respect to their masculinity. In our studies, then, liberal men may have reacted to threat with heightened political aggression in order to avoid confirming a (presumably) negative stereotype of their ideological group. Suggesting that this stereotype is, in fact, negative, accusations of femininity constitute a recurring attack line against liberal politicians, presidents, and laypeople—from both the left (Dowd, 2006; Prabhu, 2016) and the right (Fahey, 2007; French, 2015). Conversely, aggression and masculinity are widely regarded as positive political qualities in American politics (Ducat, 2004; Fahey, 2007; Katz, 2016; Messner, 2007), rendering “feminized” liberal men stereotype-incongruent in political contexts (Bauer & Carpinella, 2018). Future research should further examine the possibility that liberal men experience a form of gendered stereotype threat in the realm of politics.
Fourth, research has found that liberals become more conservative in their attitudes when exposed to system threats (a phenomenon termed conservative shift; Bonanno & Jost, 2006; Nail et al., 2009; Nail & McGregor, 2009). This raises the question of whether our findings might reflect a conservative shift among liberals rather than an increase in their political aggression per se. We believe the answer may be found in our findings regarding nonaggressive policies, such as attitudes toward Obamacare, affirmative action, and other social-welfare policies. Such stances have clear (liberal) ideological content. Thus, if masculinity threat were simply causing liberals to become more conservative, we should have observed liberals endorse such policies less under threat. We did not, however, observe any reliable effect of masculinity threat on such ideologically laden, yet nonaggressive, attitude dimensions. We therefore believe the present findings reflect a “aggressive shift” that is not reducible to a conservative shift.
We see another possible—though more speculative—explanation for the fact that masculinity threat increased political aggression more among liberals than among conservatives. Specifically, it may be that liberals and conservatives are distinguished by a differential tendency to repair masculinity in public vs. private contexts. If conservatives only see the utility of performing reparative behaviors in public, while liberals engage in such behaviors in public or private, it might explain liberals’ relatively large threat-induced increase in political aggression in the present experiments. Indeed, two of three studies (Experiments 1 and 2) measured political aggression in private, laboratory contexts. Only Experiment 3 was conducted in a public space (a park), and perhaps not coincidentally produced the largest threat effects among conservative men. Although we know of no data or theory that specifically suggests a public–private distinction along ideological lines, little is known about factors that make public vs. private performance of manhood preferable. (We thank one of our reviewers for raising these issues.)
We wish to caution that, while our masculinity threats increased political aggression only among relatively liberal men, threat failed to close the gap between liberals’ and conservatives’ overall levels of political aggression. Indeed, conservatives displayed consistently and drastically stronger support for aggressive policies and vignette responses. This could be due to people’s longstanding ideological affinities, as aggressive political policies tend to be more conservative than nonaggressive policies. Despite these caveats, political liberals’ heightened susceptibility to messages that impugn their masculinity suggests that left-leaning men should be vigilant against attempts to manipulate their politics through such means.
Limitations and Future Directions
This research has several limitations. First, our experiments did not include manipulation checks. We chose not to include such checks out of concern that doing so would have hinted at the true intentions of the research. This, unfortunately, means that we cannot know whether and to what extent the participants perceived each threat manipulation to challenge their masculinity. While our manipulations had face validity, future research should systematically measure the extent to which each type of manipulation employed in the present research is experienced as a threat to masculinity.
Second, our samples were disproportionately politically left-leaning. It is possible that we would have seen differences by threat across the political spectrum (and not only for left-leaning men) if we had had access to a greater number of highly conservative participants. To investigate this possibility, using the combined dataset, we plotted the estimated effects of masculinity threat on aggressive policy endorsement at every point on the conservatism scale. As can be seen in Figure S1, despite widening of the 95% confidence intervals due to the relatively small number of conservatives in the sample, there is no suggestion of a threat effect emerging at the highest levels of conservatism. Despite the lack of evidence for masculinity threat among conservatives, a definitive test awaits research that includes a large number of extremely conservative men.
Finally, our indicators of political aggression tend to be ones that conservatives endorse. This necessarily makes it difficult (though, as discussed above, not impossible) to tease apart aggressive responses from merely conservative ones. As such, future research should examine the extent to which threats might cause men to employ more aggressive methods to reach whatever their political ends may be. Perhaps, for instance, manhood threats would cause conservative men to embrace more aggressive means of passing gun rights legislation while also causing liberal men to embrace more aggressive means of passing social-welfare legislation. Such a study of political strategy (as opposed to outcomes) is a promising avenue for future investigation.
Conclusion
The present research illuminates the impact of manhood threat on male aggression in the political domain—specifically, men’s adoption of political views that communicate toughness, forcefulness, and strength. Contrary to our original expectations, our data suggest that liberal men may be most likely to react to masculinity threat with increased political aggression, at least on the types of political outcome measures used in the present research.
Because our participants were largely left-leaning, our findings should be replicated in samples containing a higher proportion of conservative men and using measures tapping more extreme right-wing attitudes. Nonetheless, our failure to observe an effect of masculinity threat on conservative men’s political aggression provides good reason to believe that such future work will reveal a similar ideological asymmetry in political responses to such threat. Critically, future research should examine the possible mechanisms behind such an asymmetry—including political stereotype threat, conservatives’ already elevated levels of chronic masculine concern, or ceiling effects reflecting right-leaning men’s already-strong endorsement of conservative ideological stances.
Our finding that liberal men are vulnerable to increased political aggression in the face of masculinity threat has crucial implications for the future of gendered politics in the United States, as it suggests that right-wing candidates might benefit from media strategies designed to induce masculine insecurity among liberal men. A challenge for the future is to inoculate all men from chronic and acute masculine insecurity—perhaps through concerted efforts to combat societal stereotypes and sex roles that limit what it means to be a “real man.”
Electronic supplementary material
Below is the link to the electronic supplementary material.
Acknowledgements
We thank Marie Helweg-Larsen for feedback on earlier versions of the manuscript.
Appendix
Text of “Warring Countries” Vignette
You are the president of a country named Liga [Bombola], which has been in a centuries-long conflict with its neighboring country, Bombola [Liga]. Your country is in competition with Bombola [Liga] for natural resources and claims for the rightful ownership of border territories. Recently, political tensions have escalated, and it is very likely that they could lead to war. Given the substantial military forces and powerful warfare technology that both Liga and Bombola [Bombola and Liga] possess, a war between the two countries will result in immeasurable destruction and loss of life for both sides. However, it is yet uncertain whether Bombola [Liga] positively plans to act aggressively or to initiate armed conflict. Their unusually high levels of military activity detected this past week could turn out to be nothing more than a military exercise. But, it could also indicate their intentions to start a war at any moment, in which case you must be ready to counter accordingly. Failure to do so will result in catastrophic damage to your country structurally, economically, and politically, as well as to your domestic and international reputation as a competent state leader. Yet should you choose to take up arms against Bombola in anticipation of their attacks, and it is later revealed that your actions were uncalled for or unnecessary, you will be held accountable for the fallout. The fact remains that the present predicament has been alerted as a circumstance of imminent threat to national security, thus you must take steps in preparation. There is no right nor clear answer, and not even a small decision cannot be made lightly.
Below are various questions about the decisions you think you would make as the president of Liga. Please put yourself in that position, and answer the questions according to your own opinions on each question. It can be hard to make decisions without having all the necessary information that could potentially matter, but please try to make a decision based on the information provided (re-read the paragraph again if needed to answer the questions).
Funding
Sarah DiMuccio received support from the National Science Foundation Graduate Research Fellowship and from New York University.
Data Availability
Complete data for each experiment, the code to reproduce all reported analyses, and each survey containing all administered measures are available at https://osf.io/t8gmh/.
Declarations
Ethics Approval
All studies were approved by the New York University Institutional Review Board. Informed consent was obtained from all participants.
Informed Consents
Participants completed written informed consents before participating and were thoroughly debriefed following completion of the experiment.
Human and Animal Rights
The research reported in this manuscript involved human participants and was therefore submitted to New York University’s Institutional Review Board and accepted.
Conflicts of Interest
The authors have no conflicts of interest to disclose.
Footnotes
Sarah DiMuccio is now at Catalyst.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- Avishai, B. (2020, October 29). Is American tolerance for political violence on the rise? The New Yorker. https://www.newyorker.com/news/daily-comment/is-american-tolerance-for-political-violence-on-the-rise
- Bauer NM, Carpinella C. Visual information and candidate evaluations: The influence of feminine and masculine images on support for female candidates. Political Research Quarterly. 2018;71(2):395–407. doi: 10.1177/1065912917738579. [DOI] [Google Scholar]
- Bem SL. The measurement of psychological androgyny. Journal of Consulting and Clinical Psychology. 1974;42(2):155–162. doi: 10.1037/h0036215. [DOI] [PubMed] [Google Scholar]
- Berke DS, Reidy DE, Miller JD, Zeichner A. Take it like a man: Gender-threatened men’s experience of gender role discrepancy, emotion activation, and pain tolerance. Psychology of Men & Masculinity. 2017;18:62–69. doi: 10.1037/men0000036. [DOI] [Google Scholar]
- Bonanno GA, Jost JT. Conservative shift among high-exposure survivors of the September 11th terrorist attacks. Basic and Applied Social Psychology. 2006;28(4):311–323. doi: 10.1207/s15324834basp2804_4. [DOI] [Google Scholar]
- Bosson JK, Vandello JA. Precarious manhood and its links to action and aggression. Current Directions in Psychological Science. 2011;20:82–86. doi: 10.1177/0963721411402669. [DOI] [Google Scholar]
- Bosson, J. K., Jurek, P., Vandello, J. A., Kosakowska-Berezecka, N., Olech, M., Besta, T., Bender, M., Hoorens, V., Becker, M., Timur Sevincer, A., Best, D. L., Safdar, S., Włodarczyk, A., Zawisza, M., Żadkowska, M., Abuhamdeh, S., Badu Agyemang, C., Akbaş, G., Albayrak-Aydemir, N., … Žukauskienė, R. (2021). Psychometric properties and correlates of precarious manhood beliefs in 62 nations. Journal of Cross-Cultural Psychology, 52(3), 231–258. 10.1177/0022022121997997
- Bosson JK, Vandello JA, Burnaford RM, Weaver JR, ArzuWasti S. Precarious manhood and displays of physical aggression. Personality and Social Psychology Bulletin. 2009;35:623–634. doi: 10.1177/0146167208331161. [DOI] [PubMed] [Google Scholar]
- Bosson JK, Weaver JR, Caswell AT, Burnaford RM. Gender threats and men’s antigay behaviors: The harmful effects of asserting heterosexuality. Group Processes and Intergroup Relations. 2012;15:471–486. doi: 10.1177/1368430211432893. [DOI] [Google Scholar]
- Bowles, N. (2020, February 29). The pied pipers of the dirtbag left want to lead everyone to Bernie Sanders. The New York Times. https://www.nytimes.com/2020/02/29/us/politics/bernie-sanders-chapo-trap-house.html
- Brannon R. Looking at the male role. Psyccritiques. 1976;2:795–796. doi: 10.1037/014712. [DOI] [Google Scholar]
- Bruni, F. (2017, April 1). Manhood in the age of trump. New York Times. https://www.nytimes.com/2017/04/01/opinion/sunday/manhood-in-the-age-of-trump.html
- Carian EK, Sobotka TC. Playing the Trump card: Masculinity threat and the U.S. 2016 presidential election. Socius: Sociological Research for a Dynamic World. 2018;4:1–6. doi: 10.1177/2378023117740699. [DOI] [Google Scholar]
- Caswell AT, Bosson JK, Vandello JA, Sellers JG. Testosterone and men’s stress responses to gender threats. Psychology of Men & Masculinity. 2014;15:4–11. doi: 10.1037/a0031394. [DOI] [Google Scholar]
- Cauterucci, C. (2017, February 7). Trump’s Saturday Night Live fixation reveals the cracks in his fragile masculinity. Slate. http://www.slate.com/blogs/xx_factor/2017/02/07/trump_s_saturday_night_live_fixation_reveals_the_cracks_in_his_fragile_masculinity.html
- Cheryan S, Cameron JS, Katagiri Z, Monin B. Manning up: Threatened men compensate by disavowing feminine preferences and embracing masculine attributes. Social Psychology. 2015;46:218–227. doi: 10.1027/1864-9335/a000239. [DOI] [Google Scholar]
- Cicchirillo V, Hmielowski J, Hutchens M. The mainstreaming of verbally aggressive online political behaviors. Cyberpsychology, Behavior and Social Networking. 2015;18:253–259. doi: 10.1089/CYBER.2014.0355. [DOI] [PubMed] [Google Scholar]
- Cohn AM, Seibert AL, Zeichner A. The role of restrictive emotionality, trait anger, and masculinity threat in men’s perpetration of physical aggression. Psychology of Men & Masculinity. 2009;10:218–224. doi: 10.1037/a0015151. [DOI] [Google Scholar]
- Connell RW. Masculinities. University of California Press; 1995. [Google Scholar]
- Curran PJ, Hussong AM. Integrative data analysis: The simultaneous analysis of multiple data sets. Psychological Methods. 2009;14(2):81–100. doi: 10.1037/a0015914. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Diekman AB, Schneider MC. A social role theory perspective on gender gaps in political attitudes. Psychology of Women Quarterly. 2010;34:486–497. doi: 10.1111/j.1471-6402.2010.01598.x. [DOI] [Google Scholar]
- DiMuccio SH, Knowles ED. The political significance of fragile masculinity. Current Opinion in Behavioral Sciences. 2020;34:25–28. doi: 10.1016/j.cobeha.2019.11.010. [DOI] [Google Scholar]
- DiMuccio SH, Knowles ED. Precarious manhood predicts support for aggressive policies and politicians. Personality and Social Psychology Bulletin. 2021;47(7):1169–1187. doi: 10.1177/0146167220963577. [DOI] [PubMed] [Google Scholar]
- DiMuccio SH, Yost MR, Helweg-Larsen M. A qualitative analysis of perceptions of precarious manhood in U.S. and Danish men. Psychology of Men & Masculinity. 2017;18(4):331–340. doi: 10.1037/men0000062. [DOI] [Google Scholar]
- Donnelly K, Twenge JM. Masculine and feminine traits on the Bem Sex-Role Inventory, 1993–2012: A cross-temporal meta-analysis. Sex Roles. 2017;76(9–10):556–565. doi: 10.1007/s11199-016-0625-y. [DOI] [Google Scholar]
- Dowd, M. (2006, December 13). Will Hillzilla crush Obambi? New York Times. https://www.nytimes.com/2006/12/13/opinion/13dowd.html?
- Ducat SJ. The Wimp Factor: Gender gaps, holy wars, and the politics of anxious masculinity. Beacon Press; 2004. [Google Scholar]
- Fahey AC. French and feminine: Hegemonic masculinity and the emasculation of John Kerry in the 2004 presidential race. Critical Studies in Media Communication. 2007;24:132–150. doi: 10.1080/07393180701262743. [DOI] [Google Scholar]
- Fasteau MF. The male machine. McGraw-Hill; 1974. [Google Scholar]
- Feather NT, O’Driscoll MP, Nagel T. Conservatism, sex-typing, and the use of titles: Miss, Mrs, or Ms? European Journal of Social Psychology. 1979;9(4):419–426. doi: 10.1002/ejsp.2420090407. [DOI] [Google Scholar]
- Frederick DA, Shapiro LM, Williams TR, Seoane CM, McIntosh RT, Fischer EW. Precarious manhood and muscularity: Effects of threatening men’s masculinity on reported strength and muscle dissatisfaction. Body Image. 2017;22:156–165. doi: 10.1016/j.bodyim.2017.07.002. [DOI] [PubMed] [Google Scholar]
- French, D. (2015, September 29). Victim culture is killing American manhood. National Review. https://www.nationalreview.com/2015/09/victim-culture-kills-american-manhood/
- Gallup AC, White DD, Gallup GG. Handgrip strength predicts sexual behavior, body morphology, and aggression in male college students. Evolution and Human Behavior. 2007;28(6):423–429. doi: 10.1016/j.evolhumbehav.2007.07.001. [DOI] [Google Scholar]
- Gilmore DD. Manhood in the making: Cultural concepts of masculinity. Yale University Press; 1990. [Google Scholar]
- Glick P, Gangl C, Gibb S, Klumpner S, Weinberg E. Defensive reactions to masculinity threat: More negative affect toward effeminate (but not masculine) gay men. Sex Roles. 2007;57:55–59. doi: 10.1007/s11199-007-9195-3. [DOI] [Google Scholar]
- Hansen, M. A., Clemens, J. L., & Dolan, K. (2020). Gender gaps, partisan gaps, and cross-pressures: An examination of American attitudes toward the use of force. Politics and Gender, 1–23. 10.1017/S1743923X20000690
- Hinck SS, Hinck EA, Dailey WO, Hinck RS, Hansen E, Madsen B. A tale of two parties: Comparing the intensity of face threats in the Democratic and Republican primary debates of 2012 and 2016. Argumentation and Advocacy. 2018;54:104–121. doi: 10.1080/00028533.2018.1446865. [DOI] [Google Scholar]
- Huber PJ, Ronchetti E. Robust statistics. Wiley; 2009. [Google Scholar]
- Kapur, S. (2021, January 13). “This isn’t the final chapter”: Analyst warns, again, about rise of right-wing extremists. NBC News. https://www.nbcnews.com/politics/politics-news/isn-t-final-chapter-analyst-warns-again-about-rise-right-n1253950
- Katz J. Man enough? Donald Trump, Hillary Clinton, and the politics of presidential masculinity. Interlink Books; 2016. [Google Scholar]
- Kimmel M. Manhood in America: A cultural history. 2. Free Press; 2006. [Google Scholar]
- Kosakowska-Berezecka N, Besta T, Adamska K, Jaśkiewicz M, Jurek P, Vandello JA. If my masculinity is threatened I won’t support gender equality? The role of agentic self-stereotyping in restoration of manhood and perception of gender relations. Psychology of Men & Masculinity. 2016;17:274–284. doi: 10.1037/men0000016. [DOI] [Google Scholar]
- Lance CE, Butts MM, Michels LC. The sources of commonly reported cutoff criteria: What did they really say? Organizational Research Methods. 2006;9:202–220. doi: 10.1177/1094428105284919. [DOI] [Google Scholar]
- Lee-Won RJ, Tang WY, Kibbe MR. When virtual muscularity enhances physical endurance: Masculinity threat and compensatory avatar customization among young male adults. Cyberpsychology, Behavior, and Social Networking. 2017;20(1):10–16. doi: 10.1089/cyber.2016.0418. [DOI] [PubMed] [Google Scholar]
- Lizotte MK. Gender, partisanship, and issue gaps. Analyses of Social Issues and Public Policy. 2017;17:379–405. doi: 10.1111/asap.12144. [DOI] [Google Scholar]
- Lizotte MK. Investigating the origins of the gender gap in support for war. Political Studies Review. 2019;17:124–135. doi: 10.1177/1478929917699416. [DOI] [Google Scholar]
- Maass A, Cadinu M, Guarnieri G, Grasselli A. Sexual harassment under social identity threat: The computer harassment paradigm. Journal of Personality and Social Psychology. 2003;85:853–870. doi: 10.1037/0022-3514.85.5.853. [DOI] [PubMed] [Google Scholar]
- Martens AL, Stratmoen E, Saucier DA. To preserve, protect, and defend: Masculine honor beliefs and perceptions of the 2016 presidential candidates. Basic and Applied Social Psychology. 2018;40(5):308–319. doi: 10.1080/01973533.2018.1500288. [DOI] [Google Scholar]
- Marx Ferree M. The crisis of masculinity for gendered democracies: Before, during, and after Trump. Sociological Forum. 2020;35(S1):898–917. doi: 10.1111/socf.12599. [DOI] [Google Scholar]
- McDermott ML. Masculinity, femininity, and American political behavior. Oxford University Press; 2016. [Google Scholar]
- Messner MA. The masculinity of the governator muscle and compassion in American politics. Gender and Society. 2007;21(4):461–480. doi: 10.1177/0891243207303166. [DOI] [Google Scholar]
- Nail PR, McGregor I. Conservative shift among liberals and conservatives following 9/11/01. Social Justice Research. 2009;22(2–3):231–240. doi: 10.1007/s11211-009-0098-z. [DOI] [Google Scholar]
- Nail PR, McGregor I, Drinkwater AE, Steele GM, Thompson AW. Threat causes liberals to think like conservatives. Journal of Experimental Social Psychology. 2009;45(4):901–907. doi: 10.1016/j.jesp.2009.04.013. [DOI] [Google Scholar]
- Norrander B. The history of the gender gaps. In: Whitaker LD, editor. Voting the gender gap. University of Illinois Press; 2008. pp. 9–32. [Google Scholar]
- O’Dea CJ, Chalman ST, Castro Bueno AM, Saucier DA. Conditional aggression: Perceptions of male violence in response to threat and provocation. Personality and Individual Differences. 2018;131:132–141. doi: 10.1016/j.paid.2018.04.029. [DOI] [Google Scholar]
- Prabhu, M. T. (2016, December 20). Vincent Sheheen calls own Democratic Party leaders ‘sissies.’ The Post and Courier. https://www.postandcourier.com/politics/vincent-sheheen-calls-own-democratic-party-leaders-sissies/article_2ef7983e-c6c7-11e6-b354-c7475862f330.html
- Roberts DC, Utych SM. Linking gender, language, and partisanship: Developing a database of masculine and feminine words. Political Research Quarterly. 2020;73(1):40–50. doi: 10.1177/1065912919874883. [DOI] [Google Scholar]
- Rudman LA, Mescher K, Moss-Racusin CA. Reactions to gender egalitarian men: Perceived feminization due to stigma-by-association. Group Processes and Intergroup Relations. 2013;16(5):572–599. doi: 10.1177/1368430212461160. [DOI] [Google Scholar]
- Saucier DA, Till DF, Miller SS, O’Dea CJ, Andres E. Slurs against masculinity: Masculine honor beliefs and men’s reactions to slurs. Language Sciences. 2015;52:108–120. doi: 10.1016/j.langsci.2014.09.006. [DOI] [Google Scholar]
- Saucier DA, O’Dea CJ, Stratmoen E. Hard targets: Masculine honor beliefs and motivations for muscularity. Psychology of Men & Masculinity. 2018;19(4):547–559. doi: 10.1037/men0000133. [DOI] [Google Scholar]
- Selya, A. S., Rose, J. S., Dierker, L. C., Hedeker, D., & Mermelstein, R. J. (2012). A practical guide to calculating Cohen's f(2), a measure of local effect size, from PROC MIXED. Frontiers in Psychology, 3, 111. 10.3389/fpsyg.2012.00111 [DOI] [PMC free article] [PubMed]
- Sharrow EA, Strolovitch DZ, Heaney MT, Masket SE, Miller JM. Gender attitudes, gendered partisanship: Feminism and support for Sarah Palin and Hillary Clinton among party activists. Journal of Women, Politics and Policy. 2016;37(4):394–416. doi: 10.1080/1554477X.2016.1223444. [DOI] [Google Scholar]
- Shaw, B. P. (2022). Effect sizes for contrasts of estimated marginal effects. The Stata Journal, 22(1), 134–57. 10.1177/1536867X221083901
- Simmons JP, Nelson LD, Simonsohn U. False-positive psychology: Undisclosed flexibility in data collection and analysis allows presenting anything as significant. Psychological Science. 2011;22(11):1359–1366. doi: 10.1177/0956797611417632. [DOI] [PubMed] [Google Scholar]
- Spencer SJ, Steele CM, Quinn DM. Stereotype threat and women’s math performance. Journal of Experimental Social Psychology. 1999;35(1):4–28. doi: 10.1006/jesp.1998.1373. [DOI] [Google Scholar]
- Stanaland A, Gaither S. “Be a man”: The role of social pressure in eliciting men’s aggressive cognition. Personality and Social Psychology Bulletin. 2021;47(11):1596–1611. doi: 10.1177/0146167220984298. [DOI] [PubMed] [Google Scholar]
- StataCorp . Stata statistical software: Release 17. StataCorp LLC; 2021. [Google Scholar]
- Steegen S, Tuerlinckx F, Gelman A, Vanpaemel W. Increasing transparency through a multiverse analysis. Perspectives on Psychological Science. 2016;11(5):702–712. doi: 10.1177/1745691616658637. [DOI] [PubMed] [Google Scholar]
- Stephens, B. L. (2020, January 31). Bernie’s angry bros. The New York Times. https://www.nytimes.com/2020/01/31/opinion/sanders-bernie-bros.html
- te Grotenhuis M, Pelzer B, Eisinga R, Nieuwenhuis R, Schmidt-Catran A, Konig R. When size matters: Advantages of weighted effect coding in observational studies. International Journal of Public Health. 2017;62:163–167. doi: 10.1007/S00038-016-0901-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Terwee CB, Bot SDM, de Boer MR, van der Windt DAWM, Knol DL, Dekker J, Bouter LM, de Vet HCW. Quality criteria were proposed for measurement properties of health status questionnaires. Journal of Clinical Epidemiology. 2007;60(February):34–42. doi: 10.1016/j.jclinepi.2006.03.012. [DOI] [PubMed] [Google Scholar]
- Ursachi G, Horodnic A, Zait A. How reliable are measurement scales? External factors with indirect influence on reliability estimators. Procedia Economics and Finance. 2015;20:679–686. doi: 10.1016/S2212-5671(15)00123-9. [DOI] [Google Scholar]
- Vandello JA, Bosson JK. Hard won and easily lost: A review and synthesis of theory and research on precarious manhood. Psychology of Men & Masculinity. 2013;14:101–113. doi: 10.1037/a0029826. [DOI] [Google Scholar]
- Vandello JA, Bosson JK, Cohen D, Burnaford RM, Weaver JR. Precarious manhood. Journal of Personality and Social Psychology. 2008;95:1325–1339. doi: 10.1037/a0012453. [DOI] [PubMed] [Google Scholar]
- Vescio, T. K., Schermerhorn, N. E. C, Gallegos, J. M., & Laubach, M. L. (2021) The affective consequences of threats to masculinity. Journal of Experimental Social Psychology, 97. 10.1016/j.jesp.2021.104195.
- Vescio TK, Schermerhorn NEC. Hegemonic masculinity predicts 2016 and 2020 voting and candidate evaluations. Proceedings of the National Academy of Sciences of the United States of America. 2021;118(2):1–10. doi: 10.1073/pnas.2020589118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Weaver JR, Vandello JA, Bosson JK. Intrepid, imprudent, or impetuous?: The effects of gender threats on men’s financial decisions. Psychology of Men and Masculinity. 2013;14:184–191. doi: 10.1037/a0027087. [DOI] [Google Scholar]
- Willer R, Rogalin CL, Conlon B, Wojnowicz MT. Overdoing gender: A test of the masculine overcompensation thesis. American Journal of Sociology. 2013;118:980–1022. doi: 10.1086/668417. [DOI] [Google Scholar]
- Winter NJG. Masculine republicans and feminine democrats: Gender and Americans’ explicit and implicit images of the political parties. Political Behavior. 2010;32:587–618. doi: 10.1007/s11109-010-9131-z. [DOI] [Google Scholar]
- Zirin, D. (2017, September 24). The fragile, toxic masculinity of Donald Trump. The Nation. https://www.thenation.com/article/archive/the-fragile-toxic-masculinity-of-donald-trump/
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Complete data for each experiment, the code to reproduce all reported analyses, and each survey containing all administered measures are available at https://osf.io/t8gmh/.