Introduction

Research suggests that sexism has become more covert and subtle (Benokraitis and Feagin 1995; Swim et al. 1995), but still has adverse performance consequences for women. Subtle sexism leads to decreased performance for women (Dardenne et al. 2007), particularly in stereotypically masculine domains, including mathematics (Gervais et al. 2011; Spencer et al. 1999; Vescio et al. 2005), engineering (Logel et al. 2009), and visuospatial tasks (Campbell and Collaer 2009). Subtle sexism also appears to have adverse consequences on women’s performance across many cultures. For example, it has adverse effects on the performance of women in the United States (Gervais et al. 2011; Spencer et al. 1999; Vescio et al. 2005), Canada (Logel et al. 2009), Belgium (Dardenne et al. 2007), Italy (Cadinu et al. 2005), the Netherlands (Barreto et al. 2010) and France (Croizet et al. 2004) just to name a few. Although the detrimental performance consequences of subtle sexism have been documented in the U.S., Canada, and Europe, the mechanisms for these effects have received considerably less attention. This research begins to fill this critical gap in the literature by focusing on whether situational determinants of control may explain gender differences in math performance in response to patronizing behavior, a form of subtle sexism marked by trivial niceties, but group-based inequities in valued resources (Biernat and Vescio 2002; Vescio et al. 2005).

The purpose of the present research was to examine whether contextual features that affect control contribute to gender differences in math performance that emerge from patronizing behavior. Control represents the degree to which one’s actions influence one’s outcomes (Ross and Mirowsky 1989; Rotter 1966). When women experience patronizing behavior, conceptualized as praise paired with devalued positions (Vescio et al. 2005), from a man in a stereotypically masculine domain, they may believe that their performance in those devalued positions will have little to no impact on their subsequent outcomes because the man is sexist and thus, show decreased performance. To examine these possibilities, we conducted two experiments in which men and women undergraduates from a U.S. Midwestern University received patronizing behavior (i.e., praise paired with a devalued position that did not provide an opportunity for monetary rewards) from a male leader in a stereotypically masculine domain. In this context, we manipulated situational determinants of control and measured math performance. Specifically, in Experiment 1, we decreased control by varying whether position assignments were made once at the outset (no control) or multiple times (ambiguous control) by one patronizing man. In Experiment 2, we increased control by varying whether position assignments were made by one man (ambiguous control) or multiple men (enhanced control). Together, Experiments 1 and 2 considered whether two separate, but related situational indicators of control explain gender differences in response to patronizing behavior. To derive testable hypotheses, we review research on patronizing behavior, stereotype threat, and gender differences in control.

Patronizing Behavior

Prejudice against women is not marked by uniform antipathy. Women are stereotypically perceived to be incompetent, but warm (Fiske et al. 2002) and pitied (Cuddy et al. 2007) across many cultures (Cuddy et al. 2009). For this reason, theorists have suggested that sexism is comprised of hostility and benevolence (Glick and Fiske 1996; 2001). Hostile sexism refers to traditional prejudice that is characterized by antagonistic and disrespectful behaviors, like aggression toward and competition with women. Benevolent sexism is characterized by seemingly kind, but condescending and paternalistic behaviors, like helping women with tasks that are stereotypically beyond their reach. Because sexism contains both hostile and benevolent components, it is perceived as less prototypically prejudiced (Inman and Baron 1996) and is less readily identifiable as prejudice (Barreto and Ellemers 2005). For example, Barreto and Ellemers (2005) found that Dutch college students were less likely to identify people who endorsed benevolent (vs. hostile) sexism as sexists because they were perceived as likeable. Thus, the adverse consequences of sexism may not be immediately apparent to either the perpetrators or the targets of sexism (see Bosson et al. 2010, for study with U.S. participants).

Sexist hierarchies, like other group-based inequities, are typically and most effectively maintained through acts of benevolence and paternalism rather than overt acts of hostility and disrespect (Jackman 1994; see also Glick and Fiske 1996; 2001). Benevolent and paternalistic sexism may be hard to detect from the perspective of actors (e.g., “How could I be sexist if I feel positive sentimentality and praise her?”), while having psychologically meaningful consequences for recipients (Gervais and Vescio 2007). For example, benevolent sexism caused decreased performance and feelings of incompetence for Belgian women (Dardenne et al. 2007; Dumont et al. 2010). Benevolent sexism also caused U.S. women to more strongly endorse systems of gender-based inequality (Jost and Kay 2005). Finally, implicit romantic fantasies were associated with less personal power for U.S. women (Rudman and Heppen 2003).

Importantly, the seemingly kind acts of benevolence may also mask gender inequities in the allocation of meaningful resources and have differential effects on men and women. For example, when U.S. male leaders categorized their subordinates on the basis of gender, they praised their female subordinates more than their male subordinates while assigning the same female subordinates more devalued positions that offered them opportunities to earn little or no money compared to male subordinates (Vescio et al. 2005; see also Biernat and Vescio 2002; Biernat et al. 1998). This patronizing behavior, marked by trivial niceties and a pattern of group-based inequities, creates gender differences in performance in stereotypically masculine domains that do not otherwise exist. Vescio et al. (2005), for example, found that both U.S. men and women who were given multiple patronizing evaluations by one male leader reported increased anger, but angry men performed better, whereas angry women performed worse in stereotypically masculine contexts. In the present work, we examined whether situational control contributes to gender differences in response to patronizing behavior.

Stereotype Threat

Patronizing behavior from men in stereotypically masculine domains may cause reduced performance because it is a situation that arouses stereotype threat for women. Stereotype threat is a form of social identity threat (Steele et al. 2002), which emerges when negatively stereotyped people worry that they will confirm stereotypes through their performance (Steele and Aronson 1995). In stereotypically masculine domains, women worry about confirming negative domain-specific gender stereotypes (e.g., women are inferior to men) resulting in decreased performance on math tests for U.S. women (Spencer et al. 1999), engineering tests for Canadian women (Logel et al. 2009), and visuospatial tests for U.S. women (Campbell and Collaer 2009). A considerable body of research suggests that stereotype threat and its consequences follow from disrupted cognition (e.g., decrements in working memory; Schmader et al. 2008), but fewer studies have examined the cues in the social environment that arouse threat in the first place (c.f. Cheryan et al. 2009 and Inzlicht and Ben-Zeev 2000 for studies with U.S. women; Logel et al. 2009 for study with Canadian women).

Patronizing cues in the social environment may be an important cause of stereotype threat for women. Research shows that stereotypic beliefs underlie subtle forms of sexism, including benevolent sexism (Glick and Fiske 1996) and patronizing behavior (Vescio et al. 2005). As a result, receiving praise paired with a devalued position from a man in a stereotypically masculine domain may convey to a woman that the man is sexist and that she has little control in the situation; namely, that it is unlikely that the patronizing and powerful man who controls her desired outcomes will treat her fairly regardless of how she performs.

Gender Differences in Control

We suggest that differences in control may explain gender differences in performance in response to patronizing behavior in stereotypically masculine domains. People who have control perceive the link between their actions and their outcomes (Ross and Mirowsky 1992; Rotter 1966). When control is undermined, people no longer perceive their actions as influencing outcomes. Instead, outcomes are perceived to be determined by external sources, for example powerful others or fate (Ross and Mirowsky 2003).

A close examination of U.S. men and women’s responses to patronizing behavior is consistent with the notion that it may be a situational cue that conveys to women (vs. men) that they have little or no control over their outcomes. Specifically, Vescio et al. (2005) assigned U.S. participants to low power, team member roles and led them to believe that they would be working under the supervision of one male leader in a stereotypically masculine domain. Participants believed that they would engage in three rounds of competition in which they would complete standardized quantitative and analytic problems which were adapted to appear relevant to the masculine context (e.g., football strategy) after being assigned to a valued or devalued position (where they could or could not earn money) and praised or not praised by the male leader. Asking participants to complete three rounds of competition allowed for the possibility that increased performance on the math and logic problems could secure valued position assignments in subsequent rounds. Importantly, gender differences in math performance only emerged in the patronizing condition. When praised and assigned to a devalued position women performed worse whereas men performed better than when they were not praised and assigned to a devalued position or assigned to valued positions, regardless of praise. Why might this have been the case?

Both women and men should perceive patronizing behavior as unfair. Their devalued position assignments do not match their praiseworthy actions. Consistent with this notion, both U.S. women and men reported more anger in the patronizing condition compared to the other conditions (Vescio et al. 2005). Experiencing patronizing behavior from a high power male in a stereotypically masculine domain, however, may represent a very different experience for women and men. On the one hand, female stereotypes underlie patronizing behavior (e.g., Glick and Fiske 1996; Vescio et al. 2005) and the unfair feedback came from an outgroup member in a domain in which women are anxious about confirming negative stereotypes (Shapiro and Neuberg 2007; Spencer et al. 1999). In such situations, patronizing behavior may serve as a cue to women that they are being viewed in sexist ways and will be afforded the same devalued position regardless of how they perform, undermining control and performance. On the other hand, male stereotypes do not underlie patronizing behavior and the unfair feedback came from an ingroup member in a domain in which men are positively stereotyped. As a result, it is unlikely that patronizing behavior would serve as a cue to men that they are being viewed in biased ways and will necessarily be afforded devalued positions in subsequent rounds.

Consistent with the notion that patronizing behavior undermines control for women, but not men, both men and women similarly desired to earn the money associated with success in the experiment and believed that the stereotypically masculine skills needed to succeed in the experiment were related to general life success, but men thought they would perform better than women (Vescio et al. 2005). This is also consistent with prior findings showing that in masculine domains, U.S. women tend to have lower expectations for positive outcomes, less confidence, fewer perceptions of control, and worse performance relative to men (Meece et al. 1982; Parsons et al. 1982; Ryckman and Peckham 1987; Stipek 1984). Additionally, although both men and women were angrier in the patronizing condition, anger was associated with increased performance among men, but unrelated to performance among women (Vescio et al. 2005). Subjective feelings of anger are associated with approach motivation or reparative action only when one has control to alter or escape the adverse situation through the behavioral activation system (BAS, Harmon-Jones et al. 2003; Harmon-Jones and Harmon-Jones 2011). If control was blocked for women, then women would report anger, but anger would be unassociated with BAS or reparative action.

The foregoing considerations point to the possibility that patronizing behavior serves as a situational cue that undermines control for women, but not men. Whereas men who are treated in patronizing ways may not feel stereotype threat and as a result feel control over future outcomes, women may see patronizing behavior as a cue that they are being viewed in sexist ways, and as a result perceive little to no control over future outcomes. To the degree that control is related to increased performance, then men would perform better when treated in patronizing ways and women would perform worse. To be clear, consistent with previous research in the U.S. showing few reliable gender differences in control (e.g., Ross and Mirowsky 2002), we suggest that patronizing behavior from a male leader in a stereotypically masculine domain is a situational cue that undermines control for women, not that women generally perceive less control than men across all situations. This reasoning implies, however, that women’s performance may be increased (eliminating gender differences) when contextual cues convey to women that they have some control over their outcomes. Likewise, gender differences may also be eliminated if control is curtailed for both women and men.

Hypotheses and Overview of the Present Work

To test these possibilities, we conducted a pilot test to examine how men and women perceived and responded to patronizing behavior. We then conducted two experiments in which the context was manipulated to be patronizing—all participants received praise paired with devalued positions from a male leader in a stereotypically masculine domain. All participants were exposed to patronizing behavior because previous research shows that gender differences in performance are more likely to emerge in subtly sexist (vs. non-sexist) situations (Dardenne et al. 2007; Logel et al. 2009; Vescio et al. 2005). Across both experiments, we examined motivation in the masculine context before the patronizing behavior was introduced. Specifically, we assessed performance estimates and desire to succeed in the masculine context. Because this was a stereotypically masculine domain, we expected that both men and women would want to succeed in this domain, but that women would expect worse performance compared to men. This also provides a replication of previous work (Vescio et al. 2005). Specifically, we hypothesized that women would have lower performance estimates than men (Hypothesis 1), but we explored whether men and women would report similar desires to succeed.

Next, the patronizing context was manipulated and participants had no control, ambiguous control (replicating Vescio et al. 2005), or enhanced control. In Vescio et al.’s (2005) original work, the same leader made position assignments before each of three rounds. We conceptualized this as an ambiguous control condition because women may perceive less control than men, given that they are negatively stereotyped in masculine domains. By contrast, men may see each round as an opportunity to improve performance and avoid being the recipient of future patronizing behavior. Thus, we hypothesized (Hypothesis 2) that men would perform better than women in the ambiguous control condition. This was the primary hypothesis of the present work and was tested in both Experiments 1 and 2. Compared to the ambiguous control condition, we removed control by having the patronizing leader make final position assignments for all three rounds at the outset (no control). By removing the possibility that the leader could make different position assignments after the first or second round, we curtailed the possibility that increased performance on the problems could elicit a valued position in subsequent rounds. If differences in control cause gender differences in performance, then men should decrease performance when contextual cues indicate that they have no control over future outcomes, eliminating gender differences.

To provide converging evidence that control is a critical factor, we boosted situational control in Experiment 2 to examine whether gender differences would also be eliminated when women had control. Specifically, we informed participants that different leaders would make decisions across the three rounds of competition (enhanced control). When control was enhanced, men and women believed that performance could influence decisions of future leaders, boosting the possibility that increased performance was related to valued positions in future rounds. As in Experiment 1, we hypothesized that men would perform better than women in the ambiguous control condition (Hypothesis 2). Finally, for Experiment 3, we hypothesized that women in the enhanced control condition would perform better than women in the ambiguous control condition (Hypothesis 3), suggesting that when control is possible, women’s performance can increase even in the face of patronizing behavior.

Experiment 1: Reducing Control

Pilot Study: Gender Differences in Perceived Control

If the above logic holds, then when men and women are treated in patronizing ways we would expect them to perceive patronizing behavior as equally negative, but women should perceive less control than men. We initially tested this idea in a small pilot study. Specifically, 20 undergraduates (10 women) from an introductory psychology course at a large U.S. Midwestern university participated for course credit. Participants were asked to imagine a position that they really wanted (e.g., a dream job) and that would result in desired outcomes (e.g., money, respect). They then imagined that they had to possess stereotypically masculine attributes (e.g., assertiveness, competitiveness) to acquire the position. Next, participants imagined that they were one of 10 people who interviewed for the position and that they received patronizing feedback during the interview; a male interviewer told them that their performance was impressive and that he believed that the participant would be a supportive and central member at the company, school, or on the team, but they were not awarded the position they sought. Then, participants rated the male interviewer’s behavior on six dimensions that we developed to assess negativity (negative, unfair, disrespectful, belittling, condescending, inappropriate) using 9-point scales (endpoints labeled with bipolar adjectives, e.g., 1 = extremely fair, 9 = extremely unfair). When submitted to a principle components factor analysis with a varimax rotation, a single factor solution emerged (as indicated by the scree plot and an eigenvalue >1 criterion). Therefore, a leader behavior index was created by taking the average rating of negativity items (α = .71), with higher numbers indicating more leader negativity. Participants also rated how much they agreed with five statements we developed to assess control (e.g., I could earn the position if I worked hard; attaining the desired position is out of my control—reverse scored) on 9-point scales (1 = strongly disagree, 9 = strongly agree). A control index was created by taking the average rating of the control items (α = .89), with higher numbers indicating more control. Finally, participants indicated their gender.

Participants’ ratings of leader negativity and control were submitted to separate one-way, participant gender (men or women) analyses of variance (ANOVAs). An effect of participant gender emerged on control, F(1,19) = 4.44, p < .05, η 2p  = .20, but not leader ratings, F < 1. Although men (M = 4.65, SD = 1.15) and women (M = 4.95, SD = 1.29) perceived the leader’s patronizing behavior as similarly negative, women (M = 3.50, SD = 2.07) reported less control than men (M = 5.50, SD = 2.17). Despite similarly evaluating powerful others imagined to patronize the self, women perceived less control over future outcomes than did men. Given that women already have lower performance estimates than men in masculine domains and that they experience sexism more than men, women may have been able to easily conjure images of subtly sexist and patronizing behavior, which may serve as a situational cue of stereotype threat and undermine control. Building on this initial finding, we conducted two experiments in which we manipulated control and measured performance.

Method

Participants and Design

One hundred and thirteen undergraduates (60 women) from introductory psychology courses at a U.S. Midwestern university participated for course credit. The undergraduates were recruited to participate in a study on teams through the psychology department participant pool. Participants in this pool were primarily White and aged 18–24. On arrival, participants were randomly assigned to one of two experimental conditions—the ambiguous control condition or the no control condition—with the restriction that there had to be approximately equal numbers of male and female participants in each condition. Thus, Experiment 1 used a participant gender (men or women) X condition (ambiguous control or no control) between participants design.

Procedure

Participants were brought into the lab in mixed gender groups of three to eight people and were led to believe that they would be participating in a contest where the teams would be comprised of participants in each session. Participants completed a leadership skills questionnaire, which they thought would be used to select a leader and team members and some initial questions which they thought would be shared with others to facilitate acquaintanceships among participants (e.g., year in college, age, gender, major, favorite television shows and restaurants). Participants also selected an experimental name for the session.

After participants completed both questionnaires, they were given 5 min to read a written introduction of the study and to examine sample problems. The written introduction explained that they would be participating in an academic contest involving three rounds of competition. Furthermore, the person with the highest score in a leadership skills questionnaire would be assigned to the role of team leader and all other participants would be team members. Participants read that leaders would assign team members to valued positions that would provide participants with opportunities to earn money and devalued positions that would not provide opportunities for monetary rewards.

The competition tasks were described as problems tapping basic academic skills, including: strategic planning, formal systems, and capital gains problems. In reality, the problems were comprised of math and logic problems taken from Scholastic Assessment Test (SAT) and Graduate Record Exam (GRE) study guide books. We altered the context and appearance (e.g., photographs) to create problems that appeared stereotypically masculine. For instance, a standardized GRE logic problem was altered to look like a problem with a game plan for a football team. Participants were then told that a leader and team members had been selected, but that leader and team member assignments were not going to be made public right away. All participants were next given a memo indicating that had been assigned to the role of team member and reminded that leaders had the potential to earn $25 and that team members had the potential to earn $5 if assigned to valued, competitor roles or no money if assigned to devalued, alternate roles.

Motivation

We used the motivation measures from Vescio et al. (2005) in Experiments 1 and 2. Specifically, participants were then asked to indicate their estimations of performance and their desire to succeed in the domain. To assess estimations of performance, participants reported their interest in the challenge problems and their estimates of personal performance. These two items were combined to create a performance estimates score (r = .74), with higher numbers indicating greater performance estimates. To assess desires to succeed, participants reported their desire to attain monetary rewards and perceptions of the degree that the masculine skills required for success in the experiment were important to getting ahead in life. These two items were combined to create a desire to succeed score (r = .71), with higher numbers indicating greater desire to succeed. All responses were given using 9-point scales (1 = not at all, 9 = extremely). We included this measure before the patronizing feedback to see if men and women differed in their initial performance estimates and desire to succeed in the masculine environment (Hypothesis 1).

Feedback

After completing the performance estimates and the desire to succeed items, participants received a handwritten memo from a male leader (identified by his experimental name, “PSU Guy”). The memo praised the participant (i.e., “I liked what you wrote” and “You’ll do great on the problems”) and assigned the participant to the devalued, alternate role, making it impossible for the participant to earn money. All participants received the same memo, but we manipulated control by giving feedback about the context.

In the ambiguous control condition, participants believed that there would be three rounds of position assignments from the same male leader and that they could improve their positions in later rounds through increased performance. Thus, we left control ambiguous by allowing the potential to improve positions in subsequent rounds. The ambiguous control condition was identical to the patronizing condition from Vescio et al. (2005, Study 2). In the no control condition, participants were told that they would be assigned to positions at the outset and would remain in those positions across all three rounds. Thus, we severed performance from future position assignments, decreasing control.

A separate sample of nine undergraduate students (seven women) in a classroom setting imagined a scenario in which they received patronizing behavior from a male leader in a stereotypically masculine domain paired with a description of the ambiguous control condition (i.e., the position assignments for Rounds 2 and 3 would be made by the same male leader), the no control condition (i.e., the position assignments for Rounds 2 and 3 would be the same as those made by the male leader for Round 1), and the enhanced control condition (i.e., the position assignments for Rounds 2 and 3 would be made by different leaders). Order was counterbalanced. Participants rated the degree to which they agreed (1 = strongly disagree, 9 = strongly agree) with the 3 statements derived from the pilot study assessing control. The statements included: “I could do well enough on the problems to get a playing position in Rounds 2 and 3”; “I could earn a playing position in Rounds 2 and 3 if I worked hard”; “Attaining a playing position for Rounds 2 and 3 is out of my control” (reverse-coded). Control scores were calculated for the ambiguous control condition (α = .92), no control condition (α = .98), and enhanced control condition (α = .81), with higher numbers indicating more control. An effect of scenario, F(1,8) = 61.88, p < .0001, η 2p  = .95, revealed that people perceived the least control in the no control condition (M = 1.30, SD = .89), followed by the ambiguous control condition (M = 6.37, SD = 2.14), followed by the enhanced control condition (M = 8.07, SD = .94). The differences between all of the conditions were significant, ps < .0001—.05, providing the necessary information required to interpret the manipulations as differing in situational control with the no control condition (Experiment 1) indicating the least control, the ambiguous control condition (Experiments 1 and 2) indicating moderate control, and the enhanced control condition (Experiment 2) indicating the most control.

Performance

Participants were then presented with 12 quantitative and analytic problems (see Vescio et al. 2005). As noted above, these were standardized SAT and GRE problems adapted to appear stereotypically masculine. Problems of moderate difficulty were selected; each required careful attention, but increased effort on the problems would increase performance for the majority of college students. All participants were given the same problems and were given 10 min to complete them. After participants completed each problem, however, they were asked to give their answer to the experimenter so that the experimenter could ostensibly deliver the problems to the leader.

After 10 min had elapsed, the experimenter collected the unanswered problems and remaining answer sheets. After a short break, they received a memo from the leader. In the ambiguous control condition, participants received a second memo that was consistent with the first memo with slight changes to alter believability stating, “Great job on the problems” and “Keep up the good work” paired with a devalued position. In the no control condition, participants simply received the memo that they had initially received from the leader to remind them which position the leader had assigned them to at the beginning of the competition. Once again participants were given 10 min to complete 12 problems and the experimenter ostensibly delivered answers from completed problems to the leader. After 10 min had elapsed, the experimenter told participants that they would not have time for a third round. As in Vescio et al. (2005), participants were given 1 point for each correctly answered problem to create a performance measure (α = .68), with higher numbers equally better performance (ranging from 0 to 24). After the participants were probed for suspicion, they were thanked and thoroughly debriefed.

Results and Discussion

Descriptive Statistics

As shown in Table 1, performance estimates and desire to succeed were correlated for both men, r = .60, p < .0001, and women, r = .64, p < .0001, but were unrelated to actual performance for men and women, rs -.04–.21, ps > .12. Additionally, we submitted performance estimates, desire to succeed, and performance scores to a participant gender (men or women) X condition (ambiguous control or no control) between participants multivariate analysis of variance (MANOVA). This yielded only a significant main effect of participant gender, F(3, 109) = 9.06, p < .0001, Wilk’s Λ = .79. Neither the main effect of condition, nor the interaction between participant gender and condition was significant, Fs < 1.25. To test our specific hypotheses, we submitted performance estimates, desire to succeed, and performance scores to separate participant gender X condition ANOVAs.

Table 1 Descriptive statistics and inter-correlations between measures by participant gender (Experiment 1)
Full size table

Motivation

We hypothesized that men would have higher performance estimates than women (Hypothesis 1), but did not expect gender differences on desire to succeed in the stereotypically masculine domain before patronizing behavior was introduced. To test Hypothesis 1, performance estimates and the desire to succeed measures were submitted to separate participant gender (men or women) X condition (ambiguous control or no control) between participants ANOVAs. Consistent with Hypothesis 1 and replicating previous work, a significant main effect of participant gender emerged on performance estimates, F(1,112) = 21.27, p < .0001, η 2p  = .17. Men (M = 5.88, SD = 1.78) estimated better performances than women (M = 4.93, SD = 1.71, see Table 1). Neither the main effect of condition, nor the interaction between participant gender and condition were significant for performance estimates, Fs < 1. As well, no reliable effects emerged on desire to succeed, Fs < 1.

Performance

We hypothesized that men would perform better than women in the ambiguous control condition (Hypothesis 2), but that men’s performance would decrease with men and women performing similarly in the no control condition. To test Hypothesis 2, we submitted performance to a participant gender (men or women) X condition (ambiguous control or no control) between participants ANOVA and estimated a priori contrasts testing for gender differences in performance within each condition. Overall, men (M = 10.98, SD = 3.73) performed better than women (M = 9.42, SD = 2.80) as indicated by a significant main effect of participant gender, F(1,112) = 6.10, p < .02, η 2p  = .05 (see Table 1). The interaction between condition and participant gender did not approach significance, F < 1. However, a priori contrasts revealed gender differences in performance in only one condition, as hypothesized. As shown in Table 2, consistent with Hypothesis 2, men performed better than women in the ambiguous control condition, F(1,112) = 5.28, p < .024, η 2p  = .05. No reliable differences emerged in the no control condition, F < 1. As well, we found similar effects when analyzed the performance from Round 1 and Round 2 separately.

Table 2 Performance as a function of gender and control (Experiment 1)
Full size table

If the gender differences in performance are a result of gender differences in control, then we should be able to manipulate situational control to affect performance in two different ways. First, as in Experiment 1, if we eliminate control by having position assignments made at the outset of the study, then men should perform less well and similarly to women. Second, if we enhance control by having multiple leaders make position assignments across rounds, then women should perform better and similarly to men in the ambiguous control condition. We turn attention to the latter possibility in Experiment 2.

Experiment 2: Boosting Control

Though women may generally view patronizing situations as uncontrollable and underperform relative to men, we reasoned that if we manipulated the context to include features that afforded more control, that women would perform similarly to men and gender differences would be eliminated. More specifically, in Experiment 2, we enhanced control by providing additional opportunities to attain valued positions. Harmon-Jones et al. (2003) found that when U.S. participants who were opposed to a tuition hike were told that a tuition increase was likely, they reported high levels of anger. Only those participants who had control (e.g., could sign a petition to stop the hike), however, showed increased relative mid-frontal cortical activity, which is the area of the brain associated with BAS and actions to ameliorate anger-inspiring situations. Thus, control can be situationally manipulated. Like the ambiguous control condition of Experiment 1, in the ambiguous control condition of Experiment 2 participants received repeated feedback (all patronizing) from a single male leader who made task assignments and provided feedback across all three rounds. By contrast, in the enhanced control condition, participants were led to believe that they would receive feedback and position assignments from different leaders in each round.

We hypothesized that women in the enhanced control condition would perform better than women in the ambiguous control condition (Hypothesis 3); women’s performance in the enhanced control condition would be similar to men’s performance in the ambiguous control condition. We were tentative regarding the effect that the manipulation of control would have on gender differences in the two conditions. We saw no reason to believe that the manipulation of control would not affect men, as well as women. To the degree that we enhance control among both men and women, gender differences may persist across conditions, but women’s performance should be higher in the enhanced control than the ambiguous control condition. Critically, if gender differences in control are the cause of performance differences in the ambiguous control condition (Hypothesis 2), as we have suggested, women’s performance in the enhanced control condition should be similar to men’s performance in the ambiguous control condition. We explored these possibilities in Experiment 2.

Method

Participants and Design

One hundred and thirty-two undergraduates (53 women) from introductory psychology courses at a U.S. Midwestern University participated for course credit. The undergraduates were recruited to participate in a study on teams through the psychology department participant pool. Participants in this pool were primarily White and aged 18–24. Men and women were randomly assigned to one of two conditions—the ambiguous control condition (from Experiment 1 and Vescio et al. 2005) or the enhanced control condition. Thus, Experiment 2 used a participant gender (men or women) X condition (ambiguous control or enhanced control) between participants design.

Procedure

The procedure for Experiment 2 was exactly the same as Experiment 1, with one exception. The no control condition of Experiment 1 was replaced with an enhanced control condition. In the enhanced control condition, participants were told that separate male leaders would make decisions before each of the three rounds of competition. To bolster the manipulation, feedback came from seemingly different leaders with different experimental names and with different handwriting. We did this to make clear to participants that the decisions of one leader were completely separate of the decisions of subsequent leaders.

As in Experiment 1, we used the motivation and performance measures from Vescio et al. (2005) in Experiment 2. Performance estimates (r = .85) and desire to succeed (r = .80) scores were created, with higher numbers indicating greater performance estimates and greater desire to succeed, respectively. Finally, participants were given 1 point for each correctly answered problem to create a performance measure (α = .70), with higher numbers indicating better performance (ranging from 0 to 24).

Results and Discussion

Descriptive Statistics

Performance estimates and desire to succeed were correlated for both men, r = .48, p < .0001, and women, r = .57, p < .0001, but were unrelated to actual performance for men and women, rs .01–.13, ps > .20 (see Table 3). Additionally, we submitted performance estimates, desire to succeed, and performance scores to a participant gender (men or women) X condition (ambiguous control or enhanced control) between participants MANOVA. This yielded only a significant main effect of participant gender, F(3, 128) = 5.47, p < .002, Wilk’s Λ = .88. Neither the main effect of condition, nor the interaction between participant gender and condition was significant, Fs < 1.61. To test our specific hypotheses, we submitted performance estimates, desire to succeed, and performance scores to separate participant gender X condition ANOVAs.

Table 3 Descriptive statistics and inter-correlations between measures by participant gender (Experiment 2)
Full size table

Motivation

We hypothesized that men would have higher performance estimates than women (Hypothesis 1), but did not expect gender differences on desire to succeed in the stereotypically masculine domain before patronizing behavior was introduced. Replicating Experiment 1 and consistent with Hypothesis 1, a significant main effect of participant gender emerged on performance estimates, F(1,112) = 10.10, p < .002, η 2p  = .08. Men (M = 5.92, SD = 1.62) estimated greater performance on the challenge problems than women (M = 4.96, SD = 1.74). Neither the main effect of condition, nor the interaction between participant gender and condition were significant for performance estimates, Fs < 1. As well, no reliable effects emerged on desire to succeed, Fs < 1.78.

Performance

Performance was submitted to a participant gender (men or women) X condition (ambiguous control or enhanced control) between participants ANOVA and a priori contrasts were estimated to test for condition differences in performance within gender. Two significant effects emerged from this analysis. First, a main effect of gender, F(1,131) = 3.97, p < .05, η 2p  = .03, revealed that men (M = 10.29, SD = 3.48) performed better than women (M = 9.08, SD = 3.30). Second, a main effect of condition, F(1,131) = 3.95, p < .05, η 2p  = .03, indicated superior performances in the enhanced control condition (M = 10.43, SD = 3.61) than in the ambiguous control condition (M = 9.14, SD = 3.16).

Though the condition X gender interaction did not approach significance, F < 1, the means additively combined and the contrasts revealed the hypothesized effects. Consistent with Hypothesis 3, women performed better in the enhanced control than the ambiguous control condition, F(1,131) = 4.59, p < .034, η 2p  = .03, whereas men performed similarly across conditions, F < 1. In addition, as an examination of Table 4 reveals, women’s performance in the enhanced control condition was similar to men’s performance in the ambiguous control condition. Men’s performance also increased, but not significantly (F = 1.89, p = .17). We also estimated the magnitude of gender differences in performance within condition to facilitate comparisons of effects across Experiment 1 and Experiment 2. Replicating the findings of Experiment 1 and consistent with Hypothesis 2, there were gender differences on performance in the ambiguous control condition, F(1,131) = 3.81, p < .0501, η 2p  = .03, where men performed better than women. In the enhanced control condition, however, performance did not vary as a function of participant gender, F < 1.

Table 4 Performance as a function of gender and control (Experiment 2)
Full size table

In summary, in Experiment 2, participants in the enhanced control condition were led to believe that they would be assigned positions by different leaders in three rounds. By contrast, participants in the ambiguous control condition were led to believe that all decisions would be made by a single leader. As hypothesized, women performed better in the enhanced control condition than in the ambiguous control condition (consistent with Hypothesis 3). Men’s performance did not reliably vary as a function of condition. Estimates of the magnitude of gender differences in performance within condition further revealed significant gender differences in the ambiguous control condition – men performed better than women (consistent with Hypothesis 2). There were not reliable differences in the enhanced control condition.

General Discussion

The goal of the present research was to examine the suggestion that patronizing behavior from male leaders in stereotypically masculine domains arouses decreased control for women, but not men. To test this possibility, we introduced patronizing behavior with all participants receiving praise and a devalued position on a team in a stereotypically masculine domain and measured performance across multiple rounds of competition. We then manipulated control; specifically, we included a no control condition (the leader made position assignments for all rounds at the outset) in which performance could not affect position assignments in subsequent rounds, an ambiguous control condition (the same leader made position assignments across rounds) in which performance may or may not affect position assignment in subsequent rounds, and an enhanced control condition (different leaders made position assignments across rounds) in which performance would affect position assignments in subsequent rounds, given that the decisions of the leaders were completely separate. The findings from our pilot study and two experiments provide converging evidence that situational control contributes to gender differences in performance.

In the pilot study, we that found that men and women perceived patronizing male leaders similarly and rather negatively, but women perceived less control than did men. Although the patronizing behavior was perceived as similarly negative by men and women, it may have effectively severed control among women, but not men. The control explanation suggests that men’s performance would decrease to levels similar to those of women if control was reduced. In Experiment 1, we reduced control by severing performance from future position assignments. Gender differences only emerged in the ambiguous control condition. In the no control condition, there were no reliable gender differences. Inversely, the control explanation suggests that women’s performance would increase to levels similar to those of men if control was enhanced. In Experiment 2, we enhanced control by allowing multiple male leaders to make position assignments across rounds. When participants were introduced to multiple male leaders who would make position assignments across rounds, women’s performance increased and gender differences in performance again became non-significant.

Together, the findings across experiments suggest that control is an important contributor to gender differences in performance in response to patronizing behavior in stereotypically masculine domains. When control was eliminated (Experiment 1) gender differences in performance became non-significant. By contrast, when control was enhanced (Experiment 2) gender differences in performance again became non-significant because women’s performance increased and became more similar to men’s. In these conditions, men’s performance also increased, but key to this finding is that women’s performance in the enhanced control condition was comparable to men’s performance in the ambiguous control condition. Furthermore, men and women similarly desired to succeed in the stereotypically masculine domain before the patronizing behavior was introduced, so our results cannot be explained by women’s lack of motivation in the masculine context.

Limitations

There are a few limitations to the present research. First, we did not include a non-patronizing control condition. Previous research shows that gender differences in performance emerge more in subtly sexist situations (e.g., Dardenne et al. 2007; Logel et al. 2009). Vescio et al. (2005), for example, only found gender differences in the patronizing condition when U.S. participants received praise paired with a devalued position. No gender differences emerged when participants were treated in non-sexist ways or received no praise paired with a devalued position or when participants received valued positions, regardless of praise. Because our research was designed to examine why gender differences emerged in patronizing situations, rather than whether gender differences sometimes emerge in non-patronizing situations, we only included patronizing situations in the present work. As a result, it is possible that women would show decreased control and performance, regardless of whether they received patronizing behavior or not. This seems unlikely, however, given that previous research shows that gender differences in performance are more likely to emerge when women experience subtle sexism (Dardenne et al. 2007; Vescio et al. 2005) and gender differences in control are not stable across situations (Ross and Mirowsky 2002). Additionally, this explanation is inconsistent with our results. We found that women’s performance increased when women were patronized in the enhanced control conditions. Another concern might be that women and men felt less patronized in the no control condition compared to the ambiguous control condition because the exact same patronizing feedback was provided in the former and different patronizing feedback was provide in the latter. Although the different patronizing feedback was designed to be equivalent in the control and ambiguous control conditions, future research could further rule out this possibility.

Additionally, our conclusion that control is a critical factor explaining gender differences in performance as a result of patronizing behavior was primarily supported by the hypothesized pattern of significant a prior contrasts. The two-way interaction did not approach significance in either experiment. Instead, the means additively combined such that the planned contrasts revealed the hypothesized effects. Additionally, from a statistical point of view, the effect sizes across Experiment 1 and Experiment 2 are small – participant gender explains approximately 3%-5% of the variance in performance in the ambiguous control condition and far less (essentially none) of the variance in the experimental conditions. A statistically small effect size, however, should not be confused with practically insignificant consequences. Though statistically small, these effects have many practical implications for women who find themselves in sexist masculine domains. If a single encounter with patronizing behavior can account for 3% to 5% of the variance in performance differences among men and women in a controlled experimental setting, one can easily imagine the cumulative effects across time when women are repeatedly exposed to patronizing behavior in stereotypically masculine domains (e.g., science, technology, engineering, and math, STEM contexts).

Implications and Future Directions

The present work has several theoretical implications. First, this study helps to fill a critical gap in the extant literature of stereotype threat. A considerable body of research suggests that stereotype threat and its consequences follow from disrupted cognition (e.g., decrements in working memory; Schmader et al. 2008). However, only a few stereotype threat studies have examined interpersonal cues in the social environment, which may elicit stereotype threat. For example, Logel et al. (2009) found that interacting with a man who displayed nonverbal behavior that seemed to show sexual interest (such as sitting closer) undermined engineering performance for Canadian women. Similarly, in the present work, patronizing behavior from a male leader in a stereotypically masculine domain undermined performance for women. Importantly, and extending previous research, performance decrements were eliminated for women when women interacted with multiple male leaders. Although they were initially patronized, interacting with multiple, different male leaders left open the possibility that a subsequent leader would advance them to a valued position on the basis of their performance.

Second, our findings are consistent with the suggestion that we made at the outset that subtle sexism undermines control and has implications for the BAS. Vescio et al. (2005) found that although men and women reported increased anger when they were the targets of patronizing behavior, anger was associated with increased performance for men but not women. Subjective feelings of anger are associated with approach motivation or reparative action only when one has control to alter or escape the adverse situation through BAS (Harmon-Jones et al. 2003; 2011). If control was blocked for women, then women would report anger, but anger would be unassociated approach motivation or reparative action. Consistently, in the pilot study reported here, men reported perceptions of greater control than did women in response to a patronizing scenario. Additionally, when control was decreased (in Experiment 1) or increased (in Experiment 2) gender differences in performance were eliminated. Although our work is consistent with this premise, whether patronizing behavior leads to differential BAS activation at a physiological level for men and women remains an empirical question for future research.

Finally, we focused on a sample of U.S. men and women in the current work and future research should examine whether the effects hold in different cultures. Previous research indicates that sexist behavior and environments result in decrements in performance in stereotypically masculine domains for women in the U.S. (Gervais et al. 2011; Vescio et al. 2005), Canada (Logel et al. 2009), and Europe (Barreto et al. 2010; Cadinu et al. 2005; Croizet et al. 2004; Dardenne et al. 2007), so we might expect similar effects for replications in North America and Europe. It remains less clear, however, whether these results would be applicable to other cultures, for example Asian cultures. To our knowledge, the only published study examining math performance for women in China showed no stereotype threat effects (Tsui et al. 2011). It is possible that a combination of government policy that stresses gender equality and a rigorous national curriculum focusing on superior math and science skills may reduce some of the effects of stereotype threat, patronizing behavior, or subtle sexism more generally in China. Future research is needed to further explore this very important issue across cultures.

This research also has some practical implications. For example, when men and women have equal control, anger and other related approach-oriented emotions, including determination (e.g., Harmon-Jones et al. 2011), could be aroused as unbiased motivating factors to enhance performance for everyone. For example, receiving a devalued position paired with praise could be perceived as an indicator of encouragement. People who are initially assigned to low status positions within a company may excel if they believe that their actions may help them move through the ranks to high status positions. Although they may not have initially received a valued position, praise may provide encouragement that with increased performance a better position could be acquired in the future.

Additionally, an examination of the features of patronizing behavior in the current experiments also suggests that men and women could be patronized with less damaging effects by changing the features of the situation. In our experiments, a constellation of features came together to make patronizing behavior particularly damaging – it came from a male leader in a masculine domain. The present research suggests that organizations could consider having employees receive feedback from multiple leaders. Even if women interacted with one patronizing leader, knowing that multiple leaders would be making decisions may help women to maintain control. Although we did not examine other features of the situation in the present work, changing the power dynamic, the gender of the leader, or the stereotypicality of the domain, may also be ways that organizations could eliminate gender differences in performance.

Conclusion

Although subtle prejudice may go by unnoticed by perpetrators, it is keenly felt by targets (Jackman 1994). These studies indicate that the effects of subtle sexism are not subtle to the women who are targeted. Importantly, these studies also shed light on ways to reduce the adverse effects of patronizing behavior for women. Even when patronized, women increased performance when they had control. By increasing control, men and women may perform optimally in stereotypically masculine domains.