Gender bias triggers diverging science interests between women and men: The role of activity interest appraisals
Women leave science fields at greater rates than men, and loss of interest is a key motivator for leaving. Although research widely demonstrates effects of gender bias on other motivational processes, whether gender bias directly affects feelings of interest toward science activities is unknown. We used a false feedback paradigm to manipulate whether women (Study 1) and men (Study 2) participants perceived the reason for feedback as due to pro-male bias. Because activity interest also depends on how students approach and perform the activity, effects of biased feedback on interest appraisals were isolated by introducing gender bias only after the science activity was completed. When the feedback was perceived as due to pro-male bias, women (Study 1) reported lower interest and men (Study 2) reported greater interest in the science activity, and interest, in turn, positively predicted subsequent requests for career information in both studies. Implications for understanding diverging science interests between women and men are discussed.
KeywordsInterest Motivation Discrimination Stereotypes Gender bias Women in science
Among undergraduate students who initially choose a Science, Technology, Engineering, or Math (STEM) major, attrition is far greater for women than men (AAUW 2010). Commonly cited reasons for why women switch out of these majors emphasize deficits in ability, preparation, and self-efficacy (see AAUW 2010 for review). However, when Seymour and Hewitt (1997) interviewed women who switched, the top two reasons identified for switching were a lack or loss of interest in STEM majors, and the belief that non-STEM majors held greater interest. Although women also cited other reasons such as struggles to maintain social belonging and feelings of competence, the most commonly cited reason for leaving in these interviews was based on their degree of interest toward the content and activities in STEM. This pattern has recently been confirmed in a large scale study of first semester senior undergraduates in nine institutions who were identified as having rigorous preparation in STEM (Renninger and Schofield 2012). Comparing those who chose to stay versus leave STEM, both groups reported equally high value of STEM fields, and issues of competence and problems with pedagogy similarly emerged among both groups. For both men and women, the major difference between those who stayed in STEM and those who left was interest in the STEM field relative to interest in other fields. Together, these data suggest that feelings of interest toward STEM content and activities are key and equally important for both men and women in determining whether they continue in STEM fields. The question then arises as to why women might experience lower interest than men once they engage in STEM activities. As Renninger and Schofield’s (2012) findings suggest, interest is not just the result of factors related to ability or efficacy; interest can be affected even when these factors are not (Ainley 2006; Bergin 1999; Renninger and Hidi 2011).
Interest has been identified as a positive emotion that is strongly associated with approach motivation (Izard 1977; Silvia 2008; Tomkins 1962). Interest motivates exploration, focused attention, and persistence (Berlyne 1966; Schiefele et al. 1992), thereby promoting the development of competence, skills, and knowledge about the object or topic of exploration (Berlyne 1978; Izard and Ackerman 2000). The approach motivation associated with interest also serves as a counterweight to avoidance-oriented feelings of frustration, boredom, and confusion (Katz et al. 2006). Its unique appraisal structure, functions, and the combination of attentional, cognitive, and affective components that comprise the experience of interest make it distinct from other positive emotions (e.g., happiness) and positive mood (Silvia 2008; Thoman et al. 2011).
Does gender bias affect women’s appraisals of science activity interest?
Gender bias reflecting widespread cultural stereotypes of men’s greater science competence than women remains prevalent in science. For example, Moss-Racusin and colleagues found that science faculty were more likely to recommend hiring, mentoring, and paying more to a man (John) than woman (Jennifer) when reviewing identical lab manager applications (Moss-Racusin et al. 2012). In this case, the “applicants” were not real, but research suggests that awareness or suspicion of gender bias, or stereotypes favoring men in science more broadly, can influence women’s motivation to persist in science.
Psychological explanations for gender differences in persistence have focused on how negative stereotypes toward women in science weaken performance, achievement motivations (e.g., expectations for success and value of competence) and goals, and sense of belonging in the field (see Inzlicht and Schmader 2011; Thoman et al. 2013 for reviews). Research has also examined how these stereotypes make women feel about themselves, in terms of perceptions of ability (Rosenthal et al. 2007; Smith et al. 2013) and self-esteem (Major and O’Brien 2005). Only one study to date has examined how these psychological processes triggered by gender stereotypes can affect one’s subsequent experience of interest in a STEM activity. Smith et al. (2007) found that when the stereotype that men outperform women in math was salient (vs. invalidated) at the outset of a computer science activity, achievement-oriented women were more likely to adopt performance avoidance goals before starting the activity, which was associated with lower task absorption while engaged and lower subsequent interest in the computer science activity. These data illustrate that negative gender stereotypes can change how women approach and engage with a science activity, which can then influence their subsequent ratings of interest in the activity.
Although feelings of interest toward science-related activities are an important source of motivation to persist in science domains (Renninger and Schofield 2012; Sansone et al. 2015; Seymour and Hewitt 1997), there is no research on whether gender bias can shape activity interest appraisals even when the pattern of engagement with the activity is not affected. That is, gender stereotypes might not only change how individuals feel about themselves or the goals and strategies they adopt when approaching an activity. The feelings engendered by feedback that occurs during a given activity can become part of how individuals define the activity (cf. Sansone and Berg 1993), coloring the feelings engendered by features of the task itself. In contexts where women perceive gender bias in the feedback, negative feelings may become associated with the activity, leading women to appraise the activity as relatively uninteresting.
To empirically isolate effects of gender-biased feedback on interest appraisals from gender-bias effects on how individuals approach and interact with the activity, we experimentally introduced a gender bias manipulation after activity completion. We then examined effects on subsequent appraisals of interest in the activity. Previous research has shown that interest appraisals can change after activity completion when subsequent conversations conveyed lack of attention and interest by others (Pasupathi and Rich 2005; Thoman et al. 2007, 2012). In the present studies, we examined whether experiencing gender bias similarly affects appraisals of interest after completing the activity. In this case, feedback conveying gender bias could change how women and men appraise their interest in the activity even though actual engagement does not change.
We expected that women who were hurt by gender bias following completion of a science activity would subsequently view that activity as less interesting than women who were not harmed by gender bias. Further, we tested whether this bias experience can affect interest appraisals even when accounting for potential effects on one’s expectations and value of competence in the task. As noted, these achievement-based expectancy-value variables have been shown to be affected by gender bias and can also influence one’s experience of interest (Sansone and Thoman 2005). Previous work has demonstrated that attributing the reason for negative feedback as due to gender bias can help women protect positive perceptions of competence and self-esteem that would otherwise typically be hurt by negative competence feedback (Crocker and Major 1989). Attributing negative feedback to gender bias can also cause women to devalue their competence or achievement in science (Major et al. 1998), which could in turn influence interest appraisals. However, effects on interest appraisals could occur separately from these achievement-based expectancy or value variables. That is, although effects of gender bias on interest can result from downstream consequences of effects on expectancy value variables, it is possible for women to maintain their perceptions and value of competence and self-esteem in the face of gender-biased feedback but still appraise the activity as less interesting. In fact, similar to the logic of Crocker and Major (1989), as well as self-evaluation maintenance theory (Tesser et al. 1996) and implicit self-esteem compensation theory (Rudman et al. 2007), rather than accept the implications of the feedback for one’s self concept or esteem, it may be less threatening to associate negative feelings with appraisals of the activity.
Does benefiting from gender bias affect men’s appraisals of science activity interest?
We also examined the effects of benefiting from gender bias on men’s interest appraisals, but the expected direction of the effect was less clear. Little empirical research has examined implications for those who benefit from gender bias, and no work has considered effects on feelings toward the activity. Existing research suggests possible competing hypotheses regarding the direction of the potential effect of benefiting from gender bias on men’s appraisals of science activity interest. On one hand, unfairly over-benefiting or thinking about one’s group’s privilege can trigger guilt and other negative feelings (e.g., Branscombe 1998; Mallett and Swim 2007; Schmitt et al. 2000), which could negatively affect men’s appraisals of activity interest. On the other hand, social cues that one’s group membership is a positive attribute in a given domain can increase interest in (Martin and Ruble 2004), value of (Sechrist and Stangor 2001), and future motivation for (Sherman 1981) that and related activities (Patterson and Bigler 2007). Thus, it is possible that gender-based differences in motivation to select and persist in STEM come from boosts in interest when men benefit from bias, in addition to declines for women who are harmed. Finally, potentially reconciling these competing hypotheses, it is also possible that positive effects only occur when men are unaware that the bias is operating and hurting someone else. We test these possibilities for men who benefit from science gender bias in Study 2.
Overview of studies
We developed a lab paradigm in which student participants worked alongside an opposite-sex confederate on a forensic science task, and examined whether performance feedback that was biased in favor of men affected interest in the task for both women who were hurt by (Study 1) and men who benefited from the same biased feedback (Study 2). To create the perception of pro-male bias, we created a situation that was similar to what real applicants might have experienced in the study by Moss-Racusin and colleagues. Specifically, we made it appear as if the man and woman performed equally well on the task, but the man was selected as the outstanding group member. To isolate the effects of being selected (or not) from being selected because of pro-male bias, we compared feedback that was clearly attributable to gender-bias to feedback when the reason for receiving less/more positive feedback for similar work was unknown. We also included a no feedback control condition to serve as a baseline of activity interest relative to the other conditions, in order to help interpret directions of any effects associated with receiving more/less positive feedback (Sansone 1986).
In Study 2, we also examined whether effects for men who benefited from gender bias differed when the woman ostensibly hurt by the feedback explicitly expressed her unhappiness with the biased feedback. In both studies, we included self-esteem as an outcome measure to test whether the feedback manipulation influenced students’ perceptions of themselves, in addition to or instead of perceptions of the activity. Although we expected that the feedback could directly affect interest appraisals, we also included measures of perceived competence and competence value to test an alternative hypothesis that feedback effects on interest appraisals could be explained by indirect effects on expectation and value of achievement on the activity. Support for this alternative hypothesis would suggest that gender biased feedback affects interest through previously established effects on expectancy-value processes, rather than having an independent effect.
In addition to measuring our key outcome, appraisals of activity interest, in both studies we included a behavioral measure of career interest, operationalized as the request for information on related careers. Although one’s appraisal of activity interest is the focal theoretical outcome variable in our analysis, because activity interest tends to strongly predict interest in related careers (Su et al. 2009) one important implication of effects on activity interest appraisals could be downstream effects on women’s career interest. We explore this potential implication by testing whether the feedback manipulation would influence participants’ requests for related career information through its effects on activity interest appraisals.
Women participants were randomly assigned to one of three experimental conditions: a baseline no feedback condition and two conditions where women received feedback that the male peer had been selected as the outstanding group member despite similar levels of performance. They “accidently overhead” a conversation that provided either no reason for this selection or a pro-male bias reason.1
Ninety-six women undergraduates from the participant pool participated for extra-credit. Due to technical errors (e.g. volume was too low on recorder, batteries died during message), 11 of these participants’ sessions were incomplete, and their data were therefore unable to be included. Data from six participants with completed sessions were excluded prior to analyses because of high suspicion reported in debriefing.2 The final sample consists of 79 women (84.7 % European-American, M age = 20.11).
Women experimenters ran participants in groups of two: one participant and one male confederate. Confederates3 were trained to be polite but have minimal interaction with participants. Upon starting the session, participants and confederates were seated on opposite sides of a cubical and could not see one another. There were supposedly two sessions, with the person who did best on the first task selected as “outstanding group member” and chosen to lead the subsequent group session (Major et al. 2003). Participants completed a sample of the “first” task (a science learning activity) followed by baseline activity perceptions measures. Participants then read five “forensic mystery stories designed to help students develop scientific observation and reasoning skills” and selected the clues that most likely lead the forensic expert to his conclusion. The activity was framed as a set of CSI-style learning stories in which details of each story and clues to each mystery emphasized chemical, biological, or physical science knowledge, and students needed to decide which clues were relevant to the solution. Thus, the format was designed to be interesting and the content focused on scientific knowledge and scientific reasoning.4 The ostensible male participant was supposedly participating in the same task, but using a different subset of stories. When the forensic science task was complete, the experimenter ostensibly graded them.
For all participants, the experimenter said she needed to consult her supervisor about where to hold the group session. For the no feedback condition, she did not mention task performance or any comparison between participants. For the other conditions, the experimenter also said that their performance looked “about the same” and she needed help from her supervisor to determine who should be the outstanding group member. The experimenter used the intercom to contact her supervisor (John) and appeared surprised when there was no reply. The experimenter left the lab, ostensibly to go to the supervisor’s office. Over the intercom, one of three recorded conversations between the experimenter and supervisor was “overheard”. These recordings were identified only by code numbers, so the experimenter remained blind to all but the baseline no feedback condition (when the experimenter made no mention of choosing an outstanding group member).
The recorded message began with conversation about the dysfunctional intercom and where the next session would be. For the no feedback control condition the conversation ended here. For the experimental conditions, the experimenter next said that participants’ answers were about the same. Participants then overheard the supervisor choose the male participant as the outstanding group member. In the no reason condition, no reason was provided for his choice. In the pro-male bias reason condition the supervisor said the man should lead the next activity because men usually do better than women at science. This feedback manipulation procedure removes potential effects of participants’ discomfort with working alongside the sexist feedback giver (Adams et al. 2006) because participants never met the supervisor who gave the feedback.
The experimenter returned to the room, giving no indication that her conversation had been overheard, and announced that the male confederate was the outstanding group member for this session. Participants completed post-feedback measures, including an order form with which they could request (or not) information on careers in forensic science to take with them. The confederate was then taken to a separate room to ostensibly be interviewed and participants were then given a written manipulation check (described below), followed by a structured debriefing interview beginning with a verbal manipulation check in a series of directed questions.
Following the practice activity, baseline levels of interest, perceived competence, and competence value (presented in that order), were measured using a five-point scale, with higher scores indicating more of each construct. The items were: “How interesting do you think this activity will be to you?” (interest), “How well do you think you will perform on this activity?” (perceived competence), and, “How important is it for you to do well on this activity?” (competence value). These measures, obtained before the main activity engagement period and the feedback manipulation, were included in both Studies 1 and 2 so that initial individual differences could be controlled when examining effects of feedback.
Following the activity and feedback manipulation, participants rated (in the order presented here) their perceived competence and competence value using identical items and rating scales from the pre-activity measures, worded in the past tense. Post-feedback activity interest was measured using a scale adapted from Smith et al. (2007). Individuals rated their agreement with seven items (α = .88) using five-point scales (1 = not at all, 5 = very much) that described appraisals of interest for the forensic science activity (e.g., “I would describe this task as very interesting”).
We also measured self-esteem using the 20 item (α = .74) State Self-Esteem Scale (Heatherton and Polivy 1991), which asks participants’ degree of agreement with statements (e.g., “I feel good about myself”) on a scale from 1 (Not at all) to 5 (Extremely). However, there were no effects on self-esteem in either study, and so results involving self-esteem are not presented further.
To assess subsequent likelihood of pursuing related careers, we asked participants if they would like to receive information on potential careers in forensic science (yes or no). To decrease potential demand characteristics, it was presented as an anonymous order form that the participant would enclose in an envelope to ostensibly be seen only by the secretary later (all participants who requested this information were provided it at the end of the session). We utilized this behavioral measure of career information requests to minimize shared method variance between measures of activity and career interest to ensure that correlations between activity and career interest measures were not artificially inflated by similarly constructed self-report measures.
The perceived discrimination manipulation check, completed after all other measures, included open-ended and directed questions about potential unfair treatment in general (e.g., “All procedures were conducted fairly” and “I experienced unfair treatment in this study”) and based specifically on several characteristics, including gender (e.g., “I experienced discrimination based on my sex in this study”). Because we expected that participants may not be willing to report discrimination to researchers, we disguised this questionnaire as a general form used to evaluate the lab’s effectiveness, and included the university’s anti-discrimination policy. We also interviewed participants extensively before debriefing.
The manipulation check interview confirmed that participants accurately reported the overheard conversation. Despite reporting the conversation accurately, two participants in the feedback/no reason condition reported possible gender discrimination, but said (during the interview) that they only thought about potential discrimination after seeing the gender discrimination questions on the final form. Of the participants in the feedback/pro-male bias reason condition, all but four reported discrimination. These four participants were visibly uncomfortable and anxious during their interviews (cf. Woodzicka and LaFrance 2001), and confederates confirmed that the recording played clearly. Analyses were conducted with and without these 6 participants, with no significant changes. Reported analyses included them.
Effects of feedback manipulation
Studies 1 descriptive statistics and correlations
1 Activity interest
2 Perceived competence
3 Competence value
4 Request for career information
Study 1 means by feedback condition from ANCOVA models
Request for career information
ANCOVA results indicated a significant effect of feedback condition on activity interest, F(2, 75) = 6.62, p = .002, ηp2 = .15. Women in the feedback/pro-male bias condition reported lower interest than those in the no feedback condition (p = .005) and the feedback/no reason condition (p = .003). The no feedback and feedback/no reason conditions did not differ (p = .94).
Results indicated a significant effect of the feedback manipulation on competence value, F(2, 74) = 3.28, p = .04, ηp2 = .08. Women in the feedback/no reason condition reported greater competence value than those in the no feedback condition (p = .01). Competence value for those in the feedback/pro-male bias condition was between but not significantly different from the feedback/no reason condition (p = .09) and the no feedback condition (p = .23). ANCOVA results indicated no effect of the feedback manipulation on perceived competence, F(2, 74) = 0.27, p = .76, ηp2 = .007.
Indirect effects of biased feedback on requests for career information via interest appraisals
Although activity interest was the focal outcome variable in our predictions, prior research (and the high correlation between activity interest appraisals and the behavioral measure of career interest) suggests that one implication of effects on activity interest is a consequent effect on the likelihood of seeking further information about possible linked careers. Therefore, we tested for possible indirect effects of the feedback manipulation on whether participants requested information on careers in forensic science, using regression analyses with bootstrapped confidence intervals and adjusting the analysis for the binary outcome (Preacher and Hayes 2008).5 To test the possibility of the achievement-related expectancy value variables as alternative mediators, we also included perceived competence and competence value in the model.
We used orthogonal contrast codes to test for indirect effects of the feedback manipulation to create focal comparisons within the regression analysis (Rosenthal and Rosnow 2008). This allowed us to separate the independent effects of receiving any performance feedback (relative to no feedback) (Sansone 1986) from the effects of receiving different reasons (pro-male bias vs. no reason) for the feedback. When included in the regression simultaneously, the Any Feedback Contrast code [i.e., no feedback (+2) vs. feedback/no reason (−1) and feedback/pro-male bias (−1)] and the Feedback Reason Contrast code [i.e., feedback/pro-male bias (−1) vs. feedback/no reason (+1) vs. no feedback (0)] allows for the estimate of each focused comparison while controlling for the effects of the other focused comparison. Contrast codes must be included in analyses as a full set (otherwise they lose their intended interpretation), so we tested the model that included these two feedback contrast codes as predictors, activity interest, perceived competence, and competence value as mediators, and all pre-activity variables as covariates (to control for individual differences prior to engagement and feedback). All continuous variables were centered.
Activity interest directly predicted requests for information on careers in forensic science (b = 1.94, se = .71, p = .007). There were no direct effects of either feedback contrast code. Testing for indirect effects through interest, the Feedback Reason contrast comparing the pro-male bias reason versus no reason for receiving the relatively negative feedback was significant. The bootstrapped 95 % confidence interval for the indirect effect of this contrast on requests for career information via activity interest excluded zero, 95 % CI [.01, .85]. The confidence intervals for all other indirect effects included zero.
When the reason for receiving less positive feedback on a science activity than a male student for similar work was perceived as due to pro-male bias, women reported lower interest in the science activity. This was the case even though the gender bias feedback occurred after participants’ actual performance and experience of the activity. In turn, activity interest was the strongest predictor of requesting information about related forensic science careers. These negative effects from receiving feedback favoring men were not accounted for by effects on perceived competence or competence value.
Although not a direct aim of the study, consistent with previous research (e.g. Stangor et al. 2002; Woodzicka and LaFrance 2001), we also found that no women confronted or questioned the experimenter about discrimination. Most women who experienced the pro-male bias did not report it until directly asked in an anonymous questionnaire with the university’s anti-discrimination policies, and some still did not report it then.
In Study 2, we use the same experimental approach, but focus on the experience of benefiting from pro-male biased feedback for similar work on the science activity. Procedures and measures were mostly identical to Study 1; we used the same novel science activity and feedback manipulation procedure and conditions. In Study 2, however, the woman who was ostensibly harmed by the feedback was a confederate and the man who benefited was the actual participant.
Previous research suggests competing hypotheses for how men who benefit from feedback attributable to gender-based competence stereotypes might feel toward the activity. Thinking about one’s group’s privilege can trigger guilt and other negative feelings (Mallett and Swim 2007; Schmitt et al. 2000), which could lower appraisals of interest. Alternatively, receiving feedback that one’s group membership is a positive attribute in a given domain can increase interest in (Martin and Ruble 2004), value of (Sechrist and Stangor 2001), and future motivation for (Sherman 1981) that and related activities (Patterson and Bigler 2007). Both possibilities are tested in Study 2, with the direction of effects indicating which hypothesis is correct.
Potentially reconciling these competing hypotheses about the direction of the effect, we reasoned that whether effects on interest appraisals are positive or negative could also depend on a third variable: the reaction of the woman hurt by the feedback. Perhaps one reason that men could experience benefiting from gender bias positively is because they are unaware that it injured another person. Men might regard the experience differently if they know that the woman felt hurt. If effects are due primarily to attributions for why the feedback occurred, in contrast, then the concerns voiced by the woman should not matter. To test whether women’s reactions moderated effects of the feedback, we manipulated whether the woman confederate verbally indicated that she was upset by what she saw as gender bias.
We recruited 121 male undergraduates from the participant pool who participated for extra-credit. Fourteen participants did not have completed sessions because of technical errors (e.g. volume was too low on recorder, batteries died during message). Data from another ten participants were excluded because of high suspicion expressed in debriefing. The final sample consists of 97 men (87.2 % European-American, M age = 21.2). Participants were randomly assigned to one of five conditions in a 2 (pro-male bias vs. no reason) by 2 (no confederate response vs. confederate response) design, with the added baseline no feedback group.
A female experimenter ran participants in groups of two: one participant and one female confederate. All procedures and materials except for the added confederate manipulation were identical to Study 1. Reliability for measures of post-feedback activity interest (α = .85) and self-esteem (α = .78) were similar to Study 1.
Confederate response manipulation
In the conditions with comparative feedback, the confederate response manipulation occurred when the experimenter was away, following the recorded conversation. In the response condition, the confederate said in a disgusted tone, “This is so unfair, it’s just because I’m a woman.” In the no response condition, the confederate remained quiet (the behavior displayed by all participants in Study 1). In all conditions, the confederate did not initiate interaction beyond this statement and responded minimally when participants attempted to initiate discussion.
Manipulation check interviews confirmed that participants accurately reported the overheard conversation. Of participants in the pro-male bias reason condition, all reported discrimination on the form or in the interview. However, six participants in the no reason condition reported possible gender discrimination. All were in confederate response conditions and said (during interviews) that their reports were based on her comments. Analyses were conducted with and without these six participants, with no significant changes. Reported analyses include them.
Does the effect of feedback depend on the confederate’s response?
A key empirical question in the first study to examine effects of benefiting from pro-male bias on men’s science activity interest appraisals is whether men’s reactions to the feedback differ depending on whether or not the woman expressed concerns about being harmed. Because the confederate manipulation could only occur when participants received feedback, we first conducted analyses including only those conditions to test whether the confederate statement moderated effects of benefiting from pro-male bias vs. no reason feedback conditions. Results from the 2 (confederate statement: present vs. absent) × 2 (feedback: no reason vs. pro-male bias) ANCOVAs revealed no significant interactions between the confederate statement manipulation and the feedback manipulation for any dependent variables.
Although only two of the three feedback conditions were included in this 2 × 2 analysis, results revealed a significant main effect of feedback condition on activity interest appraisals. Men in the feedback/pro-male bias condition reported greater interest than those in the feedback/no reason condition [3.57 vs. 3.23, F(1, 69) = 6.65, p = .012, ηp2 = .08]. There were no significant differences between these two feedback conditions for models predicting perceived competence, competence value, self-esteem, or requests for related career information (all ps > .10).
Further, there was a significant main effect of confederate statement on perceived competence and competence value. When the confederate acknowledged being harmed by pro-male bias (vs. when she said nothing), men reported greater perceived competence [3.17 vs. 2.84, F(1, 69) = 4.51, p = .04, ηp2 = .06] and competence value [3.34 vs. 3.01, F(1, 69) = 11.17, p = .001, ηp2 = .14] for the forensic science activity. The confederate statement manipulation did not predict activity interest or requests for related career information. Means by feedback and confederate statement conditions are presented in Table 4.
The results demonstrate that the effect of pro-male bias feedback for men does not depend on the woman’s reaction. Therefore, in all subsequent analyses we collapse across confederate response conditions to focus on effects of feedback across all three feedback conditions for men.
Effects of feedback manipulation
Study 2 means, standard deviations, and zero order correlations among outcome variables
1 Activity interest
2 Perceived competence
3 Competence value
4 Request for career information
Study 2 means by feedback and confederate statement (CS) conditions from ANCOVA models
Request for career information
Indirect effects of biased-feedback on requests for career information via interest appraisals
Using the same procedure as in Study 1, we tested effects of the feedback manipulation on requests for information on careers in forensic science via interest appraisals. We again included the alternative mediators, perceived competence and competence value. We used the same feedback manipulation codes as in Study 1, and added a confederate statement code [present (+1) vs. absent (−1)] as well as the interaction between the confederate statement and the code comparing pro-male bias versus no reason feedback. Because ANCOVA results yielded no significant interactions between the confederate statement and feedback manipulation, the confederate statement code and the interaction code were included primarily as covariates. We did, however, examine potential indirect effects involving these variables; none were significant.
Paralleling Study 1, there were no direct effects of either feedback contrast code on whether participants requested information on careers in forensic science. Only interest predicted whether participants requested information on careers in forensic science (b = 1.60, se = .54, p = .003), and results again indicated a significant positive indirect effect of pro-male bias vs. no reason on whether participants requested information on careers in forensic science via activity interest, bootstrapped 95 % CI [.03, .77].
When the reason for benefiting from more positive achievement feedback than a female student on a science activity was perceived as due to pro-male bias, men reported greater interest in the activity. Benefiting from gender bias also led to greater subsequent requests for information on related careers via effects on activity interest. Further, benefiting from gender bias positively affected men’s appraisals of interest in the activity regardless of whether the woman harmed by the feedback made her discomfort known. As with Study 1, effects involving the achievement-related expectancy-value variables did not account for men’s greater interest when they received pro-male bias feedback.
Unexpectedly, a significant positive main effect of the confederate statement emerged on men’s perceived competence and competence value, such that men reported greater competence and value of that competence when women expressed being hurt by the feedback compared to when she said nothing. We are cautious about over interpreting an unexpected effect, but this results suggest that the woman’s objecting to presumed discrimination might have made the positive social identity implications of the stereotype for men’s achievement more salient. Alternatively, we can compare means from the no feedback control condition to feedback conditions with and without the confederate statement. It appears that when the woman did not say anything, men tended to report slightly (though non-significantly) lower perceived competence and competence value. Conversely, these means were slightly higher (again, non-significantly) than the no feedback condition when the woman expressed being hurt. The statement could have shifted feelings of internalized guilt away from the men into externally-directed anger toward the supervisor, thereby protecting men’s feelings of competence and value. The present study focused on interest and this finding was not expected, however, so future work is needed to establish replication of and interpret this finding.
Similar to women harmed by perceiving feedback as due to pro-male bias in Study 1, the men who benefited from this feedback did nothing behaviorally in response. Like women who are harmed by pro-male bias, men who benefit from the same feedback may be unlikely to do anything to call attention to or protest the unfair feedback.
Across studies, results demonstrate diverging effects of pro-male bias on appraisals of science activity interest for women and men. The study design isolated effects of the pro-male bias from differential engagement on interest appraisals by timing the feedback after the science activity was completed. When participants attributed the reason for the man being chosen as outstanding group member to gender bias from the feedback giver, women reported lower interest (Study 1) and men reported greater interest (Study 2) in the science activity, compared to control conditions.
These findings have implications for understanding how STEM-related interests and choices for men and women become more divergent over time (cf. Ceci et al. 2009; Ceci and Williams 2012). If a man and woman begin a science activity with similar levels of interest, being on different ends of the same relative experience of gender bias can push their interests in opposite directions. Further, in both studies the feedback indirectly influenced whether individuals subsequently requested career-related information (negatively for women and positively for men) through its effects on activity interest appraisals. Psychological explanations for how gender bias in STEM fields affects women’s motivation to persist or quit in a domain have (implicitly or explicitly) relied primarily on motivational models based on achieving, expecting, and/or valuing competence in the domain. This emphasis has missed the potentially important role played by a second, though related, motivational route—feelings or appraisals of interest toward the activity (Sansone and Thoman 2005; Sansone et al. 2015). In the present findings, the feedback conditions did not affect perceived competence or competence value, so these variables did not account for effects on interest appraisals. Further, most studies examining motivational effects of gender stereotypes and bias in STEM focus on women harmed by the feedback. Our results suggest that divergent STEM interests between women and men may come not only from negative effects of male bias on women’s interest but also from positive effects on men’s experience of interest. This pattern is consistent with findings on “stereotype lift” that suggest men can receive a performance boost (typically a smaller effect than the negative performance effects for women, similar to the pattern observed across the present studies) from stereotype salience (Walton and Cohen 2003).
Although our studies isolated effects of the pro-male bias from differential engagement on interest appraisals and ruled out process explanations based on expectancy-value variables, more work is needed to explicate the reasons why interest appraisals changed. It is likely that these explanations differ for women and men on different sides of the feedback. For women, the most likely explanation is a self-protective mechanism (cf. Crocker and Major 1989; Tesser et al. 1996). That is, losing interest in the science activity may have been the least self-threatening option for women, who maintained positive perceptions of self-esteem, perceived competence, and competence value. But the nature of the self-protection likely differs in our study from the studies conducted by Crocker, Major, and colleagues because of key differences in study designs. In our paradigm, students were first told that their relative performance was about the same, and then subsequently received the pro-male bias feedback. In this case, for women the threat is less about their current performance and more about what it might be like to continue in this domain. Thus, a self-protective mechanism leading to lower activity interest appraisals likely functions similarly to the preemptive (rather than reactive) automatic threat defense system described and studied by Rudman and colleagues (e.g., Rudman et al. 2007). Appraising the activity as uninteresting may be an effective way to preemptively avoid future potential threats to self, which may be more likely to occur for women in gender-biased activity contexts.
An alternative explanation for the differential shift in interest appraisals derives from patterns found in classic research reconciling dissonance and self-perception explanations of attitude change (Fazio et al. 1977). In this case, for women, the experience of gender bias on an interesting science activity creates two dissonant ideas [potentially with affective–cognitive inconsistency (Chaiken and Baldwin 1981)]—the task was interesting and they were harmed by stereotype-based feedback. Thus, women change their appraisals of activity interest to be more negative. For men, on the other hand, the task was interesting and they benefited from stereotype-based feedback, which may not be dissonant beliefs. If receiving the positive feedback is not dissonant with their interest appraisal based on the experience the result could be that interest is rated more positively (Fazio et al. 1977; Stone and Cooper 2001). Although some of the original work on dissonance and self-perception used ratings of activity interest as the operationalizations of attitude (e.g., Festinger and Carlsmith 1959; Lepper et al. 1973), theories of dissonance and self-perception have rarely been linked to the study of interest experiences or appraisals as a result of gender bias. Particularly for effects of feedback after engagement, however, this theoretical lens might be useful in explaining the pattern of divergent interest appraisals and provides clear theoretical direction for future research. For example, dissonance might be created in men in circumstances where benefiting from gender-biased feedback has negative emotional consequences (Mallett and Swim 2007; Schmitt et al. 2000), although this was not the case in the present research. This lens also suggests that there could be other ways that women might resolve dissonance that does not result in more negative interest appraisals [e.g. values affirmation (Steele 1988)].
These findings also add to a growing body of work demonstrating that interest for an activity is not only influenced by experiences before and during activity engagement, but also by social feedback following activity engagement. Previous work focused on post-activity conversations about the activity (Thoman et al. 2007, 2012), whereas the present findings illustrate that the signaling of stereotypes or social beliefs regarding the (positive or negative) implications of one’s social identity (gender) for the activity or related domain can also shape subsequent interest appraisals. The boundary conditions for how and what kinds of post-activity feedback can influence interest appraisals are not yet well understood, but represents a promising direction for theoretical models linking interest appraisals to the development of interest (cf. Connelly 2011; Silvia 2006).
Results from Study 1 help explain why women who perceive gender bias as part of a science activity choose to quit that activity or domain despite maintaining positive views of self and performance. Particularly if this pattern is repeated, women might internalize the lack of interest into a personal characteristic (“this is not me”) that becomes a stable part of their self-concept (Seymour and Hewitt 1997; Major et al. 1998). Steele and colleagues (e.g., Steele 1997; Steele et al. 2002) have written extensively about this process of disidentification, whereby an individual distances the relevance of domain performance from their self-concept following social identity threats. At the phenomenological level, experiencing lower interest toward an activity or domain could indicate, co-occur with, or even precede and trigger disidentification. For example, from a self-evaluation maintenance perspective (Tesser et al. 1996), changing appraisals of interest toward a topic/activity may allow one to protect self-views of competence in the face of experiencing bias, with the implication that the individual leaves because science is uninteresting rather than feeling like she can’t do it. That is, it may be easier for women to lose interest in the activity as a means for protecting the self than to attribute feelings of disidentification to social stigma against them. Although Steele and colleagues have not written about interest appraisals, the connection between interest and disidentification seems plausible and worthy of future research.
One’s experience of interest should be a strong predictor of intentions to persist in a domain, but not the only factor, and distinguishing between these variables can have important implications for encouraging greater participation (Thoman et al. 2013). For example, cues that signal gender bias lower women’s sense of belonging, which makes them less likely to persist in STEM (Cheryan et al. 2009; Murphy et al. 2007). One implication is that a woman who perceives the field as changing to better embrace femininity should be likely to reenter the field. However, if those stereotypical cues also negatively affect her appraisal of interest in STEM, she would be unlikely to pursue STEM even if she perceived the field accepting of women, because she would lack sufficient interest in STEM activities or topics relative to alternatives. Similarly, perceived instructor sexism negatively affects women’s comfort with and perception of the instructor (Adams et al. 2006), but feelings toward an instructor are distinct from feelings toward the activity. If a student loses interest in the activity, her interest in pursuing similar activities should be low regardless of the instructor. If she maintained a positive interest appraisal toward the activity, however, and only feels uncomfortable with that instructor, she should be more likely to pursuing similar activities without the sexist instructor. Coupled with recent educational findings (Renninger and Schofield 2012) and Seymour and Hewitt’s (1997) interviews, our findings suggest that understanding the role of interest along with other motivational variables within a cohesive conceptual framework is important for this field of research (Thoman et al. 2013).
The current work is not without limitations. Across studies we did not experimentally separate participant gender from whether participants were harmed by or benefited from the feedback. This methodological choice is linked to the phenomenon of pro-male bias in STEM that motivated this research, but does limit conclusions. It is unclear whether men would respond similarly to women when harmed by perceived female bias or whether women would respond similarly to men when benefiting from female bias. Past research using artificially created group stereotypes, randomized individual assignment to group membership, and manipulated perceived legitimacy and status value of the intergroup comparisons (Martiny et al. 2012; Patterson and Bigler 2007; Schmader et al. 2001) suggests that if men faced a psychologically equivalent situation to what women faced in Study 1, results would look similar. However, because women contend with systemic threats that men do not share, the parallel suggestion of sexism against men may not be psychologically equivalent (Adams et al. 2006; Schmitt and Branscombe 2002).
In addition, although these data are the first to empirically demonstrate that gender bias occurring after activity engagement leads to divergent effects on men’s and women’s appraisals of science activity interest, further work is needed to explicate the mechanisms that lead to divergent interests. These experiences may have similar or different mechanisms, and examining both processes will lead to a better understanding of how social psychological variables influence appraisals of interest. Previous research has primarily examined how discrimination affects those who are harmed, and less frequently considered motivational implications for individuals who benefit. Although this focus on those who are harmed is justified, understanding how the same bias affects people on both sides of the equation would lead to greater understanding of broad effects of stereotypes and may provide additional insight into why cultural stereotypes are so difficult to counter.
In both studies, when pro-male bias clearly occurred, participants did not confront the experimenter. This apparent reluctance to call attention to discrimination was equally true for men who benefited (Study 2) and women who were harmed by (Study 1) the feedback. Others have observed similar reluctance to report discrimination among individuals harmed by group-based feedback (e.g., Stangor et al. 2002; Woodzicka and LaFrance 2001). Future work should further explore what constrains or encourages individuals to report socially unjust experiences, both when they are harmed by and benefit from stereotypes.
In conclusion, gender bias not only leads to fewer systematic opportunities for women in science (Moss-Racusin et al. 2012), but the awareness that one is the target of such bias can lead to diverging appraisals of interest toward science activities for women and men, and therefore diverging interest in pursuing science further. This motivational process is important to understand because interest is strongly related to women’s choices to persist or leave STEM (Seymour and Hewitt 1997; Renninger and Schofield 2012) and linked closely to identity (Renninger 2009), and because interest appraisals can be affected separately from effects on achievement motivation processes. This creates the possibility that interventions focusing on making academic activities more interesting could be an important tool in promoting greater representation of women in STEM.
Originally there were two pro-male bias conditions, attempting to disentangle stereotypes based on the individual feedback giver’s experience from general domain-based stereotypes. Participant feedback made clear that these conditions were not distinguishable; both were perceived as reflecting domain bias. Confirming that participants did not distinguish these conditions, we compared these conditions on all dependent variables and manipulation checks, finding no differences. Thus, we collapsed into a single pro-male bias condition.
In both studies, no participant reported suspicion in the no feedback condition where individuals performed the activity separately then completed study measures. Among other conditions, the number of participants reporting suspicion was evenly spread, suggesting that participants were not differentially suspicious as a function of the reason for the feedback. In addition, nearly all cases of high suspicion (in both studies) occurred near the end of the academic semester, when there is greater likelihood that participants would have heard about the study from peers or have participated in previous studies that used deception. Participants who were dropped versus retained did not differ on any demographic or individual difference measures. Analyses were re-run including participants who reported high suspicion and all substantive findings were consistent with the reported results.
In both studies we tested for but found no significant differences for any effects across confederates or experimenters.
Pilot testing indicated that the stories were moderately difficult, with at least two answers as equally likely, and that the task was perceived as relevant for learning scientific content and reasoning, characteristics needed for plausibility of the manipulation.
In this model we did not include tests for potential indirect effects of gender bias feedback on activity interest appraisals via perceived competence and competence value because these variables were not significantly predicted by the feedback manipulation (which establishes a lack of X to M relationship in a mediation model). We did conduct these indirect tests in a separate model, however, and as expected no indirect effects on activity interest were significant.
- American Association of University Women. (2010). Why so few: Women in science, technology, engineering, and mathematics, by C. Hill, C. Corbett, & A. St. Rose. Washington, DC: Author.Google Scholar
- Branscombe, N. R. (1998). Thinking about one’s gender group’s privileges or disadvantages: Consequences for well-being in women and men. British Journal of Social Psychology, 37, 167–184. doi:10.1111/j.2044-8309.1998.tb01163.x.
- Cheryan, S., Plaut, V. C., Davies, P., & Steele, C. M. (2009). Ambient belonging: How stereotypical environments impact gender participation in computer science. Journal of Personality and Social Psychology, 97, 1045–1060. doi:10.1037/a0016239.
- Connelly, D. A. (2011). Appying Silvia’s model of interest to academic text: Is there a third appraisal? Learning and Individual Differences, 21, 624–628. doi:10.1016/j.lindif.2011.04.007.
- Festinger, L., & Carlsmith, J. M. (1959). Cognitive consequences of forced compliance. Journal of Abnormal and Social Psychology, 56(2), 276–278.Google Scholar
- Izard, C. E., & Ackerman, B. P. (2000). Motivational, organizational, and regulatory functions of discrete emotions. In M. Lewis & J. M. Haviland-Jones (Eds.), Handbook of emotions (2nd ed., pp. 253–264). New York: Guilford Press.Google Scholar
- Murphy, M. C., Steele, C. M., & Gross, J. J. (2007). Signaling threat: How situational cues affect women in math, science, and engineering settings. Psychological Science, 18, 879–885. doi:10.1111/j.1467-9280.2007.01995.x.
- Patterson, M. M., & Bigler, R. S. (2007). Relations among social identities, intergroup attitudes, and schooling: Perspectives from intergroup theory and research. In A. Fuligni (Ed.), Contesting stereotypes and creating identities (pp. 15–41). New York: Russell Sage Foundation.Google Scholar
- Renninger, K. A. (2009). Interest and identity development in instruction: An inductive model. Educational Psychologist, 44, 1–14. doi:10.1080/00461520902832392.
- Renninger, K. A. & Schofield, L. S. (2012). Measuring interest: The open-ended response in a large scale survey. In M. D. Ainley (Chair), Assessment of interest: New approaches, new insights. Symposium conducted at the meeting of the American Educational Research Association, Vancouver, British Columbia.Google Scholar
- Rosenthal, R., & Rosnow, R. L. (2008). Essentials of behavioral research: Methods and data analysis (3rd ed.). New York: McGraw-Hill.Google Scholar
- Sansone, C., Thoman, D. B., & Fraughton, T. (2015). The relation between interest and self-regulation in mathematics and science. In K. A. Renninger, M. Neiswandt, & S. Hidi (Eds.), Interest in mathematics and science learning. Washington, DC: American Educational Research Association.Google Scholar
- Schiefele, U., Krapp, A., & Winteler, A. (1992). Interest as a predictor of academic achievement: A meta-analysis of research. In K. A. Renninger, S. Hidi, & A. Krapp (Eds.), The role of interest in learning and development (pp. 183–212). Hillsdale, NJ: Erlbaum.Google Scholar
- Schmader, T., Major, B., Eccleston, C. P., & McCoy, S. K. (2001). Devaluing domains in response to threatening intergroup comparisons: Perceived legitimacy and the status value asymmetry. Journal of Personality and Social Psychology, 80, 782–796. doi:10.1037/0022-35126.96.36.1992.CrossRefPubMedGoogle Scholar
- Seymour, E., & Hewitt, N. M. (1997). Talking about leaving: Why undergraduates leave the sciences. Boulder, CO: Westview Press.Google Scholar
- Silvia, P. J. (2006). Exploring the psychology of interest. New York: Oxford University Press.Google Scholar
- Smith, J. L., Lewis, K. L., Hawthorne, L., & Hodges, S. D. (2013). When trying hard isn’t natural: Women’s belonging with and motivation for male-dominated fields as a function of effort expenditure concerns. Personality and Social Psychology Bulletin, 39, 131–143. doi:10.1177/0146167212468332.CrossRefPubMedGoogle Scholar
- Steele, C. M. (1997). A threat in the air: How stereotypes shape intellectual identity and performance. American Psychologist, 52, 613–629. doi:10.1037/0003-066X.52.6.613.
- Steele, C. M., Spencer, S. J., & Aronson, J. (2002). Contending with group image: The psychology of stereotype and social identity threat. In M. Zanna (Ed.), Advances in experimental social psychology (Vol. 34, pp. 379–440). New York: Academic Press.Google Scholar
- Tesser, A., Martin, L. L., & Cornell, D. P. (1996). On the substitutability of self-protective mechanisms. In P. M. Gollwitzer & J. A. Bargh (Eds.), The psychology of action. New York: The Guilford Press.Google Scholar
- Tomkins, S. S. (1962). Affect, imagery, consciousness: Vol 1. The positive affects. New York: Springer.Google Scholar
- Walton, G. M., & Cohen, G. L. (2003). Stereotype lift. Journal of Experimental Social Psychology, 39, 456–467. doi:10.1037/0022-35188.8.131.52.