A "Rebound Effect" After Stereotype Threat?

Two studies investigated a potential cognitive mediator for stereotype threat, a phenomenon whereby the mere threat of confirming a negative stereotype results in a performance deficit. It was hypothesized that people attempt to suppress stereotypes in memory during threatening situations, consuming cognitive resources, but that the suppression is released after the threatening situation has ended. This results in a “rebound effect” and a subsequent increase in stereotyped thought. The experiments failed to find a significant stereotype threat effect when examined individually, but when the data from the experiments were aggregated aggregated, a performance deficit was found. However, because of the failure to find a significant performance deficit in any one experiment, the results to not directly bear on any potential rebound effect. Stereotype Rebound 3 A “Rebound” Effect After Stereotype Threat? Stereotype threat is a phenomenon in which the mere threat of confirming a negative stereotype in and of itself results in decreased performance on a given task. This phenomenon is relatively robust, and has been demonstrated in such groups as women taking math tests (Quinn & Spencer, 2001), African Americans taking a verbal portion of the GRE (e.g. Steele & Aronson, 1995), and even White Americans taking the Implicit Attitudes Test (Frantz, Cuddy, Burnette, Ray, & Hart, 2004). Current research makes less clear, however, exactly how stereotype threat occurs. More specifically, the question remains as to what mediating processes cause the threat of confirming a negative stereotype to decrease performance on a given task. The present study seeks to help clarify this issue by determining whether people engage in thought suppression with respect to stereotypes when performing a threatening task. The earliest research on stereotype threat, conducted by Steele and Aronson (1995), focused on explaining the black / white standardized test gap. It found that if African Americans were told that a test was diagnostic of their verbal abilities, they scored lower on the test than if they were told the test was investigating the psychological factors in solving verbal problems. Steele and Aronson (1995) explained these findings by theorizing that the diagnosticity manipulation primed the participants’ awareness of stereotypes positing African Americans’ poor mental ability, and that the awareness of these stereotypes placed an extra affective load on these participants that resulted in decreased performance. Underlying Mechanisms Subsequent research has attempted to pinpoint the underlying mechanism causing stereotype threat. Following Steele and Aronson’s (1995) assumption that stereotype threat causes a burdensome affective load, a significant portion of these studies has focused on the affective underpinnings of stereotype threat. The theory is that aversive affective states, such as increased arousal and anxiety, lead to distraction during the threatening task, and thus, Stereotype Rebound 4 decreased performance. For example, several studies have suggested that increased anxiety may be related to the stereotype threat phenomenon (Aronson, Lustina, Good, Keough, Steele, & Brown, 1999; Spencer, Steele, & Quinn, 1999). Additionally, manipulations designed to decrease anxiety and arousal seem to moderate the effects of stereotype threat (Martens, Johns, Greenberg, & Schimel, 2002). However, other studies make the link between affect and stereotype threat less clear. For example, Gonzales, Blanton and Williams (2002) found no difference in self-reported anxiety between conditions. In another study, Brown and Josephs (1999) found no evidence that words related to performance anxiety were more accessible in memory during stereotype threat, implying that anxiety may not strongly mediate the phenomenon. Other studies have yielded similar results (e.g. Oswald & Harvey, 2001, Schmader, 2002, Stone, Lynch, Sjomeling, & Darley, 1999). Thus, the general picture that emerges from the literature regarding an affective mediator for stereotype threat is that negative affect may contribute to decreased performance, but that it is by no means the only factor involved. Since anxiety does not seem to account fully for the damaging effects of stereotype threat, other research has sought to determine other possible mediators. One possible explanation is that the stereotype negatively impacts an individual’s self-confidence to perform the given task and, in a self-fulfilling nature, subsequently disrupts the individual’s ability to perform on the task (Rosenthal & Jacobson, 1966). In their original study, Steele and Aronson (1995) provided a test of this hypothesis by administering a word completion task that provided an implicit measure of self-doubt after finishing the threatening task. In support of the performance confidence hypothesis, they found that black participants in the threat condition spontaneously filled in the highest number of doubt-related words. In another study using a similar measure of self-doubt as Steele and Aronson (1995), Stone (2002) induced stereotype threat for golf performance by relating golf to intelligence for African Stereotype Rebound 5 Americans and natural ability for White Americans. Stone then administered a word completion task for both groups of golfers measuring self-doubt. Both threatened groups produced more doubt-related words than control groups. However, although the studies cited above have found potential links between stereotype threat and self confidence, a number of studies have not (e.g. Aronson, Lustina, Good, Keough, Steele, & Brown, 1999, Keller, 2002, Kray, Thompson, & Galinski, 2001, Shih, Pittinsky, & Ambady, 1999). For example, Kray and colleagues (2001) conducted an experiment at an MBA program in which one member of either a male-female or a male-male negotiating dyad were either told that a negotiating exercise was highly diagnostic of important, stereotypically male managerial skills, or that the exercise was simply illustrative of classic buyer-seller bargaining. Additionally, Kray and colleagues administered a selfconfidence to the participants across both conditions. Although female participants exhibited a performance deficit in the threat condition relative to the male participants, they showed no corresponding change in confidence levels. Thus, although performance confidence may partially mediate stereotype threat, current evidence suggests that this construct cannot fully account for stereotype threat’s damaging effects. Another body of research has examined possible behavioral mediators for stereotype threat. Some researchers have investigated whether stereotype threat damages performance by causing the target of the threat to change the amount of effort expended in the threatening task. According to this view, stereotype threat would either cause the threatened person to decrease the effort expended, resulting in decreased performance, or drastically increase the amount of effort expended, resulting in overexertion and, ultimately, decreased performance. However, a variety of studies have shown that neither of these theories is a good explanation for the negative effects of stereotype threat (e.g. Aronson et al., 1999, Gonzales, Blanton, & Williams, 2002, Keller, 2002, Keller & Dauenheimer, 2003, Smith & White, 2002). Stereotype Rebound 6 Another possible behavioral mediator of stereotype threat is self-handicapping, or an individual’s attempt to protect the self by either behaving or claiming to have behaved in such a way that a poor performance could be attributed to external circumstances (Leary & Shepperd, 1986). Applied to stereotype threat, self-handicapping implies that threatened individuals may claim more self-handicaps in response to a threatening stereotype. This hypothesis has received partial support in subsequent research. For example, Stone (2002) found that threatened individuals opt not to take practice swings when preparing for a golf task. Additionally, Quinn & Spencer (2001) found that threatened females were less able to formulate strategies for solving math word problems. However, as in the case of the other potential mediators for stereotype threat discussed above, the majority of studies investigating self-handicapping as a potential mediator have provided null results (e.g. Croizet & Claire, 1998, Keller & Dauenheimer, 2003, Kray et al., 2001, Shih et al., 1999). An additional potential mediator suggested by Smith (2004) that is not easily classified into affective, behavioral, or cognitive categories relates to the adoption of performanceavoidance versus performance-approach achievement goals. Achievement goal research posits that performance expectancies determine the type of achievement goal adopted (Elliot & Church, 1997). A high performance expectancy results in a performance-approach achievement goal, wherein the goal is to demonstrate competence, whereas a low performance expectancy generates a performance-avoidance goal, wherein the goal is to avoid demonstrating incompetence. Several researchers have found that performance-approach achievement goals generally result in positive outcomes, while performance-avoidance goals generally result in negative outcomes (e.g. Barron & Harackiewicz, 2001, Elliot & Church, 1997). Moreover, Smith, Sansone, and White (2007) have found that women with high achievement motivation who undergo stereotype threat spontaneously adopt performanceStereotype Rebound 7 avoidance goals. However, the literature linking stereotype threat to achievement goals is still in its infancy, and is unlikely to provide a complete mediator for stereotype threat. Thought Suppression as a Mediator Although the above research focuses mainly on the behavioral and phenomenological underpinnings of stereotype threat, another body of research focuses on possible cognitive factors of the phenomenon. For example, Croizet, Després, Gauzins, Huguet, Leyens, and Méot, (2004) presented an adaptation of the Progressive Matrices Test to participants with a perceived reputation for intellectual inferiority. Concurrent with the Progressive Matrices Test, Croizet and colleagues administered several autonomic measures of cognitive load. When the task was presented as a measure of intellectual ability, people belonging to stereotyped groups not only exhibited performance deficits, but also exhibited evidence of greater cognitive load than people from non-stereotyped groups. This study implies a link between a person’s available cognitive resources and stereotype threat. In another study investigating the cognitive factors of stereotype threat, Schmader and Johns (2003) found that stereotype threat reduces participants’ scores on a working memory task, even though the task was presented as being unrelated to the task triggering the stereotype threat. To explain these results, Schmader and Johns conjecture that cognitive resources might be consumed in suppressing thoughts relating to the target stereotype. Indeed, other research, such as that of Spencer (2003), indicates that targets of a stereotype might often try to suppress thoughts related to that stereotype. This implies that cognitive factors may mediate some of the negative ramifications of stereotype threat. Thus, the extra effort expended in suppressing thoughts related to a stereotype seems to be a reasonable explanation for the working memory deficits elicited in Schmader and Johns’ research. With respect to thought suppression, previous studies have shown that suppressing thoughts actually makes those thoughts more accessible in memory. Wegner and Erber (1992) Stereotype Rebound 8 demonstrated that attempting to suppress a target word in a high-stress situation resulted in longer reaction times in naming colors and more responses matching that target word when the word is primed than in a control condition. To explain these results, they proposed a model, termed the ironic process model, whereby the mind utilizes two simultaneous processes to suppress thoughts, one controlled and one automatic. The automatic process scans the contents of working memory to detect traces of the unwanted thought, which would indicate a lapse in mental control. If such a thought is detected, the controlled process, which is intentional and limited by available working memory, seeks to replace this thought with a suitable distracter thought. The implication of this model is that at any given time, a person attempting to suppress a thought should be highly sensitized to the presence of the unwanted item. Thus, when high-stress situations trigger a large cognitive load, the controlled process becomes subsequently less able to replace unwanted thoughts with distracter items and unwanted thoughts become even more accessible than if the person did not engage in thought suppression at all. This theory has since been applied to stereotype suppression. The reasoning goes that since there is a present social norm prohibiting the explicit expression of stereotyped attitudes, people who have stereotypes made salient to them have an incentive to consciously suppress these thoughts. However, according to the ironic process theory, people attempting to suppress stereotyped thoughts who are under a disruptive mental load should have these thoughts hyperaccessible in working memory, and thus when these people discontinue their thought suppression effort, they should experience a “rebound effect” where they are hyperaware of the unwanted stereotype. Consistent with this model, Macrae and colleagues (1994) show that when people try to suppress stereotyped thoughts, they subsequently express more negative stereotyped thoughts than people in a control condition. Thus, stereotype suppression may actually lead to more stereotyped thinking than may otherwise occur. Stereotype Rebound 9 These results have important implications for people in situations where stereotype threat is triggered. First, they imply that if a person engages in thought suppression during the specific task, as long as this task is sufficiently difficult and uses sufficient cognitive resources, suppression efforts should use cognitive resources that would normally be allocated to the task at hand, resulting in decreased performance. However, if the task is not sufficiently difficult, there should be no stereotype threat effect. Consistent with this prediction, O’Brien and Crandall (2003) have found that when women are administered simple math tests, which presumably do not use sufficient cognitive resources to conflict with the stereotype suppression process, they do not exhibit the stereotype threat effect. However, when the difficulty is scaled up, increasing cognitive load, performances in the stereotype threat condition decrease with respect to controls. Second, the results imply that people in a threat condition engaging in thought suppression should experience a rebound effect after the thought control process is relaxed, and thus stereotyped thoughts should be hyperaccessible after the threat condition has ended. This rebound effect could represent post-test ruminations about either the test itself or the individual’s performance on said test. However, the reason finding a rebound effect after stereotype threat is important is that it would re-emphasize the value of considering the cognitive approach when devising interventions to reduce or eliminate the stereotype threat effect. Additionally, although some of the theories regarding the underlying mechanisms for stereotype threat discussed above appear to have more empirical support than others, it is important to emphasize that these theories are not mutually exclusive. The overall picture that emerges from the literature is that no one factor fully mediates stereotype threat. Instead, the threat of confirming a negative stereotype about one’s group triggers a cascade of changes at multiple levels of analysis, including, but not necessarily limited to, the affective, behavioral, phenomenological, and cognitive. These changes then combine and interact, producing a performance deficit. In a phenomenon as Stereotype Rebound 10 complex as stereotype threat, multiple psychological approaches are not only beneficial, but necessary for devising effective interventions. Thus, this study attempted to use the fine grain of analysis provided by the cognitive approach to describe one of the proximate factors responsible for the destructive effects of stereotype threat with the ultimate goal of allowing more targeted interventions at the cognitive level. In service to the above goal, the present study sought to provide a conclusive link between stereotype threat and thought suppression, and subsequently further elucidate the means by which the cognitive mediators of stereotype threat operate. First, I will attempt to show that females engaging in a threatening math task engage in thought suppression during the task, resulting in a “rebound effect”, where unwanted stereotypes are hyperaccessible after the completion of the task. Second, I hope to show that when this thought suppression is eliminated, the negative effects of stereotype threat are alleviated. Experiment 1 Experiment 1 was designed to establish a firm link between stereotype threat and thought suppression. Participants engaged in a math task that has been framed as either having been shown to produce gender differences or as having been shown to be neutral with respect to gender. Each participant, regardless of how the math task was framed, will complete two lexical decision tasks designed to measure the activation of stereotypes related to women, one during the math task and one afterwards. If the threat manipulation is effective and thought suppression does play a role in stereotype threat, I expect to see two main patterns of results. First of all, participants in the threat condition should score lower on the math task than participants in the control condition. Secondly, participants in the threat condition should show relatively more stereotype activation after the math task is completed as compared to during the task because of the rebound effect. Participants in the control Stereotype Rebound 11 condition should show a similar amount of stereotype activation both during and after the math task. Method