A public versus private administration of the implicit association test

This research includes two experiments that examined (a) whether the assessment situation in which individuals complete an implicit measure of bias alters their responses and (b) whether the hypothesized effect of the assessment situation on implicitly assessed bias reflects socially desirable responding. Participants in Experiment 1 (N1⁄4 151) completed an IAT measuring bias toward homosexuality in either a public or a private assessment situation. Consistent with studies of explicitly assessed attitudes, implicitly assessed bias toward homosexuality was significantly lower when assessed in a public versus a private assessment situation. Participants in Experiment 2 (N1⁄4 102) completed an IAT measuring bias toward homosexuality in a public assessment situation under a bogus pipeline or no-bogus pipeline condition. Results indicated that participants’ implicitly assessed bias did not significantly differ across these conditions. The authors discuss these findings in terms of possible automatic processes affecting the malleability of implicitly assessed attitudes. Copyright # 2006 John Wiley & Sons, Ltd. Implicit measures of attitudes are a revolutionary component of research on prejudice because they have the potential to assess attitudes that people are unwilling or unable to consciously report (Greenwald & Banaji, 1995). Greenwald, McGhee, and Schwartz (1998) provided a dramatic illustration of this point when they showed that White participants endorsed positive or neutral attitudes toward AfricanAmericans on an explicit measure but exhibited negative attitudes toward African-Americans on an implicit measure. These results have subsequently been replicated (Ottoway, Hayden, & Oakes; 2001), and have been extended to include attitudes toward a variety of social groups such as Hispanics (Ottoway et al., 2001), Turks (Neumann & Seibt, 2001), elderly people (Nosek, Banaji, & Greenwald, 2002), and homosexuals (Banse, Seise, & Zerbes, 2001; Steffens & Buchner, 2003). MODERATORS OF IMPLICITLYASSESSED ATTITUDES Implicitly assessed attitudes were originally conceptualized as automatic and uncontrollable (Devine, 1989). According to this conceptualization, attitudes assessed by implicit measures should be less Received 25 June 2004 Copyright # 2006 John Wiley & Sons, Ltd. Accepted 20 October 2005 *Correspondence to: Guy A. Boysen, Psychology Department, W112 Lagomarcino, Iowa State University, Ames, IA 500113180, USA. E-mail: [email protected] susceptible to many of the factors that are known to have an unwanted influence on the attitudes assessed by explicit measures. If that were true, implicit measures would have a major advantage over explicit measures (Fazio, Jackson, Dunton, & Williams, 1995). However, an accumulating body of evidence indicates that implicitly assessed attitudes are influenced by various cognitive, motivational, and situational factors (Blair, 2002). The malleability of implicitly assessed attitudes is theoretically important because it suggests that implicit measures may be susceptible to some of the same sorts of unwanted influences to which explicit measures are susceptible. One factor that has been reliably found to have an unwanted influence on explicitly assessed attitudes is the assessment situation. People consistently report more positive attitudes toward stigmatized groups on explicit measures when they complete those measures in a public versus a private assessment situation (e.g., Blanchard, Crandall, Brigham, & Vaughn, 1994; Plant & Devine, 1998). If implicit measures are susceptible to some of the same unwanted influences that explicit measures are susceptible to, then they too should be influenced by the assessment situation, especially given its robust effects on explicitly assessed attitudes. Support for this hypothesis would provide evidence that implicit measures have some of the same disadvantages as explicit measures and that they should not be conceptualized as being the more valid of the two (Blair, 2002; Nosek, Greenwald, & Banaji, in press). EXPLANATIONS FOR DIFFERENCES IN PUBLICLYAND PRIVATELY ASSESSED ATTITUDES Explicitly measured attitudes toward stigmatized groups are more positive when assessed in a public situation than when assessed in a private situation due to people’s motivation to appear unprejudiced to others (Plant & Devine, 1989). The possibility that the assessment situation may also alter implicitly assessed attitudes raises questions about the underlying process that might produce such an effect. According to Fazio and Olson (2003), most studies of implicitly assessed attitudes are not implicit in the strictest sense because the measures require processing of explicit information and individuals are often aware of the attitudes being assessed. Thus, Fazio and Olson suggest that the measures are implicit while the attitudes may not be. According to this line of reasoning, the attitudes assessed by implicit measures may be partly under voluntary control. Consistent with this perspective, recent evidence suggests that individuals can exert some control over implicitly assessed attitudes. For example, individuals who took the implicit association test (IAT; Greenwald et al., 1998) were able to control their responses when they had experience with the measure (Steffens, 2004) and when they had been given explicit information about faking strategies (Kim, 2003). Although the ability of individuals to exert voluntary control over their implicitly assessed attitudes does appear to have limits (Asendorpf, Banse, &Mücke, 2002; Banse et al., 2001; Kim, 2003; Steffens, 2004), the fact that effortful control occurs at all suggests that people may be able to alter their responses on implicit measures in a socially desirable manner. Implicit measures were initially seen as a possible way to avoid socially desirable responding by circumventing voluntarily controlled response patterns (Greenwald et al., 1998). For example, in their review of implicit measures used in social cognition research, Fazio and Olson (2003) state that: What these various approaches have in common is that they all seek to provide an estimate of the construct of interest without having to directly ask the participant for a verbal report. Their major appeal is that these indirect estimates are likely to be free of social desirability concerns. (p. 300) 846 Guy A. Boysen et al. Copyright # 2006 John Wiley & Sons, Ltd. Eur. J. Soc. Psychol. 36, 845–856 (2006) If people can exert voluntarily control over attitudes on implicit measures, then they, like explicit measures, may be susceptible to conscious attempts to misrepresent one’s true attitudes in a socially desirable manner, i.e., faking. However, simply finding support for the hypothesis that people show more positive attitudes toward stigmatized groups on implicit measures when those measures are given in a public versus a private assessment situation does not necessarily mean that people consciously altered their responses in a socially desirable manner. Blair (2002) has proposed that the malleability of attitudes assessed with implicit measures occurs in response to changes in automatic processing. According to this perspective, the influence of the assessment situation on implicitly assessed attitudes occurs outside of people’s awareness and thus does not directly reflect conscious efforts to provide socially desirable responses. OVERVIEW OF RESEARCH AND HYPOTHESES The present research examines whether a public assessment situation alters implicitly assessed attitudes in much the same way that it alters explicitly assessed attitudes and whether the hypothesized effect of the assessment situation on implicitly assessed attitudes is the result of conscious efforts to provide socially desirable responses. To test whether the assessment situation influences implicitly and explicitly assessed attitudes in a similar way, participants in Experiment 1 completed implicit and explicit measures of bias toward homosexuality in either a public or a private assessment situation. Participants in the public assessment situation believed that the experimenter would know their level of bias toward homosexuality whereas participants in the private assessment situation believed that the experimenter would not know their level of bias. We predicted that bias toward homosexuality would be lower in the public than in the private assessment situation on both the implicit and explicit measures. Then, in Experiment 2, we considered whether the expected influence of the assessment situation on implicitly assessed attitudes reflected socially desirable response patterns that are under voluntary control. To this end, participants in Experiment 2 completed an IAT measuring bias toward homosexuality in a public assessment situation believing that the experimenter either would or would not be able to detect whether they had responded truthfully.