Awareness of implicit attitudes

Abstract

Research on implicit attitudes has raised questions about how well people know their own attitudes. Most research on this question has focused on the correspondence between measures of implicit attitudes and measures of explicit attitudes, with low correspondence interpreted as showing that people have little awareness of their implicit attitudes. We took a different approach and directly asked participants to predict their results on upcoming Implicit Association Test (IAT) measures of implicit attitudes toward 5 social groups. We found that participants were surprisingly accurate in their predictions. Across 4 studies, predictions were accurate regardless of whether implicit attitudes were described as true attitudes or culturally learned associations (Studies 1 and 2), regardless of whether predictions were made as specific response patterns (Study 1) or as conceptual responses (Studies 2-4), and regardless of how much experience or explanation participants received before making their predictions (Study 4). Study 3 further suggested that participants' predictions reflected unique insight into their own implicit responses, beyond intuitions about how people in general might respond. Prediction accuracy occurred despite generally low correspondence between implicit and explicit measures of attitudes, as found in prior research. Altogether, the research findings cast doubt on the belief that attitudes or evaluations measured by the IAT necessarily reflect unconscious attitudes.

Figure 1

Prediction scale participants used to make their predictions of their IAT score (example of Black-White IAT) in Study 1. Photos used in the actual IATs were depicted above the ends of the scales on the left and right. In Study 2, labels below the buttons were changed (see text).

Figure 2

Study 1: Mean IAT scores by condition. Higher scores mean more positive implicit attitudes towards the comparison group (i.e., regular White adult). Negative scores indicated more positive scores towards the target group (Black, Asian, Latino, celebrity, or child). Error bars are calculated from mean square errors from a 5 (target) x 2 (condition) ANOVA on IAT scores.

Figure 3

Study 3: Mean predictions of participants’ own IAT score and predictions for IAT scores of the average participant participating in the same study. Scales range from 1–7 with scores above 4 indicating more bias in favor of the comparison group (White, regular, or adult), and score below 4 indicating bias in favor of the target group (Black, Asian, Latino, celebrity, or child). All pairwise differences between predictions for self and predictions for the average participant are significant (see text). Error bars are calculated from mean square errors from a 2 (self vs. other) x 5 (targets) x 2 (self first vs. other first) ANOVA.

Figure 4

Study 4: Average within-participant correlation between IAT score predictions and actual IAT scores by condition. Error bars are calculated from mean square errors from a 2 (explanation) x 2 (experience) ANOVA on participants’ individual correlations calculated separately in a first step.

Figure 5

Average IAT score predictions (1–7 scale) and average actual IAT D scores across all four studies. Shaded areas represent the areas in which an implicit attitude would be labeled as “slightly more positive” on the predictions scales or as a “slight preference” according to conventions from the IAT webpage (www.projectimplicit.com, Nosek et al. 2006, personal communication from N. Sriram to I. Blair on July 6, 2009).

Figure 6

Mean thermometer ratings before participants made their IAT score predictions (time 1) and after they completed all IATs (time 2) across all four studies (N=430). Scales range from 0–100 with higher score indicating more positive evaluations. Ratings of “regular people” and “adults” were only assessed in Studies 3 and 4. Hence, those means are based on a smaller sample (N=275). Error bars reflect standard errors.