Changing Race Boundary Perception by Reading Narrative Fiction

Abstract

Participants read a story about a counterstereotypical Muslim woman and were then asked to determine the race of ambiguous-race Arab-Caucasian faces. Compared to a content-matched control condition, participants who read the narrative exhibited lower categorical race bias by making fewer categorical race judgments and perceiving greater genetic overlap between Arabs and Caucasians (Experiment 1). In Experiment 2, participants determined the race of ambiguous-race Arab-Caucasian faces depicting low and moderate anger. Emotion-related perceptual race bias was observed in the control conditions where higher intensity anger expressions led participants to disproportionately categorize faces as Arab. This bias was eliminated in the narrative condition.

Notes

¹Pilot testing (N=20) indicated these faces were ambiguous (M=2.25, SD=.63). Participants rated the faces on a 4-point scale: 1 (Arab), 2 (mixed, more Arab than Caucasian), 3 (mixed, more Caucasian than Arab), 4 (Caucasian).

Note. For Experiment 1, narrative n=33, synopsis n=35. For Experiment 2, narrative n=38, synopsis n=37, control n=35.

Note. df=66 for all tests.

²Although there was a significant difference in comprehension between the narrative (M=89.70%, SD=18.11%) and synopsis conditions (M=66.29%, SD=22.63%), F(1, 66)=22.01, p<.001, none of the primary results changed (i.e., genetic overlap or categorizations) after including comprehension as a covariate, and there was not a correlation between comprehension and either dependent variable in either condition.

³Pilot testing (N=20) indicated the face sets were low (M=2.29, SD=.53) and moderate (M=2.60, SD=.59) in anger intensity, F(1, 19)=10.00, p=.005. Participants were asked how much anger was present on each face, on a 4-point scale from 1 (none) to 4 (a significant amount), and 2 and 3 were not labeled. We inferred that ratings in the 2 and 3 range indicated low and moderate amounts of anger, respectively. Critical for the current study was that participants reliably detect an increasing amount of anger across the two subtle intensities, and results confirm this perception. It is also important to note that the averages do not approach 4; this indicates that the anger expressions, on the whole, were not perceived as high intensity.

⁴The same analysis was run with continuous race judgments using each individual's average race judgment score for low and moderate intensity faces as dependent measures. A 3 (between-subjects condition: narrative, synopsis, control)×2 (within-subjects anger intensity: low vs. moderate) mixed ANOVA was performed. There was a marginal main effect of anger intensity, F(1, 107)=3.79, p=.054, and a significant interaction did not emerge using continuous race judgments, F(2, 107)=0.95, p=.389. However, an inspection of cell means indicates trends consistent with significant results reported using categorical judgments. Individuals in the narrative condition exhibited the smallest change in continuous race judgments from low to moderate anger intensity (M=2.50 − 2.50), followed by the control (M=2.35 − 2.29), followed by the synopsis (M=2.44 − 2.35). Lower numbers indicate a shift toward Arab race judgments, where 1=Arab, 4=Caucasian. It should be noted that the current study's results are consistent with Hugenberg and Bodenhausen's (2004) findings in that they found not a main effect of race on anger intensity but instead an interaction with implicit prejudice, where only individuals high in implicit prejudice rated outgroup race faces as depicting higher anger intensities compared to ingroup faces of the same actual intensity.

FIGURE 1 Narrative fiction's reduction of the outgroup categorization shift for increasing anger intensity in ambiguous Arab-Caucasian faces from Experiment 2. Higher Arab-Caucasian categorization scores indicate more Arab than Caucasian categorical judgments. The potential range was from −12 (all Caucasian categorizations; 0 − 12 =− 12) to 12 (all Arab categorizations; 12 − 0 = 12). Standard error bars are depicted.

Note. For narrative vs. synopsis (df=73), narrative vs. control (df=71), synopsis vs. control (df=70).

⁵Although there was a significant difference in comprehension between the narrative (M=88.94%, SD=18.42%) and synopsis conditions (M=67.02%, SD=20.55%), t(73)=5.02, p<.001, the primary results did not change (i.e., categorizations) after including comprehension as a covariate, nor was there a correlation between comprehension and categorization difference scores in any condition.

⁶Thanks to a reviewer for this insight.

⁷Although transportation was assessed in the current study, the power was too low to test a moderation hypothesis, and the study was not designed to investigate the role of individual differences.