Differences in exposure to individuals of the same-race and another race was expected to moderate the CRE. Thus, it was important to first determine the extent to which exposure to same-race people was greater than that of the other-race people. A repeated measures ANOVA showed a main effect of exposure in White, F(3, 162) = 32.60, p < 0.001, \({\eta }_{p}^{2}\)= 0.376, and Black Participants, F(3, 78) = 16.73, p < 0.001, \({\eta }_{p}^{2}\)= 0.392. Contrasts showed that White participants had significantly more exposure with White individuals than Asian, F(1, 54) = 69.37, p < 0.001, \({\eta }_{p}^{2}\)= 0.562, Black, F(1, 54) = 61.70, p < 0.001, \({\eta }_{p}^{2}\)= 0.533, and Latino individuals, F(1, 54) = 22.07, p < 0.001, \({\eta }_{p}^{2}\)= 0.290. Additional comparisons showed that White participants reported more exposure with Latino individuals than Asian, t(55) = 3.39, p = 0.001, and Black individuals, t(55) = 4.08, p = 0.001. As for Black participants, contrasts showed more exposure with Black individuals than Asian, F(1, 26) = 47.88, p < 0.001, \({\eta }_{p}^{2}\)= 0.648, and Latino individuals, F(1, 26) = 5.92, p = 0.022, \({\eta }_{p}^{2}\)= 0.185. Additional comparisons showed that Black participants reported less exposure to Asian individuals than Latino, F(1, 26) = 15.12, p = 0.001, \({\eta }_{p}^{2}\)= 0.368, and White individuals, F(1, 26) = 38.88, p < 0.001, \({\eta }_{p}^{2}\)= 0.599.
Separate repeated measures ANOVAs were conducted on recognition accuracy (d′) comparing high-ET same-race faces to high-ET and low-ET other-race faces for White and Black participants (see Fig. 1 for means). Furthermore, supplemental analyses were conducted entering differences in exposure to same-race and the relevant other-race persons as a covariate and are presented in brackets. Planned contrasts were assessed with a Bonferroni correction alpha level of 0.0167, representing the three relevant comparisons necessary to test for the CRE for each other-race face.
High-ET same-race and high-ET other-race
In White participants, comparing high-ET White faces to high-ET other-race faces, the Hyunh–Feldt (Epsilon = 0.931, p = 0.012) analysis, used to correct a violation of sphericity, found a race of face main effect, F(2.79, 150.86) = 27.44, p < 0.001, \({\eta }_{p}^{2}\)= 0.337. Planned contrasts showed that participants recognized high-ET White faces better than high-ET Asian, F(1, 54) = 35.16, p < 0.001, \({\eta }_{p}^{2}\)= 0.394 [F(1, 53) = 14.41, p < 0.001, \({\eta }_{p}^{2}\)= 0.214], Black, F(1, 54) = 72.60, p < 0.001, \({\eta }_{p}^{2}\)= 0.573 [F(1, 53) = 34.86, p < 0.001, \({\eta }_{p}^{2}\)= 0.397], and Latino faces, F(1, 54) = 41.27, p < 0.001, \({\eta }_{p}^{2}\)= 0.436 [F(1, 53) = 43.80, p < 0.001, \({\eta }_{p}^{2}\)= 0.452]. These results suggest that the CRE occurred consistently for each other-race face. Also, it is noteworthy that the effect size decreased in a way that suggests exposure moderated the effect except in the case of Latino faces.
Another interest was whether the size of the CRE varied significantly by the type of other-race face. Thus, a repeated measures ANOVA was conducted on the difference scores between high-ET White and high-ET other-race faces. The Huynh–Feldt analysis (Epsilon = 0.870, p = 0.002), used to correct a violation of sphericity, found a main effect of CRE comparison (i.e., White Asian, White Black, and White Latino), F(1.74, 93.97) = 4.19, p = 0.023, \({\eta }_{p}^{2}\)= 0.072. Planned contrasts showed that the CRE was significantly larger for Black faces than Asian, F(1, 54) = 10.32, p = 0.002, \({\eta }_{p}^{2}\)= 0.161, and Latino faces, F(1, 54) = 4.59, p = 0.037, \({\eta }_{p}^{2}\)= 0.078, but this last difference was not a significant at the Bonferroni correction level (i.e., α = 0.0167). This suggests that discriminability of Black faces appears to be more difficult for these White participants which could be a testament to the quality of their perceptual expertise on those faces. However, exposure ratings were similar for Black and Asian faces, suggesting that these effects are either due to aspects of exposure not reflected in the scale or socio-cognitive mechanism that hinder processing of Black faces more so than Asian faces.
For Black participants, the CRE was not as prevalent as White participants. Comparing high-ET Black faces to high-ET other-race faces, analysis found a race of face main effect, F(3, 81) = 3.57, p = 0.017, \({\eta }_{p}^{2}\)= 0.117. Planned contrast showed that participants recognized Black faces better than Asian faces, F(1, 27) = 20.28, p < 0.001, \({\eta }_{p}^{2}\)= 0.429 [F(1, 24) = 5.35, p = 0.030, \({\eta }_{p}^{2}\)= 0.182], but not Latino, p = 0.205, or White faces, p = 0.449. Note that controlling for exposure to Asian faces not only reduced the size of the CRE, but also changed the difference to nonsignificant at the Bonferroni correction level (i.e., α = 0.0167). Furthermore, the low exposure to Asian individuals stands out among the other types of other-race faces for which no CRE was found. Moreover, this result is a partial replication of Gross (2009) who found that Black participants exhibited the CRE for Asian and Latino faces, but not White faces in a diverse face array.
High-ET same-race and low-ET other-race
Comparing high-ET White faces to low-ET other-race faces, there was also a race of face main effect, F(3, 162) = 6.52, p < 0.001, \({\eta }_{p}^{2}\)= 0.108. Planned contrasts showed that participants recognized high-ET White faces better than low-ET Asian, F(1, 54) = 23.64, p < 0.001, \({\eta }_{p}^{2}\)= 0.305 [F(1, 53) = 16.59, p < 0.001, \({\eta }_{p}^{2}\)= 0.238], Black, F(1, 54) = 11.64, p = 0.001, \({\eta }_{p}^{2}\)= 0.177 [p = 0.079] and Latino faces, F(1, 54) = 5.93, p = 0.018, \({\eta }_{p}^{2}\)= 0.099 [F(1, 53) = 7.79, p = 0.007, \({\eta }_{p}^{2}\)= 0.128], but not at the Bonferroni correction level. Note that exposure reduced the size of the CRE for Asian faces, eliminated the CRE for Black faces, and exacerbated it for Latino faces. The exposure effect with Latino faces was shown from a different statistical vantage point (i.e., a regression analysis) in Marsh (2021). In either case, it suggests that an increase in exposure may improve one’s ability to racially identify these racially ambiguous Latino faces in a way that hinders the processing of their individuating features. When assessing the size of the CRE by type of other-race face, there was no main effect of CRE comparison, p = 0.324. This suggests that the size of the CRE did not vary by race type in racially ambiguous other-race faces.
In addition, a 2 (Ethnic Typicality: High and Low) X 3 (CRE Comparison) repeated measures ANOVA was conducted to assess the effect that racial ambiguity has on the size of the CRE while ignoring type of other-race face. The analysis found a main effect of ethnic typicality, F(1, 54) = 35.31, p < 0.001, \({\eta }_{p}^{2}\)= 0.395, wherein the CRE was larger when comparing high-ET White faces to high-ET other-race faces (M = 1.24, SE = 0.146) than low-ET other-race faces (M = 0.619, SE = 0.144). A similar analysis was not necessary for Black participants, because when comparing high-ET Black faces to low-ET other-race faces, there was no race of face main effect, p = 0.724, thus no CRE. These findings suggest that racial ambiguity affords other-race faces enough individuating processing to mitigate the CRE in White participants and eliminate it in Black participants. One plausible explanation of these effects is that racial clarity about an other-race face facilitates access to the face’s social category and its deindividuation. In contrast, racial ambiguity limits access to the face’s social category, therein increasing the potential for the face to be individuated. Also, exposure did not moderate the CRE for high-ET and low-ET Latino faces in White participants, even though participants had the most exposure with Latino individuals than Asian or Black individuals. This race-specific discrepancy suggests either exposure’s mitigating effect could be moderated by the type of other-race and the experimental context, or that exposure ratings do not represent the same quality of contact for each race.
However, these effects occurred under circumstances that highlighted the racial category of each face. In fact, the racial identification task was expected to exacerbate the CRE at least among high-ET faces. Thus, there is a question of whether the results will replicate when presenting the same procedure without the racial identification task. Moreover, would the effect size of the CRE comparisons be noticeably smaller once the racial identification task is removed?