Categorising intersectional targets: An “either/and” approach to race- and gender-emotion congruity

ABSTRACT

Research on the interaction of emotional expressions with social category cues in face processing has focused on whether specific emotions are associated with single-category identities, thus overlooking the influence of intersectional identities. Instead, we examined how quickly people categorise intersectional targets by their race, gender, or emotional expression. In Experiment 1, participants categorised Black and White faces displaying angry, happy, or neutral expressions by either race or gender. Emotion influenced responses to men versus women only when gender was made salient by the task. Similarly, emotion influenced responses to Black versus White targets only when participants categorised by race. In Experiment 2, participants categorised faces by emotion so that neither category was more salient. As predicted, responses to Black women differed from those to both Black men and White women. Thus, examining race and gender separately is insufficient to understanding how emotion and social category cues are processed.

KEYWORDS:

• Emotional expressions
• stereotypes
• social categorisation
• intersectionality
• Black women

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹In addition to response time, our primary dependent measure of interest, we conducted an analogous ANOVA on the average percentage of incorrect responses for each gender, race, and expression combination. The 4-way interaction was significant, F(2, 164)=5.79, MSE=.002, p=.004. Further details of these analyses are available from the first author.

²An analysis with participants’ gender as a between-subjects factor revealed no main or interaction effects, and therefore this factor is not included in the analyses.

³Supplementary analyses including only non-Black participants produced the same significant interactions, and follow-up comparisons reported as significant remained significant. Analyses with a reduced sample including only White participants also produced significant interactions, and follow-up comparisons remained significant, with the exception of the pairwise comparisons for angry male vs. female targets (p=.177) and happy male vs. female targets (p=.167). An analysis with participants’ race (White vs. non-White) as a between-subjects factor revealed a participant race x target gender interaction, p=.015. Non-White participants categorised male faces faster than female faces, whereas White participants showed no target gender difference in response times. No other main or interaction effects with participant race were significant.

⁴Following the recommendations of Judd et al. (2012), we conducted supplementary analyses using a mixed model with crossed random factors, in which both stimuli and participant were treated as random factors. The advantage of this alternative analysis is that it accounts for random variation in the stimuli, but one concern about this analytic approach is that if the stimuli have been carefully selected and pre-tested to ensure they do not vary largely on factors such as attractiveness and category typicality, as in the present research, and thus are not randomly sampled, including the stimuli effect in the analyses reduces statistical power, increasing the likelihood of Type II errors. In Experiment 1, the significant gender × expression × categorisation task interaction persisted in this new analysis, as did all subsequent pairwise comparisons, but the race × expression × categorisation task interaction and subsequent pairwise comparisons did not.

⁵The 3-way interaction of target gender, race, and categorisation task for response times to correctly identified stimuli was also significant, p=.030. Consistent with previous work (Johnson et al., 2011; Thomas et al., 2014), Black women were categorised as female slower than were White women, and as Black slower than were Black men. The reported effects of emotional expression on categorisation speed occurred above and beyond these differences.

⁶Whereas previous work has varied in terms of the particular keys assigned to each category and whether task key assignments were counterbalanced or consistent across targets (Becker et al., 2007; Bijlstra et al., 2010; Lipp et al., 2015; Smith et al., 2015), future work should directly address whether these factors have a systematic impact on responses.

⁷Average incorrect responses made by each participant were also analysed using analogous ANOVAs. The three-way interaction was not significant for incorrect responses in either task, ps>.543 Details of these analyses are available from the first author.

⁸Initial analyses with participants’ gender again showed no significant main or interaction effects.

⁹Supplementary analyses including only non-Black participants produced the same significant interactions, with the exception of the gender x expression interaction for White targets in the angry vs. neutral trials (p=.082), and follow-up comparisons remained significant. Analyses with a reduced sample including only White participants produced a non-significant 4-way interaction, p=.240, and a marginal 3-way interaction in the angry vs. neutral trials, p=.111; however, significant follow-up comparisons remained significant. An analysis with participants’ race (White vs. non-White) as a between-subjects factor revealed no main or interaction effects.

¹⁰We again conducted supplementary analyses using a mixed model with crossed random factors, in which both stimuli and participant were treated as random factors. The overall 4-way interaction (p=.322) and 3-way interaction in the angry versus neutral task did not reach significance, but the pattern of significance among the pairwise comparisons remained the same, with faster response times to anger (vs. neutral expressions) on Black male faces and neutral expressions (vs. anger) on White female faces and no differences in response times to anger and neutral expressions on Black female and White male faces. In addition, there were significant gender x expression × task and race × expression × task interactions (p=.004 and p=.043, respectively), which demonstrated that in the angry vs. neutral but not happy vs. neutral task, anger was categorised faster than neutral expressions on male faces and Black faces overall, and anger was categorised more slowly than neutral expressions on female faces and White faces overall.