Figure 1 depicts the overall results by plotting the deviations from baseline that occurred as a result of specifying race, gender, or both.
Specifying race or gender
When only one or the other was specified, race and gender seemed to have orthogonal effects. On one hand, for the evaluative dimension, the LMER indicated that specifying race as Black produced a reliable negative shift in evaluative scores, β = − 0.22, t(49.0) = 2.58, p = 0.013, whereas specifying race as White produced a marginally positive shift in evaluative scores, β = 0.14, t(128.8) = 0.051, relative to leaving race unspecified. There was no reliable effect of gender designations on the evaluative dimension. On the other hand, for the potency dimension, specifying race had no reliable effect, but specifying gender as Female produced a reliable reduction in perceived potency, β = − 0.29, t(22.6) = 3.08, p = 0.005, and specifying gender as Male produced a reliable increase in perceived potency, β = 0.43, t(21.2) = 4.28, p < 0.001.
Because race only affected the evaluative dimension and gender only affected the potency dimension, these effects are depicted in Fig. 1 as wide lines forming a box in the two-dimensional semantic space of evaluation and potency. The widths of these vertical and horizontal lines correspond to twice the magnitude of the standard errors computed in the LMER. The words “Male,” “Female,” “Black,” and “White” are presented at their respective mean locations in the two-dimensional space relative to the origin represented by the baseline condition.
Specifying race and gender
If the effects of race and gender were additive, we should expect that combinations of race and gender would fall near to the corners of the semantic space formed by the separate effects of race and gender alone. Indeed, for three of the four combinations tested (Black Male, White Male, and White Female), LMERs found the predicted deviations both in evaluative scores, β = − 0.27, t(51.6) = 2.99, p = 0.004, β = 0.20, t(119.1) = 3.08, p = 0.007, β = 0.20, t(103.7) = 2.61, p = 0.010, respectively, and in potency scores, β = 0.36, t(20.2) = 2.79, p = 0.011, β = 0.41, t(24.9) = 4.50, p < 0.001, β = − 0.20, t(26.0) = 2.36, p = 0.026, respectively. These effects can be seen by the locations of the circular symbols in Fig. 1.
However, there is a fourth combination which clearly does not fit the additive model. As is evident in Fig. 1, specifying Black Female had no reliable effect on either the evaluative dimension or the potency dimension compared to the baseline condition where neither race nor gender was specified. There was no shift in the evaluative dimension, β = 0.00, t(65.7) = 0.00, p = 0.999, nor was there a reliable shift in the potency dimension, β = − 0.08, t(24.4) = 0.83, p = 0.415.
Participant identity (exploratory analysis)
Because the ratings were framed in terms of “how does society view…” (rather than “how do you view…”), it seems worth exploring how the race and gender identity of each of the participants related to their perception of how society judges people according to race and gender (McCormick-Huhn et al., 2019). Including participant race and gender in the LMER analyses supported the following conclusions.
When only race or gender was included in the social role categories, an LMER on the evaluative dimension showed that Black categories had lower evaluative scores relative to the case without race or gender, β = − 0.30, t(246.0) = 2.17, p = 0.031, and there were no reliable interactions with participant race or gender.
When only race or gender was included in the social categories, an LMER on the potency dimension showed that female categories had lower potency scores relative to the case without race or gender, β = − 0.25, t(60.4) = 2.11, p = 0.039, and male categories had elevated potency scores, β = 0.56, t(50.4) = 4.44, p < 0.001. However, the elevation of male categories on potency was smaller among male participants than among female participants, β = − 0.26, t(775.3) = 2.06, p = 0.039. This effect of participant gender on the perception of male power is consistent with the idea that female participants are more sensitive to perceived power differences between men and women.
When both race and gender were included in social categories, an LMER on the evaluative dimension showed that both White Male categories and White Female categories had elevated evaluative scores relative to the baseline judgments for social roles without race or gender, WM: β = 0.30, t(535.5) = 2.33, p = 0.020, WF: β = 0.33, t(489.9) = 2.52, p = 0.012. Black Male categories had lowered evaluative scores overall, β = -0.73, t(239.1) = 5.20, p < 0.001, but this effect was modulated both by participant race, with White participants giving higher evaluative scores than Black participants, β = 0.57, t(774.9) = 3.21, p = 0.001, and by participant gender, with male participants giving higher estimates than female participants, β = 0.55, t(774.9) = 3.15, p = 0.002. The participant race effect on the evaluative scores for Black Male categories is consistent with the idea that Black participants are more aware of societally-embedded negative evaluations of Black men than are White participants.
When both race and gender were included in the social role identity, an LMER on the potency dimension showed that both White Male social categories and Black Male social categories had elevated potency scores relative to social roles without race or gender, WM: β = 0.48, t(75.1) = 4.03, p < 0.001, BM: β = 0.65, t(38.2) = 4.30, p < 0.001. For Black Male categories, however, this effect was modulated both by participant race, with Black participants giving higher potency scores for Black Males than did White participants, β = 0.30, t(775.5) = 2.35, p = 0.019, and by participant gender, with female participants giving higher potency scores to Black Male identities than did male participants, β = 0.40, t(775.5) = 3.07, p = 0.002. This latter effect is consistent with females’ overall higher ratings of potency for males, but was only reliable for Black Male social categories in the present analysis.
Figure 2 shows how the main graph would look for each subset of participants. Overall, when the race and gender of participants were factored in, the main conclusions remained similar regarding effects of race, gender, and their combination. The exploratory analysis supports the idea that Black participants and Female participants were more sensitive to societal biases regarding both race-based and gender-based differences in perceived evaluation and perceived power. Notably, with respect to the issue of intersectional identities, in no analysis did the perception of Black Female social categories reliably differ from social categories where neither gender nor race was specified.