Fig. 5

The graph shows the interaction between the Congruency, Race, and AQ score (F (1, 37,302) = 10.73; p = .001). Specifically, it shows how the interaction between Congruency and Race changes at the various levels of the AQ score, that are: AQ = 10 (-1SD) in the upper left panel; AQ = 16 (mean) in the upper right panel; AQ = 23 (+ 1SD) in the lower left panel; AQ = 29 (+ 2SD) in the lower right panel. The graph shows that at medium–low, medium, and medium–high AQ score levels (i.e., AQ = 10; AQ = 16; AQ = 23) participants were faster in the congruent condition than in the incongruent one, with both White (AQ = 10: ratio = 0.98; p _Bonferroni < .0001; AQ = 16: ratio = 0.98; p _Bonferroni < .0001; AQ = 23: ratio = 0.97; p _Bonferroni < .0001) and Black gaze-cueing faces (AQ = 10: ratio = 0.97; p _Bonferroni < .0001; AQ = 16: ratio = 0.98; p _Bonferroni < .0001; AQ = 23: ratio = 0.99; p _Bonferroni = .009). Whereas at high AQ score level (AQ = 29, lower right panel) participants were faster in the congruent condition than in the incongruent one only with White faces (ratio = 0.97; p _Bonferroni < .0001; M = 542; SE = 15.2; 95% CI [513, 572] vs. M = 561; SE = 15.7; 95% CI [531, 593]) and not with Black ones (ratio = 0.995; p _Bonferroni = 1; M = 551; SE = 15.4; 95% CI [522, 582] vs. M = 554; SE = 15.5; 95% CI [524, 585]). The shaded areas around lines represent the 95% confidence intervals. Black asterisks are referred to the solid lines, white asterisks to the dashed lines. ** p < .01, *** p < .001