Does confidence predict face-identification accuracy for same-race and other-race faces?

Confidence in face-identification decisions can influence legal outcomes. Additionally, people are less accurate at recognizing other-race than same-race faces (other-race effect [ORE]). Although confidence predicts face-identification accuracy for same-race faces (Hahn et al., 2021; Jeckeln, et al., 2022), it is unclear whether this finding holds for other-race faces. Here, we used a comparative-confidence task (Mamassian, 2016) to examine observers’ ability to evaluate the correctness of their face-matching decisions for same-race and other-race faces. Participants (27 White/Caucasian [W/CA], 27 Black/African American [B/AA]) completed a comparative-confidence task embedded in a face-identity matching test: On each face-matching trial (24 B/AA trials, 24 W/CA trials), participants viewed three face images (two same-identity images and one different-identity image), and selected the odd-one-out. After completing two face-matching trials (2 B/AA trials or 2 W/CA trials), participants selected the trial on which they felt more confident (confidence-selected trial). There was an ORE for identification accuracy (F(1, 54) = 6.25, p = .015, η2= .10). However, the ORE did not affect the confidence-accuracy relationship: Accuracy was greater for confidence-selected trials than confidence-rejected trials for both same-race and other-race faces (p < .01). Additionally, trial difficulty (DifficultyTrial1-DifficultyTrial2) predicted B/AA participants’ confidence choice for same-race (R^2 = .51, F(1,12) = 10.59, p < .01) and other-race faces (R^2 = .69, F(1, 10) = 21.87, p < .001), and W/CA participants’ confidence choice for same-race faces (R^2 = .47, F(1, 10) = 8.712, p = .0145), but not other-race faces (p = .2519). These results demonstrate that confidence serves as a predictor of accuracy in the context of same-race and other-race face-identity matching and that confidence is driven by trial difficulty.