Dilhan Töredi¹, Eric Y. Mah², Megan Lall³, Haiyang Jin⁴, Brett D. Roads⁵, James W. Tanaka²; ¹John Jay College of Criminal Justice, ²University of Victoria, ³Adler University, ⁴Zhejiang Sci-Tech University, ⁵University College of London

The cross-race effect (CRE) refers to better recognition of same-race than cross-race faces (Lee & Penrod, 2022). The Face Space Model (Valentine, 1991) attributes the CRE to the fact that more frequently encountered same-race faces are represented with diverse features in perceptual space, making them more differentiated in memory. In contrast, less frequently encountered cross-race faces cluster around race-typical features, resulting in high perceptual similarity and reduced discriminability (Valentine, Lewis & Hills, 2016). To test the Face Space Model predictions, participants (N= 46 African, N = 45 Caucasian, N = 40 Chinese) judged the similarity of African, Caucasian, and Chinese faces by selecting the most similar faces to query faces from a 3x3 grid. Using these judgements, we generated multidimensional face spaces (i.e., psychological embedding) for same- and cross-race faces using the PsiZ platform (https://psiz.org). Contrary to the Face Space Model predictions, cross-race faces were judged as more differentiated from one another than same-race faces. However, results were less inconsistent when analyzed by participant race. Specifically, African participants showed a more differentiated face space for Caucasian faces, Caucasian participants for Chinese faces, and Chinese participants for African faces. Results were not explained by differential levels of racial contact. In a follow-up experiment, these results were replicated with Chinese (N = 35) and Caucasian (N = 25) participants and new stimuli. As a test of CRE, these stimuli were employed in a standard recognition memory paradigm with Caucasian participants (N = 37). We found a robust CRE; Caucasian participants showed better recognition of same-race faces than cross-race faces, notwithstanding these faces were not more differentiated in perceptual similarity judgements. The diverging results suggest that the Face Space Model cannot fully account for the CRE, suggesting that race-based differences in perceptual similarity cannot solely drive differences in recognizing same-race and cross-race faces.

Acknowledgements: This work was supported by an NSERC Discovery Grant awarded to James W. Tanaka (34665-58400)