Constructing a cross-race face identity triad test.

Abstract

Despite the challenges associated with cross-race face identification, there are no publicly available tests of people's ability to identify own- versus other-race faces. We introduce the Cross-Race Face Identity Triad (CR-FIT) test, designed to be challenging for individuals of varying abilities. A key methodological advantage of the CR-FIT test over other face identity matching tests is that it eliminates response bias in face-identity matching through the use of face-image triads. A triad of face images includes two images of the same person and one image of a different person; participants must select the image of the "different" person. A second advantage of the CR-FIT test is that it ensures comparable difficulty for face stimuli from two races (African American, AA, Caucasian, CA) by leveraging machine and human pre-screening of items. This prescreening assures that performance differences reflect own- versus other-race identity matching ability, rather than differences in stimulus set difficulty. Triads were pre-screened to be challenging based on the performance of a publicly available face-identification algorithm. Items were further screened for comparability of difficulty using the performance of AA and CA observers. The test yields a classic "other-race effect"- observers identify own-race faces more accurately than other-race faces, with no effect of item race. Performance for untrained participants exceeded chance, but was far below ceiling, making the test suitable for a wide range of face identity matching abilities. The CR-FIT test is publicly available in an open science platform for research purposes.