Cross-Racial Identification of Transformed, Untransformed, and Mixed-Race Faces

Abstract

Many studies report that people of one race find other-race faces difficult to remember. The generality of this observation was tested by examining the recognition memory of 2 groups of European (n = 32), and African (n = 32) subjects on same-race, other-race, and mixed-race faces, which were either transformed or untransformed in pose between initial study and recognition test. A mixed-design ANOVA (2-between × 1-within factors), using the Signal Detection Theory measures of discrimination accuracy (d') revealed the following: (1) a significant interaction between Race of Subject × Race of Face, with Europeans being more accurate than Africans at identifying untransformed white faces. and Africans also better than Europeans in identifying transformed black faces; (2) a significant interaction between Transformation interaction between Transformation Race of Subject, in that while Europeans were significantly more accurate than Africans on untransformed faces as a whole, Africans were slightly superior to Europeans in terms of accuracy on transformed faces; (3) untransformed faces were recognized better than transformed faces; (4) there were no significant main effects due either to Race of Subject or to Race of Face; (5) no significant differences were found between the 2 groups on accuracy for either transformed or untransformed mixed-race faces; (6) no significant interaction between Race of Face × Transformation. and no significant 3-way interaction between Race of Subject × Race of Face × Transformation. It was concluded that differences in ‘race’ as a variable alone may not be responsible for the differences usually found on accuracy in cross-race eyewitness identifications. Rather, differences in levels of task difficulty and individual differences in recognition strategies adopted by subjects irrespective of race of stimulus face, may be the most important variables. The implications of these results for the efficient design of cross-race identification tasks for both forensic and training purposes were discussed.