Other-race faces are given more weight than own-race faces when assessing the composition of crowds

In two experiments we examined the performance of Asian and Caucasian participants as they were asked to estimate the ethnic composition of arrays of 16 concurrently presented faces. Across trials we systematically varied the physical proportion of Asian and Caucasian faces presented in the arrays using the method of constant stimuli. The task was to explicitly indicate which group was in the majority. The position of the 16 faces within the array were continuously shuffled using a 4 x 4 moving grid to block explicit enumeration. Measures of bias and sensitivity were estimated by fitting cumulative normal distributions to individual response data. Consistent with recent findings on "ensemble" face processing, all participants were able to make group estimates quite accurately. This was true using both full-colour, non-normalised, headshots (Exp1) and centre-apertured, normalised, grey-scale images (Exp2). However, the main finding was that performance estimates from the two groups of participants did not overlap. Specifically, patterns of bias suggest that other-race faces are weighted more heavily than own-race faces (Exps 1 & 2), while sensitivity is better for groups instructed to decide if the other-race, rather than own-race, is more numerous (Exp 2). To our knowledge, these are the first demonstrations of other-race biases affecting decisions that have to be made about groups of faces.