Measuring sexual dimorphism with a race-gender face space.

Faces are complex visual objects, and faces chosen to vary in 1 regard may unintentionally vary in other ways, particularly if the correlation is a property of the population of faces. Here, we present an example of a correlation that arises from differences in the degree of sexual dimorphism. In Experiment 1, paired similarity ratings were collected for a set of 40 real face images chosen to vary in terms of gender and race (Asian vs. White). Multidimensional scaling (MDS) placed these stimuli in a "face space," with different attributes corresponding to different dimensions. Gender was found to vary more for White faces, resulting in a negative or positive correlation between gender and race when only considering male or only considering female faces. This increased sexual dimorphism for White faces may provide an alternative explanation for differences in face processing between White and Asian faces (e.g., the own-race bias, face attractiveness biases, etc.). Studies of face processing that are unconfounded by this difference in the degree of sexual dimorphism require stimuli that are decorrelated in terms of race and gender. Decorrelated faces were created using a morphing technique, spacing the morphs uniformly around a ring in the 2-dimensional (2D) race-gender plane. In Experiment 2, paired similarity ratings confirmed the 2D positions of the morph faces. In Experiment 3, race and gender category judgments varied uniformly for these decorrelated stimuli. Our results and stimuli should prove useful for studying sexual dimorphism and for the study of face processing more generally.