Abstract
It has been proposed that the categorization of a face as part of an ethnic group occurs spontaneously, whereas its individuation is effortful (Hugenberg et al., 2010). In this framework, the other-race effect (ORE) arises from a tendency to attend race-specific, as opposed to identity-specific, features. Here, dynamic Bubbles (Vinette et al., 2004) were used to investigate the time course of feature utilization during the identification and categorization of same-race (SR) and other-race (OR) faces. The stimuli consisted of 300ms movies displaying a face (8 Caucasian, 8 African-American) in which information was randomly sampled through time. On each trial, the participant (N=8, 9600 trials) had to decide which of the 16 identities was presented. The number of bubbles was adjusted such that on 15% of the trials, race-categorization errors occurred (erroneous identification of a face of the wrong ethnicity). This manipulation allowed us to reveal, using a single task, identity-specific and race-specific information. On average, the participants correctly identified 39.9% of the SR, and 27.4% of the OR, faces, replicating the ORE [t(7)=4.01, p<0.05]. We first computed static classification images (CI) showing race-specific and identity-specific visual information by performing a multiple linear regression on the bubbles’ spatial and temporal locations and accuracy at categorizing or identifying faces. Diagnostic identity-specific information was located in the eye region, whereas race-specific information was located on the left nostril and the whiter part of the eyes (Zcrit=3.98, p<0.05). We then constructed dynamic CIs separately for SR and OR faces showing the time course of information utilization for identification and categorization. We correlated each frame of the dynamic CIs with the identity- or race-specific CIs. The results show that for SR faces, identity-specific information is processed earlier and more thoroughly, whereas for OR faces, it is the race-specific information that is treated as such.
Meeting abstract presented at VSS 2014