Abstract
People typically rely on middle spatial frequencies (SF) for face recognition, and stimuli containing only low SFs can be difficult, or even impossible, to recognize (Gold et al., 1999). One explanation for this effect is that the horizontal information around the eyes/eyebrows that people rely on most for face recognition (Dakin & Watt, 2009; Sekuler et al., 2004) may not be the most informative for discrimination for low SF faces. Here we ask whether the processing of low SF filtered faces can be influenced by altering the context in which they are presented, or whether the stimulus drives processing strategy through bottom-up information. We measured N170s for low SF filtered faces presented in a 10AFC identification task. Participants were randomly assigned to one of two context conditions: face or texture. In the face condition, trials intermixed unfiltered faces with low SF filtered faces. In the texture condition, trials intermixed textures with low SF filtered faces. In both conditions, observers completed 200 trials of each stimulus types, for a total of 400 trials. Observers’ behavioural performance was similar for unfiltered faces and textures, and, as expected, unfiltered faces led to large N170s, while textures did not. For low SF filtered faces, performance was significantly reduced compared to that of both unfiltered faces and textures, but it did not vary significantly across conditions. In contrast, the EEG results for low SF filtered faces varied considerably across conditions: participants in the face condition showed strong N170, whereas those in the texture condition showed no significant N170 even though the stimuli were identical in the two conditions. Hence, the N170, but not response accuracy, was sensitive to stimulus context. These results suggest that subjects used different processes in the two conditions, even though performance was the same.
Meeting abstract presented at VSS 2012