Abstract
Research has suggested that object recognition does not follow a linear processing sequence of object detection, object categorization, and within-category identification. Rather, an object's category is retrieved as soon as it is detected (Grill-Spector & Kanwisher, 2005, Psychological Science). Here we examined whether face recognition is processed likewise, when face view is regarded as a category. We measured behavioural performance on three tasks: face detection, face-view categorization, and within-view face identification, by using the method of constant stimuli combined with a two-alternative forced-choice (2AFC) match-to-sample paradigm. The stimuli were synthetic faces (size: 3.25° × 4.49°) with 5 views for each face identity (front, 20° left, 20° right, 20° up, 20° down). The observer's task was to identify the previously-flashed stimulus between two alternatives, following brief presentation of a stimulus (duration between 13 and 133 ms) and a mask. The two alternatives were: (1) a face and a non-face (detection task), (2) two different face views of the same individual (categorization task), and (3) a face and its anti-face with the same view (identification task). Detection threshold as a function of presentation time was the point of 75%-correct performance on the psychometric function. The results showed a significantly shorter threshold duration for face detection than for face-view categorization, and only a slightly (non-significantly) shorter threshold duration for face-view categorization than for face identification. Further analysis suggested a significant difference between the psychometric functions for categorization and identification. We demonstrated that the face-view category is retrieved after face detection, and importantly, this view-based categorical analysis takes almost as long as the face identification process. We therefore concluded that the processing sequence of view-based face recognition does not follow Grill-Spector and Kanwisher's proposal on object recognition. Additional processing is essential for face-view categorization as opposed to face detection.
This work was supported by CIHR Grant #172103 and NSERC Grant #OP227224 to H.R.W.