Abstract
Normal observers process individual faces holistically, meaning that facial features are processed simultaneously over a face-wide perceptual span and that the face is represented as a single perceptual unit. Acquired prosopagnosia, an impairment of face recognition following brain damage is thought to be caused by a problem with holistic processing of faces (Sergent & Villlemure, 1989). However, previous studies have only demonstrated an impairment of interactive processing of features in patients who, in addition, present general visual integrative agnosia (e.g., Boutsen & Humphreys, 2002; Levine & Calvanio, 1989), so that direct evidence for a deficit in holistic face processing in prosopagnosia is lacking. In an experiment with a brain damaged case of prosopagnosia (PS, Rossion et al., 2003), we used two gaze-contingent techniques allowing manipulation of the amount of information that was simultaneously available. First, a gaze-contingent foveal mask, prevented the use of foveal, high resolution information, necessary for detailed investigation of the facial features, but allowed holistic processing based on lower resolution peripheral information. Second, a gaze-contingent foveal window covering all peripheral information, prevented the simultaneous use of several facial features, but allowed detailed investigation of each feature individually. In a face matching task, normal control participants showed increased errors and RTs with a foveal window in comparison to conditions with a foveal mask or with a full face. PS, on the contrary, was almost unimpaired by a foveal window relative to a full face, while she had major difficulties recognizing faces with a foveal mask. Moreover, the eye movement data confirmed the findings from previous studies, that for face recognition, PS mainly relies on the mouth region, while normal observers attend to the region just below the eyes. These data provide direct evidence for impairment of holistic face processing in acquired prosopagnosia.