Learning to Discriminate Face Views

Although visual feature leaning has been well studied, we still know little about the mechanisms of perceptual learning of complex object. Here, human perceptual learning in discrimination of in-depth orientation of face view was studied using psychophysics, EEG and fMRI. We trained subjects to discriminate face orientations around a face view (i.e. 30°) over eight daily sessions, which resulted in a significant improvement in sensitivity to the face view orientation. This improved sensitivity was highly specific to the trained orientation and persisted up to six months. Different from perceptual learning of simple visual features, this orientation-specific learning effect could completely transfer across changes in face size, visual field and face identity. A complete transfer also occurred between two partial face images that were mutually exclusive but constituted a complete face. However, the transfer of the learning effect between upright and inverted faces and between a face and a paperclip object was very weak. Before and after training, we measured EEG and fMRI BOLD signals responding to both the trained and the untrained face views. Analyses of ERPs and induced gamma activity showed that face view discrimination training led to a larger reduction of N170 latency at the left occipital-temporal area and a concurrent larger decrease of induced gamma activity at the left frontal area with the trained face view, compared with the untrained ones. BOLD signal amplitude and MVPA analyses showed that, in face-selective cortical areas, training did not lead to a significant amplitude change, but induced a more reliable spatial pattern of neural activity in the left FFA. These results suggest that the visual system had learned how to compute face orientation from face configural information more accurately and that a large amount of plastic changes took place at a level of higher visual processing where size-, location-, and identity-invariant face views are represented. The learning might be implemented through neuronal tuning curve sharpening and internal noise removing.