Dynamics of shape perception
We have been studying various aspects of shape perception with an emphasis on the dynamics of underlying computations. Recently, we investigated how the visual system determines the shape of moving objects. Results show that performance does not deteriorate when patterns are in motion: angles (moving triangles defined by three dots) can be discriminated equally well whether they are static or moving up to moderate speeds. In another study, we measured sensitivity to changes in the shape of closed contours. Surprisingly, detection thresholds for such deformations are independent of the specific shape. However, covering differing portions of the patterns revealed different computational strategies: local computation versus global pooling is shape dependent. The talk will conclude by addressing the question of how studies on simple stimuli (such as the closed contours above) are important in the study of more naturalistic patterns. It will be shown that they provide a powerful description for head shapes and (together with facial features) can be used to describe entire faces. We used such synthetic faces in a masking paradigm to explore the temporal dynamics of face discrimination. Masking by faces greatly disrupted face discrimination within a 140 ms window following stimulus presentation. Strikingly, both noise and houses had no masking effect for any delay.