Content area
Abstract
Using adaptation as a tool, this study investigated the nature of the mechanisms underlying the face representations as they develop through learning. The first part of this study examined whether experience with multiple views enhances the adaptation effect and its transfer across viewpoint change. Learning from multiple views did not produce an advantage in increasing the strength of adaptation effects over learning from a single view. These results replicate previous findings and control factors that may have contributed to previous failure in finding a multiple-view advantage.
The second part of this study explored the extent to which 2D interpolation and 3D model building are the mechanisms underlying the development of face representations through multiple-view learning. Specifically, I made use of the fact that both viewpoint and illumination change reference the 3D structure of faces. I accessed the effects of multipleview experience on the adaptation transfer across illumination change. Compared to the adaptation transfer measured at the learned view, comparable adaptation transfer across illumination change was found at a novel view centered between two learned views. These results indicate that the contribution from 3D information acquired through multiple-view learning could compensate for the cost of forming an intermediate representation. Combined, these results suggest that both 2D and 3D based mechanisms may contribute to the progression from a view-specific to a more view-transferable face representation, as faces become more familiar. The results support the use of adaptation as a tool for probing the nature of face representations.
