Abstract
When inverted, faces tend to be identified less quickly and less accurately than when upright (Yin, 1969). This inversion effect frequently is considered indicative of face-like processing, yet trained observers can develop similar effects for non-face objects (e.g., Gauthier & Tarr, 1997; Gauthier et al., 2000), a result that may be related to changes in the activation of the fusiform face area (FFA). We sought to determine the boundary conditions of learned inversion effects by training observers to discriminate amongst a relatively homogenous set of houses that differed only by their internal features. Unlike previously tested object classes, houses preferentially activate the parahippocampal place area (PPA), a region distinct from the FFA, providing us with a much stronger test of the flexibility of FFA processing. Pre-testing indicated no initial inversion effect (in terms of response accuracy). Following 10 days of training with upright houses on a 10AFC discrimination task, an inversion effect was observed: accuracy was 14% higher for upright than inverted houses. Training on inverted houses induced a reverse inversion effect: accuracy was 13% higher for inverted than upright houses. Thus, the inversion effect can be learned for houses, and appears to reflect orientation-specific perceptual learning. To determine the extent of transfer to novel stimuli, observers were trained for 5 days on 1 of 2 sets of 10 upright houses then tested on both sets. For the trained house set, practice led to improved performance and the emergence of an inversion effect. However, we found only a partial transfer of learning to the novel house set, the extent of which did not vary significantly with orientation. Future research will examine the neural correlates of training with these houses, specifically, how activation of the FFA and PPA relate to learning.
Supported by NSERC Discovery Grants 42133 & 1054994, the Canadian Research Chair Program, and an NSERC PGS-A Scholarship awarded to Husk