Abstract
Previous reports have shown (a) viewpoint-dependent time costs in object recognition and (b) that shape similarity affects these viewpoint costs. However, it remains unclear which aspects of shape similarity interact with viewpoint effects (e.g., to what extent similarity is computed from solely 2D image-based shape properties and/or 3D geometric structure). In this study we examined the relative contributions of these factors to viewpoint related time costs in order to elucidate the nature of the shape representations mediating recognition. Using a series of sequential matching experiments (same shape/different shape) we measured recognition performance for 3D novel objects or their silhouettes at the same or different viewpoints. The results showed a significant interaction between viewpoint and similarity. When participants matched objects at the same viewpoint, responses for different objects sharing the same 3D configuration (but not local features) were slower than those for different objects sharing local shape features alone. The same pattern of results was also found with silhouettes suggesting that this time cost may be attributed to the similarity of the 2D global outline rather than to internal 3D configuration per se. Further, when participants compared different rotations of objects a viewpoint-dependent time cost was found. This cost was greater for different objects sharing only local shape attributes than for different objects sharing only 3D configuration. These findings are consistent with previous proposals that recognition is mediated by 2D image-based representations. However, in addition, they further suggest that access to these representations involves the parallel operation of two perceptual mechanisms: a rapid analysis of global shape outline at a coarse spatial scale, and a relatively slower analysis of internal local shape features at a fine spatial scale. Thus, shape similarity interacts with viewpoint over different time courses depending on the spatial scale at which similarity is computed.