Representations of shape in memory.

Abstract

The question of how we recognize shapes when they are misoriented with respect to their typical orientation has not yet been adequately answered. Two kinds of theories have been proposed to address this question: One theory contends that initial access to the correct stored representation is orientation-dependent and the other theory contends that initial access is orientation-independent. Current evidence, based on naming latencies, does not distinguish between these two theories. In the present experiments, Cooper and Shepard's (1973/1982) cuing paradigm was adapted to distinguish between the orientation-dependent account and the orientation-independent account of shape recognition. A new measure of shape recognition involving figure-ground decisions was employed because naming responses may be too far removed from perceptual processing. The results of three experiments supported the orientation-dependent account of shape recognition suggesting that the representations of shape in memory specify the canonical orientation of the shapes they represent. A process of normalization is required prior to access to canonical orientation representations; however, the evidence suggests that this normalization process may not be mental rotation. Other conceptions of the normalization process, including the establishment of multiple representations, are discussed.