Perceptual Categorization and Memory

Vanderbilt University, Nashville TN

Investigators

Abstract

Abstract Palmieri, Thomas BCS-9910756 Perceptual Categorization and Memory Any time we decide that some visually presented object is a terrier rather than a collie, a bottle rather than a jar, or a tree rather than a shrub, we are making categorization decisions by comparing the perceptual attributes of an object with information about categories that have been acquired previously through experience. Developing psychological theories of perceptual categorization requires an understanding of what information is provided by the perceptual system, how that information is compared with category information that has been previously acquired, what kinds of representations are stored in memory, how category representations change with experience, and how classification decisions are made on the basis of evidence for various categories. Perceptual categorization forms a fundamental interface between basic perceptual processes and higher-level cognition. By comparing the relative abilities of various formal models to account for qualitative and quantitative aspects of observed data, this theoretical work will test well-specified hypotheses about the fundamental mechanisms of perceptual categorization. The present work focuses on the role of specific remembered category instances (referred to exemplars) in perceptual categorization as formalized by a proposed exemplar-based diffusion model (EBDM). This model combines elements of Nosofsky's generalized context model of categorization, Nosofsky and Palmeri's exemplar-based random walk model of categorization and automaticity, Logan's instance theory of automaticity, and Ratcliff's diffusion model under a single theoretical framework. According to the proposed model, categories are represented in terms of stored exemplars, evidence for a particular category is a function of the relative summed similarity of a presented item to stored exemplars, and category responses are determined by a continuous-time diffusion process driven by retrieved exemplar information. Preliminary work shows the model able to qualitatively and quantitatively account for both categorization accuracy and categorization response times under a variety of conditions with relatively few free parameters. Several empirical studies are planned to contrast the predictions of the EBDM with other competing frameworks centered around category representations based on prototypes, rules, and decision boundaries. New theoretical advancements are also outlined that specify how perceptual information might evolve overtime within a particular categorization episode.