Abstract
In categorical perception effects (CP), observes are more sensitive to differences that cross category boundaries than to differences that do not (Harnad, 1987). CP has been found for a variety of stimulus categories, including colors (Bornstein, 1987), faces (Levin & Beale, 2000), and newly learned visual stimuli (Goldstone, 1994; Livingston, Andrews & Harnad, 1998). Most research on CP, however, has asked participants to discriminate between stimuli presented sequentially, requiring participants to hold one visual item in memory until the second item is presented. In 3 experiments, we investigated whether the enhanced CP ability to discriminate between visual stimuli relies on memory. In all 3 experiments, participants were first trained to categorize visual stimuli (thin and fat fish). After training to criterion, participants were asked to discriminate between fish that either belonged to the same category or belonged to different categories using a same/different (d′) task. To minimize reliance on memory, participants in experiment 1 discriminated between stimuli presented simultaneously. No CP effects were found. In experiment 2, we manipulated salience of category membership by asking half of the participants to report category labels (thin vs. fat) while making discriminations. We also manipulated the presentation of the stimuli, simultaneous vs. sequential presentation. We found CP only when 1) participants were required to report category labels and 2) stimuli were presented sequentially, suggesting that memory for the category label is crucial in CP. Experiment 3 replicated the finding that CP is found when both of the above factors are present. We present a memory-based theory of CP discrimination that is consistent with these findings and can account for most other CP effects found in the literature.