Explaining Categorization Response Times with Varying Abstraction

We use the Exemplar-Based Random-Walk model (EBRW) to extend the Varying Abstraction Model (VAM). Unlike the VAM which is designed to account for categorization proportions, this Varying Abstraction-Based Random-Walk (VABRW) model is able to predict categorization response times. The extension is especially useful in situations where response accuracies are not very informative for distinguishing between models. Application of the VABRW to data previously gathered by Nosofsky and Palmeri (1997) provides additional evidence for the view that people use partial abstraction in category representations.