The Bayesian evaluation of categorization models: comment on Wills and Pothos (2012)

Wills and Pothos (2012) reviewed approaches to evaluating formal models of categorization, raising a series of worthwhile issues, challenges, and goals. Unfortunately, in discussing these issues and proposing solutions, Wills and Pothos (2012) did not consider Bayesian methods in any detail. This means not only that their review excludes a major body of current work in the field, but also that it does not consider the body of work that provides the best current answers to the issues raised. In this comment, we argue that Bayesian methods can be--and, in most cases, already have been--applied to all the major model evaluation issues raised by Wills and Pothos (2012). In particular, Bayesian methods can address the challenges of avoiding overfitting, considering qualitative properties of data, reducing dependence on free parameters, and testing empirical breadth.