In this paper, we develop three measures of association between concepts and features from three measures of category structure preference. These measures are total cue validity, feature possession score, and category utility. We compare these measures experimentally using stimuli from the Leuven Natural Concept Database (de Deyne et al., 2008). We find the measure developed from feature possession score, collocation, best captures human associations between concepts and features, followed by the measure developed from category utility, and finally the measure developed from total cue validity. We discuss how these results can be applied to open questions in categorization and similarity judgment such as domain differences in representation and modeling the effects of context.