Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Robert C. Mathews


This study investigated the relation between recognition and categorization and examined learning processes associated with categorization of ill-defined concepts. Three experiments were conducted, in which recognition and categorization data were simultaneously collected (Estes, 1986b). Stimuli were bar charts and letter strings that simulated symptom patterns. Category structures were defined by independent features in Experiment 1 and correlated features in Experiments 2 and 3. Three different models of categorization, exemplar models, rule models, and dual process models, were contrasted in predicting recognition and categorization performances and the learning processes used to categorize. Across the three experiments, recognition and categorization were often affected differently by experimental variables. Stimulus types in Experiment 1 and salience in Experiment 3 had significant effects on categorization, but had no effect on recognition. The main effect of duration was significant in recognition, but not in categorization in Experiment 3. Finally, in all three experiments, block effects suggested that recognition performance decreased across blocks of practice, whereas categorization tended to increase across blocks. As noted by Metcalfe & Fisher (1986), recognition is based on exemplar memories, whereas categorization depends on abstract rules but is also influenced by exemplars in some conditions. Categorization rules were related to relative feature frequency in the learning of categories based on independent features. In categorization of categories based on correlated features, biconditional or symmetry rules were used when the correlations were salient or non-salient, respectively.