FLEXIBLE TUNING OF SIMILARITY IN EXEMPLAR-BASED CATEGORIZATION

Flexibility in exemplar-based categorization was investigated in 3 experiments. Ss categorized schematic drawings of individuals' faces while they received information about the kinship relation between these individuals and previously observed exemplars. In Experiments 1 and 2, initial training was incidental, whereas an intentional learning task was used in Experiment 3. In all experiments, the kinship manipulation systematically affected categorizations. Model-based analyses showed that the effect of the kinship manipulation could be described by similarity-based exemplar models if it was assumed that Ss applied different levels of generalization as a function of the kinship instructions. These results demonstrate that categorization is a highly flexible process, and their methodological and theoretical implications are discussed.