View article

Many current models of categorization assume full knowledge of the properties of the stimulus to be categorized. To remedy this situation, it is first necessary to understand how humans categorize stimuli with missing information. To that end, two visual category learning experiments were conducted using an inverse base-rate effect paradigm. In the second experiment, transfer trials included stimuli in which a category-diagnostic present/absent feature was occluded. Response proportions showed that people tend to treat occluded features as being absent from the stimulus, suggesting a more general tendency to assign default values to features of unknown status at the time of categorization. This pattern of results could not be replicated by several computational models – EXIT, SUSTAIN, or EXALT, a modification of EXIT implementing additive similarity.