Two experiments with 150 undergraduates studied the acquisition of style recognition in a category abstraction paradigm. Ss initially classified 18 impressionist paintings, according to artist, into categories of 3-9 examples, followed by a transfer test given immediately or after a 2-wk delay. In the transfer test, Ss attempted to classify the old paintings seen during learning, new paintings by these artists, and paintings by other artists. The prediction that experience with more examples of a style would yield greater classification accuracy for new examples of that style was confirmed. Classification accuracy for new paintings was unaffected by the delay, although significant forgetting occurred for the old paintings. A measure of conceptual structure, derived from the multidimensionally scaled space, was found to be significantly correlated with transfer performance. It is concluded that the ability to recognize painting style is affected by variable manipulations found to be effective in the abstraction of information from ill-defined categories. (32 ref)