A paradigm designed to examine the degree to which verbal reports represent the subject's learning was evaluated using a concept-acquisition task. Specifically, the degree to which hypotheses verbalized during the concept-acquisition procedure represent the covert hypotheses controlling category responding was examined. Subjects were required to make category responses, verbal reports, and sorts and re-sorts (blank-trial probes) of the stimuli at intervals in a reception paradigm. By comparing the sorts made during the acquisition procedure, the re-sorts after a 1-week delay, and the sorts made by subjects who had not participated in the acquisition procedure, it was determined that the verbal reports do not accurately represent the underlying hypothesis-testing process under all conditions. Verbal reports elicited at the beginning of the acquisition procedure and at criterion accurately represented the underlying hypotheses; those elicited at intervening points did not. The ambiguity of the verbal reports was a similar function of acquisition trials.