The use of exemplars can greatly aid concept learning. However, it is unclear how learning is affected when peopleencounter incorrect exemplars. We report a study that examines this issue, wherein subjects were shown hypotheticalexperiments and were asked to indicate whether or not each was a true experiment. One group of subjects was only showntrue experiments (i.e., correct exemplars), another was only shown non-true experiments, and a third group was shownboth. After each response, some subjects received explanatory feedback, whereas others received no feedback. Subjectswere then given a posttest involving novel hypothetical experiments (comprising true and non-true experiments) and wereasked to classify each. Subjects who were shown both types of exemplars performed best on the posttest, but only if theywere given feedback. These findings suggest that people can indeed learn from incorrect exemplars, but only if they canrecognize that the exemplars are incorrect.