In a repetition of D. N. Jackson's experiment (1954) only half of 32 subjects reported previously rewarded profiles more often than punished ones. Post‐experimental interviews suggest that the effect may depend upon whether or not the subject grasps the purpose of the experiment. Another alternative to the original interpretation, based on the concept of ‘autism.’, is that rewarded profiles are reported more frequently because they are learned better than the punished profiles. A second experiment supports the assumption that the one profile of a reversible pair which is learned most completely is also the one reported most frequently in the post‐training series. The appropriateness of the experimental design for the study of figure‐ground perception is q