Retardates, primary students, and college students were given either a reversal or an intradimensional shift after either a criterion of five or twenty correct on a pre-transfer problem. An automated two-choice apparatus projected planometric color and form cues from the rear onto panels that the subject was instructed to press. Both the pre-transfer and the shift problems required S to choose one of two difficult-to-discriminate forms and ignore two easy-to-discriminate colors. Under these conditions the intradimensional shift was harder than the reversal for both retardates and primary students. The reversal was harder for the college students, but only after a pre-shift criterion of five correct. It was noted that the mechanism of attention-to-dimension or mediation was insufficient to account for these data. A supplementary mechanism was proposed. (Author)