Maier and Ellen's rats were required to learn to respond to a particular card irrespective of its position (side) after position habits were well developed. The proportion of times that the previously rewarded response was rewarded and punished, was Varied : Group A, 80 : 20 ; Group B, 50 : 50 ; and Group C, 20 : 80. Main results were : (a) in terms of the number of rats giving up the old response and the number learning the new, Group A and C were similarly superior to Group B, (b) the number of trials required by the learners to abandon the old response was the largest in Group A, and Groups B and C followed, and (c) the learners in Group B need the least number of trials to criterion and the other groups were equally inferior in discrimination. All the points except the second, according to the authors, contradicted reinforcement theory, and a perceptual factor of consistency of effect was basic for explaining the results. Similar findings were obtained by Saji et al. and Akamatsu et al.Experiment I was conducted to repeat the Maier-Ellen design using children in a slightly different situation in that two boxes with different colors were' used with a candy as reward. The reward-punishment ratios in color discrimination were the same as Maier et al. The results showed the homogeneity of the three groups in position habits (Tables 1 through 4). Only one S in each of Groups B and C failed to learn discrimination in 90 trials although they abandoned position habits early in trials. This finding may be due to the relative easiness of the discrimination compared with Maier et al.'s.The median number of trials to criterion was the least for Group B and no slgnificant difference appeared between the other two (Tables 5 & 6), while Group A showed the largest number of trials to give up the old response, and Groups B and C followed in this order (Table 7). The two points agreed with Maier et al. New measures were observed in fhe present study. In terms of the percentage of response alternations, Group B was the highest and Groups A and C were similarly low (Table 8). It was indicLted that although Group C was earliest in giving up the old response, they were yet as slow as Group A in abandoning‘position habit’Per se, i. e. whether the persisted response was previously rewarded or not.In terms of the percentage of wrong responses, Group A and C were equally low relative to Group A. In this point, Muenzinger's finding (6) that discrimination was facilitated by punishment, was suggestive.Contrary to Maier's contention, the results may be explained in terms of reinforcement theory in that Group B was most favored in discrimiaated because only in this group the particular (red) box was rewarded irrespective of position (i. e. 50 : 50), while in the other two groups, the same box was reinforced not only with respect to color but also (partially) to position (80 : 20 or 20 : 80). This explanation was confirmed in Experiment II where no previous position learning was included (Table 10). The same results may also be interpreted in terms of‘set’in that Group A and C were similarly slow in abandoning their set of position habit as revealed by the analysis in terms of the percentage of response alternations and thus they were inferior in the new discrimination to Group B. The two interpretations were of course complementary. The weakness of Maier's position seemed to lie in that although it could explain the similarity of Groups A and C but failed to show why they were both inferior to Group B.