adult males, who were estimated co be 5 yr. old and had served in a variety of learning set experiments, served alternately as demonstrator ( D ) and observer (0) for 150 problems. In design the apparatus closely approximated that described by C. L. Darby and A. J. Riopelle (Observational learning in the rhesus monkey. I. comp . physiol. Psychol.. 1959, 5 2 , 94-97), except there were four food wells instead of two on the stimulus tray. For each problem Step 1 consisted of training D to respond to the correct one of four stereometric objects to a criterion of three consecutive correct responses. 0 did not observe this training. The object to be rewarded for each problem was selected in the following manner. Four objects were placed over the food wells, and D was given three unrewarded presenc3cions. The object to be rewarded subsequently was randomly selected from those not chosen on these trials. Its position relative to the other objects was randomly determined for each problem. These procedures effectively excluded object and position preferences as reliable bases for correct responses by 0. In Step 2 0 observed D's responses for eight trials. The stimulus tray was then presented to 0 for Test Trial 1, with the placement of the objects reversed left to right relative to 0 's view during the demonstration trials. Another demonstration trial was then given, followed by observer Test Trial 2. Each S served as 0 for 75 problems. Ss were 4 hr. food deprived. For each S the 75 Test Trial 1 performances as 0 were dealt with as chree blocks of 2 5 trials. For one S %s of correct responses were 4496, 56%, and 48%; for the ocher S they were 32%, 32%, and 32%. Individually and combined their performances were significantly better than chance (@ < . 0 5 ) , wich .25 defined as chance. Very likely the method of selecting the object to be rewarded in each problem reduced the probability of 0 ' s being correct by chance on Tesc Trial 1 to less chan .25, for the rewarded object was non-preferred as well as one of four. 0s ' Tesc Trial 2 performances were about 8% better than those on Test Trial 1. Thus, these Ss learned by observing one another make correct responses in an objecr-quality discrimination situation. The superiority of the one S over