Using an observing procedure similar to Wyckoff's (1969), two normal and two mentally retarded children were studied. The children were initially trained to earn pennies by pressing a rectangular key according to a multiple variable-interval extinction schedule of reinforcement. After discrimination training, the children were allowed to produce the discriminative stimuli by depressing a second or observing key. The retarded children were different from the normal children in that they produced more of the positive discriminative stimulus (S+) than the negative discriminative stimulus (S−). In the case of the retarded children, S+ observing started high and remained high. In the case of the normal children, S+ observing was initially substantial and then rapidly diminished.