The ability of old (24 months) and young (3 months) male rats to reverse a previously acquired discrimination was compared in 5 experiments. The old rats did not need more trials to learn a position habit in a T-maze to obtain water reward, but required more trials to reverse the position habit. The old rats showed a similar deficit in a second, but not in subsequent reversals of the position habit. In a second experiment, old rats were slower in learning to operate one of two levers in an operant chamber to obtain food reward on a CRF schedule, but by the session prior to reaching criterion for acquisition they showed response rates similar to the young animals. When the rats were required to operate the alternative lever to obtain reward, the young rats emitted 70% of their responses during the first reversal session on the newly-correct lever, but the old rats only 35%. Nevertheless, the groups were similar in the number of sessions required to reach a criterion of 95% of responses on the correct lever. In 3 subsequent reversals, old and young rats did not differ nor were there differences in the number of responses in 4 extinction sessions in the rats which had received reversal training. In experiment 3 with old and young rats which had received only acquisition training, old rats emitted fewer responses than young animals during extinction. From these experiments it was hypothesized that the apparent difficulty of old rats in learning a reversal task was due to the low probability of their emitting spontaneously a novel or previously unrewarded response, and not to a difficulty in forming a new association. This hypothesis was tested in two further experiments in which rats were required to learn a brightness discrimination in a T-maze. Old and young rats which had learned and reversed position habits in the T-maze in experiment 1, did not differ in either acquisition or reversal of the brightness discrimination, suggesting that old rats do not differ from young animals in reversal tasks when the motor response requirements for the task are already within the animals' behavioural repertoire. Consistent with this hypothesis, naive old rats were slower than young rats in acquiring a similar brightness discrimination but did not differ in the reversal task.