Two experiments were conducted to examine the degree to which mentally retarded adults can be trained to improve their reaction time (RT) and movement time (MT). A discrete aiming task that had both RT and MT components was utilized in Experiment 1. Only MT improved significantly across the 15 training days, and this performance level was maintained after 5 months without practice. In Experiment 2 the extent to which RT can be trained in the absence of the spatial-temporal constraints of the aiming response was examined. Reaction time decreased significantly for both the training and training-plus-feedback groups over a 10-day training period. The findings demonstrate the role of response complexity in response-initiation processes for retarded adult workers.