Twenty-two patients were withdrawn from normal-dose, long-term benzodiazepine treatment and their data compared with those of two control groups. Patients and controls were assessed repeatedly on the Digit Symbol Substitution Test (DSST), Symbol Copying Test (SCT), Cancellation Task (CT), Auditory Reaction Time (RT) and Key Tapping Rate (KTR). A substantial and prolonged practice effect was found on all the tests except RT and KTR. Prior to withdrawal the patients did not show the performance decrement on the CT, RT and KTR customarily associated with the initial phases of benzodiazepine therapy. A rebound performance increment was observed on KTR during the withdrawal. Patients demonstrated impaired performance on tasks requiring the combined use of sensory and fine motor skills.