Four skills performance (psychomotor) tasks, including simple and choice reaction time, divided attention, and vigilance, were given to 99 chronic insomniacs to determine whether the use of midazolam 15 mg, or of flurazepam 15 or 30 mg, compared with placebo would produce next-day impairment throughout a 14-day treatment period. Tests were administered during 2 baseline days and on treatment days 1 and 2 (early interval), 7 (middle interval) and 13 and 14 (late interval). Compared with placebo, performance on all four tasks was impaired by flurazepam 30 mg. The deficits associated with flurazepam 15 mg were roughly half the magnitude of those produced by flurazepam 30 mg, but these changes did not reach statistical significance for single response measures at single time intervals. Midazolam showed no consistent pattern of performance effects; however, for divided attention tracking error, there was a significant decrement. The placebo group showed flat performance curves or improved performance as a result of learning (practice effect).