Twelve, healthy normal men aged 21-25 years received each of four treatments (triazolam placebo plus caffeine placebo, triazolam 0.50 mg plus caffeine placebo, triazolam 0.50 mg plus caffeine 4 mg/kg, triazolam 0.50 mg plus caffeine 8 mg/kg), double blind, in a Latin-Square design. Triazolam or placebo was administered at 0830 and caffeine or placebo at 1000 and 1245. On two memory tasks, administered at 1015 with an immediate recall and a delayed recall at 1230 following a 90 min nap (1030-1200), both immediate and delayed recall was impaired by triazolam. Neither caffeine dose reversed the impairments. Sleep latency and sleep efficiency were improved by triazolam and not reversed by caffeine. On a performance battery presented at 1300 most measures of performance were impaired by triazolam; in general the caffeine dose of 4 mg/kg partially reversed the effect while the dose of 8 mg/kg completely restored performance.