Subjective and objective measures of daytime sleepiness and psychomotor function were determined in normal control subjects and in epileptic patients on chronic monotherapy with phenobarbital or valproate (n = 10 in each group). All patients had primary generalized epilepsy with a normal resting EEG and were seizure-free for at least 1 year. After nocturnal polysomnographic recording, each subject was evaluated at 2 h intervals between 10:00 and 16:00 h by using multiple sleep latency tests (MSLT), a visual analogue rating scale for alertness (VARS), an anxiety scale (STAI-X1) and a battery of psychomotor tests. Nocturnal sleep parameters before daytime assessment were comparable in the 3 groups. At MSLT, patients on phenobarbital showed a shorter mean sleep latency (9.0 +/- 1.7 min) compared with the valproate group (12.5 +/- 1.3 min) and controls (12.9 +/- 1.2 min), though within-group variability was considerable. Compared with controls, patients on phenobarbital showed longer motor movement times, impaired attention (cancellation test, CT), reduced processing speed (digit-symbol substitution, DSS) and a trend towards lower critical flicker fusion threshold. Patients on valproate showed some impairment in attention and a trend towards longer motor movement time. In patients, no correlation was found between assessed parameters and serum drug concentrations, which were 19.3 +/- 1.7 micrograms/ml for phenobarbital and 85.7 +/- 4.7 micrograms/ml for valproic acid.