The relative effects of antiepileptic drugs (AEDs) on cognition are controversial. We compared the cognitive effects of phenobarbital, phenytoin, and valproate in 59 healthy adults using a randomized, double-blind, incomplete-block, crossover design. Cognitive assessments were conducted at baseline, after 1 month on each drug (two AEDs per subject), and at two repeat baselines 11 weeks alter each AED treatment. The neuropsychological battery included 12 tests, yielding 22 variables: Choice Reaction Time, P3 Event-Related Potential, Finger Tapping, Lafayette Grooved Pegboard, Selective Reminding Test, Paragraph Memory, Complex Figures, Symbol Digit Modalities Test, Stroop Test, Visual Serial Addition Test; Hopkins Symptom Checklist, and Profile of Mood States. More than one-half of the variables exhibited AED effects when compared with nondrug baselines, and all three AEDs produced some untoward effects. Differential AED effects on cognition were present for approximately one-third of the variables. Phenobarbital produced the worst performance; there was no clinically significant difference between phenytoin and valproate.