Pigeons were successfully trained to discriminate 60 mg/kg valproic acid from saline using a two-key drug discrimination procedure. When 5-80 mg/kg doses of valproic acid were administered during generalization tests the percentage of responses directed to the valproic acid-appropriate key varied directly with dose. The effects of administering the training dose of valproic acid at presession injection intervals ranging from 15 to 120 min were described by an inverted U-shaped function; the 30-min interval used during discrimination training engendered the largest percentage of valproic acid-appropriate responses. The discriminative stimulus properties of valproic acid failed to generalize to the anticonvulsant compounds phenobarbital (10, 20 mg/kg), phenytoin (2.5, 5 mg/kg), and ethosuximide (40, 80 mg/kg), indicating that not all anticonvulsant compounds share similar discriminative properties. Clonazepam (0.25, 0.50 mg/kg) and diazepam (1, 2 mg/kg), two benzodiazepines with anticonvulsant properties, produced quite different effects. The stimulus properties of valproic acid generalized to all doses of clonazepam, whereas intermediate generalization was evident with diazepam. Pentylenetetrazol (10, 20 mg/kg), chlorpromazine (5, 10 mg/kg), tripelennamine (2.5, 5.0 mg/kg), d-amphetamine (0.5, 1.0 mg/kg), morphine (1.25, 2.50 mg/kg), and imipramine (2.5, 5.0 mg/kg) induced only saline-like patterns of responding. The concomitant administration of pentylenetetrazol failed to antagonize the discriminative stimulus properties exerted by the training dose of valproic acid.