We explored the action of chronic valproic acid (VPA) on the human epileptic cortex by means of transcranial magnetic stimulation (TMS). TMS is an emerging biomarker for neurotropic drugs. We had 15 drug-naive patients with different epileptic syndromes. Interictally, we measured several TMS indexes of cortical excitability before commencing VPA and 3 months later. At that time, all patients were clinical responders to the drug, whose plasma levels were in the "therapeutic range". We then compared the two conditions, while 18 healthy subjects, of whom 12 were retested at a similar delay, acted as controls. In the pooled patients, the baseline resting motor threshold to TMS was similar to that of controls, but it increased significantly (P < 0.05) after VPA. Intracortical facilitation, another index of cortical excitability, was abnormally enhanced at baseline but decreased significantly after VPA (P < 0.05). On splitting patients according to their diagnosis, the threshold increase was significant (P < 0.05) among partial, but not generalized epilepsies. The reverse was true for changes in intracortical facilitation. TMS phenomena had no linear relation to VPA serum levels. Based on the known pharmacology of TMS effects, VPA reduced the intrinsic membrane excitability of motor cortical neurons, possibly through changes in Na+ channel activity. Then, VPA corrected a transmitter-mediated interneuronal hyper-excitability of the primary motor cortex. The former effect was best seen in partial, and the latter in generalized epilepsy patients.