2 groups of 12. (one person's data was lost).

No baseline pre-treatment data.

Each group did two trials: VPA then placebo, or vice versa. The VPA-first group did well on VPA, but the VPA-second group did not. Improvement was suggestive but slight.