Is the short-latency dopamine response too short to signal reward error?