Reward positivity: Reward prediction error or salience prediction error?