Double responding: A new constraint for models of speeded decision making