An experimental study of the effect of differential non-reinforcement of the incorrect response on the learning of the correct response in simple T-maze learning