We have an approach from stocastic process as one of the mathematical learning models. The model of Bower and Trabaso, which is Markov absorbing process, is included in it. But the data which we got did not agree with this model. So, this model was modified. The principal point of modification was that the stage of transition probabilities was changed from one to three. In this paper the modified model has been tested by expansion into reward process. It can be concluded that the process of the recognition of nature can be understood to be hypo thesis confirmation process and the rate of learning increases discretely with the progress of training.
