To discriminate the tokens of fraudulent articles which can not be distinguished by an optical sensor and to play with the tokens of the normal articles.

When a fed token 115 advances through a token path 102 by self-weight, the size and two-dimensional magnetic data of the token 115 are obtained on the basis of two or more pieces of token detection data obtained by the respective magnetic sensors 105a of a magnetic sensor group 105, then whether it is the normal article or the fraudulent article is discriminated on the basis of the size and the two-dimensional magnetic data, the advancing direction of the token 115 is set to a token return path 107 at the time of discriminating that it is the fraudulent article, and thus the token 115 of the fraudulent article is removed. On the other hand, at the time of discriminating that it is the normal article, the advancing direction of the token 115 is set to the token path 102, it is detected by a fed token sensor 103, and a game is played with the token 115 of the normal article.