This study investigated how participants would reject an initial rule when they faced positive and negative instances of an initial rule. Using eye movement data, we analyzed a perspective that indicated the type of rules that participants consider. Our experiments yielded the following results. A tendency to consider rules from the perspective that participants used for finding and confirming the initial rule was retained in the phase in which both positive and negative instances of the initial rule were given. This tendency was observed only when participants faced negative instances. We concluded that, when participants faced negative instances, they tried to change the initial rule peripherally to explain them.