1993 年 8 巻 3 号 p. 357-366
This study proposes a category formation model based on attributive differences and demonstrates its validity by psychological experiments and computer simulations. The basic suppositions of this model are a) relativity of attributive values which leads to change in attributive weights according to their differences and b) contextual change of degrees of attention toward attributes. Results of categorization experiments supported the basic assumptions above, and newly suggested the need of flexible contextual determination processes of degrees of attention, which led to the modification of the model. Simulations, given appropreate intial values of weights and degrees of attention, replicated the experimental results, indicating the validity of the modified model. Individual simulations of all subjects also demonstrated its effectiveness, but revealed problems on selection of attributes and reconstruction of representation. Finally the need of reasonable heuristics to change categorization strategy abruptly was discussed in terms of the differences between the processes in the experiments and the simulations.