中華心理學刊
民96 ， 49 卷， 3 期， 285-300

Lee-Xieng Yang（Institute of Cognitive Science, National Cheng-Kung University）

Abstract

In investigating human mental processes and mental representations, a cognitive model represents a theoretical view, provides explanations to the observed phenomena and makes predictions about an unknown future. When evaluating how well a theory can account for the phenomenon of interest, modeling is a powerful research tool. However, local (Taiwanese) psychology students have limited exposure to what cognitive modelling is, how to do implement cognitive models, and why cognitive modelling is important. This is partly due to a lack of university courses that teach cognitive modelling and partly due to the demands that modelling places on one’s skills. The purpose of this article is to provide a conceptual guideline of how to do modeling, by fitting two neural network models - ALCOVE and ATRIUM to the data from the study of Yang and Lewandowsky (2003), which tested the theoretical concept of knowledge partitioning in categorization. The modeling results show
that ATRIUM outperforms ALCOVE in accounting for the knowledge partitioning results. Some relevant theoretical-level discussions, such as the heterogeneity of categorization, are also included. 

Keywords: cognitive modeling, cateogrization, neural network