Search

The purposes of this study were twofold: (1) to test the hypothesis that a Similarity Rating Model based on a Classification Scheme of relationships between six words in physics is representative of human subjects' judgments of similarity between the six words, and (2) to test the hypothesis that the Classification Scheme could be used in a Word Association Model to generate responses in a word association test like those produced by human subjects. The Clsssification Scheme was in the form of a three-dimensional space with the six chosen physics words plotted as points in space. Pairwise similarity in the Classification Scheme was defined as distance between words. In this research stoachastic models of human response behavior were designed based on the Classification Scheme, in the form of FORTRAN computer programs. Ten independent sets of data were generated under each model. Strong and consistent correlations were found between data from the Similarity Rating Model and human data. These findings were interpreted as evidence that the Rating Model was a valid representation of human subjects' judgments of similarity, and the Classification Scheme could thus be used to generate these responses. Moderate to low correlations were found between model data and human data in word association tests, thus interpreted as neither a valid nor reliable representation of human word association behavior. (Author/EB)