1991 Volume 38 Issue 7 Pages 581-587
Relationships between 51 abstract terms and 54 concrete terms for oder-description were evaluated by three types of panel. On the assumption that the abstract terms were explanatory factors of the concrete terms, the response pattern matrix was analyzed by non-parametric multi-dimensional scaling (quantification theory III) and cluster analysis was done on the scores from the first to the fifth axis obtained by quatification theory III. Concrete terms were separated into 23 groups as a whole, but terms that coincided with classification among three types of panels and the sum of the three data sets, were only 17, less than the result obtained by the analysis of similarity matrix. It was considered as a reason that the terms were classified more distinctly and more closely, as compared with the former analysis. The resolution 1 and 2 of sample score (concrete terms) and category weight (abstract terms) obtained by quantification theory III, were profiled two-dimensionally, respectively. When they were overlapped together, the abstract terms explained well the concrete terms to which they corresponded at the positions.