Abstract
In the analysis of free-answer responses to a social survey questionnaire, some words similar in their dictionary definitions may sometimes need to be distinguished as responses of distinct (sub)cultural groups. A psychometric model called the bicultural generalized high threshold (BIGHT) model is presented to give a criterion to distinguish such words, based on a pattern analysis of responses to some multiple choice questions. This model includes parameters to characterize the respondents by their degrees of conformity to their cultural groups. The degrees are used in a maximum likelihood method to investigate the distinction of two groups corresponding to two similar words used as free-answer responses. This investigation can lead to the distinction of the words as responses of distinct groups. Three illustrative examples on the study of Japanese national characters are also given.
