Abstract
<p>This study develops a multidimensional unfolding model for measuring social attitudes from preference data collected through Best-Worst proximity judgments. The model locates respondents and statements in a common multidimensional attitude space and represents choices of the statements closest to and farthest from a respondent’s opinion through utility differences based on Euclidean distance. It also includes a statement-level distance-independent selectability parameter that captures a statement-specific tendency to be selected as closest more often and as farthest less often after respondent-statement distance is controlled. All parameters are estimated jointly in a Bayesian framework. Simulation results showed accurate recovery of statement configurations and the selectability parameter across the examined conditions. Recovery of respondent-statement distances was also generally strong but depended more on dimensionality and the amount of comparison information. Omitting the selectability parameter systematically displaced statement locations and degraded recovery of interstatement and respondent-statement distances. In an application to attitudes toward an optional separate-surname system for married couples in Japan, respondent coordinates on the first principal axis correlated strongly with a single seven-point Likert-type item (Spearman’s ρ = .848), whereas the second axis differentiated rationales among statements expressing similar levels of support or opposition. In a second application to consumption attitudes, the model produced an exploratory map of multiple value criteria without prespecifying a single support-opposition dimension. These findings show that the model can both refine the interpretation of a known attitude dimension and support exploratory mapping when the dimensional structure is not specified in advance, while separating semantic location from distance-independent selectability.</p>