The purposes of this study were: (a) to determine the optimal categorization of an instrument measuring women's exercise perseverance and barriers using the Rasch analysis and (b) to examine urban women's exercise perseverance and barriers. A 23-item barrier instrument with five response categories was administered to 479 women from a metropolitan area. The data analysis started from collapsing the original five adjacent categories into three and four categories and 11 sets of original and collapsed data were analyzed using the Rasch rating scale model. The model-data fit, category and separation statistics, and parameter estimates provided by Rasch analyses were used to determine the optimal categorization of the instrument. Instead of the original five-category construct, which had a disordered internal construct, a collapsed three-category construct (i.e., Very Often/Often, Sometimes/Rarely, and Never) was found to have the optimal categorization. The time barrier domain was found to be the most severe barrier domain, but the barrier "lack of self-discipline" was the most severe individual barrier. Rasch calibration provides a new way to construct an instrument with optimal categorization, to describe the nature of barrier items and the respondents' attribute being measured, and to develop a practical and informative scoring sheet.