This study developed and tested the accuracy of a model designed to predict adherence to a pelvic floor muscle exercise regimen by Taiwanese women with urinary incontinence. The sample was composed of 106 women treated for urinary incontinence at urban hospitals in central and northern Taiwan from April 2000 to March 2003. All participants had practiced prescribed pelvic floor muscle exercises for at least 6 weeks at the time they completed study measures, which included adherence to pelvic floor muscle exercise, self-efficacy for the exercise, knowledge of the exercise, attitudes toward the exercise, dyadic cohesion, perceived benefits of the exercise, and severity of urine loss. After stepwise multiple regression analysis, a path analysis was conducted, with significant paths retained as modifiers. Self-efficacy for pelvic floor muscle exercise strongly and directly affected adherence to the exercise regimen. Attitudes toward the exercise, dyadic cohesion, and perceived benefits of the exercise affected adherence when mediated by self-efficacy for pelvic floor muscle exercise. Severity of urine loss also directly affected adherence. Exercise knowledge affected neither self-efficacy nor adherence. The model fit the data and accounted for 40% of adherence variance. Findings affirm the significant role of self-efficacy in predicting adherence to pelvic floor muscle exercise. Thus, self-efficacy for exercise can be an indicator for nurses to tailor exercise-training programs for women with urinary incontinence. Nurses can use the study findings to develop interventions to increase women's adherence to the exercise.