Background: Few studies have examined long-term exercise adherence in older women. The purpose of this study was to assess predictors of adherence to an intervention involving walking and balance exercises. Methods: This was a randomized controlled trial with 2-year follow-up. Sedentary women (n=137) aged ≥70 randomized to the exercise intervention were evaluated in their homes. The exercise prescription included walking 30 minutes per day 5 days per week and completing 11 balance exercises twice per week. The main outcome measure was exercise adherence of the intervention group only. Results: The average number of minutes walked per week was 95.2 (SD 68.8); 17% walked the recommended 150 minutes or greater. The average number of times the balance exercises were done was 1.5 (SD 1.6) per week. Results of regression analysis for walking adherence showed clinical variables accounted for the greatest variance (17%) of all the blocks, and cognitive variables were second highest (12%). The final model explained 19% of the variance in predicting adherence to walking. Results of regression analysis for adherence to balance exercises showed health-related quality of life (HRQOL) variables accounted for the greatest variance (14%), followed by cognitive variables (12%). The final model explained 24% of the variance in predicting adherence to balance exercises. Conclusions: Adherence to exercise was below recommended goals, although this study demonstrated that sedentary women can adopt and continue regular exercise long term. Predictors of adherence varied with different forms of exercise. Individually tailored exercise interventions may be most amenable to older women.