One hundred thirty-three incontinent women in seven nursing homes were assigned randomly to a 13-week behavior therapy program for urinary incontinence or to a control group that received usual incontinence-related care. The therapy became effective after 6 weeks of training. By the final month of training, the treatment women's wet episodes had been reduced by 0.6 episodes per day, a 26% reduction over baseline. This reduction in the number of wet episodes was statistically significant, both with respect to this group's baseline levels of incontinence and in comparison with the performance of the control women. The number of wet episodes in the control group remained about the same throughout training and the 22-week follow-up period. The treatment women improved partly because they learned to request help, a response prompted and reinforced by the program. Trainees with a high frequency of incontinence during baseline, the more cognitively intact residents, and residents with normal bladder capacity responded better to this behavior therapy program.