We examined the possible mediating role of physical self-perceptions, physical self-esteem, and global self-esteem in the relationships between exercise and depression in a group of socioeconomically disadvantaged women with elevated symptoms of depression. Forty-four female residents of a low-income housing complex were randomized into a 7-week-long exercise-training group or a wait-list group. Depression, physical self-perceptions and self-esteem were measured repeatedly. Significant changes were found for depression, self-esteem, physical self-worth, and self-perceived physical condition in the exercise-training group. Intent-to-treat analyses did not alter the results. Most of the reduction in depression occurred between Week 2 and Week 4 while initial improvement in physical self-worth and self-perceived physical condition was observed between baseline and Week 2. These variables can be seen as plausible mechanisms for effects of exercise on depression.