This study examined interpersonal and ecological predictors of re-victimization of a sample of women with abusive partners. All women (N = 124) had sought refuge from a battered women's shelter 3 years earlier, and half the sample had been randomly assigned to receive free, short-term advocacy services immediately upon exit from the shelter. Results 2 years post-intervention revealed positive change in the lives of participants (C. M. Sullivan & D. Bybee, 1999), including a decrease in abuse for women who had worked with advocates. The current study examined intervention effects 3 years after the program ended, as well as other predictors of re-abuse. Nineteen percent of the original sample had experienced domestic violence between 2 and 3 years after shelter exit (65% by current partners, 35% by ex-partners). The advocacy program's effect on risk of re-victimization did not continue 3 years post-intervention. However, having worked with an advocate 3 years prior continued to have a positive impact on women's quality of life and level of social support. The risk of being abused 3 years post-shelter stay was exacerbated by a number of factors present 1 year prior, including women's (1) having experienced abuse in the 6 months before that point; (2) having difficulties accessing resources; (3) having problems with the state welfare system; and (4) having people in their social networks who made their lives difficult. Women were at less risk of abuse if, 1 year earlier, they (1) were employed; (2) reported higher quality of life; and (3) had people in their networks who provided practical help and/or were available to talk about personal matters. These findings support the hypothesis that access to resources and social support serve as protective factors against continued abuse.