The present study examines self-identified reasons for returning to abusive relationships. Ninety female residents of an urban domestic violence shelter completed structured interviews that included questions about their previous attempts to separate from the batterer and their perceived likelihood of returning in the future. The results indicated that participants appeared to underestimate their likelihood of returning to the relationship. A particularly intriguing pattern of findings emerged with respect to emotional attachment. Compared to those leaving for the first time, participants with a history of past separations were significantly more likely to indicate that they might return to the batterer because of their continued attachment. Emotional attachment also had a high rate of repeat endorsement, indicating that those who considered returning because of their attachment to the batterer were significantly more likely to have done so for this reason in the past. Clinical implications of the findings will be discussed.