Data from 59 women during a 4-month period showed that length of stay was not related to a woman's age, race, education, the length of the abusive relationship, the number or ages of her children, or her current pregnancy status. Length of stay was related to the nature of the relationship with the partner and with the destination on leaving the shelter. Women abused by their husbands, those with unknown destinations, and those who returned to their partners were more likely than others to leave in the first week. A quasi-experimental design was used to assess the impacts of the shelter stay. Twenty women who had been randomly and secretly assigned to pretest and posttest groups were interviewed during their first or third weeks at the shelters. They answered more than 400 structured questions focusing on social and psychological functioning. The degree of exposure to the shelter was found to be positively associated with several measures of recovery, particularly psychological independence from the abuser. Tables, notes, and 10 references.