CONTEXT: Violence toward pregnant women is a gross violation of human rights with adverse health consequences for the woman and potentially for her fetus; however, few studies have examined factors associated with such abuse in developing countries or with population-based data. METHODS: A sample of 2,553 ever-pregnant women aged 15-49 from one urban and one rural site in Bangladesh were surveyed in 2001 as part of a World Health Organization multicountry study. Multilevel logistic regression analysis was used to examine factors associated with physical spousal abuse of women during pregnancy. RESULTS: Urban and rural women whose mother or mother-in-law had experienced physical spousal abuse had increased odds of experiencing abuse during pregnancy (odds ratios, 2.1-3.4); increased spousal communication was negatively associated with the outcome in both settings (0.6 and 0.7). Among urban women, being older than 19, having a husband with more than 10 years of education and being from certain higher income quartiles were negatively associated with abuse (0.2-0.5); living in a community highly concerned about crime was positively associated with abuse (1.1). Among rural women, being able to depend on natal family support in a crisis was negatively associated with abuse (0.5); being in a marriage that involved dowry demands and being Muslim were positively associated with abuse (1.8 and 3.6, respectively). CONCLUSIONS: The message that a family history of spousal violence increases a daughter's risk of such abuse should be widely communicated. Further research is needed to determine whether increased couple communication reduces the likelihood of violence or whether absence of violence leads to increased couple communication.