Surveillance in the field of violence against women holds great promise as a tool to establish and track prevalence over time, identify risk groups and factors, and evaluate interventions. Appropriate surveillance systems are population-based and can be used as the basis for informed public policy formation and evaluation as well as public attitude assessment and analysis. They can also decrease research costs and can be established in all of the systems (legal, health, and social services) that interact with victims. Yet surveillance is not a perfect system. The article examines the issues with surveillance in the field of violence against women, including assumptions, prevalence variations, sensitivity and specificity issues, and safety concerns. To find and develop the most creative approaches to maximize the promise and minimize the perils of surveillance, the article recommends that: (1) all researchers in the field of violence against women spend time yearly in shelters or other victim services, sit in on support groups and/or spend a day in domestic violence court; (2) advocates look at the data, listen to adolescent girls and women talk about their own use of violence; (3) workers in the field interview victims; and (4) all advocates help plan research studies to help with safety issues and assist service agencies in planning evaluations and data systems. Notes, references