Four Florida prisons were selected for study. Data included official disciplinary reports of violence and inmates' perceptions of violence and safety. A violent environment questionnaire was developed and administered to inmates and key members of the correctional staff to elicit their perceptions of the degree of violence, privacy, and safety within the prison. The study hypothesized that assaults would occur more frequently in areas of poor surveillance, such as stairwells, bathrooms, and architectural 'blind spots.' It was further predicted that sexual assaults would be more frequent in open-dorm housing. Results indicated that at all four prisons, the prime site for assaults, particularly armed assaults, was the housing area, whether dorms, six-man cells, two-man cells, or single cells. Support spaces such as showers, baths, and dayrooms were the sites of the next largest number of assaults. A surprising finding was the frequency of assaults that occurred under direct staff supervision. It was as if the inmates' actions were a dare for official action. There appears to be no ideal type of prison design that solves the problem of violence, but reducing the 'blind spots,' such as deadend corridors and stairwells, and increasing good sight lines will help in the more efficient supervision and control of those spaces. Tabular data and 15 bibliographic entries are provided.