The clinical use of seclusion was studied prospectively on two acute treatment units of a university hospital. The incidence, frequency, duration, precipitant, and type of seclusion were documented for 59 secluded patients through 107 episodes and compared to 159 nonsecluded controls on a variety of demographic, diagnostic, and legal variables. Chronicity, legal status on admission, and race were patient characteristics correlated with the incidence of seclusion. Mental status and diagnosis were not significantly related to the incidence or the frequency of seclusion. Seclusion was used primarily to contain physical assault upon staff. There was no relationship between the precipitating event and duration of seclusion, nor any change in duration or restrictiveness of seclusion with repeated episodes. These findings differ markedly from previous reports. The clinical bias toward increased use of seclusion with committed, chronic, and black patients, as well as the independence of cause and duration of seclusion raise legal issues already being pursued in federal courts.