Minneapolis was the research site where the experiment began on March 17, 1981, and extended for 18 months, yielding 330 cases. Officers randomly applied one of the three treatments, and the involved households were then monitored for at least 6 months to collect data on a variety of outcome measures. A failure was indicated if police records (offense or arrest reports) revealed a postintervention assault. Households randomly assigned to arrest were substantially less likely to fail than those randomly assigned to being ordered from the premises. The small difference between randomly assigned removal from the premises and randomly assigned advice can be attributed to chance. In the case of failures, there was a longer time between intervention and failure with the use of arrest than with the other interventions. 7 tables, 1 figure, 21 refences.