Arrest records were searched for 274 subjects given an opportunity to remain in the program 18 months or longer; these constituted about 53 percent of the total number of such patients entered in the program between December 1972 and June 1974. The number of arrests were tallied for each of the 5 years prior to admission to treatment and for each month during treatment, up to and including the month of termination. The dropout groups had a consistently higher arrest rate than the stay-in group, not only throughout the 18-month treatment span but also during the 5-year pretreatment period. The difference was largest during the year just prior to admission. Moreover, the dropout population enters the program with a higher level of arrest activity, which remains higher than that of the group which remains even though it does drop somewhat at entry. Perhaps the nondropouts entered treatment with a 'built-in' bias toward positive response and would have improved in any kind of program. Further research should determine what would work best with both groups and should attempt to identify groups with differing potential for retention. Two references and three tables are provided.