A brief historical review of penal philosophy and of the evolution of halfway houses in the United States precede a literature review of halfway house studies. The population examined for the study were those male clients discharged in 1977 from three adult halfway houses located in St. Louis. The primary goal of these houses was the successful release of clients reflected through either permanent employment or vocational training supplemented by a grant for income. The study developed a prediction equation for each halfway house which would specify those variables associated with success or failure at each facility. Of the 74 total variables used, 21 were discrete and 53 were continuous. The significance of the variables was determined by chi square analysis or the t-test. Success was defined as leaving the halfway house for a constructive day (i.e., gainful employment) on an approved date. Nonsuccess was defined as the failure to fulfill program rules, reflected in a termination from running away or escaping with the ultimate result of the client's return to a closed correctional institution. The success rates of the three facilities differed; client characteristics, program structure, and staff had an effect on this rate. One interesting finding is that hard drug abusers (drug history was a key variable) fared much better at one halfway house than did those with no drug abuse or just marihuana abuse. Additionally, variables for measures of intelligence and spelling and arithmetic grade levels were significant predictors of success at two of the halfway house programs. Only the variable for legal status was significant among all three houses. Tables and about 30 references are appended.