This study investigated treatment retention in an institutional therapeutic community (ITC) for dually diagnosed male inmates. Twenty-five percent of admissions successfully completed the residential substance abuse treatment program. Graduates participated for an average of 9 months, while terminations participated for nearly 4 months. One-third of the terminations participated for at least 3 months and 15% participated for 6 months or more. A striking result was the relationship between retention and risk ratings and psychopathy scores. Both measures assess behavior and interaction styles over a range of situations. The presence of a single type of negative behavior (i.e. the nature of the committing offense) did not predict completion, but a rating based on multiple samples of behavior was a significant predictor of retention. Copyright © 2004 John Wiley & Sons, Ltd.