To assess the long-term course of recovery and relapse and predictors of outcome in anorexia nervosa. Method: A naturalistic, longitudinal prospective design was used to assess recovery and relapse in patients ascertained through a university-based specialty treatment program. Patients were assessed semiannually for 5 years and annually thereafter over 10–15 years from the time of their index admission. Recovery was defined in terms of varying levels of symptom remission maintained for no fewer than 8 consecutive weeks. Results: Nearly 30% of patients had relapses following hospital discharge, prior to clinical recovery. However, most patients were weight recovered and menstruating regularly by the end of follow-up, with nearly 76% of the cohort meeting criteria for full recovery. Relapse after recovery was relatively uncommon. Of note, time to recovery was protracted, ranging from 57–79 months depending on definition of recovery. Among restrictors at intake, nearly 30% developed binge eating, occurring within 5 years of intake. A variety of predictors of chronic outcome and binge eating were identified. There were no deaths in the cohort. Conclusion: The course of anorexia nervosa is protracted. Predictors of outcome are surprisingly few, but those identified are in keeping with previous accounts. The intensive treatment received by these patients may account for the lower levels of morbidity and mortality when considered in relation to other reports in the follow-up literature. © 1997 by John Wiley & Sons, Inc. J Eat Disord 22: 339–360 1997.