Treatment in bulimia nervosa is challenging, with rates of successful treatments for only about 50% of all patients. This study aimed to identify predictors of outcome through secondary analysis of data from a randomized clinical trial that compared inpatient and day hospital treatment for bulimia. Process measures included assessments of patients' in-session experiences, therapeutic alliance, and therapy-related intersession experiences (ISE). ISE measures were better predictors of outcome than pretreatment variables (e.g., social adjustment) or global therapeutic alliance. Outcome at 3 month follow-up was strongly related to the ISE dimension Recreating Therapeutic Dialogue with Negative Emotions, indicating a heightened risk of failure. Prediction of outcome by these variables showed a sensitivity of 0.86 and a specificity of 0.78, and 83% of patients could be correctly classified. These results show that certain aspects of ISE may serve as early and reliable indicators of long term treatment failure, prompting alternate treatment approaches and opening new directions of research.