Mentalization is a developmental achievement defined as the capacity to understand behavior in terms of mental states. This study investigated mentalization in psychoanalytic psychotherapy (PPT) and cognitive behavior therapy (CBT) through a secondary data analysis of findings from a randomized controlled trial for bulimia nervosa. It was hypothesized that mentalization would predict alliance and outcome in both treatments, whereas increase in mentalization was only expected after PPT. Furthermore, it was investigated whether change in mentalization predicted symptom change. A total of 70 participants with bulimia nervosa were randomized to PPT or CBT. Participants were assessed at 3 time points with the Eating Disorder Examination and the Adult Attachment Interview (rated for reflective functioning [RF]). Therapy sessions were rated with the Vanderbilt Therapeutic Alliance Scale. Higher intake RF significantly predicted better alliance, whereas no association was observed between RF and outcome. A significant interaction between time, therapy type, and RF found RF improving more in PPT than in CBT. There was a significant association between RF change and symptom change in the PPT group. The study suggests a relation between RF and psychotherapy process, whereas the relation between RF and outcome is more complex. Furthermore, PPT seems to enhance mentalization, which seems related to symptomatic improvement, suggesting that mentalization might serve as a specific mechanism of change in PPT.