Much of the recent literature on psychoanalytically oriented supervision assumes that the effective supervisor must at times address supervisee conflicts and relationship problems that interfere with role functioning. However, a vocal minority continues to assert that such a supervisory stance moves beyond teaching into “treating” the supervisee and is therefore inappropriate and potentially destructive. I consider the basis for this objection, and suggest a means of addressing it. I argue that explicit acknowledgment that the supervisor is a participant-observer with anxieties and conflicts, rather than an objective expert, creates a new context for supervision. Within this relational context, the teach–treat boundary becomes less problematic. The supervisor's role may then safely extend somewhat beyond the didactic, as long as the supervisee's learning needs remain paramount. An illustrative vignette is offered.