Nightmare frequency and self-rated distress were assessed retrospectively in two groups of chronic nightmare sufferers 30 months after treatment. In the initial phase, the image group (N = 9) learned a cognitive-behavioral technique (imagery rehearsal) for the treatment of nightmares. They were taught in one group session to: (1) record a nightmare; (2) change it (usually to something positive); and (3) rehearse the new images daily. The record group (N = 10) recorded nightmares during the first month only and learned imagery rehearsal subsequent to 3-month follow-up measurements. At 3 months and at 30 months, both groups had significantly fewer nightmares, but only the rehearsal group had less total distress. The results support the theory that nightmares are a primary sleep disorder rather than a symptom of an underlying psychiatric problem.