In Wells’ (1997) metacognitive model of obsessive–compulsive disorder (OCD), two types of metacognitive beliefs are considered central: thought-fusion beliefs and beliefs about rituals. According to the model, non-metacognitive beliefs such as responsibility and perfectionism, linked to OCD in other theories, are by-products of the perseverative thinking and behaviors (the cognitive attentional syndrome) activated by metacognitive beliefs. If this is the case, changes in metacognition should be a better independent predictor of changes in obsessive–compulsive symptoms following treatment than changes in non-metacognitive beliefs. This study aimed to test this in a sample of 108 in-patients with OCD, who completed an intensive (3 weeks) multimodal treatment package consisting of behavioral, cognitive, and metacognitive ingredients. Results indicated that obsessive–compulsive symptoms, cognitive, and metacognitive beliefs were significantly reduced during treatment, and treatment responders had larger reductions in these beliefs than non-responders. Metacognitive belief change emerged as a better independent predictor of recovery than cognitive belief change. These results add to the growing body of empirical support for the importance of metacognitions in OCD.