Metacognition is an important construct for the understanding of mental retardation and the design of educational treatment for this population. However, assessment procedures used for the testing of metacognition do not seem appropriate for these subjects. This article presents an interactive assessment procedure that uses a game-like task. Results show that this format adequately reflects low levels of self-regulation in persons with mental retardation and, more important, individual differences among these individuals in a manner not possible with more static approaches. Adaptive behaviour and learning of metacognition in a cognitive enrichment programme are used as external criteria, with significant correlations between these and interactive assessment scores. Implications for educational practice and assessment of metacognition are discussed. (Contains 5 figures, 1 note and 4 tables.)