The term ‘degrammaticalization’, originally coined by Lehmann in 1982 for a supposedly non-existent phenomenon, soon came to be applied to a number of often entirely different changes. Since such ‘counterexamples’ pose a potential challenge to the unidirectionality of grammaticalization, they have been the focus of much attention from grammaticalizationists and grammaticalization critics alike. While the former have attempted to dismiss them as insignificant, the latter have tended to over-emphasize their relevance. Much of the debate on degrammaticalization is rooted in different understandings of what degrammaticalization entails, or what it should entail. This paper proposes a descriptive framework which will restrict the number of potential examples of degrammaticalization, while at the same time subdividing them into three clearly distinguishable subtypes.