This paper discusses the problem of degrammaticalization, that is, the exceptions to the unidirectionality of grammaticalization. After analyzing the criteria that allow us to distinguish between various instances of counter-directional change, two principles underlying degrammaticalization are identified; one is related to the type of language and the other to the type of target structures in which degrammaticalization occurs. Firstly, the targets of degrammaticalization are usually closed-class parts of speech with an abstract semantic component. Secondly, the languages in which counter-directional grammatical changes occur turn out to be deprived of an elaborate fusional morphology. These findings may also have an impact on the theoretical conception of grammaticalization, some of whose definitional properties are discussed. The paper ends with a discussion of a more controversial point, namely, counter-directional changes by folk etymology rather than by etymology proper. Abstract: This paper discusses the problem of degrammaticalization, that is, the exceptions to the unidirectionality of grammaticalization. After analyzing the criteria that allow us to distinguish between various instances of counterdirectional change, two principles underlying degrammaticalization are identified; one is related to the type of language and the other to the type of target structures in which degrammaticalization occurs. Firstly, the targets of degrammaticalization are usually closed-class parts of speech with an abstract semantic component. Secondly, the languages in which counter-directional grammatical changes occur turn out to be deprived of an elaborate fusional morphology. These findings may also have an impact on the theoretical conception of grammaticalization, some of whose definitional properties are discussed. The paper ends with a discussion of a more controversial point, namely, counter-directional changes by folk etymology rather than by etymology proper.