Generative linguistics treats inflections as subspecifications of the node INFL. This means that, for instance, tense and mood morphemes count as heads, parallel to lexical heads, as far as the formal structure goes. It is hard to find any substantial arguments for this alleged similarity. Taking critical stances by Haspelmath as a starting point, I agree that classical tests like hyperonymy, distributional equivalence or status as a valence bearer will not apply, but contrary to Haspelmath, I hold that dependency does in fact apply, provided that the classical distinction between endocentrism and exocentrism is reintroduced and taken seriously. Tense and mood morphemes are obligatory and therefore heads, in the sense that they contract exocentric constructions.Semantically, verbal inflections are similar to lexically filled free adjuncts like sentence adverbials and dialogic particles. Both inflections and free adjuncts have scope over a part of the clausal hierarchy, over a predication or over a proposition. As parts of morphology, however, they are turned into exocentric co-heads of the predicate structure of the clause. This semantic similarity is not captured in the generative framework since free adjuncts are modifiers and inflections are heads. No substantial criteria can be found to show inflections to be endocentric heads, and the exocentric alternative has long since faded into oblivion. Thus, at the level of semantic scope, inflections and free adjuncts are similar, and inflections acquire head status through morfologisation, that is: as one of the heads of an exocentric construction.