This paper argues that possession is syntactically category-flexible. While it is clear that in many languages possession is mostly grounded in and operates in the nominal extended projection (Szabolcsi 1983; Kayne 1993), I show that this cannot be universal. The empirical part of this article is a case study of Äiwoo, which I argue has an inherently verbal counterpart of English ’s, an abstract transitive verb I label poss. This verb can be used by itself to form clausal possession: ‘I poss this boat’ ≈ ‘this boat is mine.’ Possessed DPs also contain the verb poss: the object of this verb is extracted, forming a relative clause. Informally, ‘my boat’ really is ‘the boati"Equation missing"’ ≈ ‘the boat that is mine.’ Given this, Äiwoo simply lacks true nominal possessives. The theoretical consequence is that possession can be mapped onto different syntactic categories in different languages. This is a welcome result, as it makes the syntax-semantics mapping as flexible as it needs to be: if possession is just a tool to assert that a certain relation holds between two entities, nothing in our theory of grammar predicts that such a notion should only be limited to a specific syntactic category.