The claim made in this work is that the structure of grammatical categories is predictable to a large extent once we know the range of possible cognitive structures from which they are derived. This claim is tested here with reference to one example, viz. the structure of predicative possession. It would seem that most of the possessive constructions to be found in the languages of the world can be traced back to a small set of basic conceptual patterns. These patterns are identified in this work and the way each of them affects the word order and morphosyntax of the resulting ‘have’-construction is described within the framework of grammaticalization theory (see Heine, Claudi, and Hünnemeyer 1991, Hopper and Traugott 1993, Bybee, Perkins, and Pagliuca 1994 for references). It is argued that much of the typological diversity that characterizes ‘have’-constructions in the languages of the world can be explained with reference to the principles of grammaticalization proposed in those works.

The present work has profited greatly from the assistance and co-operation of a number of colleagues. Most of all, my gratitude is due to Ulrike Claudi who, in addition to accompanying my research with constructive criticism, also gave me access to her unpublished paper on ‘To have or not to have: on the conceptual base of predicative possession in some African languages’ (Claudi 1986), which already contains a number of the basic notions to be discussed below.