The present paper deals with the semantics of locative expressions. Our approach is model-theoretic, using concepts from topology and linear algebra. We shall demonstrate that locatives consist of two layers: the first layer defines a location and the second a way of moving with respect to that location. There is to our knowledge no language in which locations form a morphologically independent class of words. Hence, the elements defining these layers, called localiser and modaliser, tend to form a unit, which is typically either an apposition or a case marker. It will be seen that this layering is not only semantically manifest but in many languages also morphologically. There are numerous languages in which the morphology is sufficiently transparent with respect to the layering. The consequences of this theory are manifold. For example, we shall show that it explains the contrast between English and Finnish concerning directionals, which is discussed in Fong. In addition, we shall be concerned with the question of orientation of locatives, as discussed in Nam. We propose that nondirectional locatives are oriented to the event, while directional locatives are oriented to certain arguments, called movers.