How are nouns categorized as denoting “what” and “where”?

•Directional markers are the locus of a new form of noun categorization.•Shifts between the two categories result in systematic semantic changes.•Boundedness of the referent is shown to underlie the distinction.•Similarities are established with other types of noun categorization.

Categorization is an inherent feature of human cognitive processes and systems that identifies coherent patterns in our knowledge and behavior. In language it takes the form of formally definable categories. Spatial categories are particularly known to pervade linguistic structure, and even to organize non-spatial domains. Within spatial linguistic theory there are two crucial notions, that of the Figure, the entity to be located (or the what), and that of the Ground, the entity with respect to which the location is defined (or the where). These two notions underlie all spatial functions and their cross-domain remappings. Yet, there are restrictions on which entity can function as the Figure and which as the Ground in a locative expression; notice the questionable sentence: *The house is behind the bicycle. Bearing in mind how rudimentary the Figure/Ground constellations are in language, it is not surprising that these two fundamental spatial categories find expression in language structure. I argue that the capability to function as the Figure or the Ground is impressed on the lexicon of Lokono, an Arawakan language of the Guianas. This grammatical distinction parallels other types of noun categorization, such as the mass/count dichotomy. It manifests itself in a narrow, though cognitively universal context, namely directionality. Shifts from one category to the other are possible and result in predictable semantic changes. The distinction is attested cross-linguistically and reflects the ontological properties of the referents.