The aim of this paper is to address a longstanding concern about the linguistic 'relativity' of ontological categories, and resulting limitations in the scope of ontological theories. Given recent evidence on the influence of language on cognitive dispositions, do we have empirical reasons to doubt that there are ontological categories that have universal scope across languages? I argue that this is the case, at least if we retain the standard 'inferential' approach within analytical ontology, i.e., if we evaluate ontological interpretations of L-sentences relative to certain material inferences in L. Research in linguistic typology suggests that types of entities postulated for the domain of Indo-European languages cannot capture the ontological commitments of the (much larger group of) non-Indo-European languages. Ontological category theory thus seems to have three options. The first option is to abandon the standard 'inferential' approach to ontological category theory. Alternatively, if we stay with the inferential approach, we face the following choice. Either ontology must let go of its ambitions to provide general domain descriptions for any language and settle for the more modest project of reconstructing the ontological commitments of a group of natural languages. Or else analytical ontologists should turn to linguistic typology in order to accommodate the diversity of inferential structures embedded in natural languages. I recommend and explore this third option, illustrating a strategy for how to construct a domain theory that can be used across languages. In a first step I show how linguistic research on the semantics of verbs and nouns (studies on so-called "Aktionsarten" and "Seinsarten") can be used to identify the inferential patterns of ten basic concepts of modes of existence in time and space. In a second step I show how these inferential data can be interpreted ontologically within General Process Theory, an ontological framework based on nonparticular individuals ("dynamics").
