Spatial cognition is central to human thinking, and spatial language is thus an important area of study, as it may reveal fundamental properties of human thought. Recent research has shown that spatial language is much more divergent across languages than had previously been thought, suggesting significant cultural patterning of spatial conceptualization. This article reviews spatial language cross-linguistically, sets out a typological framework for the language of space, and considers the relationship of spatial language to spatial cognition, in the context of extensive linguistic diversity in the spatial domain.Spatial cognition is at the heart of much human thinking, as evidenced, for example, in the explanatory power of spatial metaphors and diagrams, and the explication of spatial concepts has played a key role in the development of Western mathematics and philosophy over two and a half millennia (see Jammer, 1954). It is therefore easy to assume that spatial notions form a robust, universal core of human cognition; this assumption has dominated the cognitive sciences (e.g., Miller and Johnson-Laird, 1976). But naive human spatial language turns out to vary significantly across languages, both in the way it is semantically organized and the way in which it is coded. This has major implications for how we should think about universals in human thought and language.The Western intellectual tradition has provided an elaborate metalanguage for spatial concepts, but the conceptual underpinnings of naive human spatial language utilize only a small subset of these concepts. Moreover, there are many aspects of our spatial perception and motor control which utilize much richer and more precise representations of space than are coded in language. The study of the representation of space in language, and its relation to cognition, must therefore be treated as an empirical enterprise, and this article reports on the increasing body of cross-linguistic information about how different languages structure the spatial domain (see Levinson and Wilkins, 2006).