In 1997 I was invited to speak to this conference, and I discussed “Some contributions of typology to cognitive linguistics, and vice versa”, which evolved into Radical Construction Grammar: Syntactic Theory in Typological Perspective. This year’s conference theme is “Crosslinguistic Perspectives on Cognitive Linguistics”, and in this talk I revisit the question of the relation of typology to cognitive linguistics, and how it has developed in the past twenty-two years.

The central argument that I presented in 1997 is that morphosyntactic categories, including those of parts of speech such as noun, verb, adjective, or core grammatical roles such as subject, object and oblique are not just language-specific, but construction-specific. In other words, there are no universal morphosyntactic categories.

This conclusion, well supported by empirical evidence of distributional variation—also observed in corpus-based research in construction grammar, which to a great extent serves as the syntactic theory of cognitive linguistics—poses a problem for cognitive linguistic theories. Human cognition includes features that are general among humans. Theories such as Cognitive Grammar posit universal cognitive definitions of fundamental grammatical categories. These definitions are semantic, not based directly on syntactic distributional facts.

Yet the cross-linguistic analysis of semantic categories also indicates that there is not a finite set of universal semantic categories either. One cannot assume that semantic categories expressed in European languages are the most useful or the only ones that occur across languages. For example, Dryer (2013) finds that the definite-indefinite distinction is less common than distinctions based on the tracking status of referents (see also DuBois 1980). More radically, the ‘in’-‘on’ type spatial relations (Bowerman and Pederson; Levinson et al. 2003) are organized in a continuum of gradient semantic dimensions (Croft 2010), not discrete universal semantic categories.

However, there is another approach that is more promising to address this problem. Linguistic function includes construal of semantic content. One can offer a precise definition of construal and its constraints (Croft 2007, 2012). An example of construal is the set of propositional act constructions (reference, predication, modification) that package the information content of “content words”. That is, universals of noun, verb and adjective are found in the information packaging of content words in the process of verbalization (Chafe 1977a,b).

In work in progress (Morphosyntax: Constructions of the World’s Languages), I survey cross-linguistic variation and universals for constructions that encode the major functions expressed in language. From this survey, I conclude that all linguistic meaning involves information packaging of semantic content. This conclusion is valuable for understanding cross-linguistic variation, but it should not come as a surprise to cognitive linguists for whom meaning is construal (Langacker). The information packaging functions are more plausible bases for definitions of universals of grammatical categories and constructions. Nevertheless, even here there are continua, including a modification-predication continuum and a reference-modification continuum.

Constructions are comparative concepts. Recent work in typology offers robust definitions of comparative concepts for cross-linguistic comparison (Haspelmath 2010; Croft 2016). I distinguish constructions—any form expressing a function—from strategies—a subset of constructions that use a particular set of formal structures. Understanding strategies is the heart of grammatical analysis. Three types can be identified: encoding, coexpression and recruitment strategies. The first two types are more grammaticalized versions of recruitment, that is, the recruitment of a constructional form for a new function. Recruitment is, of course, another type of construal, of one function as being similar to another function. The marriage of typology and cognitive linguistics is a fruitful one.