Testable universals, the natural-kinds programme, and presupposed universals in grammatical theorizing (conference handout)

This talk will focus on the problems of presupposed universals and of empirical testing universals in general-theoretical linguistics. Paradoxically, universals of grammar have been very prestigious and prominent in linguistics since Greenberg (1963) and Chomsky (1965), but what exactly is universal in grammars is still largely unknown. Many linguists presuppose universals of various kinds (architectural universals like the syntax-morphology division, or substantive universals like phonological distinctive features or syntactic categories), but the universality of all these is very uncertain (see, e.g., Mielke 2008 on distinctive features, Haspelmath 2011 on the syntax-morphology division). The mere fact that established concepts can be applied to new data is not sufficient to corroborate the reality of these concepts if there is no clear sense of what observations would be inconsistent with them.

     To make true progress in general-theoretical linguistics, I argue that two kinds of steps need to be taken: First, if one is interested in readily testable, observable universals of the Greenbergian kind, one should test those that have been proposed. Linguists need to establish a culture of hypothesis-testing, in addition to their existing culture of generating new hypotheses. As psychologists have found out, there is no guarantee that proposed generalizations will hold up after more testing. Such hypothesis-testing will have to rely on rigorously defined comparative concepts as uniform yardsticks for objective masurement (cf. Haspelmath 2020).

     Second, if one is interested in innate architectural or substantive universals of the Chomskyan kind, which are not so readily testable (because of their sometimes very indirect effects), one should try to find ways of comparing competing proposals at least in special subdomains. Some of the 20th century Chomskyan proposals (which I call “natural-kinds programme”, following Baker’s 2001 comparisons with chemistry) are often presupposed as true (and even taught in introductory classes), but in reality, linguists do not know which of these proposals correspond to the true innate categories of the human mind. Thus, we need to construct compelling cases at least for some subdomains where alternative explanations (e.g. in terms of historical accident, or in terms of convergent cultural evolution) cannot work.

     Or alternatively, in order to demonstrate an innate grammar toolbox, one needs to establish correspondences between stimulus poverty and universals observed in languages. Arguments from the poverty of the stimulus are often invoked in general terms (e.g. Lasnik & Lidz 2016), but it is rarely clear what exactly is predicted and explained by such considerations.