Computational constraints on compositional interpretation: Refocusing the debate on language universals

We argue that the debate on language universals should be directed away from the discussion as to whether typological diversity is or is not an argument against the existence of language universals. Instead, given our growing awareness of the fact that the neural mechanisms underlying language use are the same as those underlying other cognitive functions in both humans and mammals, the central question for cognitive neuroscientists and linguists is what neural mechanisms can facilitate compositional interactions, and how the range of grammatical structures and the ability to use language creatively emerges from a much narrower range of neural mechanisms. We suggest two complementary methods for investigating these issues, one more linguistically oriented and one more computationally oriented, and present preliminary results from investigations concerning the expression of the mass-count distinction crosslinguistically, using both methodologies. These results suggest that universality in language might express itself at the deeper level of the computational operations involved in the processing of language, rather than in the results of those computations.