News & Events

Syntax that can change

Whitney Tabor

University of Connecticut

Humanities Center, Conference Room D

Saussure argued for distinguishing langue (the ideal language system) from parole (the actual speech or text that humans generate and interpret), and Chomsky argued for separating competence and performance. Both have cautioned against conflating historical language and synchronic language, emphasizing the perspective of the young child, whose experience during the crucial period of life in which syntax forms, is entirely with the latter and not the former. I agree. Yet, I suggest that language science will now benefit from becoming, newly, a historical science. Many language scientists and engineers today focus on the problem of generalization: How does a child, or a neural network trained on a finite body of synchronic data, end up generalizing from this experience to identify all and only the grammatical forms that the language system permits? I think the theory will become helpfully more precise if we address, in addition, the question, How do language users—not so much children, but teenagers, and young adults—end up also agreeing on what can be newly added to the language? This is the problem of innovation, different from the problem of generalization. I focus on phenomena of morphological innovation—e.g., “unfriend”, and grammaticalization—gradually, over historical time, language expressions can gain new grammatical uses (e.g., English “a lot”, strictly a determiner-noun sequence in 1800, is now, most often, a kind of quantifier—“a lot of people were curious”—or an adverb—“they squabbled a lot”). I describe a formal model, called a “Stretched Tree Metric Grammar”, in which language encodings lie in vector spaces, as in Large Language Models (LLMs), but unlike in current LLMs, there is a precise specification of the relationship to grammatical rules. The model not only does a respectable job of parsing and restrictively generating, but also, in the presence of an appropriately biasing social force, can innovate.

Whitney Tabor
Associate Professor of Psychology, University of Connecticut
Learn more about Whitney Tabor.

Register to attend via Zoom.