Speaker

Edward Gibson

Edward Gibson

Psycholinguist known as Ted Gibson, Professor of Cognitive Science at MIT and head of the MIT Language Lab, which studies why human languages take the grammatical forms they do and how people learn, process, and represent language. Trained as a mathematician and computer scientist before moving into linguistics at Cambridge, he has spent over three decades at MIT (since 1993) applying quantitative, experimental methods to questions linguists have traditionally settled by intuition alone.

Gibson’s central research finding is that human languages minimise the cognitive cost of dependencies between words — the pressure that explains cross-linguistic word-order patterns, the universal difficulty of center-embedded sentences, and the specific unreadability of legal writing. He has also conducted extensive fieldwork among isolate language groups in the Amazon, including the Piraha and the Tsimane, studying how number and colour vocabulary track cultural need rather than perceptual or cognitive universals. He is the author of Syntax: A Cognitive Approach (MIT Press, 2024).

Core positions

  • Dependency length minimisation. Word order across the world’s languages, and the specific unintelligibility of center-embedded sentences, follow from a single pressure to keep grammatically connected words close together. See Dependency Length Minimisation.
  • Language is not thought. Drawing on fMRI work by his MIT colleague Evelina Fedorenko, Gibson argues language is a dedicated brain system for communication, separate from reasoning, planning, and other cognition — directly opposed to Noam Chomsky‘s view that language underpins thought. See Language-Thought Dissociation.
  • Lexical copying over movement. Against Chomsky’s theory that syntactic forms are derived by literal movement from an underlying deep structure, Gibson favours the rival account in which related word forms (declarative and interrogative auxiliaries, for example) are simply separate lexical entries — a position he holds is both more learnable and better fits which auxiliaries actually alternate.
  • LLMs have solved form, not meaning. Large language models are, on Gibson’s account, arguably the best current theory of a language’s grammatical form, while still failing reliably at meaning-dependent reasoning that humans handle easily.
  • Vocabulary is invented, not innate. Concepts as basic as exact number words are cultural inventions driven by communicative necessity, not universal features of human cognition — evidenced by the Piraha’s complete absence of exact-counting words.

In the wiki