Plenary Speaker

Talk Schedule

April 19th (Sat), 16:10-17:10

Profile

Alec Marantz is a Silver Professor of Linguistics and Psychology at NYU, which he joined in 2006 from MIT, where he had been Head of the Department of Linguistics and Philosophy and the Kenan Sahin Distinguished Professor of Linguistics. Co-founder of the theory of the architecture of grammar known as Distributed Morphology, Professor Marantz is world-renowned for his work on morphology and syntax, and more recently, his work on neurolinguistics, carried out at NYU in Neuroscience of Language Lab. Within NeLLab, Marantz focuses on word structure and innovative MEG (magnetoencephalography) methods for functional neuroimaging. Integrating linguistic theory and psycholinguistic models with observed neural activity in the brain, this work explores: how the ability to use natural language is implemented in the brain; how the brain mediates the most critical aspects of our communication system; and which properties of the mind/brain facilitate the seemingly effortless human processing of language.
Website

Title

Morphemes as signs:  Reimagining Saussure for the era of LLMs and neuroimaging

Abstract

Saussure has been wrongly associated with the position that signs contain an arbitrary connection between a sound representation and a concept. In fact, he explained that signs don’t contain representations at all but rather work to create meaning and sound percepts via their contrast with other signs both within meaning and within sound spaces. What’s arbitrary is how many signs a language has, with the number of signs changing the coverage of sound and meaning space of each sign. We are used to thinking of the meaning of basic color terms being a function of how many basic color terms a language has to fill out color space and thinking of the phonological representations of words being a function of the number of phonemes a language has to fill out phonetic space; Saussure invites us to imagine that all signs create concepts and “sound images” via the same contrastive mechanisms. Meaning space has been instantiated by the space of meaning vectors in Large Language Models and both acoustic and articulatory spaces have been mapped by linguists. Morphemes as signs connect distributions in meaning space with distributions in sound spaces; exploiting LLMs one can assign numbers to meaning similarities and differences to test in experiments. Recent advances in the neurobiology of language have provided plausible locations for the neural instantiations of these meaning and sound spaces. We will discuss neurolinguistic experiments that help reveal how the distributed representations of morphemes, as described within the linguistic theory of Distributed Morphology, work in word recognition under the morpheme as sign hypothesis.



<< Back to "Speakers"