Doctoral Dissertation Research: The interaction between conceptual combination and linguistic structure

New York University, New York NY

Investigators

Abstract

The last decade of research into how the brain composes meaning has shown that a stronger driver of brain activity corresponds to the conceptual and meaning-based aspects of words rather than their abstract linguistic properties. That is, when combining the words 'blue' and 'hat' to form the more complex mental representation of 'a blue hat', the conceptual meaning of 'blue' (a color adjective, where the word it modifies is blue) seems to drive an increase in brain activity in regions associated with meaning and composition, but the word's linguistic status as a modifier is not the primary driver of this activity. This observation invites linguists to adapt existing theories about how people use language structure to include conceptual meaning. The project examines this by developing a series of studies that vary the structure of English sentences either while keeping the conceptual meaning relatively constant or while using those structural cues to vary which words can combine to form a new meaning. The primary study funded by this NSF grant measures word composition by using magnetoencephalography (MEG) to compare, for example, the neural representation of the word 'hat' in across sentences like (a) "The hat is a really pretty blue color" and (b) "The hat is near a really pretty blue lamp", where the key difference is that in (a) but not (b), the words 'hat' and 'blue' must combine to form the representation of 'a blue hat'. Words are presented visually one-at-a-time on a screen while the researchers record participants' brain activity. They then use a machine learning based approach to study at which points in the sentence the brain is still actively representing the word 'hat' after it has been presented. The researchers hypothesize that if it is necessary that the brain activates the full representation of a word in order to combine with another concept, then they should be able to accurately identify the representation for a word like 'hat' immediately after the word 'blue' is presented in sentences like (a), but not in sentences like (b). This study would be the first to provide direct evidence that such word-level re-activation can be triggered by certain sentence structures and inform psycholinguistic models of sentence processing. The project includes the training of undergraduate students in STEM research. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.