Dissertation/Thesis Abstract

Representing Relationality: MEG Studies on Argument Structure
by Williams, Adina, Ph.D., New York University, 2018, 160; 10750823
Abstract (Summary)

One of the quintessential properties of the human semantic system is its ability to flexibly combine the meanings of smaller pieces into larger wholes. However, not all smaller conceptual pieces are created equal; concepts differ in the extent to which they can drive meaning composition. Some concepts can drive semantic composition by establishing relationships with other concepts, while others cannot. For example, we know the concept labelled by “friend” can drive composition, since one cannot be a friend without being someone’s friend, while an animal can be a cat without standing in a similar relationship. We can thus divide the conceptual space of humans into two sections: relational concepts labelled by words like “friend”, and non-relational concepts like the one labelled by “cat”.

Evidence in favor of this division indicates that in early childhood humans are aware of the relationality of concepts (Smiley and Brown 1979; Mirman and Graziano 2012), and as we age, relational concepts remain extremely common in our lexicon, making up nearly half of the adult English vocabulary (Asmuth and Gentner, 2005; Gentner, 2005). Some relational words have been extensively studied by cognitive psychologists and formal linguists alike. One relatively mature set of investigations utilizes functional magnetic resonance imaging (fMRI) to investigate the neural basis of relational verb meaning, and finds that relational (i.e., transitive) verbs drive activity in left perisylvian cortical regions more than their intransitive counterparts (Meltzer-Asscher et al., 2015; Thompson et al., 2010, 2007; Bornkessel et al., 2005; Ben-Shachar et al., 2003). These investigations attribute this activity to verb-specific or event-specific information that is stored as part of the verb’s conceptual representation. Some support for this comes from Binder and Desai 2011 that holds that the left AG is a main semantic hub that specializes in event processing. However, formal linguistic investigations suggest that relationality should be independent of verbhood or eventivity; it is an independent abstract property of some lexical items which enables them to establish relationships and drive semantic composition. The neural basis of relationality and whether it can be independent of verbhood and eventivity is still relatively underexplored. As relationality straddles the boundary between syntax and semantics, disentangling the contributions of various linguistic features, such as syntactic category, eventivity, and plurality to left AG activity becomes a crucial exercise—one that a linguist is uniquely poised to address.

A main candidate region for relational processing is the left Angular Gyrus (lAG), because prior literature suggests it is sensitive to at least some of the features that a region that processes relationality would be sensitive to. In addition to being implicated in tasks that probe the argument structure of verbs, the left AG was found to be the most consistently activated region across numerous semantic tasks in a recent, large-scale meta-analysis (Binder and Desai, 2011), prompting it to be dubbed a domain-general ”semantic” hub (Bonner et al., 2013; Binder and Desai, 2011). Because of this, this dissertation reports the results of three MEG experiments and one computational experiment, and focuses on the left AG and surrounding perisylvian cortical regions, and uses Magnetoencephalography (MEG) to investigate its role in relational processing.

Chapter 2 asks whether left AG tracks relationality, the eventivity of verbs, or a word’s combinatorial context, and finds a main effect of relationality from 170–260 ms after the visual presentation of the target noun, and no other effects of the other factors, suggesting that it is indeed relationality and not eventivity (or context) that drives left AG argument structure findings.

Chapter 3 asks whether relationality effects could be driven by something other than the relationality of concepts, namely, by the quantity of concepts. In addition to being activated for numerous number-related tasks (see Dehaene et al. 2003 a.o., for a review), the left posterior perisylvian cortex has been found to be differentially active for plurals as opposed to for singulars (Domahs et al., 2012), suggesting that the lAG might track semantic information about plurality. Contrasting plural and singular nouns that vary in relationality, I replicate the relationality effect, and find no interaction with plurality, suggesting that, indeed, left posterior perisylvian relationality effects cannot be explained as a type of quantity processing.

Chapter 4 presents a filler experiment to the second experiment, which investigates the mass-count distinction and asks whether left AG activity is sensitive to countability and whether it interacts with plurality. I observed a main effect of countability (count nouns > mass nouns), in a broad swath of left hemisphere from 295–465 ms after noun presentation, and additionally observe a left frontal effect of plurality (bare nouns > plurals), paralleling what I found in Chapter 3. (Abstract shortened by ProQuest.)

Indexing (document details)
Advisor: Pylkkänen, Liina
Commitee: Barker, Christian, Bowman, Samuel R., Cimpian, Andrei, Marantz, Alec
School: New York University
Department: Linguistics
School Location: United States -- New York
Source: DAI-A 80/03(E), Dissertation Abstracts International
Source Type: DISSERTATION
Subjects: Linguistics, Psychology, Experimental psychology
Keywords: Argument structure, Computational linguistics, Magnetoencephalography, Plurality, Relationality, Semantics
Publication Number: 10750823
ISBN: 9780438634497
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest