Use this URL to cite or link to this record in EThOS:
Title: Representing meaning : a feature-based model of object and action words
Author: Vinson, D. P.
ISNI:       0000 0004 2731 5512
Awarding Body: University College London (University of London)
Current Institution: University College London (University of London)
Date of Award: 2009
Availability of Full Text:
Access from EThOS:
Access from Institution:
The representation of word meaning has received substantial attention in the psycholinguistic literature over the past decades, yet the vast majority of studies have been limited to words referring to concrete objects. The aim of the present work is to provide a theoretically and neurally plausible model of lexical-semantic representations, not only for words referring to concrete objects but also for words referring to actions and events using a common set of assumptions across domains. In order to do so, features of meaning are generated by naïve speakers, and used as a window into important aspects of representation. A first series of analyses test how the meanings of words of different types are reflected in features associated with different modalities of sensory-motor experience, and how featural properties may be related to patterns of impairment in language-disordered populations. The features of meaning are then used to generate a model of lexical-semantic similarity, in which these different types of words are represented within a single system, under the assumption that lexical-semantic representations serve to provide an interface between conceptual knowledge derived in part from sensory-motor experience, and other linguistic information such as syntax, phonology and orthography. Predictions generated from this model are tested in a series of behavioural experiments designed to test two main questions: whether similarity measures based on speaker- generated features can predict fine-grained semantic similarity effects, and whether the predictive quality of the model is comparable for words referring to objects and words referring to actions. The results of five behavioural experiments consistently reveal graded semantic effects as predicted by the feature-based model, of similar magnitude for objects and actions. The model's fine-grained predictive performance is also found to be superior to other word-based models of representation (Latent Semantic Analysis, and similarity measures derived from Wordnet).
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available