Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.584895
Title: Transformation and representation in similarity
Author: Hodgetts, Carl J.
Awarding Body: Cardiff University
Current Institution: Cardiff University
Date of Award: 2010
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
Similarity, being a psychological notion, involves the comparison of finite object representations. The specific nature and complexity of these representations is a matter of fierce theoretical debate traditionally, similarity research was dominated by the spatial and featural account. In the spatial account, similarity is determined by the distance between objects in a psychological space. Alternatively, the featural account proposes that similarity is determined by matching objects' features. Despite the empirical success of these accounts, the object representations they posit are regarded too simple and specific to deal with more complex objects. Therefore, two structural accounts have been developed: structural alignment (SA) and Representational Distortion (RD). This aim of this thesis was to further establish one particular structural account—RD—as a general framework for understanding the similarity between object representations. Specifically, RD measures similarity by the complexity of the transformation that "distorts" one representation into the other. This RD approach is investigated in detail by testing a detailed set of transformational predictions (coding scheme) within a rich stimulus domain. These predictions are tested through experiments and modelling that utilise both a) explicit measures (ratings, forced-choice), and, for the first time, b) implicit measures (reaction time, same-different errors & spontaneous categorisation). Moreover, RD is compared empirically with both traditional and alignment models of similarity. Overall, the results suggest that similarity can be best understood by transformational relationships in a number of contexts. The performance of RD in both explicit and implicit measures is made more compelling by the fact that rival accounts fundamentally struggle to describe the sorts of relationships that are easily captured by RD. Finally, it is emphasised that RD is actually compatible with supposedly rival approaches and can incorporate theoretically these accounts, both traditional and structural, under one general framework.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.584895  DOI: Not available
Share: