Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.391254
Title: Modelling serial order in behaviour : studies of spelling
Author: Glasspool, David William
ISNI:       0000 0001 3500 8107
Awarding Body: University of London
Current Institution: University College London (University of London)
Date of Award: 1998
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
Serial order in behaviour remains an interesting problem for computational modelling in psychology, especially for connectionist approaches. The 'Competitive Queuing' (CQ) approach to sequence generation has the advantage of accounting for a number of common features apparent in several different types of serial behaviour. This thesis addresses the general account which the CQ approach can give for constraints on serial errors within sequences by developing models of an acquired disorder of spelling, 'graphemic buffer disorder' (GBD). Two approaches to the development of a simple initial model of GBD into more complex models are demonstrated, and are related to the general problem of accounting for serial category constraints in sequencing. The initial CQ model of GBD is based on an existing model of speech production with minimal spelling-specific changes. A number of shortcomings are identified in the I performance of this model, in particular the inability to distinguish consonant and vowel letters, which prevents a striking feature of GBD errors - the preservation of consonant/vowel status - from being modelled. An analysis of the general problem of adding domain-specific constraints to CQ models suggests two approaches to improving the initial model. Two alternative extended models are thus advanced. The first is a development of the initial model incorporating an external template to specify consonant/vowel information. Simulations with this model demonstrate a much improved fit to :the data. The second model" develops a novel architecture, generalising the CQ approach to multi-layer networks. The model is less detailed but demonstrates the correct general features of the GBD error pattern. The relationship between the models is discussed and possible future research directions are identified.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.391254  DOI: Not available
Keywords: Psychology
Share: