ESRC grant on modelling eye movements during reading

In studies of comprehension, a widely used experimental approach involves gathering detailed records of the eye-movements people make as they are reading different sentences and paragraphs.  Underpinning this work is an abundance of evidence that eye-movement patterns can be dramatically affected by syntactic processing, ambiguity resolution and other aspects of comprehension.   However, at present there is no adequate description of the precise mechanisms by which sentence processing operations exert such control.  In the absence of an explicit model of high-level eye-control, it is impossible to use eye-tracking data for any kind of precise and quantitative evaluation of theories of comprehension.  As a step towards addressing this deficiency, the current project aims to refine the specification of the CIFER eye-control model recently developed by the authors.   This computational model has already been shown to make accurate predictions for certain eye-tracking measures and the proposed programme of work is designed to extend its scope of application, with the eventual aim of developing a tool capable of being used for the quantitative evaluation of theories of sentence processing.  In concrete terms, the proposed work involves collecting detailed eye-tracking data as participants read sentences that can currently be handled by the model.  By comparing observation with prediction it should be possible to identify aspects of the model which are in need of improvement.   Similar experiments will also be carried out using materials that fall beyond the range of the current implementation of the model.   In this case the aim will be to establish some of the features that need to be incorporated into an enhanced and more functional version of the model.  In all cases the intention would be to increase the precision of the model so that it can eventually be used in association with quantitative models of sentence processing to test a range of theories of higher level linguistic processing.   While it is anticipated that the immediate work will fall some way short of this ultimate goal, there is an excellent prospect that there will be wider spin-offs from the attempt to develop the model in this way.   In particular, the very least we expect to achieve by the end of the project is a much clearer specification of the properties, constraints and accomplishments that need to be built into a fully functional eye-control system of this kind.