Establishing coherence across sentence boundaries: an individual differences approach

Abstract

The aim of this study was to investigate the cognitive abilities that underlie coherence building. A coherence judgement task was used for this purpose. The task was comprised of four conditions that resulted from crossing coherence and cohesion (the presence of a connective), a manipulation that elicited two-way interactions in both judgement accuracy and reading times. Hierarchical Linear Modelling was then used to assess the influence of individual difference variables (vocabulary, working memory, and decoding) on target sentence reading times across the four conditions. Two of these variables, working memory and vocabulary, shared cross-level interactions with the coherent-incohesive condition – the only condition in which coherence had to be established with an un-cued bridging inference. Vocabulary acted to decrease reading times associated with the condition whereas working memory acted to increase them. These effects are interpreted with reference to lexical quality and retrieval-based knowledge access.

Keywords:

• coherence building
• reading comprehension
• individual differences
• working memory
• vocabulary

Acknowledgement

This research was completed by Stephen T. Hamilton in partial fulfilment of the requirements for the degree of Doctor of Philosophy at the University of Sussex. We would like to thank Dr. Evelyn Ferstl for her advice during the data collection portion of this study.

Funding

This work was supported by Economic and Social Research Council (ESRC) 1 + 3 Scholarship.

Notes

1. ‘All the colleagues are envious’; ‘Next week is the wedding anniversary’; ‘Nobody had forgotten the birthday’; ‘The friends are waiting at the door with a bottle of champagne’; ‘The hands are trembling’; ‘The heart starts beating faster every time the phone rings’; ‘The knees are grazed all the time’; ‘The mother is standing anxiously near the deep end of the pool’.

2. Data obtained from UK and North American participants were analysed separately to ensure reliability. All ANOVA interactions were significant for accuracy (both Fs > 13) and RT (both Fs > 5) in both groups of participants

Additional information

Funding

Funding: This work was supported by Economic and Social Research Council (ESRC) 1 + 3 Scholarship.