Reading comprehension in adults: component skills; false memories; and judgements of coherence

The aim of this thesis was to investigate some of the processes that contribute to the effective comprehension of text in an adult population. The thesis begins with an assessment of component skills that are of theoretical relevance to reading comprehension skill. Experiment One explored the relation between gist-based memory processes and reading comprehension skill. Weaknesses in semantic processing have been shown to contribute to comprehension difficulties both in childhood (e.g. Nation & Snowling, 1999), and adulthood (e.g. Perfetti, Yang & Schmalhoffer, 2007). Weekes, Hamilton, Oakhill & Holliday (2008) used the false memory (DRM) paradigm developed by Deese (1959); Roediger and McDermott (1995) to assess the relation between reading comprehension and memory processes in children. In the DRM, subjects memorise lists of semantically related words (e.g. bed, rest, awake) for later recall. During recall, it is typical to see intrusions of semantically related but non-presented items (e.g. ‘sleep’ is often falsely recalled following presentation of the above). Weekes et al. (2008) found that children with comprehension difficulties produced fewer such intrusions than did good comprehenders, suggesting that poor comprehenders have difficulty extracting the central theme or ‘gist’ from the word lists, a deficit that was attributed to weakness in semantic processing and memory. Experiment One demonstrated that this effect was not replicable in an adult population. Although there is evidence that deficits in semantic processing contribute to reading comprehension difficulties in adulthood, these appear to be too subtle to manifest themselves in the DRM paradigm. In Experiment Two, measures of vocabulary, word-level skills (orthography and decoding), working memory and verbal IQ were taken from a population of young adult readers. These measures were used as predictors of comprehension skill in multiple regression analyses. Moderate support for the Verbal Efficiency/Lexical Quality Hypothesis (Perfetti, 1985; 2007) was obtained, in that word-level skills and vocabulary size accounted for unique portions of variance in comprehension skill. Experiments Three and Four explored the processes involved in on-line reading comprehension and, specifically, in a comprehension task that demanded integration. In both experiments, subjects took part in a coherence judgement task (Ferstl, Guthke & von Cramon, 2002; Ferstl, 2006) in which they had to verify whether two sentences cohered with one another or not. Four conditions that resulted from crossing coherence and cohesion (i.e. the presence of a lexical connection), were used: Coherent and cohesive (where sentences cohered, and a cohesive tie made their coherence explicit); coherent and incohesive (where sentences cohered, but coherence had to be inferred on the basis of pragmatic information rather than lexical cohesion); incoherent and cohesive (where sentences that do not cohere were erroneously linked with a cohesive tie); and incoherent and incohesive (where sentences did not cohere, and were not erroneously linked with a cohesive tie). Typically, the paradigm elicits an interaction between coherence and cohesion in reading times for the second (target) sentence: Targets in coherent and cohesive trials are read more quickly than targets in coherent and incohesive trials; and targets in incoherent and incohesive trials are read more quickly than are targets in incoherent and cohesive trials. Experiment Three replicated this interaction, and demonstrated that variance in its size was predicted by working memory capacity, with high working memory readers showing larger interaction effect sizes than low capacity readers. The interaction was interpreted as a monitoring effect that was triggered by target sentences in the atypical conditions (i.e. incoherent and cohesive; coherent and incohesive). It was proposed that high capacity readers were better able to engage in this monitoring. Experiment Four sought to explore the semantic deficit hypothesis in relation to this effect, with the proposal that efficient semantic processes, rather than working memory capacity, contributed to variance in the size of the interaction. Performance on a semantic fluency task was found to predict unique variance in the size of the interaction effect, over and above that accounted for by working memory capacity. This finding suggests that the effect is better explained by semantic processing than by working memory capacity, and that the interaction may be better described as a semantic elaboration effect rather than a comprehension monitoring effect. The conclusion of this thesis is that reading comprehension in adult readers relies upon efficient and accurate lexical access, comprising both lower-level processes such as accurate word recognition and decoding skill, and higher-level processes of semantic elaboration and integration.