Inference Generation and Reading Disability

This thesis investigated on-line inference generation in 7 to 10-year-old children. Using the Simple View of Reading (Gough & Tunmer, 1986) as a framework to classify reading disability, poor comprehenders and poor decoders were compared with chronological age-matched typical readers.

In Experiments 1, 2, 3 and 4 a self-paced reading paradigm was employed to assess on-line inference generation. Reading times to target sentences that were consistent and inconsistent with inferred context were compared. Experiments 1 and 2 showed that children made inferences about fictional characters’ emotion-states and spatial inferences on-line, indexed by increased reading times to inconsistent target sentences. Results indicated that poor comprehenders made fewer inferences on-line than either typical readers or poor decoders.

In Experiment 3, participants generated inferences based on early-acquired general knowledge on-line but not inferences based on later-acquired knowledge. This was replicated in Experiment 4 for early-acquired knowledge, however only poor comprehenders failed to generate inferences on-line using later-acquired knowledge. There was evidence of carry-over in post-target sentences, which was less pronounced in poor comprehenders.

In Experiment 5, a self-paced listening paradigm showed that poor comprehenders generated fewer on-line inferences than typical readers did when processing spoken language, but contrary to prediction so did poor decoders.

In Experiment 6, children read short vignettes followed by a question. They then evaluated four answers varying in plausibility. Poor comprehenders were equally accurate but responded faster than either poor decoders or typical readers. This was consistent with the view that poor comprehenders apply a lower standard for coherence when processing language.

Together, the results support the view that decoding and linguistic comprehension dissociate, as predicted by the Simple View of Reading. These results contribute further evidence, using an on-line measure, that poor comprehenders generate fewer inferences than typical readers or poor decoders, and suggest this may be driven by the application of a lower standard for coherence when comprehending written or spoken language.