Reading as a problem solving activity
Reading comprehension in pupils in Years Four and Five was investigated in a series of experiments predominantly using error detection in short, narrative passages. The claim that readers of this age may exclusively accept decodability as sufficient to judge text as non-problematic was investigated. Little evidence was found in support of this claim. However evidence was obtained that consonant strings were more often detected than nonsense words and more nonsense words were detected than real words of inappropriate meaning when they were substituted into the same passages at the same points. Fewer real words of inappropriate meaning were detected when they were the same part of speech as the word they replaced than when they were a different part of speech. No evidence was obtained that significant numbers of children exclusively detected only nonsense words. However decreasing the readability of the passages significantly reduced the detection rate for real words of inappropriate meaning while the detection rate for consonant strings and nonsense words remained both higher and more stable. It was suggested that children who are asked to read passages of too low a readability for them may be more likely to exclusively employ a lexical standard of comprehension. No evidence was obtained that asking children to read or listen to a passage a second time before completing an error detection task improved their performance. Moreover no difference in semantic comprehension monitoring was found to be dependent on whether the material was presented orally or visually. Better comprehenders were better than less good comprehenders on both error detection and doze tasks. However there was no difference in the relationship between performance on prompted e.g. doze, as compared to unprompted e.g. error detection, comprehension tasks between better and less good comprehenders. Both groups performed better on the prompted comprehension tasks. Better performance was maintained on doze tasks even when the subjects were not only alerted to having to read the passage for meaning but knew they were to be asked questions on it. The extensive use of unprompted comprehension tasks with feedback was proposed as a method of closing the gap between students' performance on unprompted as contrasted with prompted measures of comprehension. Better comprehenders were better at sequencing sentences to make a story but did not perform better than less good comprehenders at recognising sentences from stories they had just read. Both better and less good comprehenders were less good at rejecting as having just been read sentences semantically congruent with the stories as contrasted with sentences semantically incongruent with the stories. This was consistent with most readers engaging in constructive processing of short stories. The results of this series of experiments were compatible with and discussed in terms of comprehension involving the construction of a mental model of what is heard or read while listening to or reading short stories. Suggestions for further experiments were made.