Improving question answering from short texts by 9- to 11-year-old children using induction tasks.

Abstract

Visually searching for verbal information is a complex activity for young readers. An eye-tracking experiment was conducted to investigate whether preactivation of different word-processing pathways by means of semantic or perceptual induction tasks could help children aged 9 to 11 to search for answers to questions in short texts (about 150 words). The type of questions asked (surface vs. inferential) and the nature of the induction tasks (semantic vs. perceptual) were manipulated. In addition, the quality of students' lexical representations was assessed using word identification and discrimination tests. Children made more errors when answering global inferential questions than local inferential questions, and more errors when answering local inferential questions than surface questions. Children's error rates were mostly unrelated to the amount of time they spent looking for answers in texts, which suggests that children did not spontaneously adapt their search time to question complexity or perceive that some questions were more complex than others. After performing a semantic rather than perceptual induction task, the accuracy rate of children's answers to inferential questions increased (81.5% vs. 72.5%), in relation to an increase of the time spent searching the texts. As expected, information-seeking times were significantly shorter for children with higher quality lexical representations. The impact of the semantic induction task was greater for children in the lower half of lexical quality scores than for those with high scores, who tended to spontaneously use effective strategies and did not appear to have any difficulty identifying the answer in the text.