The effects of content and structure clues on adult poor readers' ability to generate inferences

Penne Pacl Devery, Marquette University

Abstract

The study investigated the ability of adult poor readers to use content and/or structure clues to generate inferences. One hundred twenty-eight male and female junior college freshmen enrolled in developmental reading classes read two single-episode stories. The stories, presented one line at a time on a computer, were modified to yield differing levels of inferability at two story locations. Explicit clues, implicit clues, or no clues to deleted story information were provided. Questions asking subjects to recall exact content and infer missing information were given after the stories were read. The number and location of lookbacks, reading times, and decision-making times were also measured. A multivariate analysis of variance was conducted on the question and lookback measures. Subjects reading the original stories were better able to recall exact content than subjects reading stories in which information was deleted. Disruption of structure made recall of exact content more difficult. However, approximately half of the responses in the deletion conditions were written in the categorical language of the original story rather than in the language of the story read. These subjects were inferring structure to recall stories. Explicit clues were found to be necessary to draw inferences about the structure of stories, and subjects needed to be prompted by questions to make inferences. Subjects made limited use of lookbacks, primarily at the end of stories, as a strategy to resolve inconsistencies. Means before deletions and means after deletions were obtained for reading time and decision time and a repeated measures multivariate analysis of variance was conducted. The location of the deleted information was found to affect processing time. Readers attended more to beginnings and endings instead of the internal text disruptions. Though results indicated that readers rely more on structure to recall exact content and generate inferences, the variation among responses make some findings unstable. To allow for a more precise test of the role of structure and content in inferencing, a replication of this study should be carried out with good readers who are more attentive and consistent while reading.

This paper has been withdrawn.