Pronoun interpretation in reading

Abstract

This research investigated the comprehension of anaphoric pronouns in reading. Chapter, I explains the importance of this feature of text processing, and reviews the relevant psychological literature. One -important feature of pronoun interpretation is the fact that antecedents vary in accessibility in response to aspects of text content and structure; antecedents which are readily acessible to pronominal reference are said to be foregrounded. The most common theoretical explanation of foregrounding is the claim that foregrounded antecedents are held in working memory. The review also shows reasons why good and poor readers might differ in sensitivity to foregrounding. Experiments 1-3 investigated the effects on foregrounding of pronoun-antecedent distance and of topical continuity. It was found that distance had no effect on ease of pronoun assignment provided the intervening sentences were closely related to the antecedent. When the intervening sentences dealt with unrelated topics, pronoun interpretation became more difficult. The working memory explanation of foregrounding was investigated in Experiments 4-5, which found no evidence that foregrounding of an antecedent facilitated recognition memory for that antecedent. None of Experiments 1-5 found differences between the pronoun interpretation processes of good and poor readers. There then follows a critical discussion of the working memory theory, which argues that foregrounding may also be explicable by a theory stressing antecedent retrieval processes. Later experiments attempted to discriminate between these theories. Experiments 6-7 found that antecedent foregrounding was sensitive to the content of the pronoun sentence itself. Experiments 8-9 found that foregrounding of an incorrect antecedent did not hamper pronoun interpretation any more than neutral backgrounding. All the results of Experiments 4-9 contradict the working memory theory. It is concluded that the working memory theory of foregrounding is inadequate to explain the present results, and that an account of foregrounding based on antecedent retrieval processes is a plausible alternative.