INFERENCE IN L2 READING AS MEASURED BY THE CLOZE PROCEDURE.

摘要

This dissertation is intended as a contribution to reading theory. It attempts to provide a linguistic and psychological answer to the following question: "On what basis do readers infer meaning from context?".;Chapters 1-3 present the theoretical background to the present model. This is a synthesis of the relevant fields, and is loosely divided according to subject: logic and linguistics (Chapter 1), psychology and verbal memory studies (Chapter 2), text and discourse analysis (Chapter 3). Each make the case for a model of verbal inference based upon situational entailments.;In presenting a new model there is a clear need for empirical validation. The quantitative section of this study describes an experiment involving 800 Haitian and Hispanic subjects. On the basis of our model, hypotheses are made for the distribution of their scores on the 'cloze procedure', a test instrument widely used for the assessment of reading. Texts were randomly selected from a public library and cloze procedures constructed by the arbitrary deletion of every twelfth word. For two of our hypotheses, the results strongly confirmed the predictions of our theoretical model (X('2) < .001).;The implications of these results are believed to be important. The concluding chapters (Chapters 6 and 7) present experimental conclusions and conjectures of general interest to reading theory, materials development and syllabus design.;There are several problems with the present theories of reading. Not only are they characterized by a mistrust for the findings of the formal linguist (Smith, 1978, p. 78), but, apart from directing attention to the issue of inference, little attempt is made to characterize its basis or nature. A distinction ought to be made here between 'process models' and 'structural models'. A model of inference is proposed which attempts to reconcile the structural descriptions of linguistics with the process requirements of reading theory.