A THEORETICAL AND EXPERIMENTAL STUDY OF A RATIONAL DELETION SYSTEM FOR CONSTRUCTING TESTS OF LINGUISTIC AND COMMUNICATIVE COMPETENCE IN ENGLISH AS A SECOND LANGUAGE

The original motivation of this research was dissatisfaction with language proficiency tests to measure the language proficiency of non-native speakers of English. This research focused on both theoretical and empirical foundations for constructing a language test that is reliable, valid and practical and has diagnostic application.;Noticing a discrepancy between the standard every-nth-word deletion system and the recent developments in linguistic theory, Holtzman developed a rational deletion system using cloze procedure for testing the English language proficiency of foreign students. In 1979, Ahn and Holtzman examined the test results from Holtzman's 34 cloze passages administered to foreign students at The Pennsylvania State University since 1973. They concluded that a revised rational deletion system might lead to a new, more reliable, cloze procedure. It was also concluded that we needed to re-examine the reliability of cloze tests constructed by the standard every-nth-word deletion system since the potentiality salient distinction between function and content words is completely ignored by this system.;Thus, the following research questions were stated and examined: (1) Is cloze procedure constructed by the standard every-nth-word deletion system reliable in spite of variance attributable to the functional distinction between function words and content words? (2) Given cloze tests constructed by a rational deletion system on the basis of the functional distinction between function words and content words: (a) Would they be reliable? and, (b) Would they offer valid diagnostic information on linguistic competence and communicative competence?;The results provide strong evidence that the proportions of.;Traditional cloze procedure has been reported as a reliable and valid measure of language proficiency even though linguistic theory has been radically changed. There exist, however, some fundamental questions related both to the reliability and validity of cloze procedure. Since Carroll's pilot study of cloze procedure (1959), there has been controversy over the standard every-nth-word deletion system inspired by psychometric theory as contrasted with a rational deletion system inspired by linguistic theory.;function and content words are not stable in the cloze tests.;constructed by the standard every-nth-word deletion system. On.;the other hand, cloze tests constructed by a rational deletion system.;prove to be reliable. No significant difference was found between(, ).;the two test means PEHA S1 X(,CW=FW )and PEHA A1 X(,CW=FW).;A correlation matrix showed that the new cloze procedure correlated moderately with Dictation, Completion, TOEFL Structure and Writing, TOEFL Reading and Vocabulary, and TOEFL Total. Correlation coefficients of this order were expected because cloze tests measured something independent from what the other tests measured.;In terms of practicality, cloze tests constructed by a rational deletion system are not easier or faster method compared to the every-nth-word deletion system. However, this may be overcome with further study.