The same cloze for all occasions?

Abstract

Abstract Target language proficiency assessment has become an integral part of Second Language Acquisition (SLA) research design, with cloze tests frequently serving this purpose for reasons of practicality. Assumptions underlying the interpretation of such cloze test scores, however, are often not examined. With the goal of providing researchers with better means for drawing inferences from cloze test scores, we present an analysis of a combined dataset comprised of scores from 1,724 test takers on a frequently used English cloze test (Brown 1980). We examine variation in score distributions and reliability estimates among L2 groups, between L2 and native-speaker (NS) examinees, and for different scoring methods, and investigate the degree to which different sets of items were effective for classifying low- vs high-proficiency L2 examinees and L2 vs NS test takers. Standardized scores are provided for each scoring method so that future researchers can reference their scores to this larger set.