NATIVE AND NONNATIVE SPEAKER PERFORMANCE ON CLOZE TESTS1

Abstract

A frequent assumption of proficiency tests for nonnative speakers of a language is that they measure something that native speakers are able to do uniformly well. The cloze procedure is increasingly used to construct proficiency tests for administration to nonnative speakers of English. However, little evidence is available on the question of whether cloze tests are capable of making fine distinctions between native and nonnative speakers. The paper reports on a study comparing the performances of native speakers of English with those of nonnative speakers, on a number of cloze tests constructed by manipulating variables in cloze test design. It is shown that native and nonnative speakers perform similarly on cloze tests with respect to these variables. The results are related to the claim that cloze tests (as distinct from cloze items) measure essentially lower‐order language skills, and it is suggested that the notion of language proficiency needs to be revised. This has major implications for the construction and validation of language tests.