A Corpus-based Evaluation of a High-stakes EFL Exam

High-stakes assessments play a significant role in people’s lives, and their results greatly define individuals’ future social and financial prospects. Corpus linguistics has recently been used to inform the development and validation of such tests. This study aimed at identifying the degree of typicality of vocabulary items tested in the English proficiency subtest of the Master of Arts/Science Iranian University Entrance Exam. To this end, the vocabulary options and collocations in ۲۰ test versions were extracted, and their frequency of occurrence in the Corpus of Contemporary American English was examined using a specially written computer program. The results indicated that the frequency of the options in the academic genre was not as dominant as expected in a test designed for academic purposes. The findings also revealed some inconsistencies among the different parallel test versions in terms of their option frequencies. Furthermore, for some options and collocations, atypicality was observed as zero or close to zero instances in the corpus. The current study suggests the inclusion of frequency information from corpora and various wordlists to accompany test developers’ intuition for more robust vocabulary assessment.

کلیدواژه ها:

Corpus linguistics ، High-stakes exam ، Lexical coverage ، The Corpus of Contemporary American English (COCA) ، Vocabulary assessment