What does test bias have to do with fairness?

Abstract

This article reports on the issue of bias in school examinations as it affects langu ages other than English (LOTE) learners from different Ll backgrounds. Statisti cal methods for detecting test bias claim to be ethically neutral in that they do no more than demonstrate whether or not there are systematic discrepancies in test performance across groups. They leave unanswered the question of whether the criterion adopted as benchmark for group comparisons is fair. The determination of whether a test, or parts of a test, are biased depends ultimately on how the test construct is defined (and on who is defining it). What may appear to be an instance of bias or distortion in the measurement process may, on closer analysis, turn out to be an indication of real differences in the ability which the test aims to measure. This issue is discussed with reference to a scheme introduced in the Victorian State of Australia to compensate foreign language learners for alleged bias in favour of those with a home background in the target language on the Year 12 Victorian Certificate of Education (VCE) examinations in LOTE.