Publisher      Journal of Educational Statistics, 9
Page Numbers     93-128
Summary     Researchers have taken three general approaches to research on the question of test bias: (a) predictive validity studies in selection situations; (b) investigation of external biasing factors such as the race of examiner, test-wiseness of examinees, and speed of administration; and (c) construct or content validity studies of the internal structure of the test. The present research is focused on test item-bias methods, which are subsumed in the last category of inquiry. Item-bias methods are statistical procedures intended to test whether items function equivalently in two groups. Therefore, they address the basic validity question: Does the test (or individual items in the test) measure what it purports to measure for both groups?