Research on assessing the cultural fairness of individual test items is reviewed, with emphasis on Birnbaum's three-parameter logistic model. As defined in this review, differential item characteristics are exhibited when examinees from one group have a lower probability of answering correctly than do examinees of equal ability from another group. An item is assumed to be free of bias if the probability of responding correctly, given total score, is the same for all subpopulations. Other major methods of identifying differentially performing items and their limitations are briefly summarized: analysis of variance; transformed item difficulties; chi-square; and factor analysis. Twelve studies are reviewed. Subjects generally included black and white students; in some cases, hearing impaired students and Canadians were included. Comments by Petersen, Lord, and Ironson are also summarized. It is concluded that the three-parameter model's advantages include the relationship of the probability of a correct response to underlying examinee ability level and item characteristics; the model's ability to take guessing into account; and the fact that item parameters are sample-independent and invariant across groups. The model's limitations are also described; it is often difficult to use and impossible to apply. Sample sizes must be at least 1,000 and there should be at least 40 items to be analyzed. (GDC)