Investigations into item bias provide an empirical basis for the identification and elimination of test items which appear to measure different traits across populations or cultural groups. The Psychometric rationales for six approaches to the identification of biased test items are reviewed: (1) Transformed item difficulties: within-group p-values are standardized and compared between groups. (2) Analysis of variance: bias is operationally defined in terms of significant item by group interaction effects. (3) Chi-square: individual items are investigated in terms of between group score level differences in expected and observed proportions of correct responses. (4) Item characteristic curve theory: differences in the probabilities of a correct response, given examinees of the same underlying ability and different culture groups, are evaluated. (5) Factor analytic: item bias is investigated in terms of culture specific and culture common sources of variance, or in terms of loadings on a biased test factor. (6) Distractor response analysis: the relative attractiveness of item foils, or response sets, is investigated. The limitations and advantages of each approach in terms of the underlying assumptions, psychometric soundness, conceptual complexity, applicability to criterion referenced tests and applicability to interdependent groups are discussed. (Author/MV)