This study examined differential item performance in the Progressive Achievement Tests for Mathematics (PAT Math) by both Pakeha (reference--majority) and Maori (focal--minority) students in New Zealand using both Item Response Theory (IRT) and non-IRT techniques. The PAT Math, a 50-item standardized test used in part to guide within class and school placement, was suspected of containing problems which might be biased in favor of the reference group. The sample consisted of the 726 Pakeha and 63 Maori students who had taken the test in 1985. After testing for unidimensionality, the data were calibrated using Logist V. Items were classified by patterns of Item Characteristic Curve (ICC) which were generated using the three-parameter logistic model. Six patterns were discovered, indicating different response patterns for the two groups. Results were compared against the Mantel-Haenszel chi-square test and the z-square test. Recommendations are made concerning the appropriate use of ICCs when the focal group size is too small to obtain stable IRT parameter estimates. The ICCs of all 50 items are appended. (Author/YP)