Six approaches to assessing test item bias were examined: transformed item difficulty, point biserial correlations, chi-square, factor analysis, one parameter item characteristic curve, and three parameter item characteristic curve. Data sets for analysis were generated by a Monte Carlo technique based on the three parameter model; thus, four parameters were controlled: total score distributions, item difficulties, item discriminations, and guessing. Only the difficulty parameter was biased. Results indicated that transformed item difficulty had highest correlations with generated bias. The three-parameter item characteristic curve and the one parameter item characteristic curve were next highest, followed by factor analysis, then chi-square. Point biserial correlation functioned erratically. Results of the analysis were compared and recommendations were made on the use of each method. (Author/MH)