Data are summarized in Scheuneman's Score x Group x Response frequency table in order to investigate item bias. The data can arise from two different sampling models: (1) multinomial sampling in which a fixed sample size is used and the responses are cross-classified according to score, group, and response; and (2) product-multinomial sampling in which for each group a fixed sample size is used and the responses are cross-classified according to score and response; Data for both sampling models were analyzed using two logit models, i.e. special cases of log linear models, and results of the procedure were applied to Scheuneman's data using the program ECTA. The item was uniformly biased as shown by whites performing better than blacks in all score categories. Using a frequency table derived from Table 2 of Perline, Wright, and Wainer's nine-item scale for parole decision data, the linear logit model and the Rasch Model were found to be equivalent. Consequently, the estimates for the parameters in the log linear model yield unconditional maximum likelihood estimates for the parameters in the Rasch Model. (RL)