This presentation compares the results of a reanalysis of items in one form of a level of the 1978 SRA Achievement Series with the results from the pretest analysis. The bias analyses were performed on white or majority, black, Hispanic, male, and female samples from the pretest and two subsequent samples. Comparisons were made in terms of the consistency of identification of biased items, the direction of bias, and baseline comparisons across samples within a sex or racial/ethnic group. Results show less than expected consistency of identification; effective balancing of items in direction of bias; the importance of sample size in consistency; and inconclusive results in the baseline comparisons. (Author)