Studies carried out over a 12-year period addressed fundamental questions on the use of Rasch-based item banks. Large field tests administered in grades 3-8 of reading, mathematics, and science items, as well as standardized test results were used to explore the possible effects of many factors on item calibrations. In general, the results confirmed the practical value of Rasch-scaled items. For example, to investigate whether item calibrations were the same for different groups, reading test results from 1,400 fourth graders were calibrated separately for boys and girls, above and below average achievement levels, and inner city versus economically advantaged schools. Item calibrations for each subgroup were strongly correlated (.93-.99) with calibrations for the total group. Two other major empirical issues were the extent of test linkage to a single curriculum scale and the comparability of scores from different tests. Other questions included: (1) whether item bank calibrations can be used to scale new tests; (2) whether item measuring a different latent trait can be detected (by examining point biserial correlations and item curves); and (3) whether item calibrations are appreciably affected by test range, item discrimination, sample size, grade level, and time of year. (LPG)