This study measures the stability of performance exhibited where different computer programing classes study the same material. By focusing standard measurement techniques on the item difficulty (the proportion of students answering an item correctly), it was determined that up to two-thirds of the reliable variance of a classroom test is held in common with identical tests given in similar classes. The particular wording of the test item measuring a concept was shown to be a critical factor in knowledge assessment. Classes were given identical terms measuring common concepts and changed items measuring a different set of common concepts. The correlations between classes of item difficulties for identical items is approximately .70, whereas the correlation for changed items is approximately .35. Suggestions are made for utilizing the high correlation between identical items in instructional decision making. (Author/SD)