This study, one of a series designed to answer practical questions about performance based assessment, examined the comparability of school scores on short, nonparallel test forms. The data were obtained from mathematics tests with both multiple choice (MC) and performance assessment (PA) items. The tests were administered in a statewide testing program in grades 4 and 8, with three test forms administered in each grade. Scores based on the PA items in the three forms were equated, and the comparability of these scores at the school level was investigated. The most important practical implication of the study's findings is that school-by-form interactions should be taken into account if the school scores by form are to be used. Findings indicate that even when forms are equated, if they are short and not parallel, school scores can vary substantially across the forms. Making forms parallel and long would reduce this problem considerably. Other practical implications are discussed, including the conclusion that any estimate of possible variation in school scores is better than no estimate at all. An appendix shows the calculation of statistics used in the analyses. (Contains 18 tables, 17 figures, and 12 references.) (SLD)