The effectiveness of an optimal item selection method in designing parallel test forms was studied during the development of two forms that were parallel to an existing form for each of three language arts tests for fourth graders used in the Connecticut Mastery Test. Two listening comprehension forms, two reading comprehension forms, and two written communication forms were developed using item-response-theory-based Optimal Test Design software. The tests included mixed item formats, cluster-based sections, and passage-related items. Three item banks were available for the study. For each of the 1993 forms, test and cluster information curves were computed. Restrictions on the number of passages that could be accessed for any test meant that relatively few items were available, but weakly parallel tests, as indicated by their relative information curves, were assembled using the software. The parallelism of the tests was also portrayed by similarities among the total score distribution indices and reliabilities of the tests. The small differences in the test information between the target and new forms were not significant threats to the parallelism of the forms. (Contains 2 tables, 4 figures, and 20 references.) (SLD)