Classroom tests developed by seventh- and eighth-grade science teachers (n=23) and mathematics teachers (n=18) were analyzed by panels of content and measurement experts. The 41 participating teachers, each of whom contributed 2 tests, completed a questionnaire, an interview, and 2 measures of competence in testing. Teachers used all major item formats in their classroom tests. Science teachers favored multiple-choice items and mathematics teachers favored computation items. Faults were found in 35 percent of completion items and 20 percent of multiple-choice items on teachers' tests. Average test quality on 6 dimensions was rated 5.0 to 5.7 on 7-point semantic differential scales. Test quality was best predicted by scores on a multiple-choice measurement competency test. The sample of classroom tests is described, evaluated, and then related to teachers' training and experience, knowledge of testing, and content of test use to learn more about this pervasive, crucial, and understudied type of testing. Three tables and one figure illustrate study findings. (SLD)