This paper proposes that instructional validity (i.e., the extent to which teachers actually teach the skills that are tested) should be part of the test development process. The development of a district-wide ninth-grade math exam for General Mathematics I (GMI) in the Charleston County School District, South Carolina, is used to illustrate how a study of the instructional process conducted during the early phases of test development can benefit the curriculum and strengthen the match between instruction and assessment. The GMI curriculum was divided into four nine-week grading periods. Teachers reported on the levels of instruction and student mastery of each stated objective. Reports on the students' progress were submitted quarterly and analyzed with regard to the Rasch model, which was used to generate item difficulties and fit statistics and to link the four sets of test forms. Analysis of the data found much variability in student mastery of course objectives. Comparisons among the four tests showed significant differences which have implications for the interpretation of the correlation coefficients. Topics for further research are outlined. Appended are the objectives covered by the GMI test. (Author/EGS)