This analysis contributes to efforts to improve the use and understanding of survey data in education policy research by asking: "How different are student and teacher reports of classroom instruction?" "Do student, class, or teacher characteristics account for any of the differences?" Using National Assessment of Education Progress (NAEP) data, we compare the responses of middle-school students and their teachers to the same questions about mathematics instruction. We found low correlations and small significant mean differences between student and teacher reports; we also found that student reports are sensitive to key student and class variables, most notably to individual and class achievement, student race/ethnicity, income, parent education, and motivation. Implications for using student survey data in policy research are discussed. (Contains 3 notes, 18 tables, and 1 figure.)