This study examines the magnitude, destinations, and determinants of the departures of mathematics and science teachers from public schools. The data are from the National Center for Education Statistics' nationally representative Schools and Staffing Survey and its longitudinal supplement, the Teacher Follow-up Survey. Our analyses show that rates of mathematics and science teacher turnover, both those moving between schools and those leaving teaching altogether, have increased over the past two decades, but have not been consistently different than those of non-mathematics/science teachers. Mathematics and science teachers who left teaching were also no more likely than other teachers to take non-education jobs, such as in technological fields, or to be working for private business or industry. The data also show that, like other teachers, there are large school-to-school differences in mathematics and science turnover. High poverty, high minority, and urban public schools have among the highest mathematics and science turnover levels. In the case of cross-school migration, the data show there is an annual asymmetric reshuffling of a significant portion of the mathematics and science teaching force from poor to not poor schools, from high-minority to low-minority schools, and from urban to suburban schools. However, our multivariate analyses showed that a number of key organizational characteristics and conditions of schools accounted for these school differences in turnover. The strongest factors for mathematics teachers were the degree of individual classroom autonomy held by teachers, the provision of useful professional development, and the degree of student discipline problems. For science teachers, the strongest factors were the maximum potential salary offered by school districts, the degree of student discipline problems in schools, and useful professional development. (Contains 3 figures, 8 tables and 8 endnotes.)