This presentation focuses on research that can inform the improvement of mathematics teaching and learning at scale. In educational contexts, improvement at scale refers to the process of taking an instructional innovation that has proved effective in supporting students’ learning in a small number of classrooms and reproducing that success in a large number of classrooms. I first argue that such research should view mathematics teachers’ instructional practices as situated in the institutional settings of the schools and broader administrative jurisdictions in which they work. I then discuss a series of hypotheses about structures that might support teachers' ongoing improvement of their classroom practices. These support structures range from teacher networks whose activities focus on instructional issues to relations of assistance and accountability between teachers, school leaders, and leaders of broader administrative jurisdictions. In describing support structures, I also attend to equity in students' access to high quality instruction by considering both the tracking or grouping of students in terms of current achievement and the category systems that teachers and administrators use for classifying students.