Recent studies have observed connections among teachers' depressive symptoms and student outcomes; however, the specific mechanisms through which teachers' mental health characteristics operate in the classroom remain largely unknown. The present study used student-level observation methods to examine the relations between third-grade teachers' (N = 32) depressive symptoms and their academic feedback to students (N = 310) and sought to make inferences about how these factors might influence students' mathematics achievement. A novel observational tool, the Teacher Feedback Coding System-Academic (TFCS-A), was used that assesses feedback across 2 dimensions--teacher affect and instructional strategy, which have been shown to be important to student learning. Multilevel exploratory factor analysis of TFCS-A data suggested 2 primary factors: positive feedback and neutral/negative feedback. Hierarchical linear modeling revealed that positive feedback was related to higher math achievement among students who began the year with weaker math skills and that teachers who reported more depressive symptoms less frequently provided this positive feedback. Results offer new information about a type of instruction that may be affected by teachers' depressive symptoms and inform efforts aimed at improving teachers' instructional interactions with students.