This study examines concurrent teacher-student interaction quality and 5th graders' (n = 387) engagement in mathematics classrooms (n = 63) and considers how teacher-student interaction quality relates to engagement differently for boys and girls. Three approaches were used to measure student engagement in mathematics: Research assistants observed engaged behavior, teachers reported on students' engagement, and students completed questionnaires. Engagement data were conducted 3 times per year concurrent with measures of teacher-student interaction quality. Results showed small but statistically significant associations among the 3 methods. Results of multilevel models showed only 1 significant finding linking quality of teacher-student interactions to observed or teacher-reported behavioral engagement; higher classroom organization related to higher levels of observed behavioral engagement. However, the multilevel models produced a rich set of findings for student-reported engagement. Students in classrooms with higher emotional support reported higher cognitive, emotional, and social engagement. Students in classrooms higher in classroom organization reported more cognitive, emotional, and social engagement. Interaction effects (Gender × Teacher-student interaction quality) were present for student-reported engagement outcomes but not in observed or teacher-reported engagement. Boys (but not girls) in classrooms with higher observed classroom organization reported more cognitive and emotional engagement. In classrooms with higher instructional support, boys reported higher but girls reported lower social engagement. The discussion explores implications of varied approaches to measuring engagement, interprets teacher-student interaction quality and gender findings, and considers the usefulness of student report in understanding students' math experiences.