Seidel (Germany) will use information about student characteristics in order to determine differences in student engagement in physics instruction. She will present findings of a study in which her research team investigated how student profiles predict girls' and boys' verbal interactions with the teacher. The sample included N = 1378 students from 81 randomly selected high-school physics classrooms. At the beginning of the school year, the following student characteristics were assessed: cognitive abilities, pre-knowledge, self-concept, and interest. Four months later, the classroom discourse was videotaped. Five student profiles identified by Seidel (2006) on the basis of student characteristics were used in the analysis. These profiles were incorporated into a new analysis for the present study. Multilevel analysis indicated the highest amount of verbal interaction for girls with high-level cognitive and motivational-affective characteristics. It is argued that video analysis of teacher-student-interaction should include information about student characteristics in order to determine ways in which students influence teaching and vice versa.