A conceptual core model of the teacher stress process, assumed to be invariant, was tested on longitudinal data. A LISREL analysis confirmed the goodness of fit of the model. Teachers' relations to pupils and teachers' perception of pupils as a stress source result in general strain, and in turn lead to various stress reactions. Teachers who perceive students as stressors at the middle of the first term report general strain and impaired somatic and mental health at the end of the school year. The major part of successful coping seems to take place during the first half of the school year; later only minor changes are noticed. It is argued that the short-term perspective on coping should guide further studies on teacher stress.