This paper presents a complex causal model of teacher stress based on data received from the responses of 1,466 teachers from Malmo, Sweden to a questionnaire. Also presented is a method for treating the model variables as higher-order factors or higher-order theoretical constructs. The paper's introduction presents a brief review of teacher stress. The conceptual model is also described. The second section describes data and methods used for creating the model, including population, variables, index construction, higher-order theoretical constructs, and model specification. Results discussed in the third and fourth sections show that: (1) individual personality characteristics, rather than biographical ones, are associated with stress variables; (2) teachers' relationships with pupils are by far the most important source of stress; and (3) social support is ineffective in alleviating reactions to stress exposure. Another conclusion discussed is that a latent variable measuring a syndrome of work dissatisfaction and work overload is very strongly connected to teachers' health as measured by mental disorders and psychosomatic symptoms. Appendices list factor items used in the method. References are included. (JM)