Karasek (1979) proposed that job demands and decision latitude interact to cause psychological strain. Main effects of these job variables on strain have been often found, but the predicted interaction between them has been less consistently demonstrated. We argue that this lack of support results from inadequate specification and operationalization of the independent variables. In particular, most empirical tests of the interaction have been based on a general measure of decision latitude which encompasses a wide range of job properties including control, task variety and learning opportunities. Yet, theoretically, it is control which is the crucial factor. In this study we used more focused measures of demands and control, with a sample of 1451 manufacturing employees, and found clear evidence of the predicted interaction effect. Parallel analyses using a measure of decision latitude rather than of job control did not show an equivalent effect. It is recommended that greater attention be paid to congruence between theory and measurement.