The scientific environment might influence university researchers' job designs. In a principal-agent model, researchers must choose between substitutable tasks, publishing or teaching, according to their individual abilities and the scientific and pedagogical context that exists in their universities. This proposed model shows that scientific production can increase, regardless of researchers' abilities, if the scientific environment favours agglomeration effects. The authors test these predictions using an original data set of French economics professors that reveals their individual investments in both teaching and publishing. The econometric results confirm that the tasks conflict and that the scientific context affects researchers' investments in each task.