<p>Strategic delegation models predict that the incentives which principals set for their agents will depend crucially on whether strategies are substitutes or complements. We report a laboratory experiment to test this key prediction. Results show that, as predicted, principals choose competitive incentives for their agents with strategic substitutes, but contrary to prediction, principals do not set cooperative incentives in the game with strategic complements. It turns out that agents behave more cooperatively with strategic complements than predicted. This may explain why principals do not (have to) set cooperative incentives in this case.</p>