Principal-agent models in which the agent has access to private information before a contract is signed are a cornerstone of contract theory. We have conducted an experiment with 720 participants to explore whether the theoretical insights are reflected by the behavior of subjects in the laboratory and to what extent deviations from standard theory can be explained by social preferences. Investigating settings with both exogenous and endogenous information structures, we find that agency theory is indeed useful to qualitatively predict how variations in the degree of uncertainty affect subjects’ behavior. Regarding the quantitative deviations from standard predictions, our analysis based on several control treatments and quantal response estimations shows that agents’ behavior can be explained by social preferences that are less pronounced than in conventional ultimatum games. Principals’ own social preferences are not an important determinant of their behavior. However, when the principals make contract offers, they anticipate that social preferences affect agents’ behavior.