This paper reports on a laboratory experiment which investigates the impact of institutions and institutional choice in constant-sum sender-receiver games. We compare individual sender and receiver behavior in two different institutions: A sanction-free institution which is given by the bare sender-receiver game and a sanctioning institution which in addition offers the receiver the opportunity to (costly) sanction the sender after receiving feedback on the senders private information. We conduct the experiment in two phases: First, individuals are randomly assigned to an institution, and second they can choose the institution themselves.We find that sanctioning takes place predominantly after the receiver has trusted a lie by the sender. Those who are responsible for sanctioning are also responsible for truth-telling in excess with respect to models of rational payoff-maximizing agents.Thereby, the sanctioning institution exhibits more truth-telling. Most importantly, agents who sanction reveal preference for the sanctioning institution while the other subjects almost exclusively opt for the sanction-free institution. As a consequence, both institutions typically coexist in the second phase of the experiment and the sanctioning institution exhibits a higher level of truth-telling and lower aggregate material payoffs.To offer an explanation of our experimental findings, we formalize preferences for truth-telling as psychological payoffs and analyze the sender-receiver game as a dynamic psychological game à la Battigalli and Dufwenberg (2006). We demonstrate that standard models of social preferences are not able to explain observed sanctioning behavior and excessive truth-telling. Explicit psychological costs of lying and the exposition to a lie, however, are able to fill this gap. To this end, we model deontological and consequentialistic preferences for truth-telling and evaluate their respective explanatory power.