Breach remedies serve an important role in protecting relationship-specific investments. The theoretical literature predicts that some commonly used types of breach remedies protect too well, in the sense that they induce over-investment. The two driving forces behind this result are the complete insurance against potential separation that breach remedies may provide, and the possibility to prevent breach by increasing the damage payment due through the investment made. The question remains whether these two motives, and thus the derived overinvestment result, indeed show up in practice. In this paper we report on an experiment designed to address this issue. Three different remedies are studied, viz. liquidated damages, expectation damages and reliance damages. In line with theoretical predictions we find that over-investment does not occur under liquidated damages. In case of expectation damages the full insurance motive indeed appears to be operative, but due to fairness considerations it leads to (slightly) less overinvestment than predicted. Reciprocal behavior reduces the working of the breach prevention motive predicted for the reliance damages case. Overall, over-investment indeed occurs, but is less serious than theory predicts.