Competition authorities around the world have adopted leniency programs creating incentives for cartel members to come forward and provide information sufficient for cartel prosecution. We conducted a laboratory experiment simulating an infinitely repeated 4-player Bertrand game with homogeneous goods. The experiment allowed us to determine the effect of penalty reduction, extent of penalty, and detection rate on cartel formation and leniency application. The different detection and penalty combinations effectively deterred cartel formation and reduced prices. Although leniency application rates are low, there is indication that full immunity (no penalty) is more attractive than options with reduced penalties.