We study individual behaviour in a repeated linear public good experiment in which, in each period, subjects are required to contribute a minimum level and face a certain probability to be audited. Audited subjects who contribute less than the minimum level are convicted to pay the difference...