An experimental analysis of voluntary, binary contributions for step-level public goods/bads is presented. Alternatively, the situation is presented as the provision of a public good or the prevention of a public bad. From a strategic point of view these presentations are equivalent. In early periods of the 20 round experiments, behavior is indeed observed to be similar in both cases, but after about 5 periods differences start to occur, that grow larger. A simple learning model is developed that replicates the patterns in the experiments. Extrapolation beyond 20 periods show that the pattern observed reflects an equilibrium selection.