Adaptive Learning versus Punishment in Ultimatum Bargaining

Adaptive learning and punishment are highly prominent competing explanations for ultimatum game behavior. We report on an experiment that considers each theory in stand-alone form, so that one does not rely on the other in any substantial way. Our data exhibits patterns for which punishment can account but learning by itself cannot. Initial play varies substantially- and systematically-across variations on the ultimatum game, and this leads to differences in later play as well. Hence a complete theory of ultimatum game behavior will have to predict initial conditions as well as describe the influence of repeated play.