Learning the probability of success in a simple motor task from sequences of success and failure


Successful decision making depends on accurate knowledge of probabilities of outcomes: should I really try to climb that sheer rock wall? We examine how people learn their probability of success by trial and error in a simple motor task and compare their trial by trial estimates of success to those of an optimal updating model (Bayesian updating). Human performance deviates from optimal but is qualitatively consistent with Bayesian updating.