I am interested in finding rules of collective behavior that are both general and correspond closely to data. I will discuss our recent finding that a decision-making theory based on a Bayesian approach with probability matching as a decision rule gives a model valid in the animal collectives tested.