On the Seriousness of Mistakes

Several authors have recently contended that modern statistical theory provides a powerful argument in favor of the view that if scientists accept or reject hypotheses at all they do so only in a behavioral sense—i.e., in a sense which reduces “accepting P ” to “acting on the basis of P relative to an objective O ”. In this paper, the argument from statistics in favor of a behavioral view is outlined; an interpretation of two statistical procedures (Bayes method and signifigance testing) is offered which does not entail a behavioral analysis of “accepting a hypothesis”; and the conclusion that non-behavioral analyses of belief are compatible with the application of current statistical theory in the sciences is tentatively advanced.