A Refutation of the Neyman-Pearson Theory of Testing

No other part of orthodox Anglo-American statistical theory has been the object of more criticism during the past decade than the Neyman-Pearson theory of testing statistical hypotheses (henceforth labelled by 'NPT'). Much of the criticism is concerned with the abuses of the theory-or rather, with the hybrid versions of it in current use-in the social and biomedical sciences, and is thus not really directed at the theory itself. The more theoretical attacks do not qualify as refutations because they are based on views of statistical inference that are fundamentally alien to NPT.1 Of course, this is not to say that they are of no value or that they do not raise serious difficulties for the theory.2 In particular, Hacking's Received 8 November 1972