Toward a Statistical Theory of Reasoning

Abstract

It is argued that the observed variability of responding in reasoning tasks might usefully be described by mathematical models based on stochastic processes. The data of a number of experiments employing Wason's selection task are reanalysed and it is shown that selection probabilities of individual cards are statistically independent. This is consistent with a class of simple stochastic models and renders conventional “insight” explanations of the data unparsi-monious. A provisional stochastic model is formulated and subjected to a limited parametric test with reasonably satisfactory results. Some general directions for future research along these lines are suggested.