‘The concept of correlation in adults’ comes of age

Abstract

Jan Smedslund’s most cited publication is a landmark study of illusory correlations published in 1963. In two experiments, nurse students received decks of patient cards featuring the presence or absence of a specific disease along with the presence or absence of a specific symptom. Nearly all participants reported that the symptom was associated with the disease, so that the symptom would be useful for diagnostic purposes, although it occurred equally often in patients with and without the disease. Smedslund concluded that lay people’s concept of correlation was severely deficient, as the participants of his studies attended mainly to the present-present cell of a 2 x 2 contingency table. The finding was widely cited by Smedslund’s contemporaries as an instance of human irrationality in lay statistical thinking. Later research has modified these conclusions by showing that perceived correlations are also dependent on expectations, cell frequencies, and the way data are presented to subjects. We find it perhaps ironical that Smedslund, who has later claimed that human rationality is a basic assumption for psychological research, and that fallacies in thinking cannot be empirically established, was among the first to demonstrate a basic shortcoming in people’s ability to perceive statistical independence in a series of observations.