Statistical judgement: A further test of the representativeness construct

Kahneman and Tversky's (1972) construct of ‘representativeness’ as a mediator of statistical judgements has been subjected to several recent criticisms. The present study allows test of the hypothesis that individuals differ in their use of the representativeness heuristic. Tasks are devised which permit measure of two types of judgemental error attributed to use of the heuristic. One is a misperception of Bernouilli sequences, in which subjects expect sequences to have shorter run lengths than would actually occur by chance. The other error is a tendency to give insufficient weighting to the evidence of larger samples. The tasks were administered in two experiments by on-line computer. Characteristic errors of both sorts are found in two experiments, but in neither case was there any evidence of a correlation between them. It appears, then, that even a modified form of the representativeness hypothesis, that might be reconciled with other critiques, is lacking in empirical support.