Another look at Linda

Abstract

Margolis (1987) argues that poor performance on the classic Linda conjunction problem is the result of misinterpretation created by multiple ambiguities in the problem statement acting in concert. He suggests three wording changes to rectify this situation: The addition of (1) a betting scenario to prevent the misinterpretation of “probable” as “believable” or “plausible,” (2) a clarifying phrase to prevent misinterpretation of the response alternatives as mutually exclusive, and (3) survey information about the likelihood of personality profiles such as Linda’s to ensure that subjects view the small difference between the extremely small probabilities to be judged as relevant. The effects of these three factors were examined in a between-subjects factorial design. Margolis’s hypothesis was not supported. Only the addition of the response-alternative clarifying phrase significantly improved performance. The results are discussed in terms of Gigerenzer’s (1991) normative-issues argument that the Linda problem is not a frequency probability problem but rather a single-trial, subjective probability problem. © 1993, Psychonomic Society, Inc.. All rights reserved.