Conversational Implicature, Conscious Representation, and the Conjunction Fallacy

• Published in: Social cognition Vol. 9; no. 1; pp. 85 - 110
• Main Authors: Dulany, Don E., Hilton, Denis J.
• Format: Journal Article
• Language: English
• Published: New York, NY Guilford 01.03.1991; Guilford Press
• Subjects: Behavior. Attitude; Biological and medical sciences; Discourse analysis; Fundamental and applied biological sciences. Psychology; Implicature; Interpersonal Behavior; Psychology. Psychoanalysis. Psychiatry; Psychology. Psychophysiology; Social psychology; Human; Social perception; Probability; Conscienciousness; Representation; Conversation; Social cognition; Judgment
• ISSN: 0278-016X; 1943-2798
• DOI: 10.1521/soco.1991.9.1.85

This study examined judgments in four of Tversky and Kahneman's (1983) conjunction tasks, applying Gricean principles of conversational implicature and an analysis of the subjects' conscious representations. Conversational inference is itself a form of judgment under uncertainty, and hearers often venture interpretations of a speaker's intention, constrained by assumptions embodying rules of conversation. For a conjunction effect to be a fallacy, we argue, subjects must interpret the key conjunct extensionally; fallacious reasoning consists of deficient mental operations on one's own mental contents. We therefore assessed interpretations of the conjunct with reports or induced them with elaborative information, distinguishing extensional interpretations from those that absolve the judgment of fallacy. In Experiment 1, subjects most often formed absolving interpretations of the conjunct where they were most likely to judge the conjunction more probable than the conjunct. In Experiment 2, the conjunct was most often given the absolving interpretation that evidence was insufficient for saying more. Experiment 3 elaborated these results with experimentally induced interpretations. In Experiment 4, reports of representativeness were strongly related to conjunction effects and to reports of nonextensional interpretations, and there was no evidence of conjunction fallacies. Over these experiments, we estimate the incidence of genuine conjunction fallacies as between 0% and 38%, in contrast with the 85% and 90% Tversky and Kahneman reported. In our interpretation, the representativeness heuristic and conversational assumptions together work predominantly to frame conscious representations of the conjunct that absolve the conjunction effect of fallacy.