Linda conjunction fallacy experiment

The Linda conjunction fallacy experiment is Daniel Kahneman and Amos Tversky’s most famous and controversial demonstration of the conflict between representativeness and probability. A richly drawn description makes one outcome feel psychologically right, while the logical structure of the alternatives makes that same outcome necessarily less likely. The experiment showed that even educated participants can neglect a rule they know when an appealing narrative fit captures their judgment.

Design and Characterization

Linda was presented as a thirty-one-year-old, single, outspoken, and very bright woman who had majored in philosophy. As a student, she had been deeply concerned with discrimination and social justice and had participated in antinuclear demonstrations. The description was deliberately vivid enough to evoke a recognizable social and political identity; audiences in the 1980s often laughed because they immediately inferred that Linda had attended the University of California at Berkeley, then famous for radical and politically engaged students.

Participants received a list of possible descriptions of Linda’s life. These included her being an elementary-school teacher, working in a bookstore while taking yoga classes, being active in the feminist movement, working as a psychiatric social worker, belonging to the League of Women Voters, working as a bank teller, working as an insurance salesperson, and being both a bank teller and active in the feminist movement. The list made the age of the example visible, particularly through its reference to the League of Women Voters and to a feminist “movement,” but the underlying judgment problem remained readily recognizable.

Representativeness Versus Probability

The critical comparison was between Linda as a bank teller and Linda as a bank teller active in the feminist movement. Almost everyone judged the second description to be a better fit for the Linda portrait. The added detail made the story more coherent: the stereotype of a bank teller did not naturally include feminist activism, whereas the fuller description seemed to explain the political commitments already attributed to Linda.

Probability, however, imposes an inescapable relation between the two alternatives. The set of feminist bank tellers is wholly contained within the set of bank tellers. Every feminist bank teller is a bank teller, but not every bank teller is a feminist bank teller. Therefore the probability of the conjunction cannot exceed the probability of the broader category. Adding a condition to an event can only reduce, or at most leave unchanged, its probability.

The experiment thus created a sharp tension between two modes of judgment. Representativeness asked which description resembled the portrait most closely; probability asked which event included the larger set of possibilities. Participants treated the more representative account as more likely, allowing narrative coherence to defeat a logical relation that was directly relevant to the question.

Experimental Discovery and Results

The initial experiment used a between-subjects design. Some participants ranked outcomes by resemblance and others by likelihood, with only one of the two critical alternatives included in a given list. The average rankings were nevertheless identical: “feminist bank teller” ranked above “bank teller” both as a resemblance judgment and as a likelihood judgment.

The researchers then used a within-subject questionnaire containing both alternatives. They expected participants to notice their logical relation and assumed that a special test would be unnecessary. An assistant distributed the questionnaire to people signing out of the laboratory before payment. After about ten forms had accumulated on a tray, Kahneman glanced at them and discovered that every respondent had ranked “feminist bank teller” as more probable than “bank teller.” The discovery produced a vivid shock, preserved in his memory through the gray metal desk and the positions of the people around him, and led him to call Tversky immediately. Their experiment had placed logic and representativeness in direct opposition, and representativeness had prevailed.

The result was not confined to inattentive or statistically inexperienced participants. In the extended study, 89 percent of undergraduates violated the probability rule. When the same questionnaire was given to doctoral students in the decision-science program at the Stanford Graduate School of Business, all of whom had taken advanced courses in probability, statistics, and decision theory, 85 percent made the same error. Repeated efforts to remove the difficulty produced the stark version of the problem, in which participants had to choose directly between Linda as a bank teller and Linda as a bank teller active in the feminist movement. At several major universities, approximately 85 to 90 percent of undergraduates selected the conjunction.

The controversy was intensified by the participants’ reactions. When Kahneman asked a large undergraduate class whether they realized that they had violated an elementary logical rule, a student answered, “So what?” A graduate student who had made the same choice explained that the question had seemed to ask for an opinion. These responses revealed that recognizing the authority of logic and experiencing a judgment as an opinion did not guarantee that the relevant rule would guide the decision.

The Conjunction Fallacy and System 1

Kahneman and Tversky introduced the term conjunction fallacy for the error of judging a conjunction of two events to be more probable than one of its constituent events. In the Linda experiment, the conjunction was Linda’s being a bank teller and active in the feminist movement, while the broader event was simply Linda’s being a bank teller.

The error was interpreted as a failure of System 2. Participants had a clear opportunity to detect the probability rule because the two alternatives appeared in the same ranking task, yet they did not use that opportunity. Their judgments were guided instead by the rapid, associative process later called System 1. The portrait activated a compelling image of Linda, and the conjunction matched that image more closely than the bare occupational category did.

The fallacy remained attractive even after it was identified. Stephen Jay Gould described knowing the correct answer while still hearing an insistent inner voice objecting that Linda could not merely be a bank teller and that the description had to be taken seriously. The example makes the point that insight does not automatically dissolve an intuitive impression. Like the Müller-Lyer illusion, the Linda judgment can continue to feel persuasive even when the person understands why it is wrong.

Variations, Correction, and Significance

The wording and arrangement of the task affected performance. In the original eight-outcome version, only 15 percent of a comparable group of graduate students correctly judged the feminist-bank-teller option to be less probable. The two critical alternatives were separated by an insurance-salesperson item, encouraging participants to assess each independently rather than compare them directly. In the shorter version, the alternatives were placed in explicit opposition, which mobilized System 2 and allowed 64 percent of graduate students in the social sciences at Stanford and the University of California at Berkeley to give the logically correct answer. Even then, a substantial minority retained the fallacy.

This contrast showed that the error was sensitive not only to knowledge but also to task structure. Advanced training did not reliably protect participants when the alternatives were encountered as separate, plausible descriptions. Direct comparison made the relevant inclusion relation easier to notice and therefore improved performance. The experiment consequently became evidence both about a cognitive bias and about the conditions under which deliberate reasoning can correct an intuitive response.

The experiment’s wider importance lies in the way it separated psychological plausibility from logical probability. It helped establish that Humans do not consistently behave like the rational agents assumed by economic theory. Their judgments are shaped by the information made salient at the moment, by the ease with which a story can be imagined, and by the coherence of a stereotype. The Linda problem therefore remains a compact demonstration of how a richer description can feel more probable precisely because it is richer, even though each added condition narrows the set of possible cases.