Disjunction Fallacy

The disjunction fallacy is the error of judging a more specific subset of a category as more probable than the broader, inclusive category that contains it. Formally, if X is a subset of Y, the fallacy occurs when someone claims that X is more likely than Y. This violates the logical rule that a disjunction (X or something else within Y) cannot be less probable than one of its components.

Logical Form:

• Claim X is a subset of Y.
• Claim that X is more probable than Y.

Examples:

1. Mr. Pius wears a crucifix and quotes the Bible. It is concluded that it is more likely Mr. Pius is Catholic than that he is Catholic or Muslim. This is incorrect: while the description increases the plausibility of Catholicism, the broader category “Catholic or Muslim” must be at least as probable.
2. Bill is 6'9" tall and can dunk a basketball with ease. It is concluded that Bill is more likely to be a professional basketball player than a professional basketball player or a jockey. Again, the disjunction includes the subset, so it is at least as probable.

The fallacy often arises because the options are mistakenly treated as mutually exclusive alternatives, or because the description primes the narrower category, making it seem more representative.

Experimental Findings and Robustness

Experiments using both direct probability judgments and willingness-to-bet formulations have found the disjunction fallacy under both framings. When participants are given a personality description (e.g., “Oded”) and asked to rank the likelihood of various academic majors, they frequently rate a specific major (e.g., physics) as more probable than the broader field that contains it (e.g., natural sciences). In the reported problems, fallacy rates average around 64%, and in some cases exceed 80%. Although a betting framing slightly reduces the fallacy in most cases (8 out of 10 in one study), the decrease is small (typically less than 10 percentage points), and the reported rates remain above 50%. Persistence under a betting framing challenges an explanation based solely on conversational implicature or misunderstanding; it does not rule out every contribution from wording.

The Role of Representativeness and Category Level

The reported findings also depend on how representative the categories seem. When both a basic-level category (e.g., “physics”) and its superordinate (e.g., “natural sciences”) are highly representative of the given description, participants often prefer the narrower, more specific category. This mirrors classic findings with the conjunction fallacy, but the effect here depends on the perceived typicality of the categories. For foil pairs that are unrepresentative (e.g., “literature” and “humanities” for the Oded description, which fits a physics student), the pattern reverses: people more often select the broader disjunction, and the fallacy rate drops sharply—often well below 50%, averaging around 35% across studies. In these tasks, apparent typicality helps explain which level of category participants favor.

This pattern aligns with the representativeness heuristic and feature-based similarity accounts. A specific category like “physics” may share more compatible features with a description, increasing its similarity and perceived probability, whereas an incompatible specific category would accumulate more dissimilar features, making it less similar than its broader, more abstract superordinate. Thus, probability judgments track similarity rankings, and the disjunction fallacy appears when a narrow category is both representative and more richly matching than its inclusive alternative.

No Logical Exception

There is no logically valid situation where a subset is more probable than the set that contains it. The disjunction fallacy is always an error of reasoning.