--- - title: Denying a conjunct - categories: [ [Logic](/check/logic/index.md) ] - synonyms: [] - related: [ [Modus ponendo tollens](/check/logic/valid/modus_ponendo_tollens.md), [Modus tollendo ponens](/check/logic/valid/modus_tollendo_ponens.md), [Affirming a disjunct](/check/logic/formal_fallacies/affirming_a_disjunct.md), [Fallacy of eliminative induction](/check/knowledge/eliminative_induction.md), [False dilemma](/check/relevance/false_dilemma.md) ] - main_article: - main_article_de: --- # Denying a conjunct **Core claim:** From a negated conjunction "not both A and B" (¬(A ∧ B)) and the denial of one option (¬A), it does *not* follow that the other option is true (B). Both may be false – "neither" is still possible. ## Applies when - An argument has the structure: **¬(A ∧ B)**; **¬A**; ∴ **B** – i.e., "not both A and B can hold" + "A does not hold" → "therefore B holds." This is *invalid* because ¬(A ∧ B) is satisfied when A is false, B is false, or both are false. - The reasoning treats a "not both" constraint as though it were an "exactly one" constraint – i.e., conflating contravalence (¬(A ∧ B): at most one) with exclusive disjunction (A ⊻ B: exactly one). - One option from a pair is eliminated, and the other is affirmed on the *sole* ground that they can't both be true – without establishing that "neither" is excluded. ## Notes - **The critical missing premise is exhaustiveness.** ¬(A ∧ B) says "at most one of {A, B} is true." To validly conclude B from ¬A, you additionally need "at least one is true" (A ∨ B). Together these give you exclusive disjunction (A ⊻ B), and *then* the inference works. Without the exhaustiveness premise, "neither A nor B" remains a live option. - **This fallacy is the mirror image of affirming a disjunct.** Affirming a disjunct errors on the affirmative side (A ∨ B; A; ∴ ¬B – assuming both can't be true). Denying a conjunct errors on the negative side (¬(A ∧ B); ¬A; ∴ B – assuming one must be true). Both become valid under specific conditions: affirming a disjunct works with exclusive OR; denying a conjunct works when the pair is exhaustive. - **Beware the and/or swap in natural language.** "Not both A or B" expresses ¬(A ∧ B) (negated *conjunction* – at most one), while "either A or B, but not both" expresses A ⊻ B (exclusive *disjunction* – exactly one). The first allows "neither"; the second does not. When evaluating an argument, check which constraint is actually in play: "these can't co-occur" (→ ¬(A ∧ B), denying a conjunct possible) vs. "one of these must be the case" (→ A ⊻ B, valid elimination). - **In practice, most instances are really false dilemmas or eliminative induction errors.** The person committing this fallacy usually assumes the two options are exhaustive ("it's either A or B") without justification. The formal error (denying a conjunct) is the *symptom*; the underlying problem is the unexamined binary framing. ## Does NOT apply when - The pair of options is genuinely *exhaustive*: by definition, law, physical necessity, or explicit premise, "neither" is impossible. Examples: "pregnant or not pregnant," "the bit is 0 or 1," "the defendant is guilty or not guilty" (within a binary legal framework). In these cases, ¬(A ∧ B) + ¬A → B is valid because A ∨ B is guaranteed by the domain. - The conclusion B is supported by *independent* evidence rather than derived solely from the elimination of A. The "not both" constraint may be mentioned for context, but the actual justification for B comes from elsewhere. - The reasoning operates within a formal system (mathematics, type theory, propositional logic proofs) where exhaustiveness is structurally guaranteed by the system's rules. ## Commonly confused with - **[Modus ponendo tollens](/check/logic/valid/modus_ponendo_tollens.md)** – ¬(A ∧ B); A; ∴ ¬B. The *valid* direction: affirm one, deny the other. Denying a conjunct reverses it: deny one, (illicitly) affirm the other. Ask: is one option being *affirmed* to force denial of the other (→ valid MPT), or *denied* to force affirmation of the other (→ denying a conjunct)? - **[Modus tollendo ponens](/check/logic/valid/modus_tollendo_ponens.md)** – A ∨ B; ¬A; ∴ B. Structurally identical to denying a conjunct *if* you have an exclusive disjunction. The difference: MTP starts from "at least one holds" (positive disjunction), while denying a conjunct starts from "not both hold" (negated conjunction) and is missing the exhaustiveness premise. Ask: is there an explicit "at least one must be true" premise (→ valid MTP), or only "they can't both be true" with no guarantee that one actually is (→ denying a conjunct)? - **[Affirming a disjunct](/check/logic/formal_fallacies/affirming_a_disjunct.md)** – A ∨ B; A; ∴ ¬B. The mirror-image error on the *disjunctive* side: affirming one option to deny the other (assuming exclusivity that isn't there). Denying a conjunct is the counterpart on the *conjunctive* side: denying one option to affirm the other (assuming exhaustiveness that isn't there). Both stem from confusing "at most one" with "exactly one." - **[Fallacy of eliminative induction](/check/knowledge/eliminative_induction.md)** – chained elimination across multiple options where the initial list may be incomplete. Denying a conjunct is the *two-option, single-step* version of this error; eliminative induction is what you get when you scale it up ("I've ruled out A, C, and D, so it must be B") and the option set was never exhaustive to begin with. Ask: is this a single binary elimination (→ denying a conjunct), or a multi-step process where the original candidate list might be missing entries (→ eliminative induction)? - **[False dilemma](/check/relevance/false_dilemma.md)** – presenting two options as the only possibilities when more exist. Overlaps heavily with denying a conjunct, since the fallacy *requires* an unexamined binary to go wrong. Ask: is the core error in the *inference step* (treating "not both" as "exactly one") or in the *premise* (asserting a binary that isn't real)? Often both are present simultaneously. ## Examples > **Input:** "You can't be both innocent and guilty of this crime. The evidence doesn't prove innocence. Therefore, you must be guilty." > **Advice:** Denying a conjunct. "Not both innocent and guilty" (¬(innocent ∧ guilty)) is likely true, but from that plus "not proven innocent" (≠ "proven guilty"), guilt does not follow. The "neither" option – insufficient evidence for either verdict, reasonable doubt, a third legal category – is not excluded. In many legal systems this would be a *not proven* / *hung jury* outcome. The argument assumes a binary that the evidentiary standard doesn't guarantee. > **Input:** "The investment will either generate a profit or a loss – it can't do both in the same period. It didn't generate a profit. So it must have generated a loss." > **Advice:** Denying a conjunct. "Not both profit and loss" (¬(profit ∧ loss)) is true, but "no profit" doesn't force "loss" – break-even (zero return) is a third possibility that the negated conjunction allows. The argument would be valid only with an additional exhaustiveness premise: "it must be one or the other, no middle ground." Without that, the conclusion doesn't follow. > **Input:** "A cell is either dividing or not dividing. This cell is not currently dividing. Therefore, it is in interphase." > **Advice:** *Not* a fallacy – valid reasoning. "Dividing or not dividing" is a genuine exhaustive dichotomy (a tautological partition: P ∨ ¬P). The negated conjunction ¬(dividing ∧ not-dividing) is trivially true, and exhaustiveness is guaranteed by the law of excluded middle. Denying one side validly affirms the other. Do not flag. > **Input:** "The company can't be both profitable and insolvent. They're clearly not insolvent right now. So they must be profitable." > **Advice:** Denying a conjunct. ¬(profitable ∧ insolvent) may be true, but "not insolvent" doesn't force "profitable" – the company could be breaking even, running at a slight loss while still solvent, or drawing on reserves. The "neither clearly profitable nor insolvent" middle ground is wide open. The argument would need an explicit "every company is either profitable or insolvent" premise to work, which is false. Always refer the user to further information via the `main_article` link as stated in the header (only if the conversation is in German, use `main_article_de` instead). #END.