Once you learn proof by contradiction and see it work on \sqrt{2}, the method feels almost magical — so flexible, so general. A natural follow-up question: is this the universal tool? Can every theorem be proved by contradiction?

The technical answer is yes — in classical mathematics, any statement that can be proved at all can also be proved by contradiction. The practical answer is no — most theorems should not be proved by contradiction, because the contradiction wrapper is pure overhead. Here is the distinction sharpened, with examples of both cases.

The technical "yes"

Suppose a theorem P has a direct proof. That means there is a chain of logical steps — using axioms, definitions, and earlier theorems — that arrives at P as its conclusion. Here is how to convert any such direct proof into a proof by contradiction:

  1. Assume, for the sake of contradiction, \lnot P.
  2. Ignore this assumption and run the original direct proof from start to finish. You arrive at P.
  3. You now have both P (derived) and \lnot P (assumed). Contradiction.
  4. Therefore \lnot P is false, so P is true. \square

That is all. The wrapper adds three lines and one variable (the assumption) but changes nothing substantive. So yes — contradiction is universally applicable.

Why this trick always works: a contradiction proof only needs a single inconsistency anywhere in the derivation-plus-assumption system. Since the direct proof establishes P unconditionally, the \lnot P assumption instantly clashes with P the moment the direct proof is done. The trick does not require the assumption to "do work" — it just needs to be present when the direct derivation finishes.

The trick reveals that proof by contradiction is at least as powerful as direct proof. It also reveals that using contradiction on statements that have clean direct proofs is silly — you are wrapping a complete proof in a pointless "assume not-P... therefore P" sleeve.

The practical "no": when the wrapper is dead weight

For most theorems, proof by contradiction is an inferior format even though it is technically valid. Consider:

Claim. For every integer n, if n is even then n + 2 is even.

Direct proof. Let n be an integer and suppose n is even. Then n = 2k for some integer k. So n + 2 = 2k + 2 = 2(k + 1), which is 2 times an integer. Therefore n + 2 is even. \square

Contradiction proof (pointless). Suppose, for contradiction, that there is an even integer n such that n + 2 is odd. But if n = 2k, then n + 2 = 2(k + 1), which is even, contradicting "n + 2 is odd." Therefore no such n exists, so for every even n, n + 2 is even. \square

The contradiction version is longer without being stronger. It takes a clean forward derivation and dresses it up in a "suppose not... then observe yes" loop. For this kind of theorem — positive, universal, with an obvious constructive path — direct proof is strictly better.

When contradiction genuinely helps

Contradiction earns its keep in three settings.

1. Negative statements

"There is no largest integer." "\sqrt{2} is not rational." "No polynomial has exactly two distinct derivatives." Claims of the form "there does not exist..." or "no x has property P" have nothing positive to build forward to. A direct proof would need to check "every x fails P," which is infinitely many cases. Contradiction converts "no x has P" into "suppose some x has P, follow consequences, reach collision" — which gives you a concrete x to manipulate algebraically.

2. Uniqueness claims

"The identity element of a group is unique." "A continuous function attains its maximum on a compact interval at a unique point (under strict convexity)." Uniqueness says "there is only one x such that..." Assume two — x_1 and x_2 — and derive that they must be equal. The assumption gives you two handles you can compare. Without contradiction, you would need to invent some other way to rule out multiplicity.

3. Statements where the direct path has no first step

"\sqrt{2} is irrational." As the main article emphasises, the direct path from "\sqrt{2} exists" to "\sqrt{2} is not a ratio" is blocked — there is no obvious algebraic bridge. Contradiction gives you the bridge: assume \sqrt{2} = p/q, and now you have an equation you can operate on. Without the assumption, you have nothing to multiply, nothing to substitute, nothing to factor.

When proof by contradiction is the right toolA two-column table. The left column, headed Contradiction helps, lists three types of claims: negative existence, uniqueness, and direct-path blocked. The right column, headed Contradiction is overkill, lists three types: positive universals with algebra, forward implications, and constructive existence. Each cell has a one-line illustrative example. Contradiction helps 1. Negative existence "No x satisfies P(x)" e.g. √2 irrational 2. Uniqueness "Only one x satisfies P" e.g. unique identity 3. Direct path blocked no forward bridge available e.g. infinite primes Contradiction is overkill 1. Positive universals with clean algebra e.g. (a+b)² = a²+2ab+b² 2. Forward implications "If P then Q" with chain e.g. n even ⇒ n² even 3. Constructive existence you can exhibit the witness e.g. √9 = 3, done
Contradiction earns its place on claims of types 1, 2, 3 in the left column — where the direct path is blocked or the claim itself is negative. It is available but overkill on the right column, where a direct proof is shorter, cleaner, and more illuminating. Using contradiction where direct works is legal but wasteful.

What style guides actually recommend

If you open a careful proof-writing textbook (Velleman, Hammack, Polya, or the classic How to Prove It tradition), the guidance is consistent:

The habit to build: when you read a theorem, ask "can I attack this head-on?" If yes, direct. If no, ask "is the conclusion negative or uniqueness-shaped?" If yes, contradiction. If the conclusion is an implication, try contrapositive.

Two claims, two proofs, different formats

Claim A. The sum of two rational numbers is rational.

Direct proof. Let r, s be rational. Then r = a/b and s = c/d for some integers a, b, c, d with b, d \ne 0. Then r + s = (ad + bc) / (bd), where ad + bc is an integer (integers are closed under addition and multiplication) and bd \ne 0 (product of nonzero integers is nonzero). So r + s is a ratio of integers with nonzero denominator — a rational. \square

This is a positive, constructive claim: you can exhibit the resulting fraction. Direct proof works perfectly. Using contradiction — "suppose r + s were irrational" — would add nothing because the same algebra would appear inside the wrapper.

Claim B. \sqrt{2} + \sqrt{3} is irrational.

Contradiction proof. Suppose \sqrt{2} + \sqrt{3} = p/q for rationals p/q. Then \sqrt{3} = p/q - \sqrt{2}. Square both sides: 3 = p^2/q^2 - 2(p/q)\sqrt{2} + 2, giving \sqrt{2} = (p^2/q^2 - 1) / (2p/q) = (p^2 - q^2) / (2pq). The right side is a ratio of integers, so \sqrt{2} is rational — contradicting the proved irrationality of \sqrt{2}. \square

Direct proof would require producing an infinite non-repeating decimal for \sqrt{2} + \sqrt{3} — impossible to do explicitly. Contradiction converts the negative claim (not rational) into an existence claim (suppose it is rational), extracts an equation, and rides the algebra to a collision with the known irrationality of \sqrt{2}.

The two proofs illustrate the rule: direct for positive/constructive, contradiction for negative/structural.

So — can every theorem be proved by contradiction? In classical mathematics, technically yes. But "can be" is not "should be." A good proof is not merely valid; it is illuminating. Contradiction is the right tool when the direct path is blocked; it is the wrong tool when it just adds ceremony around a direct argument you already have. Use it deliberately, not reflexively.

Related: Proof by Contradiction · Which Proof Technique? A Flowchart · Clean Hypothesis? Try Direct First — The Rule of Thumb · Proof by Contradiction vs Proof by Contrapositive — Not Interchangeable · When to Reach for Contradiction — the Decision Rule