Disproof by Counterexample

For centuries Europeans were certain that "all swans are white" — until explorers reached Australia and found a black one. That single bird sank the rule for good. It is the same move that topples a confident "my app works for everyone": one crashing user is all it takes. In mathematics this weapon has a name.

Suppose someone announces a bold rule: "every odd number is prime". They check 3 — prime. 5 — prime. 7 — prime. Three for three! The rule looks unstoppable. And yet it takes just one quiet number, 9 = 3 \times 3, to bring the whole thing crashing down. Nine is odd, but not prime — so the rule is false, full stop.

That single demolishing case is a counterexample, and it is one of the sharpest tools in all of mathematics. Many statements begin with "for all…" — they claim something holds in every case. To disprove such a statement you never have to wrestle with every case. You hunt for one specific case where it fails, hold it up, and you are done. No matter how many examples supported the claim, a single genuine counterexample wins.

The method

A "for all" statement has the shape "for every x, P(x) is true". To knock it down:

find one specific x for which P(x) is false;
check that your x really does satisfy the hypothesis (it must be a case the claim actually talks about);
check that P(x) really is false for it — an honest, verifiable failure, not a near miss.

That is the entire job. One counterexample is enough — you do not need two, or ten, or "most" cases. And there is a clever place to look: the failure often hides at a small or edge case — 0, 1, 2, a negative number, or a fraction — precisely the cases people forget to try.

A universal ("for all") claim is disproved by a single counterexample — one case where it fails.
Once you have found one counterexample you are done — there is no need to check any more.
You cannot prove a "for all" statement by listing examples, no matter how many — examples can only disprove it.
A good counterexample is often a small or edge case: 0, 1, 2, a negative number, or a fraction.

Worked example 1 — "all prime numbers are odd"

The claim: every prime number is odd. Nearly all of them are — 3, 5, 7, 11, 13, \dots go on forever, all odd. But scan back to the very first prime and there it sits: 2. It is prime (its only factors are 1 and itself), and it is even.

Counterexample: 2. It satisfies the hypothesis (it is prime) and breaks the conclusion (it is not odd). One number, and "all primes are odd" is dead. This is the classic case where the counterexample lives right at the edge — the smallest prime of all.

Worked example 2 — a formula that fools you

Euler noticed something delightful: the expression

n^2 - n + 41

seems to spit out primes forever. Try it: n=1 gives 41 (prime), n=2 gives 43 (prime), n=3 gives 47 (prime)… and it keeps producing primes all the way up to n=40. Forty straight successes! Surely "n^2-n+41 is always prime" must be true?

No — and you can spot the crack with a little algebra rather than luck. Put n = 41:

41^2 - 41 + 41 = 41^2 = 41 \times 41.

That is 1681, and it is plainly divisible by 41, so it is not prime. Counterexample: n = 41. Forty confirmations counted for nothing against one failure.

Worked example 3 — a plausible algebra claim

The claim: "squaring a number always makes it bigger", i.e. x^2 > x for all x. Test a few whole numbers and it looks solid: 3^2 = 9 > 3, 10^2 = 100 > 10.

But squaring does not always grow a number — head for the fractions, the cases people skip. Take x = \tfrac12:

\left(\tfrac12\right)^2 = \tfrac14, \qquad \tfrac14 < \tfrac12.

Squaring made it smaller. Counterexample: x=\tfrac12. (And x = 1 also breaks it, since 1^2 = 1 is not greater than 1 — another edge case worth remembering.)

Disproving and proving are not mirror images — they are wildly lopsided, and this trips up almost everyone:

One counterexample refutes; a million examples never confirm. A single failing case disproves a "for all" claim outright. But no pile of supporting examples — ten, a thousand, a billion — can ever prove a "for all" claim. Confirming examples only build suspicion, never certainty; proving "for all" needs a general argument that covers every case at once.
Your counterexample must genuinely satisfy the hypothesis. To disprove "all primes are odd" you must offer a real prime that is even (that's 2). Offering 4 — even, but not prime — proves nothing, because 4 was never a case the claim was about.
The failure must be real and checkable, not a "roughly" or an "almost". If you claim x is a counterexample, the reader must be able to plug it in and watch the statement fail with their own eyes.

Yes — spectacularly. In 1769 Euler conjectured that you can never write a fourth power as the sum of three fourth powers (a cousin of Fermat's famous result). Mathematicians believed it for two hundred years. Then in 1966 a computer search coughed up a counterexample:

27^5 + 84^5 + 110^5 + 133^5 = 144^5,

toppling the five-power version of the same conjecture in a single line. Other patterns are even sneakier: some are true for the first billion whole numbers and only fail afterwards — the first counterexample to one notorious conjecture is a number with hundreds of digits, far beyond anything a person could stumble on by hand.

This is exactly why mathematicians insist on proof rather than evidence. A run of a billion successes is not a theorem; it's just a billion cases where nature has not yet sprung its trap. The counterexample may be lurking just past the last place anyone has checked — which is why you will soon meet proof by contradiction and other ways to nail a claim down for every case at once.