Mathematics & Logic · Logic & proof

Why examples are not a proof

Watch a formula produce 40 primes in a row and then fail, then learn to prove a claim for every case by direct argument, by induction and by the contrapositive.

4 short lessons · about 17 minutes · 3 sources · updated November 2026

What you will learn

  • Explain why several successful examples don't establish a universal claim.
  • Use one valid counterexample to disprove a universal claim.
  • Explain a direct proof that the sum of two even integers is even.
  • Tell arbitrary allowed values apart from specially chosen examples.

Lesson 1 of 4 · 4 min

The formula that gives 40 primes, then fails

Before you readIn 1742 Christian Goldbach guessed that every even number greater than 2 is the sum of two primes. Computers have confirmed it for every even number below 4 quintillion. Where does it stand today?

Correct.Not quite. The answer is It is still unproved.

It is still unproved, even after that many checks. This lesson shows why testing cases, however many, is a different job from proving a claim.

Try this number recipe. Pick a whole number from zero up. Square it, add the number itself, then add 41. Zero gives 41. One gives 43, and 2 gives 47.

All three results are prime. A prime is a whole number greater than 1 that only 1 and itself divide evenly. Keep going and the primes keep coming. Leonhard Euler described this recipe in 1772, and it gives a prime for every input from zero to 39. That's 40 primes in a row.

So it's tempting to say the recipe always gives a prime. Does it?

That guess is a claim about every whole number from zero up, and they never run out. Forty passes leave infinitely many inputs untested. Any one of them could still break the pattern.

Mathematicians hold themselves to this. In 1742 Christian Goldbach wondered whether every even number greater than 2 is the sum of two primes. Computers have confirmed it for every even number below 4 quintillion. It still isn't proved.

A proof needs an argument that covers every allowed input. Testing cases, however many, is a different job.

Check yourselfThe recipe n squared plus n plus 41 gives a prime for every input from zero to 39. What do those 40 results establish?

Correct.Not quite. The answer is Only that those 40 inputs give primes.

The calculations settle the 40 cases you checked. They can't rule out a failure among the inputs you haven't checked, and a single failure would disprove the claim that the recipe always gives a prime.

Now try 40. Its square is 1,600. Add 40 and 41, and you get 1,681. That equals 41 times 41.

You can see why without a calculator. Forty squared plus 40 is 40 lots of 41. Adding one more 41 makes 41 lots of 41.

So 1,681 has a divisor other than 1 and itself, which means it isn't prime. This one counterexample disproves the claim that every input gives a prime. The claim promised no exceptions, so 40 earlier successes can't rescue it.

A counterexample has to pass two tests. It must meet the original conditions, and it must break the conclusion. Here 40 is an allowed input, and its result is composite. For an integer greater than 1, composite just means not prime.

The same example also proves a different statement, that at least one allowed input produces a composite result. So one example can prove that something exists, even though it can't prove a general claim about all inputs.

Try to recallWhy is the input 40 enough to disprove the claim that n squared plus n plus 41 always gives a prime?

Forty is an allowed input, a whole number from zero up, and the recipe turns it into 1,681, which is 41 times 41 and so isn't prime. One failure on an allowed input contradicts a claim that allows no exceptions.

Forty primes in a row made a guess worth testing, and the input 40 sank it. Use examples to explore. To prove a claim about every case, you need an argument that covers them all. To disprove it, you need one valid counterexample.

Lessons in this topic

  1. The formula that gives 40 primes, then fails

    A simple recipe turns out a prime number 40 times in a row and then breaks. See why no run of examples proves a claim about every number, and why one counterexample disproves it.

    Essentials · 4 min

  2. How one short proof covers every even number

    Use the definition of an even number to prove, in three lines of algebra, that any two even integers add up to an even one.

    Essentials · 4 min

  3. How induction proves endless cases in two steps

    Add up the odd numbers and you keep landing on perfect squares. Induction proves that this never stops, using one starting case and one rule for getting to the next.

    Deeper · 4 min

  4. Proving a statement by turning it around

    A claim about nine playing cards is hard to check head on and easy to prove backwards. Learn the contrapositive, how it differs from the converse, and how to build a proof on it.

    Deeper · 5 min

Three quick stories from this topic

Odd numbers add up to squares

Add up the odd numbers from 1 and every running total is a perfect square

One plus 3 is 4. Add 5 for 9, then 7 for 16, and it never stops. Each new odd number is exactly the L-shaped border of tiles that grows one square into the next size up.

Checking totals is not enough

However many totals you check, there are always more. Proof by induction settles all of them with two short arguments.

One start and one step

First, 1 is 1 squared. Then, if the first n odd numbers total n squared, adding the next one, 2n plus 1, gives the square of n plus 1.

Forty primes, then a failure

One simple formula gives a prime number 40 times in a row, then fails

Square a whole number, add the number itself, then add 41. Every input from zero to 39 gives a prime. Input 40 gives 1,681, which is 41 times 41. Forty successes in a row proved nothing about the 41st.

The input 40 breaks the pattern

It gives 1,681, which is 41 times 41, so the result isn't prime.

Match evidence to the claim

This one example disproves "every input works" and proves "at least one input fails."

Goldbach's unproved guess

A 1742 guess about primes has been checked up to 4 quintillion and is still unproved

Christian Goldbach wondered whether every even number greater than 2 is the sum of two primes. Computers have confirmed it for every even number below 4 quintillion. Mathematicians still call it a conjecture, because infinitely many cases remain.

Why the checks do not count

A claim about every even number has no last case. Any untested number could be the one that breaks it.

What a proof would need

An argument that covers every even number at once, the way a few lines of algebra prove that two even numbers always add up to an even number.

Sources

This topic is part of From everyday numbers to proof, a guided path through related topics.

  • If-then statements and the mistake of reversing them

    Learn the one way an if-then statement can be false, why reversing it makes a different claim, and what you commit to when you deny one.

    4 lessons · 20 min

  • Sets and Venn diagrams

    A set is a collection with a clear rule for what belongs. Learn union, intersection and complement, draw them as overlapping circles, and use them to count survey answers and build search queries.

    4 lessons · 16 min

  • Truth tables and the logic of and, or, not

    Logic builds complicated statements from simple ones joined by and, or and not. Learn to test any of them with a truth table, spot two that say the same thing, and see how Boole turned reasoning into algebra.

    4 lessons · 16 min

  • When the average is the wrong summary

    When one number stands in for a whole set, which number you pick changes the story. Mean, median and mode tend to pull apart when the shape is lopsided.

    4 lessons · 20 min

  • Linear versus exponential growth

    Linear growth adds the same amount at each step and exponential growth multiplies by the same factor. See why they look alike at first, how decay and compound interest follow the same rule, and where the number e comes from.

    4 lessons · 20 min

  • The seven bridges of Königsberg

    Turn a famous walking puzzle into points and lines, then count the bridges at each piece of land to prove the walk impossible without trying a single route.

    4 lessons · 17 min

All mathematics & logic topics · Take the quiz