One Name for Each Thing

Counting follows three rules. Say the number words in their fixed order — one, two, three, never one, three, two. Give every object exactly one word: none skipped, none counted twice. Then the last word you say does double duty — it is the answer to how many in all.

Here is the surprising part: the route does not matter. Count the row left to right, right to left, or hopping around — as long as each thing gets exactly one word, you always land on the same last word. Try it on your fingers and see.

The heart of counting is one-to-one correspondence: each object pairs with exactly one number word, and each word with exactly one object — nothing left out, nothing used twice. Two more principles ride along. Stable order: the counting words come in one fixed sequence, every time. Cardinality: the final word states the size of the whole set, not a label for the last object touched.

These are separate skills, and they can come apart. Many young children can recite the words and point in perfect rhythm, yet re-count the whole row when asked how many — they have the first two principles but not yet the third. The count is finished only when the last word becomes the answer.

Strip counting to its skeleton and it is a pairing between two sets: your objects on one side, the numbers 1 through n on the other. A pairing that misses nothing on either side and repeats nothing is called a bijection. To say a set has five members means it can be put in bijection with {1, 2, 3, 4, 5}.

That definition explains the order-irrelevance you already trust. Counting the row in a different order builds a different bijection — but between the same two sets, so it must end at the same n. The size belongs to the set itself, not to your route through it.

Line up your toys. Touch one toy, say one. Touch the next, say two. One touch, one word. No toy gets skipped. No toy gets two words.

When you stop, the last word you said tells how many toys there are in all. Say it out loud: five.

Formally: a set S is finite with cardinality n when there exists a bijection f : S → {1, …, n}. For the definition to be honest, n must be unique — and it is: an induction argument shows {1, …, n} and {1, …, m} admit a bijection only when n = m, which is essentially the pigeonhole principle. Counting a set is constructing such a bijection by hand, one pairing per touch.

Write S ≈ T when some bijection S → T exists. The relation is reflexive, symmetric, and transitive — an equivalence. A cardinal number is what all mutually equinumerous sets share; the numeral you announce is just the canonical representative {1, …, n} of that class.

Hume's principle — the number of Fs equals the number of Gs exactly when the Fs and Gs correspond one-to-one — is the pivot of Frege's logicism: added to second-order logic, it derives full arithmetic (Frege's theorem). Whether it is a definition of number or a substantive axiom remains the neo-logicist battleground of Wright and Boolos.

Cantor carried the same bijection criterion where fingers cannot follow: ℕ ≈ 2ℕ and ℕ ≈ ℚ, yet no bijection reaches ℝ. Equinumerosity outruns intuition precisely because it never leans on counting order. A child pairing spoons with words and Cantor pairing ℕ with ℚ perform the same act at different scales.