Your essay's title is a question. This page is its answer — and you can check every number on it yourself. The short version: twelve is not a tradition. It is the smallest number that nearly works, and "nearly" can be measured to three decimal places.
From W2/W4 you have: an octave is ×2, a perfect fifth is ×3/2, and equal temperament multiplies by the same factor every step. That is everything this page needs. Nothing below is new machinery — it is those three facts pushed until they break.
Start on a note. Go up a perfect fifth (×3/2), twelve times. You pass through every note and land back on your starting letter, seven octaves higher. That is the story everyone tells.
Check it. Twelve fifths against seven octaves:
They miss. Not by a rounding error you can wave away — by … cents, which is about a quarter of a semitone, and comfortably audible. This gap has a name: the Pythagorean comma. It is an exact rational number, printed above, and it has been there since the first person stacked twelve fifths.
two notes a comma apart — the beating you hear IS the gap
So the real question is not "why twelve". It is: since the circle never closes, where do we put the error? Every tuning system in history is an answer to that one question. Twelve-tone equal temperament's answer is: smear it evenly, so it never piles up anywhere.
Divide the octave into N equal steps. Each step multiplies by 21/N. Now ask: how close can N steps get us to a pure fifth?
A fifth is ×3/2. In "steps of the octave" that is log2(3/2) = 0.5849625… of an octave. So we want a fraction k/N close to that number — and the best k for a given N is just the nearest whole number to N × 0.5849625.
Try N = 5, then 7, then 12. At 12 the beating nearly stops. That is 1.955 cents, and your ear is telling you what the arithmetic says.
Lower is better. Gold dots are record holders: an N that beats every smaller N. Those are the only numbers that were ever worth considering.
Out of the first sixty candidates, only a handful ever set a record:
| N | fifth = k steps | cents | error vs 701.955 | what happened to it |
|---|
Now look at what the shortlist is telling you. 2 and 3 are jokes. 5 and 7 are real systems — slendro and pelog-ish spacings, and plenty of music lives there. 12 is the first one where the error drops below the threshold most listeners notice on a sustained tone. 41 and 53 are better — 53 is nearly perfect — and almost nobody uses them.
Which means twelve wins on a criterion that is not accuracy. It is the first N that is good enough while still being small enough to build an instrument for: twelve keys per octave fits a hand; fifty-three does not. That sentence is a whole argument unit for your essay, and it is falsifiable — someone could reasonably argue 19 or 31 for other reasons, and people do.
Here is the part worth the whole page. Those record holders are not a coincidence, and you do not have to search for them. There is a procedure that generates them, in order, from the number 0.5849625 alone.
Take the number. Peel off the whole part. Flip the remainder. Repeat.
That gives the continued fraction of log2(3/2):
Stop at each stage and you get a fraction — the convergents. Each one is, provably, the best approximation possible with a denominator that small:
| convergent k/N | meaning | error (cents) |
|---|
7/12 is on that list. Twelve is not an accident of culture — it is a denominator of a convergent of log₂(3/2). The same procedure hands you 41 and 53 next, which is exactly what the search in section 3 found by brute force.
One honest footnote you should keep, because it makes the argument stronger rather than weaker: N = 29 sets a record in section 3 but is not a convergent. Record-holders and convergents almost coincide but not exactly — the in-between ones are called semiconvergents. An essay that mentions this is an essay written by someone who actually looked.
The theorem is Lagrange's: if a fraction p/q approximates an irrational x better than every fraction with denominator ≤ q, then p/q is a convergent of x's continued fraction. So "best rational approximation with a small denominator" and "convergent" are almost the same idea — which is why a procedure invented for number theory happens to predict which musical scales people built. Nobody designing a lute was doing number theory. They were doing trial and error on a search space whose winners number theory can name in advance.
And log₂(3/2) is irrational — that is why the circle of fifths can never close. If it were rational, say p/q, then stacking q fifths would land exactly p octaves up, and the comma would be zero. The comma exists because an exponential equation 2a = 3b has no whole-number solutions other than the trivial one. The Pythagorean comma is a fact about prime numbers wearing a costume.
Equal temperament buys "every key works" by making every interval except the octave slightly wrong. Here is the bill, in cents:
| Interval | Pure ratio | 12-TET | Off by | Audible? |
|---|---|---|---|---|
| Octave | 2 / 1 | 2.0000 | 0.00 | exact — the one thing we refuse to compromise |
| Perfect fifth | 3 / 2 | 1.4983 | −1.96 | essentially not |
| Perfect fourth | 4 / 3 | 1.3348 | +1.96 | essentially not |
| Major third | 5 / 4 | 1.2599 | +13.69 | yes — this is the one you can hear |
| Minor third | 6 / 5 | 1.1892 | −15.64 | yes |
the slow wah-wah on the second one is those 13.69 cents
That is the trade in one sentence: we protected the fifth (and gave up almost nothing), and we paid for it in the third. A barbershop quartet, which has no frets and no keys, drifts back toward the pure third automatically — because nothing is forcing them not to. That is evidence you can point at.
Three units, and this page hands you the evidence for all three. Write the claim yourself — the claim is where your voice is.
| Unit | The evidence on this page | The honest limitation |
|---|---|---|
| 1 · What is the question | The circle of fifths does not close: 531441/524288, exactly. Not an approximation, not a measurement error. | — |
| 2 · Why twelve | 7/12 is a convergent of log₂(3/2). The shortlist in §3 was found by brute force; §4 generates the same list from the number alone. | Convergence explains why 12 is a candidate. It does not explain why 12 rather than 41 — that needs the hand-size / instrument argument, which is historical, not mathematical. |
| 3 · What twelve costs | The cents table in §5: fifth −1.96, third +13.69. Your blind-listening test is the independent evidence. | Your listening test is a pilot with family-sized N — say so. It shows a preference exists, not how large it is in the population. |
A claim nobody could disagree with is not a claim. If yours cannot be argued against, it is a description.
Everything here is about one interval — the fifth. That is the traditional way to pose the question, and it gets you 12, but it is a choice. Ask instead "which N best approximates the fifth and the major third together" and the answer changes: 19 and 31 start looking excellent, which is exactly why those systems have real advocates. This page gives you the standard argument, honestly and checkably. It does not give you the only argument — and noticing that is the difference between reporting a result and understanding one.