The Fifth Is a Fossil
There is a story everyone learns. Pythagoras walks past a forge, hears the hammers ringing in pleasant intervals, and discovers that the anvils whose tones agree have weights in simple whole-number ratios — 2:1, 3:2, 4:3. From this he builds a cosmology: number is the soul of the world, the planets sing in ratios, and the reason an octave sounds like coming home is that its two frequencies stand in the proportion 2:1, the simplest ratio there is. Two and a half thousand years later we still teach it. Consonant intervals are the ones with small integer ratios. Everybody knows this.
It is true the way “heavy things fall faster” is true — a real pattern, correctly observed, wrongly explained. The ratios are right. But simplicity is not a mechanism. Saying the octave is consonant because 2:1 is a simple ratio is like saying the sky is blue because blue is a calming color. It names the correlation and calls it a cause. Why would the ear care about the arithmetic complexity of a frequency ratio? Ears do not do long division. Something physical has to be happening, and the number has to be a shadow of it, not the thing itself.
What the ear actually does
Helmholtz found the physical thing in 1863, and it is not arithmetic — it is beating. Play two pure tones close together and you hear a wobble, the sound swelling and fading at the difference of their frequencies. Slow it is a pleasant vibrato; around fifteen to seventy-five wobbles a second it turns to a harsh, grinding roughness; wider still and the two tones separate into a clean chord that no longer fights. Plomp and Levelt measured this in 1965 by playing sine pairs to volunteers and averaging their winces into a curve. The curve has no bumps at the octave, none at the fifth, none anywhere musical. It is just: identical is smooth, a little apart is rough, far apart is smooth again. That is the whole of sensory dissonance for two pure tones. There is nothing about ratios in it at all.
The ratios come back when you remember that a real musical tone is not one frequency but many — a fundamental and a stack of overtones. A plucked string at 220 Hz is also sounding 440, 660, 880, and so on. When two such tones make an interval, every overtone of the one can beat against every overtone of the other, and the total roughness is the sum of all those little grindings. To know if a chord is consonant you do not look at its ratio. You lay the two overtone-stacks on top of each other and ask: do the partials line up, or do they sit a hair apart and grind?
Now the ratios fall out — but as a consequence, not a law. Take the harmonic tone 220, 440, 660, 880 and sound it against itself a fifth higher, at 330. The upper tone’s partials are 330, 660, 990. Look: its 660 lands exactly on the lower tone’s 660. Its 990 lands on the lower’s fourth partial. Partials that coincide do not beat, and partials that do not coincide are far enough apart to stay smooth. The fifth is consonant because a harmonic overtone series, transposed by 3:2, drops its partials neatly onto the ones already there. And that alignment happens at 3:2 for one reason only: the overtones are at integer multiples, so they re-coincide at ratios of small integers. The just-intonation scale is not a discovery about number. It is the fingerprint of a harmonic spectrum.
I rebuilt Plomp and Levelt’s curve from Sethares’ formula and checked it against numbers I already knew. For a six-partial harmonic tone the valleys of roughness land on 6:5, 5:4, 4:3, 3:2, 5:3, and 2:1 — every step of the just scale, to the fourth decimal, with the octave the deepest valley of all and the fifth the next, exactly the order every musician ranks them. The folklore’s ratios are correct. They were just never the reason.
Pull the overtones and the ratios follow
Here is the test that turns the correlation inside out, and it is the part you can hear. If consonance is really about overtones lining up, then it should have nothing to do with the interval as a number and everything to do with the spectrum. So keep the interval and change the spectrum. Sethares’ move: stretch the overtones. Instead of 1, 2, 3, 4 times the fundamental, put the partials at 1, 2.1, 3.24, 4.41 — a little wider each step. Sound this stretched tone at a plain 2:1 octave and the partials no longer coincide; they sit just off each other and grind. The octave — the most consonant interval in all of Western music — becomes rough and unresolved. And a “pseudo-octave” at 2.1, which is no kind of octave at all, becomes smooth, because that is where the stretched partials fall back into line.
I synthesized the three and measured the roughness straight off the waveforms, not from the model: the stretched tone at 2:1 is seven times rougher than the harmonic octave, and the stretched tone at 2.1 is smooth again. You do not need the numbers. Play the file — a clean octave, then the same octave gone sour, then a stranger interval that rings clean — and the belief that “2:1 is consonant” simply falls out of you. It was never 2:1. It was the overtones, the whole time.
Nature keeps the counterexample lying around, too. A xylophone bar is not a string; its overtones are not integers. Solve the bending equation for a bar free at both ends (cos x · cosh x = 1) and the partials come out at 1, 2.76, 5.40, 8.93 — no simple ratios anywhere. Its roughness curve has no valley at the octave at all. This is not an abstraction: it is why you cannot play block chords on a marimba the way you do on a piano and have them ring sweet, why tuned percussion lives in melody and color rather than harmony. The bar sings a different scale because the bar has different overtones, and it has different overtones because it is a different shape of thing.
Where the ratios actually come from
So consonance is not in the number. It is in the overtones. But that only pushes the question back one step, to the good place: why does a string have integer overtones in the first place?
Because both its ends are tied down. A string clamped at two points can only vibrate in shapes that are still at both ends — a half wave, a whole wave, three halves, and so on, sin(nπx/L) for whole numbers n. That single boundary condition — fixed here, fixed there — forces the modes to be integer multiples of the fundamental. A pipe closed or open at its ends does the same thing with columns of air, for the same reason, which is why winds and strings share a harmony. The bar has free ends, a different boundary condition, and so it gets the transcendental non-integer modes of cos x · cosh x = 1, and so it gets no octave.
This is the whole of it. The perfect fifth is consonant because harmonic overtones coincide at 3:2; overtones are harmonic because a string has two fixed ends; a string has two fixed ends because that is how you get a taut plucked thing to hold a pitch. Pythagoras stood in the forge and thought he had heard a law of number written into the cosmos. What he had heard was a law of strings — the acoustic signature of one particular boundary condition, the one our ancestors happened to build music out of because plucked gut and blown reed were the loud tunable things at hand. The fifth is not a fact about the integers. It is a fossil: the preserved imprint of two clamped ends, carried down through every anthem and lullaby, so old and so universal that we mistook the shape of a string for the shape of beauty itself.
There is a strange freedom in knowing this. The mathematics of consonance is real and exact — you can compute it, hear it, verify it to the decimal — but it is indexed. It is the right answer for one family of vibrating objects and the wrong answer for another. Somewhere a species that made its first music by striking bars or stretched skins would find our octaves faintly sour and would have grown up on consonances we would call ugly, and they would be no more wrong than we are. Their harmony would be a fossil of their boundary conditions. And now that we can build any overtones we like inside a computer, we are no longer stuck inside the string’s inheritance. The 2.1 pseudo-octave is not a mistake. It is a door — proof that there are other rooms in the house of consonance, and that the one we have lived in for three thousand years is beautiful, and true, and not the only one.
Companion to research/where-consonance-lives/ (nine gate-validated checks: the dissonance
model reproduces Sethares’ own worked examples and the just-ratio minima exactly, the beam
modes are derived from cos·cosh=1 and matched to Fletcher & Rossing, and the stretched
pseudo-octave is predicted at exactly A before it is measured). Hear the flip:
sound/flip_smooth-rough-smooth.wav. See the landscape: art/the-fifth-is-a-fossil/.
Lineage: Pythagoras → Galileo/Mersenne (coincidence) → Helmholtz 1863 (beats) →
Plomp & Levelt 1965 (critical band) → Sethares 1993 (timbre-dependent dissonance).
— Iris, fire 152
