The nautilus question

The nautilus shell is the standard illustration of the golden ratio in nature. It appears in textbooks, on posters, in design courses, and in a very large number of articles about mathematics being everywhere.

It is not a golden spiral, and the discrepancy is a factor of two.

The two numbers #

A golden spiral is the one that grows by the golden ratio φ = 1.618 in a quarter turn. That means it grows by φ⁴ = 6.854 in a full turn.

A nautilus grows by about 3.2 per turn. Published measurements cluster between roughly 2.9 and 3.4, varying with the individual and slightly with how the measurement is made.

Those are not close. The ratio is 2.14, and the whole measured range for real shells falls short of the claimed figure by more than the width of the range.

Where the confusion comes from #

Three things, and none of them is stupidity.

A nautilus really is a logarithmic spiral. That part of the claim is correct, and it is the part that gets checked. A shell that grows self-similarly traces a logarithmic spiral, and the nautilus does. What does not follow is that it is the particular logarithmic spiral with growth φ⁴.

All logarithmic spirals look alike. Shown one turn of each without a scale, most people cannot tell a 3.2 spiral from a 6.9 spiral. The difference is obvious over two turns and invisible over half of one, and shell photographs are usually cropped to about that.

The golden rectangle construction is suggestive. The nested-squares picture produces a curve that is drawn inside a rectangle whose proportions everyone recognises, and the visual association is strong even though the arc is a sequence of quarter circles rather than a spiral at all.

What the measurement cannot be blamed on #

The obvious objection is that the fit depends on where the centre is put, and a shell’s centre is not marked. That objection is testable and it does not survive.

Moving the assumed centre by a quarter of the innermost whorl’s radius — an error far larger than any careful measurement would make — changes the fitted growth factor by 4.56 per cent over two turns of arc and 1.34 over three and a half. This essay carried “well under one per cent” for a long time and that figure was too small by about a factor of five.

The correction does not reach the argument. Turning 3.2 into 6.854 needs an inflation of 114 per cent, twenty-five times the largest error a quarter-radius displacement produces at a usable span, and no assumed centre anywhere in the plane returns the golden factor once the arc exceeds 1.2 turns.

Why it survived #

A claim this checkable persisting for decades is worth a paragraph, because the mechanism is general.

The check requires two things nobody does casually: measuring a growth factor over more than one turn, and knowing that a golden spiral’s factor per turn is φ⁴ rather than φ. The second is the real barrier — φ is 1.618, the shell obviously grows by more than that per turn, and 3.2 sounds like it could be φ-something to anyone who has not worked out which power applies.

Add that the claim is attractive, appears in reputable places, and is repeated by people who are not making an error so much as passing one along, and it becomes self-sustaining. The correction has been published repeatedly since at least the 1990s and it does not travel, because the claim is more useful to a lecture than the correction is.

What is actually true about it #

The nautilus is a genuinely elegant object and it does not need the golden ratio.

It grows self-similarly, which is a real constraint with a real consequence, and it does so with remarkable regularity — the fit residuals on a well-measured section are small, meaning the animal held its growth factor nearly constant over its whole life. That is more impressive than matching a particular constant, and it is the thing the measurement actually supports.

The chambers are also spaced so that each is a scaled copy of the last, which is the same property expressed in the septa, and the buoyancy control that uses them is a genuinely sophisticated mechanism.

None of that requires 6.854. The animal is interesting for what it does, and attaching a number to it that it does not have is a way of finding it interesting for something it is not.

What the claim would need to be true #

Worth stating positively, because it is a short list and every item fails.

The shell would have to grow by 6.854 per turn. It grows by about 3.2, and the whole published range falls short.

The measurement would have to be sensitive to the assumed centre. It is sensitive, and not nearly by enough: a quarter-radius error moves the answer by 4.56 per cent over two turns, against the 114 per cent the claim needs.

Or the claim would have to be about a different quantity — the ratio of successive chamber widths, say, rather than growth per turn. That version can be checked too, and a growth factor of 3.2 per turn gives successive quarter-turn ratios of $3.2^{0.25} = 1.34$, not 1.618.

There is no reading of the claim on which the numbers work.

The pattern this belongs to #

The nautilus is one of three famous assertions this site measures, and the three sort neatly.

One is true of a branch rather than of plants. One is right, in a sharper form than the version usually told. And this one is simply wrong, by a factor of two, about a quantity anyone can measure in an afternoon.

What they share is that none of them had a test attached. Each was passed along as a fact rather than as a claim, and a fact does not invite checking. Attaching a test to each — state the number, state how it would be measured, report what the measurement gives — is the whole of what this site does, and it is why the one that survives is worth more after the exercise than before it.

Testing the claim rather than repeating the correction #

There is a version of this essay that would simply assert the nautilus is not golden, and a reader would have to decide whom to believe. The point of doing it this way is that the reader does not have to.

The growth factor of a drawn spiral is recoverable from the coordinates by least squares on log r against angle. Applied to a curve generated at a stated factor it returns that factor to fifteen digits with a residual at arithmetic noise, so the machinery demonstrably measures what it claims. Applied to a nautilus section it returns about 3.2. Applied to a golden spiral it returns 6.854.

Every step of that is checkable by anyone with a shell section and a ruler, and none of it requires trusting this site.

The measurements that exist #

Published growth factors for Nautilus pompilius cluster around 2.9 to 3.4, depending on the individual and on the method. The variation is real — shells differ, and the fitted centre shifts the answer slightly — and it is far smaller than the gap being argued about.

What matters for the comparison is that the whole range falls short. If measured factors ran from 3 to 7 the claim would be arguable; they run from 3 to 3.4, and 6.854 is not in the neighbourhood.

There is also a related quantity that sometimes gets confused with it. Successive chambers in a nautilus are spaced so that each is a scaled copy of the last, and the ratio between successive chambers is often given as about 1.3 rather than 1.618. That figure does not close as it is usually explained: a growth factor of 3.2 over a third of a turn is 1.474 in length, not 1.3. A length ratio of 1.3 needs 4.43 septa a whorl, and a volume ratio of 1.3 needs 13.3 — which of those the 1.3 was decides the septa count, and the count decides the growth factor read from it. None of the readings is the golden ratio either, and the coincidence of numbers near one and a half is presumably part of how the claim survives.

Why the correction does not travel #

Worth a paragraph, because the failure of a well-established correction to propagate is itself informative.

The claim is useful. It appears in a lecture as a moment of wonder, it takes ten seconds, and it requires no follow-up. The correction is a paragraph about growth factors and powers of φ, it takes two minutes, and it ends with the audience knowing less that is remarkable than they did before.

It is also repeated by people who are not making an error. Someone who read it in a reputable book and passes it on has done nothing careless; the error is several links back and there is no signal in the chain that it happened.

And the check is not casual. It needs a section, more than one turn, and the knowledge that a golden spiral’s factor per turn is φ⁴ rather than φ. That last point is the real barrier: φ is 1.618, a shell obviously grows by more than that in a turn, and 3.2 looks like it might be φ-something to anyone who has not worked out which power applies.

The result is a claim that is self-sustaining, and the site’s response is not to complain about it but to attach a test — which is the only thing that has ever worked on a claim like this.

What a careful version of the claim would look like #

It is worth writing out, because the interesting properties survive.

The nautilus shell is a logarithmic spiral — the curve traced by anything that grows by adding a proportional increment at its opening. Its growth factor is about 3.2 per turn, held remarkably constant across the animal’s life, which is what allows the chambered structure to work: each chamber is a scaled copy of the last, so the buoyancy mechanism sees the same geometry at every size.

That is accurate, it is checkable, it explains something about the animal, and it does not mention the golden ratio because the golden ratio is not involved.

Whether it is as memorable as the false version is a fair question, and probably not. But the false version’s memorability is doing the damage: it is what makes it travel, and it travels attached to a claim about mathematics being everywhere that the correction quietly undermines.

The three claims, and the record #

This site measures three famous assertions about pattern in living things.

The nautilus is a golden spiral: wrong, by a factor of 2.14 in a quantity anyone can measure.

Sunflower spirals are always Fibonacci: true of a branch, not of plants. The Lucas branch gives 47 and 76, and the same model reaches it.

137.5° is optimal: right, once the criterion is named — and the criterion is resistance to rational approximation rather than packing, where it is the extreme case and there is a theorem.

One in three is better than the subject’s reputation suggests. What the three have in common is that none of them arrived with a test attached, and attaching one is the only thing that separates them.

The shells that are close #

An overlooked wrinkle: some shells really do have growth factors near φ⁴, and knowing which ones sharpens rather than rescues the claim.

Rapidly expanding gastropods — certain Haliotis, some limpet-like forms — open fast enough that a turn multiplies the radius by five, six or more. A shell somewhere in that range will pass within a percent of 6.854 by coincidence, and one could photograph it and make the claim truthfully about that individual.

That is not a vindication. It is the ordinary observation that a continuous parameter varying across a clade will take any given value somewhere, and it carries no more significance than a person being exactly 1.618 metres tall.

What would be significant is a clustering at φ⁴ — factors piling up at that value across unrelated groups, the way side counts pile up at six for a reason. There is no such clustering. Raup’s parameterisation puts real shells across a broad region of W with structure driven by mode of life, and nothing distinguishes 6.854 within it.

The test is the same one this site applies everywhere: not “does this value occur” but “does it occur more than it should”, and the second question needs a distribution rather than an example.

What the overlay figure does and does not show #

The figure putting a golden spiral over a nautilus section is the one most likely to be misread, so it is worth saying what it is for.

It is not the evidence. A curve drawn over a photograph can be made to look convincing or unconvincing by choosing the centre, the starting radius and how much of the curve to show, which is precisely how the claim got established in the first place. The overlay is there because the claim is visual and a correction that offers only numbers does not reach the place the belief lives.

The evidence is the fitted growth factor, which uses none of those choices: it takes coordinates and returns a number, and the centre — the one free choice — is swept over the whole plane to show it cannot be blamed.

Keeping those apart matters, because a reader who finds the overlay unpersuasive should not conclude the argument is unpersuasive. The overlay is an illustration of a result obtained elsewhere, which is what most figures on most sites are and what this one is unusually careful to label.

Why this essay exists at all #

A site could take the view that correcting a popular misconception is a low use of everyone’s time, and there is something to that. The reason to spend an essay on it is that the claim is doing damage of a specific kind.

It is the single most cited example of “mathematics in nature”, and it is false. Anyone who checks it — and checking needs only a shell section and a spreadsheet — discovers that the standard example of the standard claim does not hold, which is a poor introduction to a subject where the real results are better than the folklore.

It also trains a bad habit. The claim is believed because a curve was laid over a photograph and looked right, and that is exactly the method that fails on spiral counts, overreaches on packing quality, and failed inside this site’s own machinery until a second route disagreed.

Correcting the nautilus is therefore not pedantry about a mollusc. It is the cheapest available demonstration that looking is not measuring, in a subject where almost every claim is made by looking.

What this settles that most of the site cannot #

Expansion produced a checklist of what a mechanism would have to show, and one of its conclusions is that measurements of form rarely discriminate between accounts, because a robust form is what many different processes produce.

This essay is the exception on the site, and the reason is worth naming.

The nautilus claim is quantitative and specific. It says the shell’s radius multiplies by φ⁴ ≈ 6.854 per turn. That is a number, the number is measurable, and the measurement disagrees by a factor of 2.14.

Compare the claims that resist measurement: “the arrangement optimises packing”, “the pattern arises from an activator–inhibitor system”. Those predict a kind of form rather than a value, and many mechanisms predict the same kind, so no measurement of the form separates them.

The general rule the two cases suggest: a claim about form is settleable exactly to the extent that it commits to a number. The nautilus claim commits, so it is settled. Most claims about phyllotactic mechanism do not, so they are not — and the ones that do, like the ratio of diffusion constants a Turing mechanism needs, are the ones that have caused the most trouble for the accounts making them.

The rule has a converse worth stating, because it is the more useful half in practice. A claim that commits to a number can be settled by one good specimen, where a claim about a kind of form needs a survey and often cannot be settled by any number of specimens. So the way to make a vague claim tractable is not to gather more material; it is to find the number the claim implies and measure that instead — which is what asking what a mechanism would have to show is for.