The angle is not the object
Almost every account of phyllotaxis is organised around 137.5°. It is the hook, the title of the popular articles, and the thing a reader is expected to come away with. This site’s foundation phase, for all its scepticism, was organised the same way: several essays measure claims about the angle, and the field that explains where it comes from is called “where the angle comes from”.
After a phase spent on cylinders, forks and frequencies, that organisation looks like a mistake — not because the angle is unimportant but because it is a derived quantity, and building an account around a derived quantity puts the explanation in the wrong place.
Four things now known about the number
Set out flatly, the phase’s findings about 137.50776° are these.
It is at no fork. Every branch point of the lattice tree sits at a rational divergence with denominator — 5/14, 15/38, 37/98, 99/258 of a turn. The golden angle is the accumulation point of that sequence and is not a member of it. The forks are exact and none of them is this.
It is the limit of one path. Keeping the larger count at every fork from gives divergences converging on it. One different choice, at the first fork, gives 99.502° instead. Both are limits of paths; the tree has infinitely many others.
It is not what the geometry favours at fine scales. At a rise typical of a seed head, a divergence chosen arbitrarily gives consecutive Fibonacci counts about one time in seven, and the share falls as patterns get finer.
It is not optimal packing. The disc essays established that three packing criteria give three different winners and none is 137.5°. The cylinder adds that the equilateral — best-packed — lattices sit at the forks, which are rational and are not this angle.
What survives from the foundation phase is the arithmetic claim, and it survives intact: the golden angle is the hardest number to approximate by rationals, scoring 0.4377 against 0.3306 for the best of 1,500 sampled. That is a real distinction and it is the reason a golden-divergence lattice never develops the radial rows a near-rational one does.
What the number is a symptom of
Put those together and the angle stops being an explanation and becomes a consequence.
A pattern that starts coarse is at a low fork, where there is very little to choose between. As its rise falls it passes fork after fork, and at each it keeps one of its two families. If it never jumps, its divergence tracks a converging sequence, and the limit of that sequence is a specific irrational number determined entirely by which daughter was taken each time.
Take the larger every time and the limit is 137.508°. Take the smaller once and it is 99.502°. Take the smaller twice and it is something else with no name.
So 137.5° is what “never jumped, starting from the coarsest fork” evaluates to. It is a summary of a history, and quoting it without the history is quoting an answer without the question.
That also explains why it is so nearly universal in real plants while being geometrically unremarkable at the scales where plants are observed. Everything is coarse before it is fine. A shoot with three leaves is at the top of the tree where the Fibonacci branch is most of the available space, and thereafter it is following a path rather than choosing an angle.
What “the golden angle” is doing in the popular account
It is worth being fair about why the number took over, because it is not simply an error.
It is memorable, it is a single quantity, and it connects to a constant most readers have already heard of. An account built on it can be given in a paragraph, and the paragraph is not false — plants really do have divergences very near 137.5°, and the number really is the hardest to approximate by rationals.
The trouble is what the paragraph then has to do with everything else. Counts that change with radius become an anomaly. Lucas phyllotaxis becomes an exception. Whorled arrangements become a different subject. Rising phyllotaxis becomes a developmental curiosity. Each of those is a normal feature of the plane, and each has to be explained away when the account has only one coordinate in it.
An account organised around the tree has the opposite property: the famous case takes a few sentences to locate and everything else is already in the picture. That is the ordinary trade between a memorable summary and a usable one, and this subject has been paying it for a century.
What the central object should be
The tree. Or equivalently the plane it is drawn in — divergence across, rise up, regions labelled by parastichy pair, forks where three regions meet.
The advantages are not rhetorical.
It contains the counts, which are what is observable. Nobody measures a divergence angle on a plant directly; they count parastichies and infer. The plane’s regions are the counts.
It contains the second parameter. Every statement about phyllotaxis has an implicit rise in it, and organising around the angle hides the rise, which is why seed head counts appear to change for no reason and why the same species shows different pairs at different ages.
It makes the history explicit. A pattern is a path through the plane, and paths are the right object because the outcome depends on where a pattern came from.
It has room for the cases the angle-centred account treats as anomalies. Lucas phyllotaxis is not an exception; it is a neighbouring branch. Whorled patterns are not a different subject; they are the regions where the counts share a factor. Bijugate patterns, which the angle-centred account struggles to place at all, are simply the regions with a common factor of two.
Rewriting three familiar claims
The reframing is worth testing against the sentences people actually say.
“Plants use the golden angle because it packs seeds most efficiently.” Both halves fail. Packing is optimised at the forks, which are rational; and no packing criterion at fixed size picks this angle. What is true: a golden-divergence lattice is the one that never develops radial rows as it fills, which is an asymptotic statement rather than a contest at a size.
“Sunflowers have 34 and 55 spirals.” True of an annulus of a head of a size. The head passes through 13/21, 21/34, 34/55 and 89/144 on the way out, at radii the cylinder’s ladder predicts with nothing fitted.
“The golden angle appears throughout nature.” The angle appears in one branch of one tree, in organisms that are on that branch. The tree is what appears throughout nature, in the weaker sense that anything spacing elements on a growing surface has one.
Where the number is still the right thing to quote
The argument is about how an account should be organised, not about whether the number should ever be mentioned, and there are places where it is exactly the right quantity.
When the question is about approximation. “Why do near-rational divergences produce radial rows and this one does not” is a question about continued fractions, the answer is about the golden angle specifically, and no amount of tree structure replaces it.
When comparing two measured plants. A recovered divergence is a single number and two of them can be compared. That is what the recovery is for, and reporting a fitted angle with its uncertainty is a perfectly good way to summarise a specimen.
When the pattern is at a fixed rise. On a stem there is one lattice, one pair, and one angle, and the tree collapses to a point. Everything the plane adds is about how patterns change, and a pattern that is not changing does not need it.
So the complaint is narrower than “stop quoting 137.5°”. It is that the number is a good summary and a bad organising principle, and that a subject taught through summaries has ended up unable to accommodate its own ordinary cases.
What this does to the site
Awkwardly, it means the foundation phase’s organisation is partly wrong, and it is worth being explicit rather than quietly renaming things.
The field called “where the angle comes from” is well named for what it contains — the dynamical model, which does produce the angle as an output — but it treats as the destination something that is one coordinate of a two-dimensional picture. The essays in it are not wrong; the frame is narrower than the subject.
The correction is not to rewrite them. It is that the cylinder field now exists alongside, and the two together say the thing neither said alone: a pattern is a point in a plane, growth is a path through it, and the angle is a coordinate of the path’s limit.
If the site were being built again from nothing, the plane would be the first figure and the angle would arrive in the fourth or fifth essay as a consequence. It was not built that way because the disc is the famous case and the disc is where anyone starts. Which is, in miniature, exactly the mistake the subject as a whole has made for a century.
What a reader should do with a photograph
The reframing has a practical form, and it is short enough to be worth stating as a procedure.
Given a picture of a sunflower, a pine cone or a stem, the questions in order are:
What geometry is it? A stem has a constant rise and one pair. A cone’s rise changes slowly along its axis. A disc’s changes as . That decides whether “the count” is a single number at all.
Where was the count taken? On anything but a stem the answer is required, and on a disc it should be given as a fraction of the radius.
Are the counts coprime? If not, the pattern is whorled and no divergence angle can be recovered from them. That is a fact about the object rather than a failure of the observer.
Which branch? Consecutive Fibonacci, consecutive Lucas, or neither — and neither is common, taking nearly half the parameter space at fine rises.
Only then does a divergence angle come into it, and it comes in as an inference from the counts rather than as an observation. Which is the reversal this essay is arguing for, stated as four questions instead of an argument.
The subject’s other reorganisations
This is not the first time phyllotaxis has been reorganised around a different object, and the history is short enough to be useful.
Schimper and Braun, in the 1830s, organised it around fractions: a leaf arrangement was 2/5 or 3/8, meaning so many turns per so many leaves. That is a description of a rational approximation to the divergence, it is what a person counting leaves actually observes, and it collapsed when patterns turned out not to be exactly rational.
The Bravais brothers, in 1837, replaced fractions with a lattice and an irrational divergence, which is essentially the modern object.
Van Iterson, in 1907, replaced the lattice with the plane of lattices and drew the branching diagram — which is what this essay is arguing for, a hundred and nineteen years ago.
The popular account then reorganised around a constant, which is a step back past all three, to something less structured than the fractions were. That is the unusual part of the history: the simplification did not come from the research literature abandoning the plane, since it never did.
What is still missing
Being clear about the boundary of the claim, since this is the field’s last essay.
The tree says what patterns are available and how they connect. It does not say what moves a pattern along a path, and nothing on this site does: the dynamical model supplies motion but cannot reach the fine end, and the mechanism field establishes that a stationary ring supplies a spacing and no arrangement at all.
Nor does anything here bear on the biology in the way that would settle it. The list of what a mechanism would have to show has six items, and this collection establishes two of them and part of a third. The items with the real evidence behind them — that the parts exist, and that interfering with them changes the pattern in the predicted way — belong to microscopy and to ablation experiments, not to figures.
So the honest position at the end of this phase: the form is now well described and its structure is a tree rather than a number. Why a plant is where it is on that tree is a question with a partial answer — continuity from a coarse start — and why it moves at the rate it does is a question about growth that has not been asked here at all.
What it cost to find this out
A note on process, since the site’s phase records are meant to be honest about what was intended.
This phase’s plan named four pieces of work: a continuous formulation of the dynamical model, Lewis’s law, cylinders and cones, and Turing’s morphogen treatment. Three of the four were built roughly as planned. The first was attempted, failed, and was replaced by something that reaches further — the tree, which needs no dynamics at all.
Nothing in the plan anticipated this essay. The reframing came out of the cylinder work as a by-product: once the plane exists and the forks in it are exact, the angle’s position in the account changes whether or not anyone intended it to. The plan asked for cylinders because most real phyllotaxis is on a stem, which is a good reason and a much smaller one than what the cylinders turned out to be for.
That is worth recording because it is the second time this site’s most useful finding has been a consequence of doing an ordinary thing carefully rather than of pursuing an interesting one. The first was the counts changing with radius, which needed nothing but pointing an existing counter at successive annuli.
The sentence to replace the number
If a reader takes one thing from this collection it should not be 137.5°.
It should be this: a phyllotactic pattern is a lattice with two parameters, its observable content is a pair of counts, and as one parameter falls the pattern moves through a branching structure whose forks are exactly computable. Which branch it is on is a fact about its history. The famous angle is the limit of the branch that most plants happen to be on.
That is longer than a number and it is checkable in every clause, which the number never was.