On the eighteenth of June 1858, Charles Darwin opened a letter from a naturalist collecting specimens in the Malay archipelago and found in it the theory he had been refusing to publish for twenty years.
Alfred Russel Wallace had written it at Ternate, in a fever, in a few days. He had no idea what Darwin had in his drawer. He was asking for an opinion. Darwin wrote to Lyell that same day: if Wallace had had his manuscript sketch of 1842 in front of him, he could not have made a better short abstract of it, and even Wallace's terms now stood as chapter headings in Darwin's own book.
The resolution was the joint reading at the Linnean Society on the first of July, Darwin's extract and Wallace's essay presented together, neither author in the room. And then the detail that every historian of science tells, because it is too good not to: the Society's president, reporting on the year 1858, recorded that it had not been marked by any of those striking discoveries which at once revolutionise a department of science.
The Pattern, Once You Look
Newton and Leibniz built the calculus within a decade of each other and spent the rest of their lives, and a good deal of the eighteenth century, fighting about it. The Royal Society appointed a committee in 1712 to arbitrate the priority claim; the report was drafted, anonymously, by Newton.
Non-Euclidean geometry arrived three times. János Bolyai in Hungary and Nikolai Lobachevsky in Kazan reached it independently in the 1820s and 30s, neither knowing of the other. When Bolyai's father sent the work to Gauss, Gauss replied that he could not praise it — because to praise it would be to praise himself, having reached the same results decades earlier and published none of them. The young man never really recovered.
Oxygen was isolated by Scheele and by Priestley and interpreted by Lavoisier, in overlapping years. Conservation of energy was formulated by Mayer, Joule, Colding and Helmholtz within about a decade, a case Thomas Kuhn made the subject of a famous essay precisely because the simultaneity is so hard to explain away. Sunspots were reported in 1611 by Johannes Fabricius, by Galileo, by Christoph Scheiner and by Thomas Harriot — four observers, one year, one newly available instrument. Bell and Gray filed at the patent office on the same day in February 1876.
What "In the Air" Actually Means
The ordinary explanation is sociological and it is almost certainly right as far as it goes. Discoveries have prerequisites. The telescope reaches a certain quality and sunspots become visible to everyone who owns one. Geology establishes deep time, Malthus supplies the arithmetic of population, animal breeding supplies the analogy, and the voyages supply the distributions — and then natural selection is available to any careful person who assembles the pieces. Darwin and Wallace were reading the same books and looking at the same problem in the same decade.
Notice, though, what that explanation actually says. It says the discovery was constrained into existence by the state of the field: given those prerequisites, the next step was not one of many, it was nearly the only one available. The sociological account and the argument of this book are not rivals. They are the same claim, and the sociological version has the advantage of being uncontroversial.
What it adds is the mechanism. Convergent evolution has physics doing the constraining — optics, acoustics, fluid dynamics. Multiple discovery has the structure of the problem doing it, plus whatever the field has already established. In both cases the search is undirected and the destination is not.
The Objection, and It Is Serious
Multiples make a satisfying story, and satisfying stories in the history of science are usually hiding something. Three cautions belong in the record before this chapter is allowed to support anything.
Hindsight flattens. Darwin's theory and Wallace's were not identical. Wallace did not frame selection at the level of the individual in the way Darwin did, and the two later diverged sharply — Wallace excepting the human mind from the mechanism entirely. Calling them the same discovery is a judgement, made afterwards, by people who know how the argument turned out. The more finely you look at any multiple, the less multiple it becomes.
Nobody counts the singletons. Merton's 264 are cases that were noticed. There is no comparable catalogue of discoveries made once and never approached by anyone else, because the absence of a second discoverer is not an event and does not get recorded. A collection of multiples tells you multiples happen. It cannot tell you what fraction of discovery they represent, and the denominator is exactly what the argument needs.
Priority incentives manufacture the appearance. When a field is hot, several laboratories publish quickly and near-simultaneously by design. That is competition around a visible finish line, not independent arrival, and it is not the same phenomenon as a letter from Ternate.
A Multiple in the Record, Dated to the Week
Every case above is historical, which makes it safe and slightly inert. Here is one from this month, in which the author of this book is one of the two parties — reported not as a credential but because a multiple you were standing inside is the only kind anyone can describe from the inside.
The Riemann zeta function satisfies a functional equation relating its value at s to
its value at 1 − s. Take the logarithmic derivative of that equation and
you get an identity about ζ′/ζ. On 8 September 2026 the Lean
mathematical library Mathlib merged such an identity —
logDeriv_riemannZeta_one_sub, pull request 43252, in a new file written for it —
contributed by Terence Tao. The same identity had already been proved, in Lean, in this series’
own book4/ZetaReflection.lean, under a different name. The two statements do not look
alike: different right-hand sides, different special functions, no shared line. Applying the
digamma duplication formula and then the reflection formula at (1+s)/2 carries one into
the other exactly. One theorem. Two routes.
The priority question is not interesting and it is already settled: the merge is the merge, it is recorded as a supersession in Book 4 · § 12.7, and nothing in this series claims otherwise. What is interesting is the second thing, and it is the octopus retina again.
sorry, compiled 11 September 2026. Two lineages arrived at one device, and the
one that came out with the wider field of view was not the one that arrived first.And the test this chapter sets for the gallery has to be run on this case before any other. By that test, it is not independent discovery in the strong sense, and it should not be reported as one. Both parties were working in the same library, against the same version, on a gap that the library itself makes visible to anyone who opens the file. The prerequisites were not merely shared — they were the same file. This is Merton’s constrained-into-existence case in its purest available form, and the right reading is the sociological one this chapter has already conceded is usually correct: given Mathlib in September 2026, that theorem was standing there for whoever looked. Which is not a smaller claim than convergence. It is the claim — the valley delivered the water twice in one week, and the only reason we can see it happen is that this particular valley keeps a version-controlled record of every drop.
The older case is larger and it has the same shape. In 1949 André Weil conjectured, for varieties over finite fields, a set of statements about counting solutions, and observed that they had the form they would have if a certain cohomology theory existed — a specification for a machine nobody had built. Bernard Dwork proved the first of them, rationality, in 1960, by p-adic analysis and no cohomology at all, while Grothendieck was building the cohomology that would shortly prove the same statement a second way. Pierre Deligne closed the hardest of them in 1974, by a route Grothendieck did not regard as the right one. And Grothendieck’s own intended route — the standard conjectures on algebraic cycles, stated 1968–69 — has still not been walked; it is open in 2026, fifty-seven years on, and no one can say whether it goes through. A destination reached twice by two routes, with a third route visible, named, and unwalked. The history is in Book 7 · André Weil; the mathematics, with the two positivity statements set side by side, is in Book 4 · § 12.8.
Why This Chapter Is the Test for the Gallery
Everything above is the method this book owes its own gallery.
Sixteen figures, from traditions that mostly could not read one another, arriving at related structures — recurrences, ratios, ladders of proportion. Read carelessly, that is a marvel. Read properly, it is a claim with exactly two candidate explanations, and they are the same two the historians use on Darwin and Wallace.
Transmission. Did the prerequisites travel? Mesopotamian astronomy reached Greece and India. Indian numerals reached Europe through al-Khwārizmī and Fibonacci. Gerbert carried the abacus back from Catalonia. Where a channel existed, resemblance is inheritance and proves nothing about structure.
Constraint. Where no channel can be shown, and the same structure appears anyway, the explanation has to be that the structure is small — that there were not many places the thinking could go.
The honest position is that this book cannot decide the question for every figure, and does not pretend to. What it can do is ask it every time, name which of the two it thinks applies, and mark the ones where it does not know. That is why each portrait in the gallery carries a tag rather than a verdict, and why the Saptarishi — who are identified with the seven stars themselves — are treated as stronger evidence than a resemblance nobody has dated.
One More Thing Multiples Tell You
If a result can be found twice, it cannot be suppressed once.
The Jesuit prohibition of indivisibles in 1632 did not end the mathematics of the infinitely small. It ended it in Italy. The work continued in France with Fermat, Roberval and Pascal, and then in England, and arrived at the calculus anyway within two generations — twice over, in Newton's hands and in Leibniz's. Descartes suppressed Le Monde himself and the heliocentric world came out regardless, through other people, in other cities, in other words.
That is not a consoling thought about the triumph of truth. It is a structural observation, and a narrow one: censorship acts on channels, and a result that the state of a field makes available is available through many channels at once. What suppression reliably destroys is not the idea but the person — Bruno, Galileo's last years, Bolyai's confidence. The valley delivers the water. It has never had anything to say about the fate of a particular drop.
Sources: W. F. Ogburn and D. Thomas, “Are Inventions Inevitable?”, Political Science Quarterly (1922); Robert K. Merton, “Singletons and Multiples in Scientific Discovery”, Proceedings of the American Philosophical Society (1961); Thomas S. Kuhn, “Energy Conservation as an Example of Simultaneous Discovery”, in The Essential Tension (University of Chicago Press, 1977); on Darwin and Wallace, the Linnean Society's own records of the 1 July 1858 reading; on Bolyai and Gauss, Gauss's letter to Farkas Bolyai of 1832.
Next: ideas converge because problems have few solutions. So do shapes. What the Wax Knows is the physical case — honeycomb, basalt and foam arriving at one angle by three unrelated mechanisms — and the door down into the formal work.