What Is a Coincidence

Everyone knows what a coincidence is. Two things line up and nothing connected them. You think of your brother and the phone rings and it is your brother. You learn a word on Tuesday and hear it four times before Friday. Two strangers at a party turn out to have been born in the same small hospital in the same week. The word coincidence is what we say to mark the alignment and, in the same breath, to deny that anything holds it together.

There are two respectable positions and they have been arguing for a century. The debunker says the alignment is nothing: with enough people, enough days and enough categories, astonishing matches are not merely possible but obligatory, and what needs explaining is not the match but your surprise at it. The believer says the alignment is something: a sign, a resonance, a hint that the world is more tightly woven than the debunker will admit.

Notice that both of them agree about the important thing. Both take it as settled that there are exactly two shelves to put a pair of events on. Either one event bears on the other, or the pair is empty. The debunker puts almost everything on the second shelf. The believer moves a few items back to the first. Neither questions that there are two shelves.

That is the assumption this piece is about, and it is worth saying at the start that the framework does not side with the believer. It removes the shelf argument entirely, and then it takes away the thing the believer was arguing for.

Two ordinary cases the two-shelf picture cannot hold

The multiples. In 1922 William Ogburn and Dorothy Thomas published a list of 148 cases in which a discovery or invention was made independently by two or more people. Calculus, twice. Oxygen, three times. Natural selection: Darwin’s essay extract and Wallace’s Ternate essay were read in sequence before the Linnean Society of London on the first of July, 1858, neither man in the room. The telephone: Alexander Graham Bell’s patent application and Elisha Gray’s caveat were filed at the United States Patent Office on the same day, the fourteenth of February, 1876, and the order within that day has never been settled and probably never will be.

Put those on the shelves. Causation will not take them: Wallace had not read Darwin’s notebooks, and Gray had not seen Bell’s drawings. Chance will not take them either — not at that density, not at that timing, and not 148 times in one appendix that its own authors described as incomplete. Robert Merton spent a career on this and the word he ended up with was multiples, which is a name for the pattern rather than an explanation of it.

And look at what historians actually say when they explain a multiple. They say the conditions were met. That sentence is not a causal claim and it is not a chance claim. It is a third thing, in ordinary use, every day, by people who would deny there is a third shelf if you asked them directly.

The coincidence that is only yours. Two guests at a party discover they were born in the same small hospital. For them it is a coincidence. For the host, who invited them both because they grew up in the same town, it is not a coincidence at all — it is nearly the point. Same pair of facts, same evening, same room. One reading has the property and the other does not.

So being a coincidence is not a property of the pair. It is a property of a reading of the pair. The two-shelf picture cannot express that, because both of its shelves are descriptions of the events themselves, and the events did not change when the host walked over.

The historians already use the third relation. They just call it “the conditions were met” and assume it is loose talk.

What the foundation says instead

The third relation is not an improvisation. It is the oldest thing in the document, and until today the daily series had never once cited it.

Start with S1. When the unconditioned is conditioned, a circle appears, and every point on its circumference is a condition. The circumference is infinite and no enumeration closes it. Conditions are not events and they are not located in the sequence of events; they are the plane the sequence is drawn on.

Then P6, which is the primitive doing the real work here, and which is three denials in a row:

P9 completes it. Conditions condition — a real relation, neither causation nor agency nor inertness. Read that as three refusals too. Not causation, so it does not push. Not agency, so it does not intend: no wanting, no strategy, no preference. And not inertness, which is the refusal everyone skips. A condition is not a passive backdrop. The relation is real and it does something that has no other name.

One more piece is needed before this assembles, and it is the other never-cited element. R1: there is exactly one past — one totality, one universal wavefunction, not several disconnected ones. Without R1, “the same condition” would be a phrase with no referent across two lives that never touched. With R1, Bell’s workshop and Gray’s are not two worlds that happen to rhyme. They are two regions of one settled totality, and a condition on the circumference is not local to either.

The relation is vertical, not horizontal

Now the assembly. Part III gives every artifact a complex value, z = σ e^iφ. The magnitude is surprisal. The phase φ is which condition the artifact serves. And φ is explicitly relational — fixed by which idea has the actualizer, not by the mark and not by the maker.

So the system already has a well-formed way of saying that two marks stand in a relation without either one touching the other. They share a phase. Neither is upstream of the other. Both are downstream of the same point on the circumference.

Which gives the definition. A coincidence is the reading of two artifacts that share a phase, by a host who has just become able to read it. First marked construction: that is a new formulation, not a numbered result. It follows from P6, P9 and Part III, so it sits in the consequence register, and it should be cited as consequence and not as grounds.

Two things, not one

The definition has two halves and the ordinary word fuses them. Pull them apart and most of the century-long argument stops being interesting.

• The alignment What it isTwo artifacts serving one condition. A structural fact about the record. WhenUndated. It does not begin when you see it, and it holds in the overwhelming majority of cases that no one ever sees. Depends on you?No.
• The noticing What it isA host becoming able to read that phase. P7: naming does not create the condition; it makes the relationship conscious. WhenDated precisely. Tuesday, 4:15pm, the moment the second guest said the name of the hospital. Depends on you?Entirely. This is the whole of it.

Second marked construction, and I want to be exact about its status. This split is built on the same pattern as the evaluation-and-reaching split in Part III, which separates a job that can succeed from a job that cannot. But that split is derived for backward-facing feelings, from R2 and R9, and I am not claiming it transfers. The parallel is structural and I am borrowing its shape, not its derivation. New, and unnumbered.

With the halves separated, both respectable positions turn out to be right about one half and wrong to think it was the whole thing. The debunker is describing the noticing, correctly, and then concluding that because the noticing is in you the alignment is nothing. The believer is describing the alignment, also correctly, and then concluding that because the alignment is real the noticing must have been provided for. Neither conclusion follows. They are separate facts about separate things.

Take the word you learned on Tuesday. Linguists call this the frequency illusion, a term Arnold Zwicky coined in 2005 for something widely reported and, as it happens, barely studied — there is almost no experimental literature on it, which is worth saying rather than dressing it up as a measured effect. The debunker’s account is complete and I have no quarrel with it: the word’s rate in the world did not change, your ability to register it did. But look at what that account requires. It requires that a real change occurred — in the reader, dated to Tuesday. The debunker files this under error because it is not a change in the world. Under the split it is not an error at all. It is the noticing, doing exactly the job the noticing does, and the fact that the word was already in circulation is the alignment, which was there and unread.

Thick conditions and thin ones

The multiples and the party are both alignments, and they are obviously not the same size. Third marked construction, new and unnumbered: conditions differ in how much they condition — call it thick and thin — and the ordinary word covers the whole range without distinguishing.

A thin one: the same hospital, two guests. Real, structural, and almost nothing else in either life stands on it. A thick one: by 1876 the conditions for a working telephone were met — and P6 forbids the obvious misreading of that sentence. Being met did not cause anyone to do anything. It did not push Bell toward the Patent Office. What it means is narrower and stranger than that: a great many makeable artifacts served that condition, and the two men were standing where such artifacts get made.

Stated as something checkable: a thick condition is one that many artifacts serve, so the probability that exactly one host reads it is low. That is why multiples cluster in time and why priority disputes are so bitter and so beside the point. Both men were right about what they had. Neither was right about whose it was, and R12 has already settled that: ownership is a category error and the manifestation is what is accountable.

What it costs

Here is the bill, and it is larger than it looks.

A condition does not intend. That is not a hedge inside P9, it is P9: no wanting, no strategy, no preference. So the believer’s structural claim survives whole — yes, the alignment is real; yes, it is denser than chance; yes, you were reading something that was actually there. The one part anyone wanted does not survive: that it was for you.

The correlation is real, most of it is unread, and none of it was sent.

A message requires a sender with a preference about a recipient. There is no such party available in the primitives. The nodes cannot supply one: He and She do not exist and cannot act, and any sentence in which they do is frame-language and has to be marked. A condition cannot supply one either, by P9. So the sense that a coincidence means something is half right in a way that is worse than being wrong, because the half that survives is the half that cannot comfort anyone.

I have older writing on this site that does not hold to this line. A piece from September 2025 on synchronicity describes these alignments as disclosures, and says they feel like gifts. The structural half of that piece survives the foundation intact — correlations within one universal state, mostly unregistered because attention is narrow. The word gift does not survive it. A gift has a giver.

And the obvious objection is a good one, so it should be stated in its strongest form rather than answered from a distance: is this not just a more elaborate unfalsifiable story, one that explains alignments by positing a layer built to explain alignments? R14 says every idea is biased toward itself, orientation and prejudice always together, and it exempts nothing, including this. The honest answer is yes — this framework has a prejudice in favour of itself and cannot be talked out of it from the inside. What it can offer instead is a cost it is willing to pay: it gives up the message, which is the thing that would have made it attractive. A story built to comfort would not have done that.

Where this is unfinished

Two open problems sit directly under this piece and it closes neither.

The first is Part VIII item three, the φ trajectory. This whole essay leans on a host becoming able to read a phase — that is the entire noticing half. But the foundation says the idea fixes φ and φ determines which idea is readable, which is sound as a dynamical system and circular as a definition. Every sentence I have written about the noticing is drawing on credit the document has not yet extended. Stated so someone can hold me to it: closing it takes a formulation in which φ evolves under a rule stated in the primitives, so that “which condition is readable now” is a consequence of where the host has been rather than a restatement of what it can read. Until that exists, the noticing half of this piece is description, not derivation.

The second is item eleven, and it is the first question a philosophy student asks: if a condition is not an event and has no date, what kind of thing is it? Is this Platonism with new vocabulary? The system has no published answer, and the foundation is explicit that one should not be improvised. So: no answer here. The relation is asserted as real, its relata are not characterised, and anyone who finds that unsatisfying is not making a mistake.

On an ordinary Tuesday

The practical residue is small, which is usually a good sign.

• Separate the two questions before you answer either. Is there an alignment here? is a question about the record and can often be settled. Was it for me? is a question about a sender, and there is not one. People who feel foolish about noticing coincidences and people who feel chosen by them have made the same move: they answered both questions with one word.
• Take the multiples seriously about your own work. If two other people are converging on the thing you are building, that is evidence the condition is thick, which is a reason to move and a reason not to feel robbed. Another person landing on it from a direction you never took is confirmation, not theft.
• Do not name the condition too quickly. Noticing an alignment is cheap. Saying what it is a condition for is a real claim and usually the first one to be wrong. The party guests have a genuine alignment and almost certainly a false account of what it is about.

The debunker and the believer have been arguing about which of two shelves to use, and the reason the argument never ends is that the thing they are arguing about is not on either. It is underneath, holding both of them up, undated, doing no pushing and meaning nothing by it.

• William F. Ogburn and Dorothy Thomas, “Are Inventions Inevitable? A Note on Social Evolution,” Political Science Quarterly 37, no. 1 (March 1922): 83–98. jstor.org
• Robert K. Merton, “Singletons and Multiples in Scientific Discovery: A Chapter in the Sociology of Science,” Proceedings of the American Philosophical Society 105, no. 5 (13 October 1961): 470–486. jstor.org
• The Linnean Society of London, on the joint reading of 1 July 1858. Three items were read in sequence; it was not a co-authored paper. linnean.org
• Encyclopaedia Britannica, “Elisha Gray,” on the filings of 14 February 1876. The claim that Bell was hours earlier rests on the Patent Office cash blotter, which records fee posting rather than delivery. britannica.com
• Arnold Zwicky, “Just Between Dr. Language and I,” Language Log, 7 August 2005 — the coinage of “frequency illusion.” Reported and observational; there is essentially no experimental literature on the construct itself. languagelog