A canonical node, FORMING on its face — in this project's own sense: published and live, not yet load-bearing, climbs in public. This is the single home for the finding; everything else references it (the never-twice rule applied to our own ideas). The primary source is Aran Jones speaking, in conversation with Tom Cassidy, 12 December 2020 — wt-aran-forensics/docs/corpus/talk-bollocks/part-2.md, lines 22–86 (a PRIVATE transcript; it is quoted here and may never be published in any form). The account is Aran's, not Tom's, and everything below that is not a quotation is assembly. Where this document extends Aran's account rather than reporting it, it says so in place. Earlier estate documents said this finding "exists nowhere" and was "blocked on Tom's own account of it" — both were written before the Talk Bollocks corpus was persisted on 2026-08-29. The block is lifted, but the source is testimony, and §3 is the ledger of what that testimony does and does not support.
1. What actually happened
Some time around mid-2019 — dated only by Aran's own approximation eighteen months later — the broadcaster Jeremy Vine "got himself into hot water because he had some Twitter storm" about Welsh. Aran is explicit that Vine "wasn't actually trying to say anything nasty at all… He'd just being slightly misunderstood and hadn't expressed himself perfectly. And then he said, sorry." Vine had also "had somebody on his show previously with that hoary old chestnut: I went into a pub and they all stopped speaking English and started speaking Welsh."
That pub story matters to the finding, so Aran's dismantling of it is part of the specimen, not a digression. The myth is a confirmation-bias artefact of how monoglots hear bilingual code-switching: "when you hear one word, of a language you understand your brain will tell you that conversation was in that language." The monoglot hears "fridge", concludes the conversation was in English, then notices it is in Welsh and concludes the speakers switched because of him — without ever asking "what was this magical tool you use to listen to them before you went into the pub?" This is exactly the class of baseless story the finding is about: the kind that, before the sentence was taught, got airtime unchallenged.
Aran went to London expecting hostility from the production side — "'Jeremy has got this bee in his bonnet and there's this Welsh guy coming down, they'll probably burn our studio'" — and got the opposite:
"and basically everybody, I was introduced to said something to me in Welsh. they'd all prepped a word or two in Welsh."
That reception is part of the specimen rather than colour, because it reversed a position Aran himself had held:
"we would always say, and it's certainly in terms of language activism, you'd go. Yep, fine. Classic tokenism - hell with you. That's just not helpful. And we've been utterly wrong about that. Every single one of those people, their effort to speak two or three words of Welsh to me immediately formed an emotional connection that would otherwise not have been there."
And the connection "did not require them to have learned anything else."
Then the teaching itself — and the quantity is the point, so it is carried exactly. "So then I taught him how to say one sentence on radio" — the sentence being built was "I want to learn to speak Welsh today because…", the segment was cut short by the studio ("they were making cut the throat signs at me"), and Aran's own precise accounting is:
"He learned 70% of a sentence or one complete sentence and the beginning of another one."
Less than one full sentence. That is the whole stimulus.
The claimed effect, in Aran's words, with his own dating inline:
"He hasn't had anybody - this is about, must be about a year and a half ago - he's not had anybody on his show since then slagging off the Welsh language."
"Eighteen months" is Aran's approximation, not a measured interval — the transcript says "must be about" and this document carries it that way.
2. What the mechanism plausibly was
Aran states most of the mechanism himself; the last paragraph of this section is extension, and is marked.
A minimum viable stimulus — less than one sentence, produced not received — created a first-hand relationship where there was none. The relationship did not change what Vine believes about Welsh. It changed the filter applied to incoming stories about it:
"there's just been a little switch there… He's had a positive experience with Welsh. He's talked a little bit about it. He can say a few words. And now when somebody comes in and says, Oh, they're all bastards, I want to tell him my story about why I hate them. His filter goes Oh, okay. Give me some evidence."
And Aran rules out, himself, the mechanism it is not:
"He's just going to think 'really?" — not "Oh, I'm not sure Aran would like this, I'd better not."
The filter is epistemic, not social deference. That distinction is load-bearing: a deference filter is a per-relationship favour, excludable and fragile; an evidence filter is a changed epistemic state that runs on every input, whoever brings it. Aran is explicit that a "legitimate story about something that did go wrong… I'm sure he'd run with that" — the filter demands evidence, it does not suppress a topic. "Suddenly the sort of baseless negative story becomes a little bit harder to get through."
The generalisations are also Aran's own. First the opinion-former version: "if we could reach 20 or 30 opinion formers in London and give all of them one sentence, then I think the kind of media discussion you get of Wales and Welsh potentially changes dramatically." Then the explicitly conflict-shaped version, and "dampening effect" is his phrase:
"if you can teach group A how to say one sentence, In language B and group B to say one sentence in language A, you probably immediately got a significant dampening effect because you've got real emotional connections..."
[ASSEMBLER — extension, not report.] Read in CE's vocabulary: the payoff structure of the move is what makes it a specimen. The cost was minutes of one person's time. The payoff arrived roughly eighteen months later, through third parties who were never present when the move was made — the would-be guest with the baseless story who no longer gets through, and every listener who never hears it — and it is non-excludable: nobody can be kept out of it, including people hostile to the thing that produced it. The move did not change any payoff within the existing game; it changed how one participant treats incoming players — the payoff matrix itself. That last clause is the rig's own reading (§4), and it is what makes the specimen more than a nice story about outreach.
3. What the evidence does and does not support
This section is a deliverable, not a disclaimer.
What it is. One man's uncontrolled observation of one broadcaster's programme, recounted once, in a private conversation, roughly eighteen months after the event — and recounted by the person who taught the sentence, who therefore holds the strongest possible prior in its favour. Aran is a careful witness by internal signals — he corrects his own past position on tokenism, he is precise and self-deprecating about the quantity taught ("70% of a sentence"), he approximates his dating out loud, and he rules out the flattering deference reading of his own effect — but careful testimony is still testimony.
What is missing, named. There is no counterfactual — no matched programme, no matched period. There is no base rate: no record of how often Vine's programme ran Welsh-bashing items before the visit, so "none since" has no denominator. There is no independent check — no evidence anyone audited the programme's output, then or since; the claim rests on Aran's monitoring, of unknown completeness. Absence of the item class is also overdetermined: editorial fashion, the Twitter-storm chastening itself, or chance could each produce eighteen quiet months without any filter change. And the effect has one data point on one person.
What would have to be true for this to count as evidence rather than testimony. At minimum: a programme-output audit over a window either side of the visit (the base rate and the after-rate); more than one instance of the intervention with the same shape (the "20 or 30 opinion formers" version, run and logged, would be exactly that); and some probe of the mechanism distinguishable from topic-avoidance — Aran's own criterion serves: a legitimate, evidenced story should still run.
What the finding is good for now, honestly. It is a mechanism hypothesis with one consistent observation, told by a witness who anticipated and closed the obvious wrong reading. That is precisely what a FORMING node is for: strong enough to shape what a rig must be able to express, nowhere near strong enough to lean weight on.
4. What a rig would have to contain for the mechanism to occur at all
This is the part the conflict channel waits on. The fork itself is not decided here — NOTES-ce-rig-v1-world-design.md §13(b) names it as genuinely Tom's: "whether to add the pairwise conflict-cost channel that would render the Jeremy Vine specimen fully rather than by the friction reading." What follows is only the inventory: what the world needs before a Jeremy-Vine-shaped move can even be expressed, piece by piece, with the rig's current position on each.
- A payoff that arrives rounds later. Native. A terrace level tended in round 2 pays every agent every round to the horizon (§10: "Timing and non-excludability: yes, natively").
- Non-excludability. Native. The dividend d·L is paid to everyone unconditionally, defectors included.
- Third parties who were not present when the move was made. Needs only the newcomer channel already designed in §10: two agents entering at round 13 with no history, drawing d·L on a terrace built by strangers. One sentence of tending in round 3; a stranger's round 16 goes better because of it.
- Something that plays the role of a hostile action. Missing, and missing on purpose: §4.2's non-coercion proof banned every hostile action from the action set. Without one, "conflict" cannot fall, because it cannot exist. This is the gap in one sentence: the world has friction but no conflict, so a dampening move can only be rendered as friction removal.
- A per-pair cost that a cheap act can reduce. Missing. §10 says the rig currently renders the specimen as "a payoff that removes friction… the same arithmetic but not the same phenomenon: the real specimen changed how people treat each other, i.e. the payoff matrix itself." The trajectory grades carry some of that weight — cooperation changing the game later players play — but the pairwise term does not exist.
§10's own closing position on the missing pieces: "it is one paragraph of design if Tom wants it, and it survives non-coercion (friction falls on everyone, never targets defectors)." Whether he wants it is the ruling this document exists to make possible.
One reading this inventory supports, offered as a reading and not a design: the filter distinction in §2 cuts at exactly the fork's joint. If the filter were social deference, friction-removal would be the faithful rendering — a per-relationship discount. Because the filter is epistemic and runs on strangers, the faithful rendering is a change to the pairwise term itself, which is the thing §10 says the rig cannot currently express.
5. The community-regeneration thread
The connection is already made in §10 of the world design, in its own words: the unit of a regeneration coefficient is community participation time, and the rig's cost unit is the tended agent-round — ΔL per agent-round of tending is the world's regeneration coefficient, denominated in participation time. And a community is a learner: the trajectory of (L, grades climbed) is a learning curve for the ten-agent community as a single learner.
The Vine case is that thread's instance that already happened — at the scale of one broadcaster rather than one valley. The participation time spent was minutes; the regeneration bought was a durable change in the environment every later participant plays in, arriving through people who were never present and excludable by nobody. If the community-regeneration doctrine is right about its unit, this is what a very high regeneration coefficient looks like in the wild: the specimen the rig would need a conflict channel to reproduce.
Source: wt-aran-forensics/docs/corpus/talk-bollocks/part-2.md (Aran Jones, in conversation with Tom Cassidy, 12 December 2020) — PRIVATE, quoted here, never to be published. Downstream retellings exist in command-surface/rbf/mission.md and the Method Pod scripts; they are retellings, not sources. Rig context: configuration-economics/src/content/open/NOTES-ce-rig-v1-world-design.md §§4.2, 10, 13(b). Cited as: The Jeremy Vine finding, Configuration Economics, docs/canonical/the-jeremy-vine-finding.md.