A Million People Listening

On the night of the twenty-third of June, in a small town in the Spanish highlands, men walk barefoot across a bed of glowing coals. It is midsummer's eve; the town numbers about six hundred, and the arena built for the night can hold five times that, filled with families home from the cities. The coals are raked flat while the town watches. Then, one at a time, the walkers cross — and most cross carrying someone on their back — a wife, a mother, a friend — because that is how it is done here, and has been for as long as anyone's grandparents can say. Nobody hurries. The whole thing is over in less than an hour, and then it is over for a year.

In 2011 a team of researchers came to the fire with heart monitors, and they wired more than the walkers. They wired the crowd. The obvious prediction was that everybody's pulse would ride the spectacle — fire is fire. What they found instead has kept the study cited for fifteen years: across the ritual's half hour, the hearts of spectators related to a walker rose and fell with his — sharing the fine-grained shape of his arousal, and peaking when their own walker crossed. Strangers in the same crowd, watching the same fire from the same distance, showed no such coupling. And since nobody was touching anybody, the coupling, as the authors put it, must have an informational basis: it travelled on attention alone.

Everyone watched. Only some listened. And the listening was not a decision anyone made — it ran on relation, in the body, below the reach of intention. Attention, it turns out, has a physiology, and the physiology keeps accounts of who belongs to whom.

Keep that finding nearby, because this essay is an audit, and the thing it audits is the boldest line in the manifesto this series extends — the one a fair reader is most entitled to disbelieve, because it sounds like poetry pretending to be a claim:

One person listening is a practice. A million people listening is building reality.

What follows is the evidence, priced honestly: what it supports, what it cannot support, and the warning it carries, which the manifesto does not say out loud and this essay will.

Start in the lab, with the cleanest question: what makes a group of people smart?

In 2010 researchers at MIT and Carnegie Mellon took nearly seven hundred people, sorted them into small groups, and ran each group through a battery of deliberately unlike tasks — brainstorming, moral reasoning, negotiating over limited resources, solving puzzles. The finding that made the paper was structural: group performance was consistent across the tasks. A group good at one thing tended to be good at the others, which means something general was present — a collective intelligence, as measurable for a group as intelligence is for a person, and as predictive of how it would do on the next unrelated thing.

The surprise was what predicted it. Not the average intelligence of the members, and not the intelligence of the smartest member — both correlations were real but weak. What predicted a group's intelligence was, first, how its members scored on a strange little instrument borrowed from autism research: photograph after photograph of nothing but the eye region of a face — thirty-five of them, by the paper's own count — each with four words for what the person behind those eyes might be thinking or feeling. And second, the evenness of the talking. Groups where the conversation was shared moved together; groups where a few voices dominated were, measurably, dumber than the sum of their parts. The paper also reported that groups with more women did better, and traced the advantage through the eyes test — not chromosomes but social sensitivity, wherever it lives.

Read the two predictors together and they have a common name. A model of other minds, and room made for those minds to speak: the laboratory's operational definition of a smart group is a listening group.

An audit reports the audit trail, and this one is genuinely mixed. The finding was pushed on hard, twice: one critique argued the collective factor might be a statistical artifact; another ran three studies of its own and found the members' ordinary intelligence explaining most of group performance — with the turn-taking and women effects failing outright. The counter came in 2021, a synthesis of twenty-two studies and 1,356 groups: the factor held, social perceptiveness held as its clearest ingredient, and a new ranking emerged — how a group collaborates mattered roughly twice as much as who was in it. What that synthesis did not retest was turn-taking itself, so the evenness-of-talk result rests on the original studies and stands contested; and a fair audit notes that the friendliest data comes largely from the originating labs' own network, which critics have not failed to mention.

Two things survive all of it, and they are the two this essay needs. The eyes test kept predicting performance even in groups that collaborated only by text, where nobody could see anybody's eyes — so it is not the eyes; it is whatever the test measures, the standing habit of modeling the person on the other side of the words. And process beat composition: how the group listened to itself outweighed who it was made of, two to one. Individual intelligence is famously hard to move — you are largely issued it. A group's process is not a property of anyone's skull. It can be redesigned this afternoon — the original paper says so itself, noting it would seem "much easier to raise the intelligence of a group than an individual."

Second claim, and this one the audit partly fails.

The famous version comes from Framingham, Massachusetts, where a heart study had tracked thousands of people and their social ties for decades. Two researchers, Nicholas Christakis and James Fowler, mined it and reported that things nobody thought contagious moved through the network like weather — obesity, smoking cessation, happiness — in clusters reaching, in their phrase, three degrees of separation: your friend's friend's friend, a person you may never have met, measurably shifting your odds. The headline number said a friend's becoming obese raised your own chance by more than half.

The critiques arrived fast and they were serious. One team showed the same statistical methods would "detect" the social contagion of acne, headaches, and height. Others proved something deeper: in observational data, contagion and homophily — the fact that like people flock together — are, in the reviewers' word, generically confounded. You cannot fully untangle "my friends made me so" from "I chose friends already like me," not with better math, not with more data. The strong three-degrees claims may be unprovable from observation, permanently.

The confound is worth one concrete sentence, because it is not pedantry. If your friends and you all grow heavier over a decade, the data cannot say whether their habits reached you or whether the thing that made you friends — the same work and its hours, the same tastes, the same closing-time kitchens — was reaching all of you at once. Likeness chooses company and company transmits likeness, and in recorded observation the two are one knot. The authors saw the objection coming and built an ingenious test for it — influence should run along esteem, so one-way friendships should carry it one way, and in their tables they seemed to — but the audit found the test's numbers too noisy to hold its categories apart, and its ordering exactly what likeness alone would predict.

What survives is smaller and better attested. In 2010 the same pair reanalysed a classic laboratory public-goods experiment — 240 strangers, groups reshuffled every round so that no two players ever met twice, which cuts the knot by design. Generosity cascaded. Each unit a player gave drew further giving from people who had merely played alongside her partners, and then alongside their partners — two steps out in the plain game, a third where the rules let players punish free-riders — the original gift roughly tripling through the chain before the effect decayed to insignificance. The radius is modest, the sums are small, and meanness travelled by exactly the same rails, which this essay will need again shortly. But the mechanism is real: what you do to one person alters, measurably, what they later do to people you will never meet — a small, audited fact that half the world's contemplative traditions asserted on no evidence at all.

Third claim: the glue. This is where the ritual scientists come in, and where the audit finds the line's missing warning.

The firewalk study was not a one-off. Dimitris Xygalatas, who helped wire that Spanish crowd, spent years at the Thaipusam festival in Mauritius, where devotees of Murugan pierce their bodies with needles and skewers, shoulder heavy bamboo structures, and drag carts hooked into the skin of their backs for upwards of four hours before climbing barefoot to the god's temple, watched by everyone they know. Afterwards, handed about two days' wages in coins and an anonymous chance to give some to the temple, the people from the extreme rite gave more than the people from the same religion's milder prayer service — and so, remarkably, did those who had merely watched the ordeal without undergoing it. Suffering witnessed bound no less hard than suffering borne; the watchers, in fact, gave slightly more. His book-length conclusion is that ritual is a social technology, and that the senselessness is the engineering: precisely because the acts purchase nothing and cost plenty, the cost is unfakeable, and what it buys is a cohesion that mere agreement never reaches.

Harvey Whitehouse has spent a career sorting that cohesion into two modes. Rituals of the first kind are frequent and calm — the weekly service, the daily office — and they bind lightly and scale widely; whole civilisations run on them. Rituals of the second kind are rare and searing — the initiation, the fire, the front line — and they produce something else, which he calls fusion: the point where the group's boundary and the self's boundary stop being distinguishable, where "we" is no longer a manner of speaking. On the standard measures — which record declared willingness to fight, die, and sacrifice, not the acts themselves — fusion is among the strongest predictors social science owns of what a person says they would give a group, and the correlation has held across a meta-analysis of seventy-eight studies.

And here is the warning the manifesto leaves unsaid. Whitehouse did not build that finding on choirs. Some of its strongest data came from 179 Libyan revolutionaries surveyed after the 2011 war. Nearly all were fused with their families, and just as many with their battalions; forced to choose between the two, a third chose the battalion — and among the men who had shared front-line combat rather than support work, nearly half. The detail that matters most for this essay is the narrowness: fused to family, fused to battalion, and fused to the ordinary Libyans who supported the same revolution — one per cent. The ordeal binds the people who shared it, not the cause in the abstract. The same architecture shows up in football ultras, fused by decades of shared ecstasy and grief over matches that decide nothing. The synchrony machinery passes no judgment on its cargo. The same coupling that ties a village's hearts to a firewalker ties a battalion; a million people listening built cathedrals, and the same million, differently pointed, built pogroms. The line under audit is true the way a lever is true. It moves what it is set against, and it does not ask what that is.

One more voice belongs in the room, because he is the senior author of the group-intelligence paper and spent a book widening it. Thomas Malone's point is that collective intelligence is not a new invention awaiting deployment. Hierarchies, democracies, markets, communities — and, in his least governable fifth category, ecosystems, where the decision rule is raw survival — are ancient machines for thinking with more than one head, each distinguished by how the group decides; and how a group decides is, seen close up, a rule for what it attends to. What is new, he argues, is only the connective tissue — the most important contribution of computers for a long while will not be artificial intelligence but hyperconnectivity, minds joined in new ways at unprecedented scale. Which sharpens the wager of the present decade: whether the new tissue connects people into something that thinks, or merely into something that reacts, at scale, faster than any of its members can think at all.

So the audit returns three findings and a condition.

A group's intelligence is mostly a listening measure — that one stood its replication. Influence travels a few real steps beyond your sight — the modest version survives; the grand version may be forever unprovable. And collective attention, at ritual intensity, welds people into something that acts as one body — for anything.

The condition is the part the traditions already knew, and earlier essays in this series kept finding it from other directions: collective attention was never left raw. It was pointed — at a consecrated object, not at whatever glitters; bounded — the fire is one night in June, not a feed; and relational — the town's pulse that night is not a mass of strangers staring at the same bright thing. That is just an audience, and audiences are cheap now — the last century built them to planetary size and kept mistaking them for communion. Billions have watched the same broadcast at the same moment, and the coupling ran one way, from the bright thing to each separate room; an audience shares an object without sharing each other, which is why it disperses without residue and can be reconvened tomorrow for anything else. What the village has is the other topology. A web of pairs — each heart tuned to its own walker, each walker carrying his own passenger — overlapping until, for one bounded night, the town has a single pulse and knows it.

That is what the boldest line is actually claiming. Not that a crowd is wise — crowds are the counterexample. That listening, practiced one relation at a time and allowed to overlap, is the load-bearing structure of any reality humans hold in common; and that the instruments we build for attention decide whether what overlaps is a web or a mob.

Reality gets built the way the fire gets crossed. Carrying someone, and carried.