Evidence

The case against sworn allegiance does not rest on temperament. Seventy-five years of experiments show what group identity does to perception and reasoning, and the membership ledgers of the world’s parties, churches, and unions record a collapse in formal enrolment that nobody organized. What that collapse means is contested — including by the scholars who assembled the numbers — and their disagreement is on this page too, because a page that hid it would not be worth reading. Everything below is cited in full in the Sources.

Part one: what joining does to a mind

The pull of the unanimous room

In 1951, the psychologist Solomon Asch sat volunteers in a room with seven confederates and asked the group to match the length of a line — a task so easy that, alone, people erred less than one percent of the time. Then the confederates began, unanimously, to give the wrong answer. Under that pressure, subjects went along with the visibly false verdict on 32 percent of trials, and three in four caved at least once.¹ The finding has been replicated across seventeen countries; the pull is human, not American, and it is stronger where the culture prizes the group more.²

Two details matter, and the first cuts against the reading this page would prefer. Asch sorted the yielders into three kinds and found the most alarming kind the rarest: a very few came genuinely to see the lines as the majority did, but most kept their own perception intact and concluded they must nevertheless be wrong, and a third group complied in full awareness, simply to avoid standing out.¹ Most of the conformity, in other words, was public compliance with private judgment held back — which is close enough to what this site recommends that it is worth sitting with rather than skipping. What Asch establishes is not that a room reaches in and rewires the eye. It is that a unanimous room can reliably buy the answer, cheaply, from people who know better; and that almost nobody finds out which of the three they are until the room goes quiet.

Second, the spell had a cheap antidote: when a single confederate broke ranks and told the truth, conformity collapsed from 32 percent to about 5.¹ The unanimous room is fragile. It takes one unjoined voice — and, as the Practice page has to admit, someone willing to be the first one.

Teams from a coin toss

Perhaps people defer to groups that matter — faiths, nations, old loyalties with real content? In 1970, Henri Tajfel removed the content. He sorted schoolboys into “groups” by trivia — ostensibly their taste for one abstract painter over another, effectively at random — with no meetings, no leaders, no history, no stakes. Members then divided money between anonymous others. They favored their meaningless in-group at once; more striking, they often preferred maximizing the gap between the groups to maximizing what their own side actually received.³ Group identity, at its bare minimum, was already enough to bend fairness — before a single belief, grievance, or anthem had been attached. Everything a real allegiance adds is added on top of that reflex.

Smart is no defense

The comforting theory is that education inoculates: sharpen the reasoning and the tribe loses its grip. Dan Kahan and colleagues put it to a test. Given a tricky statistical problem about a skin cream, the most numerate subjects did best, as expected. Given the identical numbers relabeled as a gun-control study, that advantage disappeared: numeracy stopped paying where the answer threatened the team, and the distance between partisans was widest among exactly the people best equipped to close it.⁴

Two caveats are owed here, and they are owed loudly, because this is the finding this site would most like to be true. It is a single study, and a high-powered preregistered replication did not reproduce the numeracy interaction specifically.⁵ And Kahan’s own remedy is not disaffiliation; it is trusted messengers inside groups — the researcher whose result this page leans on prescribes the opposite of what this page prescribes. What survives replication is broader and duller: partisan identity biases reasoning, and it does so at roughly equal strength on both sides.⁶ That is enough to carry the argument. It is not enough to carry the claim that intelligence makes it worse.

Related work on deliberation shows the group setting itself radicalizes: people who talk mainly with their own side drift toward more extreme versions of the side’s view than most members held going in.⁷

The mirage across the valley

What the joined mind does to its rivals has now been measured at scale. In a large United States study, partisans on each side imagined roughly twice as many people with “extreme” views on the other side as surveys actually find there. The distortion grows with devotion: the most politically engaged hold the most inaccurate pictures of their opponents, while the politically disengaged are about three times more accurate. Heavy news consumers carry nearly three times the distortion of occasional readers, and frequent political posters carry more than the silent.⁸ Meanwhile the temperature keeps rising: the share of American partisans holding a very unfavorable view of the other party climbed from under a quarter in 1994 to a majority of both parties by 2022, with large majorities of each side calling the other immoral and dishonest.⁹ Distortion and exposure travel together, in other words.

Which one drives the other, this data cannot say — it is a single cross-sectional survey, and the reading that hostile people simply consume more of the machinery fits it exactly as well. Two disclosures follow from that, since the letterhead test applies to this page’s own suppliers as much as to anyone’s. The organization that produced the study exists to reduce polarization, which is a letterhead. And “the disengaged are more accurate” has a duller available explanation: people who are not paying attention give less extreme estimates about everything, which scores well on a measure whose true answer is moderate. A neighbouring misperception has been measured in the peer-reviewed literature, and it is worse — partisans overestimate the demographic composition of the other party by an average of some 342 percent, imagining that a third of Democrats are gay or lesbian where the figure is nearer one in sixteen.¹⁰

Eric Hoffer named an adjacent engine seventy-five years ago: “The less justified a man is in claiming excellence for his own self, the more ready is he to claim all excellence for his nation, his religion, his race or his holy cause.”¹¹ Adjacent, not identical — Hoffer is describing borrowed self-esteem, not misperception of opponents.

The pattern across these findings is one mechanism seen from several angles. Group identity does not merely bias conclusions; it re-tasks perception, fairness, intellect, and imagination to the group’s defense. One honest limit on all of it: none of these experiments involved swearing anything. Asch’s confederates were strangers, Tajfel’s groups were a coin toss, Kahan’s partisans were surveyed rather than sorted. What the laboratory establishes is that mere categorization is already enough. That an oath adds more is this argument’s own inference — plausible, and conceded as an inference rather than smuggled in as a result.

Part two: the quiet exodus

None of this has escaped notice. Without a slogan or a leader — fittingly — people across the democracies have been walking out of formal membership for two generations. Whether they walked or were pushed differs by ledger, and this page marks the difference where it matters. The warning is old; George Washington used his farewell to caution “in the most solemn manner against the baneful effects of the spirit of party,” which “agitates the community with ill-founded jealousies and false alarms” and “kindles the animosity of one part against another.”¹² What is new is that the walking away is now visible in almost every ledger of membership that exists.

Politics. In the United States, a record 45 percent of adults now call themselves political independents — up from 33 percent in 1988, and 18 points clear of either party, each of which is left with 27 percent.¹³ Nearly half of independents say they often wish they had more parties to choose from.⁹ That headline number is also the most contested in American political science, and the objection is a good one: most self-described independents lean, and leaners vote about as reliably as weak partisans, which is why one standard work is titled The Myth of the Independent Voter.¹⁴ Read the 45 percent as a measure of what people are willing to call themselves, not of how they vote. It is still a fact worth having — the label has become something to shed — but it is a fact about identity, not about behavior. Across Europe the enrolment story is starker, because the baseline was higher: parties lost roughly half their absolute membership between 1980 and the late 2000s; in Britain, France, and Italy the losses run to a million or more members each; and by the end of that period members averaged under 5 percent of national electorates.¹⁵

Faith. Worldwide, the religiously unaffiliated grew by 270 million people between 2010 and 2020, reaching 1.9 billion — 24.2 percent of humanity — driven not by birthrates (the unaffiliated have few children) but by adults switching out of the faiths they were raised in.¹⁶ In the United States, the unaffiliated have risen from 16 percent of adults in 2007 to 29 percent today,¹⁷ and formal congregation membership — a distinct, stricter measure of joining — fell below half the population in 2020 for the first time in the eight decades it has been tracked, reaching 45 percent by 2023 against 70 percent as recently as 1999.¹⁸ One correction from the same institution that supplied the trend: Pew now reports the Christian share of American adults roughly stable since 2019, and titled the report saying so “Decline of Christianity in the US Has Slowed, May Have Leveled Off.”¹⁷ A curve that flattens is not a curve that reverses, but a page calling this an ongoing exodus has to say that the ongoing part is, at present, in question. Two other findings belong here rather than in a footnote: the American “nones” are substantially a political phenomenon — a reaction against the religious right, which is a herd exit rather than a set of private conclusions¹⁹ — and Pew’s own projections have the unaffiliated shrinking as a share of world population over coming decades, because the switching that grew them is a one-generation effect and their fertility is the lowest of any group.¹⁶

Labor. American union membership has fallen to 10.0 percent of workers, half its level of 1983.²⁰ This is the line on which the exodus framing is weakest, and pretending otherwise would be dishonest: union approval stands at 68 percent, near a sixty-year high,²¹ and close to half of non-union workers say they would vote for a union if given the chance.²² Density fell against demand, not with it. That makes it evidence about labor law, employer opposition, and sectoral shift — and only very weak evidence about what workers concluded on their own. It is included here because leaving it out would be worse.

Civic life. Measured across the two decades to the mid-1990s, the civic ledger told the same story at every line: attendance at any public meeting on town or school affairs down from 22 percent (1973) to 13 percent (1993); League of Women Voters membership off 42 percent since 1969; Red Cross volunteers off 61 percent since 1970 — while league bowling fell 40 percent even as the number of bowlers rose 10 percent. People kept bowling. They stopped enrolling.²³ These are Putnam’s figures and they are thirty years old; they are cited because they are the canonical series, not because they are current.

What the exodus does not prove

Honesty requires several concessions, and this page makes them rather than hiding them.

First, the scholars who assembled these numbers mostly read them against this site. Robert Putnam reads the civic decline as a loss, and in one sense he is plainly right: the leagues and lodges carried friendship, trust, and mutual aid, and nothing has fully replaced them. Peter Mair — co-author of the European party figures quoted above — read the same collapse as democratic hollowing: electorates and elites withdrawing from one another, leaving parties that no longer need members, are no longer accountable to them, and leave an opening for whoever offers the crowd a substitute.²⁴ If Mair is right, the exodus on this page is a symptom rather than a cure, and the honest response is not to trim him but to print him. This site’s answer is narrow and should be stated as such: that the fault he describes is real, and that re-enrolling in institutions that stopped listening is not a repair for it.

Second, exits have their own herds. Leaving a church because one’s new tribe sneers at churches is just a change of uniform, and — see the “nones” above — a measurable share of the American exodus is exactly that.

Third, the direction of the causal arrow is not established anywhere on this page. Falling enrolment is compatible with people concluding something, with institutions becoming less worth joining, and with law and employers making joining harder. All three are probably true at once, in different proportions per ledger.

What the numbers do show is that hundreds of millions have found formal allegiance dispensable. They do not show that any of them found judgment. That part cannot be measured, because it only happens one person at a time.

The works that argue against this page — on collective action, on exit, on what membership actually delivers, on whether independents are independent — are gathered in their own section of the Sources rather than left for a critic to supply.