Norms
The Rules Nobody Wrote Down
Introduction
Count the rules you followed today. Not the laws - the other ones. You stood in a queue in the order you arrived. You left a gap between yourself and the person ahead at the counter. You did not read over anyone's shoulder on the train, did not sit directly beside a stranger in an empty carriage, did not begin a conversation with the person next to you in a lift. You spoke to a colleague differently from how you speak to a friend. Not one of these is written anywhere, no authority enforces them, and violating any of them carries no penalty a court would recognize. You followed all of them anyway, without deciding to.
This is the overwhelming majority of social order. Law is the thin, visible, expensive layer on top; norms are the substrate. They govern more of your behavior than legislation does, they operate without any enforcement apparatus, and they are largely invisible until someone breaks one - at which point the reaction is immediate and out of all proportion to the material harm, which is itself the clue to how they work.
The reason norms are worth understanding as a system rather than as a list of customs is that they explain things nothing else does: why a rule everyone privately dislikes can persist for generations, why some rules collapse in months after standing for centuries, why laws that contradict norms are almost unenforceable, and why the cheapest societies to live in are the ones where you can predict strangers.
A Norm Is a Conditional Preference
The most useful definition comes from Cristina Bicchieri, and its value is that it is precise enough to be wrong. A social norm exists when enough people prefer to follow a rule on the condition that they believe most others follow it, and believe most others expect them to follow it. Both beliefs are load-bearing. The preference is conditional - it is not that you want to do the thing, it is that you want to do it given what you think everyone else is doing and expecting.
This distinguishes a norm from two things it is often confused with. A habit is unconditional: you brush your teeth whether or not anyone else does. A moral conviction is also unconditional: someone who thinks cruelty is wrong does not revise that on learning most people disagree. A norm sits between them, and its dependence on beliefs about others is exactly what makes it fragile in a specific way. Change what people believe about each other and the norm can move without a single person changing what they personally value.
The definition also explains the strange intensity of reactions to violations. Someone who jumps a queue has cost you a minute or two. The anger is far larger than that, because what is actually threatened is the expectation that other people are predictable - and predictability is the thing the norm was providing. Punishing the violator is not about the minute. It is about repairing the belief.
What People Do Versus What People Approve Of
Two different things get called norms and they can point in opposite directions. A descriptive norm is what people actually do. An injunctive norm is what people believe ought to be done. Where the two agree, behavior is stable and hard to shift. Where they diverge, there is leverage - and messages that ignore the distinction routinely backfire.
Robert Cialdini demonstrated this in a national park where visitors were stealing petrified wood. A sign reading "many past visitors have removed petrified wood, changing the natural state of the forest" increased theft substantially compared with no sign at all. It was meant as condemnation. It was received as information: apparently this is what people here do. A sign focused only on the disapproval, without the prevalence, reduced theft. The lesson generalizes badly far beyond parks. Any campaign that emphasizes how widespread a problem is - cheating, tax evasion, low turnout, harassment - is broadcasting a descriptive norm in favor of the behavior it opposes.
The constructive version is the same mechanism run forward. Telling households how their energy use compares with similar neighbours reduces consumption among the heavy users, which is one of the most replicated behavioral interventions there is. It also raises consumption among the light users, who learn they have room - the boomerang effect - unless the message adds explicit approval for the low number. Descriptive information alone moves everyone toward the average, including the people who were doing better than it.
Who Enforces a Rule With No Enforcer
Norms hold because violations are punished, usually not by anyone official. The punishments are small and social - a look, a withdrawal of warmth, a story told to someone else - and they are effective, because humans are unusually sensitive to reputation and to exclusion.
But this creates a puzzle. Punishing someone is costly: it takes effort, it risks retaliation, and the benefit of an upheld norm goes to everyone in the group whether or not they helped. The rational move is to let someone else do it. This is the second-order free-rider problem, and if it went unsolved no norm could survive. Experimental work by Ernst Fehr and Simon Gächter shows it is solved by the fact that people simply will punish at a cost to themselves, reliably and even in one-shot anonymous settings where no future benefit is possible. The willingness appears to be a disposition rather than a calculation.
Gossip does most of the heavy lifting in practice, and its reputation does not match how much work it does. Talking about absent third parties is what converts a private observation into a shared one, which is what makes reputation possible at a scale beyond direct experience. It is also cheap: it costs the speaker almost nothing and imposes real costs on the violator. Robin Dunbar's argument that language itself was shaped substantially by this function is speculative in its strong form, but the narrower claim - that gossip is a norm-enforcement technology rather than a social vice - is hard to argue with once you look at what it actually does.
Everyone Privately Disagrees and Publicly Complies
Because norms run on beliefs about others rather than on personal preference, a norm can survive long after almost nobody supports it. Each person complies because they believe everyone else supports it; everyone else complies for the same reason; and the belief is never tested because compliance is all anyone observes. Social psychologists call this pluralistic ignorance, and once you know the shape you find it everywhere.
The classic demonstrations are mundane. College students systematically overestimate how comfortable their peers are with heavy drinking, and drink more to match a norm that mostly does not exist. Employees stay late because they believe leaving on time is judged, while their colleagues stay late for the same reason and judge nobody. In each case, correcting the belief - simply showing people the actual distribution of private opinion - changes behavior, which is what distinguishes this from ordinary conformity.
The consequential version is political. Timur Kuran's account of preference falsification describes what happens when expressing a private view is costly. Everyone conceals. Everyone concludes from the silence that they are alone. The regime looks far more solid than the distribution of opinion beneath it. This is why sudden collapse is the characteristic failure mode. When something makes concealment stop working - a crowd large enough that individual risk falls, a single public defection that goes unpunished - the revision cascades, and a settlement that looked permanent for decades can go in weeks. Eastern Europe in 1989 is the standard example. The striking fact about it is not that observers failed to predict the timing. It is that they were systematically wrong about the strength of what they were watching, in the exact way this mechanism predicts.
The Rules You Are Not Allowed to Calculate About
Some norms are not merely rules against doing something. They are rules against considering it. Philip Tetlock's work on taboo trade-offs found that when people are asked to weigh a sacred value against a monetary one - what a friendship is worth, what price a child's safety - they do not produce a low answer. They react with anger at the question, and then engage in what Tetlock calls moral cleansing, going out of their way to reaffirm the value they were asked to price.
The mechanism is that entertaining the trade-off is itself the violation. To calculate is to concede that the thing can be traded at all, and the norm's entire function is to place it outside that category. This is why introducing money into a domain governed by a sacred norm can degrade behavior rather than improve it. The well-known case is the Israeli daycare centres where a fine for late pickup increased lateness. Parents who had been governed by an obligation were converted into customers buying a service.
This has a direct practical consequence for anyone designing policy. A great deal of public argument fails not because one side's numbers are wrong but because one side is making an argument of a category the other regards as illegitimate. Cost-benefit analysis applied to a domain someone holds sacred does not read as cold-blooded but correct. It reads as evidence of bad character, and the more rigorous the analysis, the worse the reaction.
Societies Differ in How Much Deviation They Tolerate
Michele Gelfand's research programme proposes that cultures vary systematically along a dimension of tightness: how strong their norms are and how little deviation they tolerate. Tight societies have clearer rules and stronger sanctions; loose ones have more permissiveness and more variation in behavior. Her account links tightness to a history of collective threat - invasion, density, disease, resource scarcity - on the reasoning that coordination under threat rewards conformity.
The framework is genuinely useful and the evidence is thinner than its popularity implies. The threat-history explanation rests substantially on correlations across countries where the sample is small and confounds are abundant, and a high-profile application of the theory to national responses to the coronavirus pandemic has been contested on both data and specification. Treat tightness as a real and measurable descriptive dimension, and the causal story about where it comes from as an interesting hypothesis rather than an established finding.
What holds regardless of the causal account is the trade-off, and it is the part worth carrying. Tight arrangements deliver order, coordination, and low ambiguity, at the cost of innovation, tolerance, and adaptability. Loose ones deliver the reverse. Neither is better in general, and the same society can be tight in some domains and loose in others - which is why comparisons that treat whole countries as points on a single scale usually mislead.
How Norms Change
Because norms rest on beliefs about others, they change when those beliefs change - which is why the successful cases look so different from what a rational-persuasion model would predict. Consider smoking indoors, drink-driving, and seatbelts. In each, the information about harm was available and widely known for years or decades before behavior moved. Information was not the binding constraint.
What moved them was a combination: a visible minority willing to bear the cost of open non-compliance, a change in law that made the new behavior common enough to observe, and then the beliefs catching up. The sequence matters. Law works here not primarily by punishing but by coordinating - it makes it common knowledge that a new expectation exists, which is precisely the ingredient a conditional preference needs. Notice also that the change ran through disgust and social embarrassment rather than through argument. Drink-driving stopped being a joke before it stopped being common.
The other reliable pattern is that norms change fastest where non-compliance is visible and slowest where it is private. This is the honest reason that public-facing behaviors shift within a generation while private ones - how households divide domestic work, how people treat family members - move far more slowly despite stated attitudes changing quickly. There is no audience, so there is no enforcement, so the conditional preference has nothing to be conditional on.
When Law and Norm Disagree
A law aligned with an existing norm barely needs enforcing. A law that contradicts one is close to unenforceable at any realistic cost, because the state cannot supply enough observers to replace the millions of ordinary people who normally do the watching for free. American alcohol prohibition is the standard illustration and the standard lesson: the rule was clear, the penalties real, and the norm entirely unmoved, so the activity relocated rather than stopped and the main durable product was an organized criminal industry.
The instructive comparison is with the cases where law did change behavior, because they were not cases of law overpowering a norm. Seatbelt and indoor-smoking laws arrived when a substantial minority already agreed and the majority was uncertain, and their function was to resolve the uncertainty - to make it publicly known which way things had settled. Law is a powerful coordinating device and a weak coercive one, and most disappointment about legislation not working comes from expecting the second.
The same logic sets the limit on the other side. Where a norm is strong and a law is absent, behavior can be tightly governed with no legal backing at all - which describes most of professional life, most of family life, and essentially all of ordinary courtesy. The choice is not between regulation and freedom. It is between a rule you can see and argue with and one you cannot.
The rules that govern most of your life were never voted on, are not written down, and are enforced by people who are not paid to enforce them - which makes them cheap, powerful, and almost impossible to appeal. Their weak point is also their strength: they run on what everyone believes everyone else believes. That is why they can outlive their own support for generations, and why, when the belief finally breaks, they can go faster than anything backed by force ever does.
