Crowds
in a Hurry

The Whole Thing in One Page

The usual picture of a crowd is a person with the person removed. Put enough bodies together, the story goes, and judgement drains away. Individuals become anonymous, emotion becomes contagious, a leader supplies the impulse and the mass moves. It is an old picture, assembled when European elites were frightened by strikes, revolutions and mass politics. It survives because it flatters everyone watching from outside.

The trouble is that crowds do not behave like that.

A crowd is neither a separate mind nor a pile of people behaving alone in parallel. It is coordination under partial information, temporary but real. Everyone brings a history, companions, purposes and expectations into it. Once there, each person has less direct information, more dependence on others and a new field of visible behaviour. People read one another for what is happening and what is permitted. They may begin to think of themselves less as separate customers, travellers or strangers and more as supporters, worshippers, residents, protesters or survivors. That shared identity can alter trust, intimacy, fear, disgust, obligation and the capacity to coordinate.

A gathering also changes phase. Arrival, waiting and dispersal alter what people can see and where pressure gathers. The same event can be orderly at noon and dangerous at its exit.

It does not make every member identical. Crowds contain friends, families, stewards, organisers, opponents, bystanders and several overlapping groups. Their actions remain selective. In the St Pauls disturbance in Bristol in 1980, attacks on police property could become acceptable to participants while attacks on other targets were challenged. The boundary moved; it did not disappear. A crowd can generate a new norm in minutes, but somebody must demonstrate it, others must interpret it and the action must fit what they believe the gathering is for.

Crowd experience can also outlast the event. At pilgrimages and demonstrations, acting with people who embody the same identity can make that identity feel more real, supported and possible. The memory is not merely of having watched something. It is of having enacted a version of the group in public. Sometimes the change fades when ordinary life supplies no reinforcement. Sometimes it becomes a new commitment, ritual or expectation carried home.

Then there is the part psychology cannot repeal. Bodies occupy space. Routes have widths. Flow into an area can exceed flow out. People at the back may not see distress at the front. At high density, movement can become involuntary and ordinary forward motion can produce lethal compression. Calling the result panic or a stampede turns a failure of design, information and management into a story about defective victims.

Authorities are inside this system too. A steward's message can restore information. A police line can protect a route. Indiscriminate force can give previously divided people a common fate and, when experienced as illegitimate, create the hostile unity it assumed was already there. Management changes the crowd it faces.

Numbers therefore do not reveal one hidden human nature. They change the conditions under which human nature operates. The same processes can produce singing, worship, mutual aid, democratic pressure, exclusion, violence or crush. To understand the outcome, ask who came, which identity became active, what norm was visible, what people could know, what the space allowed and how other groups treated them.

That is the book.

Why You Should Care

A packed railway platform has two realities at once. From above, it is a flow problem: bodies entering, waiting and leaving through spaces of fixed width. From inside, it is a social problem: commuters judging whether a delay is normal, reading staff, following familiar routes, making room for a child, resenting a queue-jumper and deciding whether the person beside them is merely close or properly one of us. The same square metre belongs to engineering and sociology. Ignore either and you misunderstand the crowd.

That matters because crowds are ordinary. You enter them at stations, concerts, football grounds, festivals, demonstrations, religious gatherings, sales, funerals and city celebrations. Most pass without incident, which makes them easy to treat as background. Yet they are among the few settings in which strangers coordinate at speed, emotion becomes publicly legible and a private preference can turn into collective power. The things a society celebrates, fears and permits are unusually visible when people assemble around them.

Crowds also change scale faster than institutions can think. Ten people waiting at a door can negotiate. Ten thousand cannot all see the door, hear the same message or know why movement has stopped. Each person acts on a small local sample while helping create the condition everyone else must interpret. A harmless decision at the rear can become pressure at the front. A rumour can matter because official information is absent. A few people running may signal danger, excitement, a train or nothing worth copying. Numbers make information social before they make it accurate. The event also changes phase: arrival, waiting, peak activity and dispersal create different dependencies even when many of the same people remain.

The old answer was to distrust the mass. It gave officials a simple story: civilised individuals enter, primitive impulses emerge, control must be imposed from outside. That story is still available whenever a protest becomes tense or a crowd disaster needs a culprit. It is also a poor guide. People in emergencies commonly help one another. Collective violence, when it occurs, has targets and limits. Dense gatherings can feel joyful rather than aversive when proximity expresses belonging. Police tactics, barriers, announcements and route design do not merely respond to behaviour; they alter the identities, expectations and movements from which behaviour comes.

Get the model wrong and the consequences are practical. Treat a diverse gathering as a hostile block and you may unite it against you. Withhold information because you expect panic and you remove the material people need for orderly action. Count total capacity while ignoring a narrow exit or local pocket and the average will reassure you until it is useless. Describe lethal compression as a stampede and attention shifts from inflow, barriers, monitoring and communication to the supposed savagery of the dead.

Get it right and a crowd becomes readable. You can distinguish physical closeness from shared identity, dissent from disorder, urgency from irrationality, emotion from mindlessness and coordination from obedience. You can see why the same people may regulate one another in one setting and confront police in another. You can design safer spaces, communicate through trusted relationships and notice when ordinary movement is becoming a system nobody inside can control.

It also changes what counts as evidence about people. The calm queue, the shared chant, the person lifted over a barrier and the refusal to join an attack are not exceptions to crowd behaviour. They are crowd behaviour. A theory that notices only the broken window has selected its conclusion before it begins.

The promise of this book is not a formula for predicting every gathering. Crowds are too varied, evidence is too setting-specific and human beings retain choice. The promise is a better set of questions. By the end, you should be able to look at a mass of people without inventing a mass mind, while still seeing what numbers have changed.

The Core Ideas

A Crowd Is Coordination Under Partial Information

Imagine joining a queue whose front you cannot see. The people ahead move, so you move. They stop, so you stop. You are making choices, but much of your evidence is somebody else's choice. Extend that chain through hundreds of people and the result can look purposeful from above without anyone possessing the whole plan. No group mind has appeared. Nor are these independent individuals happening to do the same thing. Their decisions have become connected: coordination under partial information.

The first distinction is physical. A physical crowd is a number of people gathered in the same space. They may share nothing beyond a platform, pavement or deadline. A psychological crowd exists when some of those people understand themselves as members of the same group in that situation. A packed concourse may contain commuters, away supporters, stewards, police, families and people trying to pass through. One headcount, several possible crowds.

That distinction matters because co-presence changes conditions before it changes identity. Sightlines shorten. Sound becomes noisy. Personal movement depends on neighbouring movement. An individual can no longer inspect the whole situation and must use other people as sensors. If those ahead stop, those behind infer a cause they cannot see. If a route opens, the first few movements advertise it. Each person responds to a local field while helping create the field for everyone else.

This is a coupled system: each person is reading and changing the situation for others. Someone pausing at a corner can make the next person swerve, slow those behind and eventually obstruct the route they meant to take. No extra brain is required. A queue forms because people recognise a relation among arrival, entitlement and position, then make that relation visible by standing in it. A cheer begins with a few voices, becomes evidence that cheering is appropriate and gains force from the response it has produced. A blockage grows because each person arriving from behind sees occupied space rather than the invisible reason it is occupied.

Temporary does not mean shapeless. Crowds inherit architecture, schedules, policing, tickets, social divisions, memories and the purpose of the event. A funeral crowd and a bargain-sale crowd can occupy the same square and produce different distances, voices and obligations. People are not reacting to density alone. They are interpreting what the gathering is and who the others are.

A crowd also changes through phases. During arrival, people orient to entrances and companions. While waiting, they establish positions and expectations. A door opening, performance ending or route closing can reverse flows within seconds. During dispersal, the social purpose may be over while the physical crowd is at its largest and most mobile. The same headcount therefore describes different systems at different times. A still photograph cannot show whether demand is accumulating, clearing or about to change direction.

This is why averages mislead. The average person may have enough space while one pocket is compressed. Most participants may be calm while a small group confronts an opponent. The whole may appear to move towards a gate although many individuals are resisting the pressure around them. Aerial footage turns variation into a surface. Analysis has to put the differences back.

The system model also prevents romance. Shared identity can increase care and coordination, but it can draw a sharper line around outsiders. Local imitation can spread a useful route or an ugly act. Self-organisation can solve a problem or conceal a dangerous condition from managers. Emergence means an outcome arises from interaction. It does not mean the outcome is wise.

The right opening question is therefore not, what does this crowd do? Ask what system has formed. Who is co-present? Which relationships matter? What can each part see? Which actions are changing the options for others? What physical limits are tightening? The crowd has no mind of its own, but numbers have built a situation no member controls alone.

People Bring Histories In

Crowd theories often begin too late. The camera finds people already assembled and treats them as interchangeable units. Yet the most important causes may have arrived hours, years or generations earlier.

People rarely enter major gatherings as detached strangers. They come with friends, relatives, colleagues, fellow supporters, congregations, campaign groups, travel parties and informal leaders. These small groups affect pace, spacing, route choice and responsibility. A parent does not evacuate as an isolated pedestrian if a child is missing. Friends slow to stay together. Organised protesters carry agreed aims and stewarding practices. Football supporters know songs, reputations and old disputes before anyone reaches the turnstile.

Selection matters too. A crowd is partly made by who chooses to join. The people attending a climate march are not a random slice of the city; nor are worshippers at a pilgrimage, shoppers at a launch or residents confronting a police operation. This was the strength of convergence theories that replaced the old group mind: similar dispositions and purposes help bring people together. Floyd Allport's sharp reply to Le Bon was that aggregation amplifies what individuals bring rather than creating a new collective mind.

It was also too neat. Existing motives do not dictate a single action, and a gathering can create possibilities no participant could enact alone. A person may arrive angry but refuse violence because it violates the group they identify with. Another may arrive merely curious and become committed after experiencing support. A peaceful supporter may confront police after an indiscriminate intervention changes who appears to be under attack. Selection explains who is available for influence. It does not finish the story of what the situation makes meaningful.

History enters through relationships between groups. A police uniform is not a neutral patch of colour to someone whose neighbourhood has experienced repeated searches. A barrier can mean routine safety at one event and exclusion at another. The same order to move may be read as sensible coordination, humiliating control or evidence of danger, depending on who gives it, how it is given and what happened before. Meaning is compressed history.

E. P. Thompson's study of eighteenth-century English food crowds made the same correction to an older economic picture. Food disturbances were not reflex spasms produced by hunger. Participants acted through ideas about fair prices, customary obligations and the proper conduct of sellers. Their actions could be forceful, but they were organised by a moral economy. The grievance did not merely supply energy. It supplied a standard.

The history is unequal. Some participants possess radios, maps, authority, vehicles or protected routes; others possess only what they can infer locally. Some can leave without losing status or safety. Others experience departure as betrayal. Some are watched as citizens exercising a right. Others are watched as a problem before they have acted. These differences shape the same physical gathering from within.

Composition therefore belongs beside density on any serious crowd map. Count age, mobility, companions, language, familiarity, purpose and group relations, not because demographics mechanically predict behaviour but because they alter needs and interpretations. A hundred schoolchildren, a hundred commuters and a hundred organised campaigners are not three samples of the same substance.

Numbers amplify what people bring, but they also rearrange it. The crowd begins before assembly in invitations, grievances, memories and relationships. It continues after dispersal in stories about what happened and who stood with whom. The event is temporary. Its ingredients are not. The same headcount assembled under another history would not produce the same field of meaning.

Shared Identity Changes the Self

The stranger beside you has not changed, but your relationship to them can change in seconds. An interruption that affects you both, a shared chant or recognition of the same colours can make their presence feel different. The important shift is from someone nearby to someone on your side. Social identity is the working definition of who counts as us here and what follows from that membership.

People carry many identities. Which one guides action depends on the setting and on contrast. Commuters who ignore one another on an ordinary platform may become Londoners after an attack, passengers during a breakdown or residents when an official dismisses their neighbourhood. Nothing personal has vanished. The level at which the person understands the self has shifted.

Once another person counts as one of us, their reaction can carry more weight. Their alarm may concern a danger that threatens us both; their confidence may suggest help is available. Expected support can make it easier to speak, share resources or attempt something together. Even proximity can feel less intrusive. The same closeness that would be unwelcome from an unrelated stranger may be part of why a supporter or worshipper came. The change is in the relationship, not the number of centimetres between bodies.

This is why dense gatherings are not uniformly unpleasant. Older personal-space accounts predicted that closeness would produce stress because strangers were violating a fixed boundary. Field studies at particular mass gatherings found a more conditional pattern. Proximity among people understood as fellow members could be welcomed, and shared identity was associated with reported safety, comfort or positive emotion where being together expressed the point of attending. The same density among hostile or unrelated people can feel invasive. Bodies are equally close; the social space is different.

The Magh Mela, a month-long Hindu pilgrimage gathering in northern India, makes this more than a theory of brief excitement. Research with pilgrims linked shared identity to intimacy, expected support and the sense of being able to live their faith together. Later measurements found stronger Hindu identification and more frequent prayer among attendees. The event offered public confirmation of a way of life: other people were there to practise it too. An identity could be carried home stronger than it arrived.

A study of Pakistani applicants to the Hajj visa lottery provides a different kind of evidence. Compared with unsuccessful applicants, those who attended later reported greater commitment to some global Islamic practices and more favourable attitudes towards equality across ethnic, sectarian and religious lines, alongside reduced attachment to some local practices. The lottery helps distinguish the effect of going from the desire to go among these applicants. It cannot tell us which part did the work: travel, ritual, contact with others and the status of having made the pilgrimage came together. This is evidence about that whole experience, not a promise about any gathering.

Shared identity also helps explain resilience in emergencies. Common danger can create a new group among strangers. When people expect others to support them, coordinated aid can become easier to attempt and sustain. Survivors hold doors, pass information, assist injured people and organise movement. This does not happen in every emergency, and conflict does not disappear. It does happen often enough to make universal panic a bad starting model.

The limit belongs inside the idea. An enlarged us can create a smaller them. Solidarity within a political crowd may accompany hostility towards police, counter-protesters or stigmatised outsiders. A group can validate courage, care and restraint, or cruelty and exclusion. Shared identity increases the capacity to act together. The content of that identity decides what together is for.

Norms Form in Public

A crowd often enters a situation without a complete script. A queue for a familiar till usually supplies a recognisable script. Far fewer know the correct response when the doors stay shut, a siren sounds, a speaker is removed or police block the route. Uncertainty does not empty people of standards. It makes standards something they must discover from one another.

This was the central advance of emergent-norm theory. Ralph Turner and Lewis Killian argued that collective behaviour is organised by definitions of the situation that develop through interaction. People watch conspicuous acts, the responses they receive and the absence of objection. One person climbs a barrier. Others see whether the act is condemned, ignored or followed. A chant is offered. Its uptake makes support public. What looked like private possibility becomes a candidate norm.

The process is selective. Not every loud person becomes a leader and not every action spreads. An act must be intelligible in the gathering's terms. Existing identities and purposes help decide who is credible, what counts as evidence and where the boundary lies. This solves a problem left by the language of contagion. If behaviour merely infected neighbours, anything could spread equally. In real crowds, some actions travel and others are rejected.

Stephen Reicher's study of the St Pauls disturbance in Bristol in April 1980 remains influential because it examined the limits, not merely the violence. The confrontation followed a police raid in a neighbourhood marked by racial tension and conflict with authority. Within the event, attacks directed at police could acquire legitimacy for participants. Yet action was not indiscriminate. In the analysis, damage to some targets was criticised or prevented because it did not fit the developing definition of what the crowd was doing. The norm changed from ordinary street conduct, but it retained an object and a moral boundary.

This is common beyond disorder. At concerts, audiences regulate talking, filming, dancing and care for anyone who falls according to genre and local expectations. At demonstrations, marshals, speakers and veteran participants model how to occupy space, respond to provocation and treat passers-by. At commemorations, silence becomes enforceable because thousands display it together. The force of the norm comes partly from visibility: each participant can see that others expect it.

Visibility creates a trap called pluralistic ignorance. People may privately doubt an action while assuming everyone else accepts it because nobody objects. Silence then looks like agreement and helps reproduce the norm many dislike. The reverse is possible too. One credible refusal can reveal hidden dissent and make further refusal easier. Numbers increase conformity in some settings because the social evidence is thick; they also make norm change dramatic when the evidence shifts.

Leaders matter, but not as hypnotists. A successful crowd leader defines who we are, what is happening and what action expresses the group. The claim must fit participants' experience and survive what others do. A megaphone can amplify a proposal, not manufacture legitimacy. Informal figures can matter more than formal speakers if the crowd trusts their knowledge or recognises them as representative.

Norms can emerge quickly without being arbitrary. They are built from prior values, visible action, shared interpretation and practical success. The crowd is not freed from social rules. It becomes a place where rules are exposed, tested and sometimes rewritten in public.

Information Travels Through Others

Inside a large crowd, nobody has the view used to describe it afterwards. The organiser has cameras and radio channels. A helicopter sees movement without hearing conversation. A person near the front can see the barrier but not the inflow behind. Someone at the rear sees people advancing and reasonably assumes there is room ahead. Crowd behaviour grows from these partial views.

Other people therefore become an information network. Direction of gaze can identify a disturbance. A sudden turn suggests a route or threat. A stationary line signals blockage, though not its cause. Speech, gestures, phones, signs, chants and uniforms carry interpretations. The network is quick because it is distributed and unreliable because each node observes only a fragment.

This is where imitation earns a place, but the word needs discipline. Copying another person's movement may be efficient when information is scarce. Several people using one doorway can make it look like the right exit, even when those watching cannot see what lies beyond it. Others waiting can make waiting look prudent. Yet visible popularity can produce an information cascade in which later people follow earlier choices rather than independent evidence. The crowd may converge on a route because it is good, because it was first or because alternatives are invisible.

Emotion travels through the same interpretive process. The old metaphor of contagion imagines fear or anger passing like an infection, beneath thought. Facial expressions, voices and bodily arousal can influence others, but people also ask whose emotion this is and what caused it. Fear shown by a trusted steward facing the gate means something different from theatrical fear in a rival group. Anger can spread when observers share the judgement that an action was unjust. It has not become airborne.

Information also changes identities. A police charge seen on a screen at one end of a protest may be interpreted as an attack on people like us elsewhere. Images from another city can make participation seem possible, popular or necessary. Waves of protest and disorder are therefore more than duplicated grievances. One event supplies evidence about what others believe, what authorities will do and whether action can work. Social media accelerates that movement, but the mechanism predates the phone: reports, songs, newspapers, travellers and rumour have long connected gatherings.

The hardest problem is meta-perception, what one group believes another group thinks of it. Supporters may expect police to see them as hooligans; police may expect supporters to see any intervention as hostile. Each anticipates the other's suspicion and adopts a posture that confirms it. A crowd can escalate around beliefs about beliefs before either side intended the final conflict.

Clear information helps only when it is credible, timely and usable. A public-address message heard at the front but not the back can worsen coordination. An instruction without a route is not a solution. Reassurance from an authority that has lost trust may be read as concealment. Communication is a relationship, not sound delivered through a loudspeaker.

This is why withholding information to prevent panic is often perverse. People do not become calm because they know less. They fill the gap with local evidence, which may be sparse, delayed or misleading. The practical task is to build enough trusted channels that different parts of a crowd can act on the same situation. In numbers, ignorance is rarely individual. It is structured by where people stand and who they believe.

Space Sets Hard Limits

A crowd can forgive discomfort, coordinate movement and care for strangers. It cannot negotiate with geometry.

Every venue is a system of entry, circulation, holding and exit. Its capacity is set by the weakest part, not by the total floor area. A wide field served by one narrow gate is a narrow-gate venue. A platform safe at noon may become unsafe when a delayed train adds arrivals while departures stop. Barriers that organise one flow can trap another. The dangerous quantity is often not how many people are present but where they are, which way they are moving and whether inflow exceeds outflow.

Pedestrian flow has a simple vocabulary. Density is people per unit of area. Speed is how fast they move. Flow is how many pass a line in a given time. The relationships are not fixed across cultures, ages, purposes or layouts, but the broad pattern is intuitive. At low density, people choose pace and direction. As density rises, avoidance becomes harder and speed falls. At greater compression, movement can be transmitted through contact. A person may be carried sideways or forwards without choosing to move.

Local conditions matter more than a single average. Two crowds with the same density can differ if one moves in one direction and the other crosses, if one consists of connected groups and the other of individuals, or if speed varies sharply. A camera that reports the mean can miss a pocket beside a barrier where forces are accumulating. Danger is dynamic.

Crowd crush is often misnamed a stampede. The animal image suggests frantic running and trampling. In many lethal incidents, people die from compressive forces and asphyxia in a mass too dense for individual escape. Those behind may continue ordinary forward movement because they cannot see that people ahead are immobilised. Once a pile-up or blockage forms, the information needed to stop inflow is least available to those supplying it.

Bethnal Green in 1943 shows social and physical causes joining. During an air-raid warning, people entered an unfinished Underground station used as a shelter. A fall on the staircase developed into a lethal crush in which 173 people died. A recent reanalysis of inquiry evidence challenges the familiar account of irrational panic at unfamiliar rocket fire. Urgent shelter-seeking made sense in wartime. What people entering could not see was the obstruction below. The staircase joined people acting on different information into one physical emergency.

Evacuation is similarly social. People seek companions, use familiar exits, follow staff they trust and delay while interpreting ambiguous signals. These actions can slow movement, assist vulnerable people or prevent separation. Models that treat every pedestrian as an identical particle may expose bottlenecks yet miss the reasons routes are chosen. Models that celebrate solidarity may miss the point at which contact forces remove choice. Safety needs both.

The management lesson is unsentimental. Monitor local density and flow. Keep routes legible. Control arrival before the front becomes trapped. Communicate to the rear as well as the people already in distress. The Sports Grounds Safety Authority formalises the weakest-link principle through entry, holding, normal exit and emergency exit capacities, with physical condition and management quality also assessed. Other venues need their own competent assessment. Once people cannot control their movement, exhorting them to behave sensibly is no substitute for changing the conditions that have trapped them.

Authority Helps Make the Crowd It Faces

The fantasy of crowd control puts authority outside the event. The crowd is the moving object; police, stewards and organisers are the hands on the controls. In practice, every barrier, order, uniformed formation and intervention becomes part of the situation participants are interpreting.

This is clearest in protest and public-order policing. A demonstration may contain committed activists, first-time marchers, families, legal observers, journalists, passers-by and a small number seeking confrontation. If police recognise those differences, facilitate lawful aims and target any intervention narrowly, participants can retain distinct relations to the event. Many have reason to distance themselves from violence and regulate people who threaten the gathering.

If police treat everyone as a single hostile category, the structure can change. An indiscriminate containment, charge or use of force creates common experience across previous differences. People who did not identify with confrontational participants may come to see both themselves and them as targets of the same illegitimate action. The boundary of us widens. Resistance that would previously have seemed improper can become defensible. Authority has helped produce the unity it claimed to be responding to.

The elaborated social identity model was built to explain this process. It does not say police cause every conflict or that crowd members lack agency. It says collective action develops through intergroup interaction, especially where one side has the power to impose a category on the other. Expectations, legitimacy and perceived possibility change through the encounter.

Research around England supporters at the Euro 2004 football championship offered a useful within-tournament contrast. The same broad category of fans behaved differently across contexts. In match cities, communication, facilitation and intervention against specific offenders were associated with self-regulation and less conflict. In other settings, more antagonistic and less differentiated relations were associated with broader opposition. The authors interpreted the pattern through intergroup interaction. The evidence is observational, so it does not prove that one policing style produces the same effect elsewhere. The label England fan did not carry one behaviour around Portugal. Interaction changed what the identity meant.

Authority includes more than police. Event managers decide routes, ticketing, alcohol provision, timings, barriers and information. Stewards decide whether a rule is explained or barked, whether a person in distress is recognised and whether a local warning reaches control. Speakers define opponents and acceptable tactics. Media and officials later decide whether an event is remembered as celebration, disorder or disaster. These judgements affect the next gathering before it assembles.

Power also determines whose account becomes official. After Hillsborough in 1989, supporters caught in lethal compression were blamed through false claims about their conduct. The 2012 panel cleared Liverpool supporters entirely and endorsed the Taylor Inquiry's identification of failed police control as the principal cause. The 2016 inquests found unlawful killing, and the 2025 IOPC and Operation Resolve report again excluded supporter conduct as a cause. The first crowd theory did political work: it redirected attention from management towards the character of the people who suffered.

Good management is not passive. It distinguishes groups and behaviours, communicates early, enables legitimate purposes, monitors changing physical conditions and acts proportionately when intervention is needed. It also recognises that trust is an operational resource. A technically correct instruction from a distrusted source can fail; a relationship built before pressure rises can make the same information usable.

The person trying to interpret a stalled queue or a line of police has only part of the picture. Authorities help determine both what they can know and what happens next. A useful explanation, a viable route or a selective intervention can preserve room for different choices. Blanket treatment can remove that room and give previously divided people a common grievance. The officials have not discovered a crowd with a fixed nature. Their decisions have become part of the crowd's next decision.

How It Actually Works

The mass becomes a problem

In 1895 Gustave Le Bon published Psychologie des foules, translated the next year as The Crowd. Europe had been shaken by revolutions, the Paris Commune, strikes, expanding cities and the arrival of mass electorates. The crowd was not new. Markets, pilgrimages, carnivals, armies and food protests were ancient. What felt new to elites was that ordinary people assembled in public could become a political actor.

Le Bon supplied that fear with a psychology. In his account, individuals entering a psychological crowd lost conscious personality, descended intellectually and became suggestible. An unconscious racial inheritance took over. Emotion passed by contagion, leaders supplied images and repetition, and the mass became capable of heroism but especially destruction. The educated man could be reduced to the same mental level as the supposed barbarian beside him.

The theory was sweeping, memorable and barely constrained by systematic observation. Le Bon classified juries, parliaments, religious groups, criminal mobs and electorates as crowds, often moving from a political judgement to a psychological mechanism without evidence between them. Yet he understood one problem that later theories had to solve: collective conduct cannot be explained by adding together independent private choices. People in numbers do become mutually oriented. Le Bon's error was to make that coordination proof of mental decline.

He was not alone. Scipio Sighele, Gabriel Tarde and others had already treated crowds, imitation and criminal collectivity as central problems. Le Bon's prose won the inheritance. His model fitted official interests because it made disorder a property of the assembled people. If the crowd was irrational by nature, authorities needed neither to understand its grievance nor to examine their own part in the encounter.

The individual strikes back

Floyd Allport's Social Psychology of 1924 attacked the group mind at its weakest point. A crowd contains nervous systems, but none belongs to the crowd. Only individuals perceive, feel and act. His compressed answer was that “the individual in the crowd behaves just as he would behave alone, only more so”. Existing preparation and emotion are intensified; concerted action becomes possible; no collective soul is required.

This cleared away metaphysics and opened convergence accounts. People with related grievances or preferences select the same event, recognise allies and express tendencies that were present before assembly. A racist mob is not manufactured from innocent minds by proximity. A demonstration is not inexplicable if participants already oppose the policy. Composition matters.

But Allport's solution flattened interaction. A choir is made of individual voices, yet its harmony cannot be explained by examining each singer alone. People alter timing, volume and pitch in response to others. In a crowd, new information, relationships and possibilities arise from co-presence. The person who would privately complain may publicly act once support becomes visible. The peaceful marcher may change position after police treat everyone nearby as an enemy. Behaviour can be continuous with prior values without being pre-written.

Rules appear in motion

After the Second World War, sociologists returned interaction to the picture. Ralph Turner and Lewis Killian's emergent-norm theory argued that unusual collective situations create uncertainty about what conduct fits. Participants mill, communicate, watch conspicuous actors and infer a working definition of the situation. New norms may arise because a suggestion is followed, an act is not punished or a response proves effective.

This was a major correction. Collective conduct became meaningful and regulated rather than a collapse into instinct. It also raised hard questions. Why does one proposal become normative while another dies? Why do some people count as credible? Why does action spread towards selected targets rather than everywhere? Existing relationships, identities and power had to return.

Historians were making a parallel move. E. P. Thompson's 1971 essay on the moral economy of the eighteenth-century English crowd rejected the image of food riots as spasms triggered by price. Participants acted through shared expectations about customary prices, obligations and the right to subsistence. They could seize grain, set a price and return money to the seller. The conduct was illegal and sometimes violent. It was not without a rule.

The historical correction matters because records are political. Officials often described disorder through the moment control failed, while participants understood it through the grievance, custom or provocation that preceded the report. A crowd history beginning with the broken window will discover a psychology of breaking windows.

Watch the gathering, not the myth

Clark McPhail pushed sociology towards detailed observation. Rather than treating a gathering as one actor, he broke it into assembling, clustering, queueing, chanting, moving, watching, dispersing and the many people who did none of the dramatic things later used to name the event. His 1991 book The Myth of the Madding Crowd attacked both the primitive group mind and the assumption that a crowd was merely a collection of pre-formed dispositions.

At a demonstration, people may listen, talk, leave early or confront police. News footage selects movement and conflict because stillness is poor television. The event can then be remembered through its most visible five minutes. Observation restores what the clip omits: how many did what, for how long, in relation to whom?

It also shows coordination at small scales. People arrive in groups, orient to companions, negotiate passage, copy gestures and create temporary lines. Collective behaviour does not require all members to perform the same act. It requires interdependent conduct, which can include complementary roles. One person speaks, others listen; some form a line, others pass supplies; one cluster confronts while another restrains it.

McPhail's approach was strongest on what bodies did and weaker on why a shared identity made one action meaningful. The next shift joined observation to a social account of the self.

St Pauls and the identity turn

On 2 April 1980 police raided the Black and White Café in St Pauls, Bristol. The neighbourhood had experienced racial discrimination, poor housing and conflict with police. The operation became a confrontation and then a disturbance lasting hours.

Stephen Reicher studied accounts of the event with an unusual question: where did the action stop? Classical theory expected anonymity and contagion to loosen all restraint. Convergence expected participants to reveal tendencies they already possessed. The observed pattern was more selective. Police became a legitimate target for many participants within a developing definition of resistance, and actions against some other property were challenged. People did not cease to judge. They judged as members of a group whose relation to police had become central.

The explanation drew on social identity and self-categorisation theory. A person can understand the self as a unique individual and as a member of social categories. When a group identity becomes salient, behaviour follows the norms and interests attached to that shared self. This does not mean mechanical conformity. People debate what the group stands for, who represents it and what action fits. Identity provides a field of agreement within which conflict about meaning continues.

Reicher's later study of the 1990 poll-tax demonstration in Westminster developed the intergroup part of the model. Police action could change the crowd's boundaries and the perceived legitimacy of resistance. The elaborated social identity model made crowd behaviour a developing relation between groups, not a property carried by one side into the street.

That is a deeper break from Le Bon than saying crowds are nice. The old theory placed civilisation in the individual and pathology in the mass. The identity approach asks how social definitions organise both restraint and violence. A crowd can become generous or brutal without becoming mindless. The question is what conduct expresses the active us.

From riots to ordinary gatherings

For years crowd psychology studied the events officials already considered problems. Religious gatherings, music festivals and celebrations posed a different question. Why would people actively seek the noise, closeness and inconvenience that were supposed to make crowds unpleasant?

At the Magh Mela, researchers tested a small part of that puzzle with an ambiguous recording. Thirty-two male pilgrims heard the same mixture of sounds. Some were told it came from the pilgrimage, others from a city. Those given the pilgrimage description reported a better mood and less discomfort, and listened longer. The sound had not improved. Its apparent source had changed. This tested discomfort, not hearing safety. It nevertheless made interpretation audible: what the listener thought they were hearing helped shape how it felt.

The Hajj supplied a related distinction between crowding and the experience of being crowded. In a 2012 sample of 1,194 pilgrims at the Holy Mosque, stronger identification with fellow pilgrims altered the relationship between density and reported safety. This measured feeling safe, not protection from a crush. Put beside the sound experiment, it makes the same useful separation: physical conditions matter, but their meaning is not fixed by a sound meter or headcount.

Researchers also followed movement. A controlled field study on the University of Sussex campus tracked 280 pedestrian trajectories. Under the tested shared-identity condition, participants moved more slowly, travelled farther and kept closer proximity than comparison pedestrians. This was ordinary walking, not escape at dangerous density. Staying with others can mean accepting a slower route or less personal space. Those choices change the trajectories that a pedestrian model must explain.

The work also returns joy to a field long preoccupied with trouble. Singing with others makes participation audible: a private liking becomes something people express together. Research on collective effervescence, the term associated with Émile Durkheim, asks how shared identity and recognition help make such occasions emotionally powerful. The aim is to explain the experience without explaining away the pleasure.

The gathering changes phase

Consider a concert that has just ended. For the audience, the common task has finished. For crowd management, a different task has begun. People who arrived over several hours may now want to leave within minutes. Some search for companions while others head for transport. A queue that worked on arrival may obstruct people moving the other way. This is an illustrative sequence, not the reconstruction of a particular event: the crucial change is in demand and direction, even before anyone's mood changes.

A football crowd also changes between kick-off and the final whistle, although many of the people remain. A protest can fragment during dispersal. A station crowd can become a stranded-passenger group after a cancellation. Naming an event once does not explain its changing purposes, relationships or movement.

Good observation therefore follows transitions: who arrives, what accumulates, what trigger changes direction, how communication reaches the rear and what happens when the common purpose ends. Crowd behaviour is made through sequence. The same intervention can be helpful early, irrelevant later and dangerous after the flow has reversed.

Emergencies without the panic script

The classic disaster story is each for themselves. Survivor accounts repeatedly complicate it: people seek companions, pass information and help someone who cannot move unaided. In studies of the 2005 London bombings, survivors described solidarity with strangers as well as fear. Shared danger could make concern extend beyond the people they already knew. Nobody had to become fearless first. Expected help made it easier to act, while the act of helping gave others reason to expect it.

This is not a claim that fear disappears. People can flee, freeze, compete, misread danger or protect their own group at others' expense. Delay is often as serious as overreaction because ambiguous warnings are discounted, routines continue and people seek confirmation. The useful correction is narrower: uncontrolled mass panic is not the default explanation, and calling an outcome panic can conceal the conditions that structured it.

Bethnal Green is a case in point. On 3 March 1943, during a wartime air-raid warning, thousands sought an Underground shelter through a staircase that could not safely absorb the inflow. A fall became an obstruction and 173 people died in the compression. The event was long described as panic triggered by unfamiliar rocket fire. Analysis of all 85 declassified witness transcripts from the original inquiry, representing 79 witnesses, found that only a small minority reported that misperception and that urgent movement was intelligible in a setting of real threat. Ordinary expectations about entering shelter, poor information about the blockage and inadequate physical arrangements combined lethally.

This kind of account does not remove psychology in favour of engineering. It explains why people moved as they did and why those behind could not know what those ahead were suffering. It also relocates responsibility. If managers can monitor the whole system and participants cannot, the burden of preventing overloading cannot be assigned to individuals at the rear.

Bodies, flows and bottlenecks

A second research tradition treated the crowd as pedestrian movement. John Fruin's Pedestrian Planning and Design in 1971 gave planners a practical language of levels of service, circulation and capacity. Later models borrowed from traffic flow, particle systems and behavioural heuristics. They reproduced familiar patterns: lanes forming in opposing streams, alternating passage at bottlenecks, stop-and-go waves and breakdown at high density.

Video from the Hajj allowed researchers to examine dangerous transitions in a real mass gathering. Work by Dirk Helbing, Anders Johansson and Habib Zein Al-Abideen identified stop-and-go movement and highly irregular displacement under extreme density. Moussaïd, Helbing and Theraulaz later showed how simple visual rules for choosing direction and speed could generate both efficient everyday patterns and dangerous breakdowns in simulation.

A simulation shows what can follow from its rules and geometry, not what caused a historical disaster. Change the assumed behaviour and the result may change too. Observed movement, local conditions and competent judgement must decide whether a model is useful for a particular event.

The practical question is where a moving crowd will run out of room next. Entry, holding, normal exit and emergency exit all matter; the sports-ground framework uses the lowest of their assessed capacities. Design and management interact. A wide exit is useless if the route to it narrows. A safe standing area can become unsafe if inflow is uncounted, communication fails or a barrier creates a local pocket. The event is not safe because the total number fits on a plan.

When the record blames the people

Hillsborough showed that a crowd disaster continues after the pressure has stopped. On 15 April 1989, Liverpool supporters entering the Leppings Lane end of Sheffield Wednesday's ground were channelled through a tunnel towards central terrace pens that were already dangerously crowded. An attempt to relieve congestion outside transferred more people inside without an adequate system for monitoring or redirecting the flow. Fences and pens restricted escape. Ninety-seven people ultimately died as a result, and hundreds were injured.

The first influential account made the supporters part of the cause. Claims about drunkenness, ticketlessness and disorder turned a crowd-management failure into a defect of the crowd. Families and survivors then spent decades forcing the record back towards the evidence. The Hillsborough Independent Panel concluded that no Liverpool supporters were responsible in any way and supported the Taylor Inquiry's finding that failure of police control was the main cause. The 2016 inquests found that the 96 people whose deaths they examined were unlawfully killed. Andrew Devine's death in 2021 was subsequently recognised as the 97th fatality. The IOPC and Operation Resolve report published in December 2025 found no evidence that supporter behaviour caused or contributed to the disaster and documented police efforts to deflect blame.

This matters to the history of crowd science because description governs prevention. If the causal word is hooliganism, the remedy is stronger control of supporters. If the causal sequence is unmanaged admission, dangerous enclosures, poor monitoring, delayed recognition and failed response, the remedy moves to design, information, capacity and command. A false psychology can protect a failed system for years.

From control to management

Public-order practice has slowly absorbed the same shift. The control model begins with threat, builds visible force and seeks compliance. The management model begins with differentiation: who is present, what are their legitimate aims, which risks are developing and which specific behaviour requires intervention?

The Euro 2004 research made this distinction operational. Before intervening, police needed to know which conduct the supporters themselves would accept, which they would reject and how an intervention would affect that boundary. Liaison could supply information in both directions: explaining restrictions to supporters while telling police what the gathering meant to those inside it. Self-regulation was something an operation could preserve or undermine, not a pleasant bonus after order had been imposed. The comparison did not establish that low-profile policing always works; it showed why the meaning of an intervention belongs in the assessment of its consequences.

Four practical principles followed from this work: learn the identities and norms in the crowd, facilitate legitimate purposes, communicate throughout the event and differentiate people and actions. These principles have influenced public-order approaches in several settings, though implementation and outcomes vary. They turn legitimacy from a moral extra into part of operational effectiveness.

Crowd safety reaches the same conclusion by another route. Managers must anticipate, monitor and control crowding risk because they possess the overview, gates, barriers, staffing and communication systems. Participants still have duties and agency. They do not possess the control room.

The shift from control to management makes a practical demand: understand the gathering before trying to change it. That requires watching the space and listening to the people, including those whose account is easiest to dismiss.

How we know

Crowd research has an awkward evidence base because dangerous events cannot be reproduced at full scale. The field combines archival records, interviews, surveys, observation, video trajectories, natural experiments, small controlled studies, simulations and comparisons across policing operations. Each sees a different layer.

Case studies such as St Pauls or Bethnal Green can identify sequence, meaning and limits but cannot establish universal frequencies. Pilgrimage studies give rare access to mass gatherings, yet their findings travel only with care across religions, countries and event designs. Controlled field walking studies isolate mechanisms while simplifying history, risk and authority. The Hajj visa lottery strengthens causal inference for applicants, but attendance remains a bundled exposure. Simulations reveal consequences of assumptions rather than replacing observation. Official records may preserve the view of the institution being challenged, and retrospective memories can change.

The strongest conclusions in this book are therefore convergent: crowds display selective norms; shared identity can alter relations and coordination; helping is common enough to defeat universal panic models; density and bottlenecks impose hard constraints; and management affects both physical and social dynamics. Exact thresholds, effect sizes and predictions remain setting-dependent.

What People Get Wrong

“A crowd makes people irrational”

The old model treats reason as a possession that leaks away with personal space. Large numbers can impair judgement, but the mechanism is usually more ordinary. People have partial information, shortened sightlines, uncertain signals and actions whose consequences depend on strangers. They may reach a poor conclusion while reasoning from the evidence available where they stand.

Crowd conduct also remains patterned. Protesters select targets, worshippers sustain rituals, supporters observe codes and people in emergencies search for companions. Even collective violence has boundaries that a theory of mental collapse cannot explain. St Pauls mattered because participants could accept attacks on police while rejecting attacks that did not fit the event's developing meaning.

None of this makes crowds wise. A group can adopt a false belief, reward cruelty or follow an ineffective route. The correction is that error requires an explanation. Ask what people could see, whom they trusted, which identity and norm guided them, and what options the setting allowed. Calling the result irrational ends inquiry at the moment it should begin.

The judgement is also applied unevenly. A military parade, royal procession or packed victory celebration may be described as discipline, loyalty or shared pride, while an unfamiliar protest is called mob behaviour. The observer often decides which collective feeling counts as reason before examining the process. A useful theory must explain orderly and disorderly gatherings with the same machinery rather than reserve psychology for crowds it dislikes.

“Anonymity removes restraint”

Masks, darkness, uniforms and large numbers can make a person harder to identify. It is tempting to turn that fact into a general law: anonymity weakens self-awareness, accountability disappears and forbidden impulses emerge.

Research on deindividuation did not support that clean sequence. A major meta-analysis found that conditions associated with anonymity and group immersion often increased responsiveness to situational group norms rather than producing one direction of conduct. Concealment can support aggression where aggression is expected and approved. It can support disciplined solidarity where restraint is valued. A masked medic, football supporter and rioter are not released into the same behaviour by the same covered face.

Personal accountability still matters, and surveillance or identification can alter risk. The mistake is to treat accountability as the only source of restraint. People also regulate themselves through identity, loyalty, morality and concern for how fellow members judge them. Sometimes the crowd makes standards more visible than solitude does. Anonymity changes the conditions of action. It does not reveal a universal creature underneath.

“Emotion spreads like a virus”

The contagion metaphor is attractive because feeling can move quickly through a gathering. One section cheers and another answers. A few frightened faces alter the mood. Anger appears to pass across a line.

But speed does not prove mindless transmission. People attend more to those they see as relevant, credible or similar. They infer what the display concerns. A scream at a concert, a scream near an exit and a scream from a rival group carry different information. Anger grows when members share the judgement that an act was unjust. Joy grows when others' response confirms that the event realises something the group values.

Mimicry, sound, arousal and close contact can contribute, and nobody needs to narrate a full appraisal before reacting. The correction is selectivity. The same display can alarm one observer, amuse another and leave a third unmoved because their relationship to the person and event differs. Treating emotion as infection hides the identities and meanings that give it a route. It also encourages the dangerous belief that information should be withheld lest fear contaminate everyone.

“Panic is what kills people”

Panic is often assigned after a disaster rather than demonstrated during it. The word can mean intense fear, urgent escape, selfish competition, irrational action or any movement that ended badly. A concept covering all five explains little.

Survivor studies commonly find fear alongside affiliation, delay, information-seeking and help. People search for relatives, follow familiar routes, assist strangers and wait for confirmation. These actions can aid survival or create delay. They are not evidence that everyone became calm, and they do not guarantee cooperation. They show that emergency behaviour is mixed and socially organised.

The panic account is especially weak when ordinary action enters a failed physical system. At Bethnal Green, people sought shelter during a real air-raid warning. A fall and blockage on the entrance stairs, limited information and continued inflow produced lethal compression. Calling the movement panic made urgency sound causeless and the victims sound defective. The useful questions concern warning, route, visibility, management and pressure. Fear may be present. It should not be allowed to do all the causal work.

“A crowd crush is a stampede”

Stampede suggests a mass running wildly and trampling anyone who falls. Some disasters do involve rapid flight, but many lethal crowd incidents develop through compression in a mass where individuals have lost control of movement.

As density rises, people have less room to avoid contact. Pressure can travel through bodies. A person can be pushed while trying to stop, carried away from an intended route or unable to expand the chest freely. Those farther back may continue moving because they cannot see distress at the front. Their local experience is a slow queue, while the same inflow is becoming force at a barrier.

There is no single density number that diagnoses every dangerous crowd. Direction, speed variation, group structure, geometry, surfaces and local pockets matter. Average space can conceal the critical point. The correction changes responsibility. Animals are blamed for stampeding. Systems are examined for admission, capacity, bottlenecks, monitoring, barriers, communication and emergency access. The people trapped in pressure are usually the least able to alter it.

“Troublemakers infect everyone else”

The agitator story gives collective conflict a compact cause. A small deviant minority provokes respectable people, emotion spreads and the crowd follows. Remove the bad actors and the mass returns to normal.

Some people do arrive seeking violence, and a narrow intervention can prevent harm. What does not follow is that their conduct automatically generalises. Others assess whether the act fits the group's purpose, whether the actor belongs, how opponents respond and what support is visible. Crowds can restrain, isolate or hand over someone whose action threatens the gathering. They can also come to support resistance that earlier seemed unacceptable when events alter its legitimacy.

This was visible in observational studies of England supporters at Euro 2004. Differentiated policing and facilitation were associated with preserved boundaries between ordinary supporters and offenders, while more antagonistic interactions elsewhere were associated with broader opposition. The infection model ignores the social judgement between example and uptake. It therefore invites indiscriminate control, which can erase the distinctions managers needed.

It also confuses visibility with representativeness. The person throwing an object creates footage; the thousands not throwing one create background. Repetition of the image can make a rare act look like the crowd's settled norm. Denominator, duration and response from other participants are part of the evidence.

“The police stand outside the crowd”

Maps often draw participants as one block and police as a line around it. The picture becomes a theory: the crowd supplies behaviour and police react.

Every intervention enters the crowd's information and identity system. A clear explanation can make a restriction legible. Facilitation can confirm that lawful participants are being treated as such. An indiscriminate containment or charge can create a common fate across people who previously had little in common. If that action is experienced as illegitimate, resistance can become more acceptable and mutual support more likely. The same sequence may then confirm police expectations that the crowd was hostile from the start.

Authority is not responsible for every escalation, and proportionate force may be necessary. Crowd members retain agency and accountability. The correction is causal inclusion. Police, stewards, organisers and venue managers affect routes, information, categories, trust and physical conditions. They are participants with unusual power and unusually broad visibility. Any explanation that places them outside the system has removed one of its strongest causes.

Use It

Find the psychological crowd

A headcount tells you how many bodies share the space. It does not tell you how many groups are operating inside it.

Begin by asking who understands whom as one of us. The answer may follow tickets, colours, faith, neighbourhood, occupation, shared danger or nothing at all. It may shift during the event. Two sets of supporters in one stadium are physically co-present and psychologically separate. Stranded passengers who began as private commuters may acquire a common identity when a failure affects them together. A demonstration can contain several organisations, unaffiliated participants and passers-by whose relations cannot be inferred from the aerial shot.

This changes prediction. Shared identity can increase trust, accepted closeness, communication and willingness to assist, but only within the boundary that has become meaningful. People outside it may be ignored or opposed. Do not ask whether the crowd is cohesive. Ask where cohesion lies, what defines it and whether the boundary is stable.

Separate meaning from mechanics

When a crowd becomes intense, social and physical explanations compete for attention. Resist the competition. Run both.

The social questions are about purpose, identity, legitimacy and information. Why are people here? What do they believe the event means? Which conduct fits that meaning? Whom do they trust? The physical questions are about density, flow, routes, timing and contact. Where are people entering and leaving? Which area is filling? What prevents movement? Can those at the rear see what has changed at the front?

Neither side can substitute for the other. A shared identity may make closeness welcome and support mutual aid, yet it cannot widen a gate. Compression may remove control of movement, yet it does not explain why people entered, stayed or chose one route. A good account should be able to say that conduct was socially reasonable and physically dangerous at the same time. Bethnal Green is easier to understand once that sentence becomes available.

Follow the phase change

A crowd at arrival is not the same system as the same crowd at departure. Map the sequence rather than naming the event once.

During assembly, people search for entrances, companions and positions. Waiting stores demand and establishes local expectations. A delay, opening or route change can release that demand faster than the next space can absorb it. At dispersal, the shared purpose may weaken just as flows reverse, groups split and transport nodes receive a surge.

Ask what changed immediately before the behaviour you are trying to explain. Did information arrive? Did an exit open? Did police move? Did the event end? Did one group become separated? A still image shows occupancy. A causal account needs direction and timing.

This lens prevents hindsight from flattening the event into its loudest minute. It also improves management. An instruction useful during entry can obstruct exit; a barrier that organises a queue can trap a reversing flow. Plan for transitions, not only steady states.

Watch the act and its reception

In numbers, private choices become public evidence. A person steps into a queue, raises a hand, sits down, runs, boos or refuses an order. Others now know that the action is possible. Their response supplies a second piece of evidence about whether it fits.

Look at proposal and reception separately. A chant without uptake is not a crowd norm. A barrier crossed by one person while others object is different from a crossing that draws approval. Silence may mean consent, uncertainty, fear of standing out or failure to hear. The visible surface has to be interpreted through the relationships beneath it.

Then ask which standard the act proposed. It may concern fairness, dignity, loyalty, mourning, enjoyment, non-violence or resistance. Leaders can offer a definition, experienced members can model it and opponents can sharpen it. None can install it by volume alone. Uptake depends on whether participants recognise the actor, claim and action as representative of who they are.

The same lens works in a meeting. Apparent unanimity may be pluralistic ignorance: each person privately doubts the direction but treats everyone else's silence as support. One credible objection can reveal a thinner norm. This does not excuse harmful conduct. It identifies the standard through which conduct was approved, challenged or contained.

Treat management as part of the system

Every instruction changes more than movement. It says who has standing, whose purpose is legitimate, what the authority believes is happening and whether people are being treated as distinct agents or one suspect block.

The practical sequence is differentiation, facilitation, communication and proportionate intervention. Learn the groups and aims present. Make lawful or expected activity possible where it can be made possible. Explain restrictions early enough for people to alter plans. Address specific dangerous conduct without erasing the difference between the actor and everyone nearby. These are not guarantees of harmony. They preserve sources of self-regulation that indiscriminate treatment can destroy.

Trust matters because information arrives through a relationship. A direction from a steward known to understand the event can be usable where the same words from an unknown voice are dismissed. Trust can form under pressure, but planning to acquire it only in a crisis is a poor substitute for earning it beforehand. Managers who regard communication as an announcement system discover too late that audibility and credibility are different resources.

Design for local pressure, not average space

The total attendance is one number in a moving system. It is rarely the number that fails first.

Trace the route as a chain: arrival, search, ticket check, entrance, circulation, viewing or waiting area, normal exit and emergency exit. Check where demand can exceed capacity, rather than assuming one attendance figure makes every part safe. Then look for changes over time. A delayed opening can store people outside. A performance ending can reverse flow. A closed route can redirect a stream into a junction. A barrier can solve queue disorder and create a pocket of compression beside it.

Average density can look acceptable while one local area becomes dangerous. Monitor where speed falls, direction conflicts, pressure builds or people stop controlling their own movement. The rear needs information about the front because inflow often comes from people who cannot see the blockage they are feeding. A message without a viable alternative route merely describes the problem.

For a participant, this is a way of noticing danger, not a rescue technique. Awareness of exits and competent local instructions helps, but individual alertness cannot make an overfilled venue safe. Managers have the larger responsibility because they control counting, routes, barriers, staff, cameras and communication. Overview creates responsibility.

The limits

Crowd science offers mechanisms, not a universal forecasting engine. A pilgrimage, station, protest and nightclub differ in selection, identity, architecture, authority and risk. Mechanisms help you ask better questions; they do not identify everyone's intentions or tell you how the next minute will unfold.

The concepts can also be abused. Calling a crowd a temporary system does not distribute responsibility equally through it. Organisers, police and venue operators possess powers and information that participants lack. Shared identity can support care, but it can also organise exclusion or violence. Self-regulation is valuable, not magical. Formal intervention remains necessary in some conditions.

Online networks borrow the language of crowds while lacking co-presence, shared movement and physical pressure. Some mechanisms travel, especially visible norms, identity and information cascades. The full model does not. Keep the body in the definition.

The one thing to keep

Return to the railway platform. From above, it still looks like an area occupied by bodies. From within, each person has a different piece of the situation: a glimpse of the stairs, a message from staff, a companion to find, a reason to trust or distrust the movement ahead. The crowd is what happens as those partial views meet one another and the limits of the space.

You should now notice more than the person who shouts or pushes. Notice the people making room, waiting for someone, objecting to an act or following a route they cannot see to its end. Notice, too, the decisions made beyond their sight: admission, barriers, announcements, delayed departures. Some of the most powerful causes of crowd behaviour are nowhere near the face in the photograph.

This neither excuses every act nor makes collective life a machine without responsible people. It restores the missing relationships. A crowd has no brain above its members, but being among others changes what each member can know and do. Once that becomes visible, the ordinary achievement is harder to overlook. Strangers often manage to share the same space, and sometimes much more than space, without anyone possessing the whole plan.

Terms

Physical crowd. People gathered in the same place, whether or not they understand themselves as one group. A station concourse may be one physical crowd containing several groups.

Psychological crowd. People who share an active social identity within a gathering. The category concerns a social relation among participants, not a headcount, mood or organisation.

Collective behaviour. Interdependent action by people whose conduct is oriented to others. It includes queueing, chanting, helping, moving, resisting and dispersing.

Classical crowd psychology. The tradition associated with Gustave Le Bon, which treated crowds as producing lost personality, suggestibility, contagion and intellectual regression. Modern evidence rejects that general model.

Convergence theory. The view that people with similar motives or dispositions gather and express what they already brought. Selection matters, but convergence alone cannot explain new norms or identities created during assembly.

Emergent-norm theory. Turner and Killian's account of collective situations in which participants infer appropriate conduct through interaction, visible examples and responses. It restored meaning and regulation to collective behaviour.

Social identity. A definition of the self as a member of a social category, carrying expectations, interests and norms. People possess several identities, and different settings can make different memberships guide action.

Self-categorisation. Classifying oneself and others at a particular social level, such as individual, supporter, resident or survivor. It explains how the operative self can shift without personal identity disappearing.

Salience. The degree to which an identity or distinction becomes relevant in a situation. Contrast, events, language and treatment by other groups can make one membership more salient than another.

Ingroup and outgroup. The people understood as us and them within a given categorisation. The boundary affects trust, influence, expected support and judgement, and can change during the event.

Shared social identity. People seeing themselves and one another as members of the same group. It can increase intimacy, accepted proximity, communication, support and capacity for coordination.

Common fate. A shared experience imposed across prior differences, such as danger or indiscriminate treatment. Common fate can create or widen social identity, although similar exposure does not guarantee unity.

Collective resilience. The capacity of a group to coordinate, support members and adapt under pressure through shared identity and social resources. It corrects universal panic models without promising harmony in every emergency.

Ingress and egress. Movement into and out of a site. The phases have different directions and information needs; a route that works for ingress may fail when egress reverses or combines flows.

Norm. A socially understood standard of appropriate conduct. In crowds, norms can be inherited, contested or formed during events, and they regulate restraint and aggression as well as orderly behaviour.

Referent informational influence. Influence from people treated as representative fellow members. Their conduct helps define what the group believes and how members should act, especially when the situation has no settled script.

Pluralistic ignorance. A condition in which many people privately reject a position while each assumes others accept it. Public silence then appears as consensus and can preserve a norm with little private support.

Information cascade. A sequence in which later decisions follow earlier visible choices instead of independent evidence. Cascades can reveal a useful route or concentrate people around a mistaken option.

Meta-perception. A belief about how another person or group sees one's own group. Hostile meta-perceptions can produce defensive conduct on both sides and help create the antagonism each expected.

Deindividuation. A family of theories linking anonymity, immersion or reduced self-awareness to altered conduct. Evidence fits increased responsiveness to group and situational norms better than a uniform release of primitive impulses.

Moral economy. Shared expectations about legitimate exchange, obligation and fairness. E. P. Thompson used the concept to show that historical food crowds acted through customary standards rather than blind hunger.

Social identity model of crowd behaviour. Reicher's account in which people act through a salient shared identity and infer situation-specific norms selectively from fellow members. It explains patterned action within crowds.

Elaborated social identity model. An intergroup account of crowd conflict and change. It examines mismatched definitions, unequal power, perceived illegitimacy, indiscriminate treatment, common fate and the widening of collective opposition.

Differentiated policing. Public-order practice that distinguishes groups, intentions and acts rather than treating everyone present as one threat. It supports targeted intervention and preserves incentives for participant self-regulation.

Density. The number of people within a unit of area. Average density is an incomplete safety measure because local concentration, direction, movement, body contact and geometry determine what people experience.

Pedestrian flow. The number of people crossing a line or passing through a route during a period. Safe operation depends on how demand, speed, width and downstream capacity interact.

Bottleneck. A point whose usable capacity is lower than the demand reaching it, such as a gate, stair, turn or search lane. Queues are manageable only while storage and inflow remain controlled.

Stop-and-go wave. Alternating movement and stoppage travelling through a dense crowd. It signals unstable flow and can emerge far from the obstruction because delay propagates backwards through local interactions.

Crowd crush. Dangerous compression in which contact forces and restricted breathing or movement can injure or kill. The term directs attention towards density, inflow, barriers, information and management rather than animal imagery.

Safe capacity. The attendance a venue can accommodate under stated conditions. Sports-ground guidance takes the lowest assessed entry, holding, normal exit or emergency exit capacity, accounting for physical condition and safety management.

Go Deeper

The current overview

John Drury, “The Psychology of Crowd Behavior,” Annual Review of Psychology 77 (2026), 339-364. This is a direct route into the modern social-identity approach and the research it has generated on density, emotion, mass gatherings, emergencies, collective action and public-order policing. Drury writes as a principal contributor to the field, which gives the review coherence and also a viewpoint. Read it for a current, integrated account of how shared identity changes relations among people in crowds. Annual Reviews issued an erratum on 26 February 2026, so consult the publisher's corrected record alongside it. Then use a physical crowd-science text to recover the geometry and engineering that a psychology review cannot cover in equal depth.

The observational correction

Clark McPhail, The Myth of the Madding Crowd (Aldine de Gruyter, 1991). McPhail replaces the dramatic crowd as a single actor with close description of gatherings, behaviours and changing participation. The book is technical in places and its theoretical machinery has aged, but its empirical discipline remains bracing. Read it to learn how much disappears when an event is named after its loudest episode. It will make you count who is acting, who is watching, how long an act lasts and how behaviour varies across the same space. That method is a permanent defence against footage selected to prove a mob.

The applied argument

Stephen Reicher and Clifford Stott, Mad Mobs and Englishmen? Myths and Realities of the 2011 Riots (Constable & Robinson, 2011). This is the easiest next read here for someone more interested in the argument than the research apparatus. Written quickly after the English riots, it applies social-identity and intergroup analysis while public explanations were hardening. It challenges accounts based on broken character, automatic contagion and an undifferentiated criminal mass. Read it for the connection between theory, policing and political blame. Its speed and proximity to the event are strengths and cautions: treat it as a forceful interpretation to compare with later evidence, not the final history of August 2011.

The physical system

G. Keith Still, Introduction to Crowd Science (CRC Press, 2014). Still approaches crowds through risk analysis, routes, density, flow, congestion, decision support and the control room. It is a useful complement to the psychological books because it forces every claim about meaning back into a space with widths, timings and capacities. Worked examples and practical tools make it closer to a professional manual than a narrative read. Use it to understand how event design and live monitoring convert crowd safety from a vague duty into an operating system, while remembering that models require local observation and competent judgement.

Notes and Sources

The notes follow the order of the book. They distinguish findings that recur across methods from results tied to one event, country or research design. Crowd psychology and crowd dynamics overlap, but they are not one settled discipline with a single predictive model. The coordination-under-partial-information frame used in this book is a synthesis: social identity, norms and information explain meaning and coordination; pedestrian research and safety guidance explain movement, capacity and compression; event phases explain why the same gathering can change without a new population.

Frame and definitions

Physical and psychological crowds. The distinction between physical co-presence and a psychological crowd organised by shared social identity follows the modern social identity approach. Templeton, Drury and Philippides (2018) state and test the distinction in pedestrian movement; Drury's 2026 review supplies a later synthesis. It does not imply that every gathering contains a single shared identity. Studies and theory repeatedly find nested and competing categories, small companion groups and people who never identify with the wider gathering. The manuscript therefore avoids treating one headcount as one social unit.

Coordination under partial information. The phrase is the book's organising description rather than a named consensus theory. It captures a secure common feature across social and physical work: each person's movement, visible reaction and interpretation alters conditions for others, while routes, density, event phase and management constrain the choices available to each person. The claim is explanatory, not mathematical. It does not assign equal causal power or responsibility to participants, police, organisers and venue operators.

Selection and pre-existing relations. Convergence accounts correctly stress that people bring motives, dispositions and relationships into gatherings. Floyd Allport's 1924 formulation is quoted briefly in the manuscript: “the individual in the crowd behaves just as he would behave alone, only more so”. The original text has been checked. The manuscript does not use the line to deny change during gatherings. It places selection beside interaction because neither alone explains the evidence.

Classical theory, convergence and norms

Le Bon and the classical model. Gustave Le Bon published Psychologie des foules in 1895; an English edition, The Crowd: A Study of the Popular Mind, appeared in 1896. His account linked anonymity, suggestibility, contagion, unconscious racial inheritance and intellectual regression. The manuscript treats the theory in its political setting and does not credit Le Bon with inventing every element. Scipio Sighele, Gabriel Tarde and other late nineteenth-century writers had already examined imitation, criminal crowds and mass society. Stephen Reicher's 1996 centenary assessment is useful on the gap between the model's continuing practical influence and its weak theoretical support.

Emergent norms. Ralph Turner and Lewis Killian's Collective Behavior of 1957 proposed that participants in uncertain situations develop working definitions of appropriate conduct through interaction. Later research has refined rather than preserved the first formulation intact. Identity, prior norms, relationships and power help explain whose example becomes influential and why uptake is selective. The manuscript therefore avoids the stronger claim that a new rule can be installed by any conspicuous actor.

Moral economy. E. P. Thompson's 1971 essay on the eighteenth-century English crowd challenged stimulus-response accounts of food disorder. His cases showed action organised through expectations about subsistence, customary obligations and legitimate market conduct. The evidence concerns particular English settings and a disputed historical concept. It supports the narrower point used here: hardship did not explain conduct without participants' ideas of fairness and entitlement.

Historical crowd conduct and observation

Counting the event. Clark McPhail's observational programme, brought together in The Myth of the Madding Crowd (1991), separates gatherings from the dramatic acts later used to name them. The manuscript draws on his attention to assembling, clustering, queueing, chanting, watching, moving and dispersing, and to variation across people and time. It does not adopt his whole theoretical system. The durable lesson is methodological: identify denominator, duration and participation before treating a visible act as representative.

Selection in records and footage. Official reports, newspapers and television often concentrate on conflict, obstruction and property damage because those moments are institutionally salient and visually legible. This is not a claim that all reporting is biased in the same direction. It is a warning about sampling. A five-second clip cannot establish how many people acted, what preceded the act, how others responded or whether the action became normative.

St Pauls and the social identity model

The event. Police raided the Black and White Café in St Pauls, Bristol, on 2 April 1980. The raid developed into hours of confrontation in a neighbourhood marked by racism, deprivation and conflict with police. Stephen Reicher's 1984 analysis used participant accounts and historical context to ask why crowd action had particular targets and limits. His evidence supported a social identity account in which conduct was patterned by a changing definition of the group and its opponent.

What the case can establish. St Pauls is a single retrospective case and cannot show how frequently the same mechanism occurs. Its importance is diagnostic: indiscriminate contagion predicts weak boundaries, while the evidence showed selective legitimacy and restraint. The manuscript uses the case to defeat a universal model, not to claim that every disturbance has the same moral boundary or relation with police.

The poll-tax demonstration. Reicher's 1996 study of the 31 March 1990 demonstration in Westminster developed the intergroup sequence later formalised in the elaborated social identity model. The account links police categorisation and action, participants' perceptions of legitimacy, common fate and a widened basis for resistance. The causal language remains conditional. Indiscriminate action can produce broader opposition where it is experienced as illegitimate; it does not do so automatically, and crowd members retain agency.

Mass gatherings, density and identity change

Magh Mela. The North Indian Magh Mela research programme used surveys, observation and longitudinal designs to study hardship, intimacy, support, health and identity at a large Hindu pilgrimage. Tewari and colleagues in 2012 reported longitudinal and comparison evidence on well-being. Shankar and colleagues (2013) tested the same ambiguous sound with 32 male pilgrims; one failed the source check and was excluded from analysis. The experiment established a framing effect, not identity as its sole cause. Interviews examined the meanings attached to the sounds. Pandey and colleagues in 2014 studied identity and physical hardship. Khan and colleagues in 2015 linked shared identity with self-reported health, and their 2016 longitudinal study examined changes in Hindu identification and religious practice. These studies share a setting and overlapping research programme, so they are not independent replications across cultures.

Hajj and density. Alnabulsi and Drury's 2014 study found that social identification moderated relations between crowd density and reported safety at the Hajj. Their 2020 work with Vignoles and Oogink examined self-change at the same type of gathering. These studies support the claim that social meaning can alter the experience of proximity and expected support. They do not show that identity can cancel dangerous density or that a positive subjective experience proves physical safety.

Hajj lottery. Clingingsmith, Khwaja and Kremer used Pakistan's lottery for Hajj visas as a natural experiment. Among applicants, attendance increased some global Islamic practices, reduced some local practices and was associated with more tolerant attitudes across ethnic groups, sects and religions, as well as more favourable attitudes towards women. Random allocation strengthens causal identification for this applicant population. The result should not be universalised to all pilgrims, all countries or all mass gatherings, and the intervention includes travel, ritual, contact and status rather than crowd density alone.

Walking together. Templeton, Drury and Philippides tracked 280 pedestrian trajectories in a controlled comparison. Participants in the psychological-crowd condition walked more slowly, travelled farther and maintained closer proximity than people in the comparison condition. The experiment supports a behavioural signature of shared identity under the tested conditions. It does not provide a universal flow coefficient or establish what happens under dangerous compression.

Emotion and collective experience. Hopkins and colleagues linked shared identity with positive experience and collective effervescence, while later work examined recognition, validation and solidarity in crowds. The manuscript preserves the distinction between bodily influence and appraisal. Voices, expressions and arousal matter, but the spread and meaning of emotion are selective. Durkheim's term collective effervescence supplies a historical description, not a sufficient modern mechanism.

Emergencies and the panic model

Helping and collective resilience. Drury, Cocking and Reicher's studies of emergency survivors and the 2005 London bombings documented solidarity, helping and shared fate. Drury and colleagues' 2016 earthquake study linked emergent identity and observed support with support provided. These findings converge with wider disaster research against universal each-for-themselves accounts. They do not establish that cooperation always dominates, that strangers receive the same aid as relatives, or that shared identity guarantees survival.

Panic as an unstable category. The literature uses panic for several different things: strong fear, rapid flight, selfish competition, irrational judgement or a bad outcome. Quarantelli, Clarke, Drury and others have criticised the term when it substitutes a label for a demonstrated mechanism. Proulx and Sime found that prompt, directive and explanatory information improved evacuation performance in their Underground-station study. The manuscript does not claim that panic never occurs. It claims that the word should not be used as the default cause without evidence of the specific perception and behaviour alleged.

Delay and affiliation. Evacuation research also finds underreaction, routine exit choice, confirmation-seeking and efforts to remain with companions. Donald and Canter's study of the King's Cross fire showed how familiar-route use could contribute to fatal outcomes. These behaviours may be socially intelligible while physically dangerous. That conjunction is central to the book's model.

Bethnal Green

Deaths and sequence. The Bethnal Green Underground shelter disaster occurred on 3 March 1943. A fall and blockage developed on the entrance staircase during an air-raid warning, and 173 people died. The recent analysis used all 85 declassified transcripts from the original Dunne inquiry, representing 79 witnesses, and distinguished movement towards shelter from the later crush.

What the witness analysis found. Barr, Drury, Butler, Choudhury and Neville found that only a small minority of witnesses reported mistaking British rockets for German bombs. Perceptions were calibrated to a wartime context in which attack was plausible, and most people were following a socially expected practice of urgent shelter-seeking. The proposed model combines density, limited information about the obstruction and normatively expected ingress. The authors present this as a model for a particular class of flight-crush disaster and call for further testing. The manuscript retains that boundary and does not apply the result unchanged to slow ingress, festival surges or every crowd disaster.

Physical crowd dynamics

Density, speed and flow. John Fruin's Pedestrian Planning and Design supplied an influential practical vocabulary for pedestrian space, circulation and levels of service. Later empirical and modelling work shows broad relations among density, speed and flow, but exact curves vary with population, direction, motivation, geometry, culture and measurement. The manuscript therefore gives no universal danger threshold.

Simulation and observed dynamics. Helbing, Farkas and Vicsek's 2000 model showed how specified interaction rules could generate clogging and other escape patterns. Helbing, Johansson and Al-Abideen's 2007 analysis of Hajj footage described stop-and-go waves and irregular movement in dangerous density. Moussaïd, Helbing and Theraulaz in 2011 modelled how simple pedestrian rules could reproduce ordinary and hazardous patterns. These works show mechanisms and signatures, not a complete causal account of any named disaster. The manuscript avoids taking simulation labels such as panic as evidence that panic was observed.

Crowd crush. Compression and restricted breathing are central mechanisms in many lethal crowd incidents. The word stampede can collapse running towards safety, running towards an attraction, crowd collapse and crush into one animal metaphor. Barr and colleagues' 2024 paper and crowd-safety literature support the distinction. The book does not claim that trampling never contributes or that all crowd deaths have one physiological pathway.

Groups in movement. Social groups may keep closer spacing, wait for companions, select routes together and assist one another. Templeton and colleagues' work argues that pedestrian models should represent small groups and psychological crowds rather than treating departures from isolated-pedestrian rules as noise. Social explanation must still stop where involuntary contact forces remove individual choice.

Hillsborough

The disaster and responsibility. On 15 April 1989, lethal overcrowding developed in the Leppings Lane terraces at Hillsborough. Ninety-seven people are now recognised as having died as a result of the disaster. The Taylor Inquiry identified failure of police control as the main cause. The Hillsborough Independent Panel's 2012 report concluded that no Liverpool supporters were responsible in any way. The 2016 inquests concerned 96 deaths and found unlawful killing; supporter behaviour did not cause or contribute to the dangerous situation. Andrew Devine died on 27 July 2021 and a separate inquest recognised his death as the 97th fatality.

The 2025 investigations. The Hillsborough Disaster: The Report of the IOPC and Operation Resolve Investigations, published on 2 December 2025, found no evidence that supporter behaviour caused or contributed to the disaster. It found fundamental failures in match planning and response, and substantial evidence of subsequent efforts to deflect blame. The manuscript uses Hillsborough to show that causal language affects both institutional accountability and future prevention. It does not compress the disaster's full legal, medical or political history into a crowd-science example.

Policing and Euro 2004

Knowledge-based policing. Reicher, Stott, Drury and colleagues set out four linked principles: education about crowd identities and norms, facilitation of legitimate aims, communication and differentiation. The principles arise from the elaborated social identity model and public-order research. They are not a claim that communication replaces preparedness or that targeted force is never required.

England supporters in Portugal. Stott and colleagues compared supporter and police relations across contexts at Euro 2004 and later published a quantitative analysis of public order, policing and crowd psychology. The evidence linked facilitative, differentiated practice with self-regulation and lower conflict in studied settings, while antagonistic relations could widen opposition. The comparison is informative because supporter category and tournament were held more constant than in cross-country anecdotes. It remains observational and context-dependent, so the manuscript uses association and process language rather than a universal effect size.

Deindividuation, information and emotion

Anonymity and norms. Postmes and Spears' 1998 meta-analysis did not support a general release of antinormative impulses under deindividuation conditions. Effects were better understood through responsiveness to situational and group norms. Concealment, surveillance and individual accountability can still alter behaviour. The correction is directional: anonymity has no single behavioural destination.

Information gaps. Other people's movements and reactions provide local evidence, but popularity can become self-reinforcing when independent information is unavailable. The manuscript uses information cascade in this general sense and does not attribute every coordinated movement to a formal cascade model. Public-address messages, stewards and trusted representatives matter because reach, credibility and an available route must coincide.

Meta-perception. Work on intergroup crowd dynamics shows how beliefs about an opposing group's beliefs can shape posture and interpretation. The manuscript treats this as one escalation route, not as proof that every conflict begins in mutual misunderstanding. Material interests, strategy, prejudice and deliberate aggression remain possible causes.

Current safety guidance and evidence limits

Safe capacity. The Sports Grounds Safety Authority's sixth edition of the Guide to Safety at Sports Grounds, first published in 2018, is the edition used here; its four-capacity framework was checked against current SGSA guidance on 5 September 2026. It assesses entry, holding, normal exit and emergency exit capacities. The lowest calculation governs. Holding capacity is also affected by assessments of physical condition and safety management. The manuscript identifies this as a sports-ground framework, not a universal formula for every venue. HSE guidance separately supports phase-specific planning, monitoring and management responsibility. Neither replaces competent venue-specific assessment.

Current synthesis. Drury's 2026 review covers social identity research on density, emotion, mass gatherings, emergencies, collective change and policing. It should be read with the publisher's erratum of 26 February 2026. The original studies and official records identified above provide the support for this book's central claims; the review is an integrated account of one major approach, not a substitute for those sources.

Prediction and external validity. Dangerous full-scale events cannot be reproduced experimentally, while laboratory studies strip away history, authority and real risk. Case studies can reconstruct sequence but not prevalence. Field surveys can measure relations in rare settings but face selection, self-report and cultural limits. Natural experiments strengthen causal inference within their assignment process. Simulations reveal consequences of assumptions. The book therefore presents a toolbox of mechanisms and questions rather than a universal crowd forecast.

Bibliography

Primary texts, original research and official evidence

Allport, Floyd H. Social Psychology. Boston: Houghton Mifflin, 1924.

Alnabulsi, Hani, and John Drury. “Social Identification Moderates the Effect of Crowd Density on Safety at the Hajj.” Proceedings of the National Academy of Sciences 111, no. 25 (2014): 9091-9096. doi:10.1073/pnas.1404953111.

Alnabulsi, Hani, John Drury, Vivian L. Vignoles, and Sander Oogink. “Understanding the Impact of the Hajj: Explaining Experiences of Self-Change at a Religious Mass Gathering.” European Journal of Social Psychology 50, no. 2 (2020): 292-308. doi:10.1002/ejsp.2623.

Barr, Dermot, John Drury, Toby Butler, Sanjeedah Choudhury, and Fergus Neville. “Beyond ‘Stampedes’: Towards a New Psychology of Crowd Crush Disasters.” British Journal of Social Psychology 63, no. 1 (2024): 52-69. doi:10.1111/bjso.12666.

Clarke, Lee. “Panic: Myth or Reality?” Contexts 1, no. 3 (2002): 21-26. doi:10.1525/ctx.2002.1.3.21.

Clingingsmith, David, Asim Ijaz Khwaja, and Michael Kremer. “Estimating the Impact of the Hajj: Religion and Tolerance in Islam's Global Gathering.” Quarterly Journal of Economics 124, no. 3 (2009): 1133-1170. doi:10.1162/qjec.2009.124.3.1133.

Donald, Ian, and David Canter. “Intentionality and Fatality During the King's Cross Underground Fire.” European Journal of Social Psychology 22, no. 3 (1992): 203-218. doi:10.1002/ejsp.2420220302.

Drury, John, Rupert Brown, Roberto González, and Daniel Miranda. “Emergent Social Identity and Observing Social Support Predict Social Support Provided by Survivors in a Disaster: Solidarity in the 2010 Chile Earthquake.” European Journal of Social Psychology 46, no. 2 (2016): 209-223. doi:10.1002/ejsp.2146.

Drury, John, Holly Carter, Chris Cocking, Evangelos Ntontis, Selin Tekin Guven, and Richard Amlôt. “Facilitating Collective Psychosocial Resilience in the Public in Emergencies: Twelve Recommendations Based on the Social Identity Approach.” Frontiers in Public Health 7 (2019): 141. doi:10.3389/fpubh.2019.00141.

Drury, John, Chris Cocking, and Stephen Reicher. “Everyone for Themselves? A Comparative Study of Crowd Solidarity Among Emergency Survivors.” British Journal of Social Psychology 48, no. 3 (2009): 487-506. doi:10.1348/014466608X357893.

Drury, John, Chris Cocking, and Stephen Reicher. “The Nature of Collective Resilience: Survivor Reactions to the 2005 London Bombings.” International Journal of Mass Emergencies and Disasters 27, no. 1 (2009): 66-95. doi:10.1177/028072700902700104.

Drury, John, and Stephen Reicher. “Collective Action and Psychological Change: The Emergence of New Social Identities.” British Journal of Social Psychology 39, no. 4 (2000): 579-604. doi:10.1348/014466600164642.

Helbing, Dirk, Illés Farkas, and Tamás Vicsek. “Simulating Dynamical Features of Escape Panic.” Nature 407 (2000): 487-490. doi:10.1038/35035023.

Helbing, Dirk, Anders Johansson, and Habib Zein Al-Abideen. “Dynamics of Crowd Disasters.” Physical Review E 75 (2007): 046109. doi:10.1103/PhysRevE.75.046109.

Health and Safety Executive. Managing Crowds Safely. INDG142. Current online guidance, consulted 5 September 2026. https://www.hse.gov.uk/pubns/indg142.htm

Health and Safety Executive. “Put Crowd Controls in Place.” Current online guidance, consulted 5 September 2026. https://www.hse.gov.uk/event-safety/crowd-management-controls.htm

Hillsborough Independent Panel. The Report of the Hillsborough Independent Panel. London: The Stationery Office, 2012.

Home Office. “Determinations and Findings of the Hillsborough Inquests.” 27 April 2016. Accessed 5 September 2026.

Hopkins, Nick, Stephen D. Reicher, Sammyh S. Khan, Shruti Tewari, Narayanan Srinivasan, and Clifford Stevenson. “Explaining Effervescence: Investigating the Relationship Between Shared Social Identity and Positive Experience in Crowds.” Cognition and Emotion 30, no. 1 (2016): 20-32. doi:10.1080/02699931.2015.1015969.

Hopkins, Nick, Stephen Reicher, Clifford Stevenson, Kavita Pandey, Shail Shankar, and Shruti Tewari. “Social Relations in Crowds: Recognition, Validation and Solidarity.” European Journal of Social Psychology 49, no. 6 (2019): 1283-1297. doi:10.1002/ejsp.2586.

Independent Office for Police Conduct and Operation Resolve. The Hillsborough Disaster: The Report of the IOPC and Operation Resolve Investigations. London: Independent Office for Police Conduct, 2025.

Khan, Sammyh S., Nick Hopkins, Stephen Reicher, Shruti Tewari, Narayanan Srinivasan, and Clifford Stevenson. “Shared Identity Predicts Enhanced Health at a Mass Gathering.” Group Processes & Intergroup Relations 18, no. 4 (2015): 504-522. doi:10.1177/1368430214556703.

Khan, Sammyh S., Nick Hopkins, Stephen Reicher, Shruti Tewari, Narayanan Srinivasan, and Clifford Stevenson. “How Collective Participation Impacts Social Identity: A Longitudinal Study from India.” Political Psychology 37, no. 3 (2016): 309-325. doi:10.1111/pops.12260.

Le Bon, Gustave. The Crowd: A Study of the Popular Mind. London: T. Fisher Unwin, 1896.

Liverpool Football Club. “A Statement from the Family of Andrew Devine.” 28 July 2021. Family statement and accompanying club statement reporting the coroner's conclusion. https://www.liverpoolfc.com/news/announcements/439184-a-statement-from-the-family-of-andrew-devine

Moussaïd, Mehdi, Dirk Helbing, and Guy Theraulaz. “How Simple Rules Determine Pedestrian Behavior and Crowd Disasters.” Proceedings of the National Academy of Sciences 108, no. 17 (2011): 6884-6888. doi:10.1073/pnas.1016507108.

Pandey, Kavita, Clifford Stevenson, Shail Shankar, Nick Hopkins, and Stephen Reicher. “Cold Comfort at the Magh Mela: Social Identity Processes and Physical Hardship.” British Journal of Social Psychology 53, no. 4 (2014): 675-690. doi:10.1111/bjso.12054.

Postmes, Tom, and Russell Spears. “Deindividuation and Antinormative Behavior: A Meta-Analysis.” Psychological Bulletin 123, no. 3 (1998): 238-259. doi:10.1037/0033-2909.123.3.238.

Proulx, Guylène, and Jonathan D. Sime. “To Prevent ‘Panic’ in an Underground Emergency: Why Not Tell People the Truth?” Fire Safety Science 3 (1991): 843-852. doi:10.3801/IAFSS.FSS.3-843.

Quarantelli, E. L. “The Nature and Conditions of Panic.” American Journal of Sociology 60, no. 3 (1954): 267-275. doi:10.1086/221536.

Reicher, Stephen. “The St Pauls Riot: An Explanation of the Limits of Crowd Action in Terms of a Social Identity Model.” European Journal of Social Psychology 14, no. 1 (1984): 1-21. doi:10.1002/ejsp.2420140102.

Reicher, Stephen. “‘The Battle of Westminster’: Developing the Social Identity Model of Crowd Behaviour in Order to Explain the Initiation and Development of Collective Conflict.” European Journal of Social Psychology 26, no. 1 (1996): 115-134. doi:10.1002/(SICI)1099-0992(199601)26:1<115::AID-EJSP740>3.0.CO;2-Z.

Reicher, Stephen, Clifford Stott, John Drury, Otto Adang, Patrick Cronin, and Andrew Livingstone. “Knowledge-Based Public Order Policing: Principles and Practice.” Policing 1, no. 4 (2007): 403-415. doi:10.1093/police/pam067.

Shankar, Shail, Clifford Stevenson, Kavita Pandey, Shruti Tewari, Nick Hopkins, and Stephen D. Reicher. “A Calming Cacophony: Social Identity Can Shape the Experience of Loud Noise.” Journal of Environmental Psychology 36 (2013): 87-95. doi:10.1016/j.jenvp.2013.07.004.

Stott, Clifford, Otto Adang, Andrew Livingstone, and Martina Schreiber. “Variability in the Collective Behaviour of England Fans at Euro 2004: ‘Hooliganism’, Public Order Policing and Social Change.” European Journal of Social Psychology 37, no. 1 (2007): 75-100. doi:10.1002/ejsp.338.

Stott, Clifford, Otto Adang, Andrew Livingstone, and Martina Schreiber. “Tackling Football Hooliganism: A Quantitative Study of Public Order, Policing and Crowd Psychology.” Psychology, Public Policy, and Law 14, no. 2 (2008): 115-141. doi:10.1037/a0013419.

Templeton, Anne, John Drury, and Andrew Philippides. “Walking Together: Behavioural Signatures of Psychological Crowds.” Royal Society Open Science 5 (2018): 180172. doi:10.1098/rsos.180172.

Tewari, Shruti, Sammyh S. Khan, Nick Hopkins, Narayanan Srinivasan, and Stephen Reicher. “Participation in Mass Gatherings Can Benefit Well-Being: Longitudinal and Control Data from a North Indian Hindu Pilgrimage Event.” PLOS ONE 7, no. 10 (2012): e47291. doi:10.1371/journal.pone.0047291.

Thompson, E. P. “The Moral Economy of the English Crowd in the Eighteenth Century.” Past & Present 50 (1971): 76-136.

Turner, Ralph H., and Lewis M. Killian. Collective Behavior. Englewood Cliffs, NJ: Prentice-Hall, 1957.

Turner, John C., Michael A. Hogg, Penelope J. Oakes, Stephen D. Reicher, and Margaret S. Wetherell. Rediscovering the Social Group: A Self-Categorization Theory. Oxford: Basil Blackwell, 1987.

Syntheses, methods and guidance

Drury, John. “The Role of Social Identity Processes in Mass Emergency Behaviour: An Integrative Review.” European Review of Social Psychology 29, no. 1 (2018): 38-81. doi:10.1080/10463283.2018.1471948.

Drury, John. “The Psychology of Crowd Behavior.” Annual Review of Psychology 77 (2026): 339-364. doi:10.1146/annurev-psych-012125-121447.

Drury, John. “Erratum: The Psychology of Crowd Behavior.” Annual Review of Psychology (2026). doi:10.1146/annurev-ps-77-022626-200001.

Fruin, John J. Pedestrian Planning and Design. New York: Metropolitan Association of Urban Designers and Environmental Planners, 1971.

Haghani, Milad. “Empirical Methods in Pedestrian, Crowd and Evacuation Dynamics: Part I. Experimental Methods and Emerging Topics.” Safety Science 129 (2020): 104743. doi:10.1016/j.ssci.2020.104743.

Haghani, Milad. “Empirical Methods in Pedestrian, Crowd and Evacuation Dynamics: Part II. Field Methods and Controversial Topics.” Safety Science 129 (2020): 104760. doi:10.1016/j.ssci.2020.104760.

McPhail, Clark. The Myth of the Madding Crowd. New York: Aldine de Gruyter, 1991.

Reicher, Stephen. “‘The Crowd’ Century: Reconciling Practical Success with Theoretical Failure.” British Journal of Social Psychology 35, no. 4 (1996): 535-553. doi:10.1111/j.2044-8309.1996.tb01113.x.

Reicher, Stephen, and Clifford Stott. Mad Mobs and Englishmen? Myths and Realities of the 2011 Riots. London: Constable & Robinson, 2011.

Sports Grounds Safety Authority. Guide to Safety at Sports Grounds. 6th ed. London: Sports Grounds Safety Authority, 2018.

Still, G. Keith. Introduction to Crowd Science. Boca Raton, FL: CRC Press, 2014.