The mechanics of crowd control: anticipation, preparation, prevention

In his 1997 book ‘Dominance Without Hegemony’, historian Ranajit Guha recounted how Mahatma Gandhi, “perhaps India’s foremost ideologue of self-discipline”, created an “elaborate” set of rules about how people should behave around him as they travelled the country. To Gandhi, Guha wrote, a haphazard crowd was “unmanageable”, “uncontrollable”, “undisciplined”, and ultimately entailed a “mobocracy”. Gandhi demanded “sacrifice, discipline, and self-control” and deplored the “instances of noise and confusion”. Yet he often suffered what Guha called a frustration “deeply felt”.

As anthropologist Ajay Gandhi wrote in a 2013 paper, “For nationalist reformers, proper comportment signalled a people’s capacity to be modern and responsible. The ‘rogues’ … who could not maintain decorum had to be expelled from official nationalism.”

Feet below head

On July 2, 2024, a crowd crush in Mughal Garhi village in Uttar Pradesh’s Hathras district, at the end of a conclave organised by self-proclaimed spiritual guru Suraj Pal Singh, left 121 people dead. Subsequent media reports suggested the local police had granted permission for 80,000 people to gather whereas some 2.5 lakh turned up. Even as Mr. Singh departed the venue in his car, “thousands of devotees shouted and ran towards the vehicle, crushing others still seated,” Reuters said citing a subsequent first information report. “Some people were trampled after falling in an adjacent field of slush and mud.”

On December 4, 2024, actor Allu Arjun made an appearance at the Sandhya Theatre premises in Hyderabad. As the throngs of people who had gathered for a film’s premier rushed towards him, a woman was trampled underfoot and her eight-year-old son was critically injured. Members of the duo’s family later filed a criminal complaint against Mr. Arjun and his security detail, the theatre’s owner, and the film’s makers. According to an officer investigating the incident, the actor’s security guards blocked exit points from the screening hall to stem the tide of people flowing towards him.

On January 29, 2025, a crowd crush during the Maha Kumbh Mela in Prayagraj, Uttar Pradesh, left at least 50 dead and more than a hundred injured. At the 1954 Kumbh, more than 800 people died in a crowd collapse; similar events claimed the lives of 39 at the Nashik Kumbh in 2003, 36 at the Allahabad Kumbh in 2013, and 10 at another Nashik Kumbh in 2015. The 2021 Kumbh Mela in Haridwar is widely believed to have been a ‘superspreader’ event after the Indian government allowed it to go ahead despite the COVID-19 pandemic.

Before the 2025 Kumbh began, Prime Minister Narendra Modi had said it would happen under the gaze of 2,700 cameras outfitted with artificial intelligence (AI) models able to spot unsafe instances of crowding and alert the authorities, so they could ameliorate the risk, in real-time. Reuters reported that the “software running the cameras” sends updates to a control centre staffed by more than 500 people when it “detects a surge in any one section of the festival city, a fire, or if people cross barricades they are not supposed to.” After the incident, State chief minister Yogi Adityanath commissioned an independent judicial inquiry.

On February 15, a ‘stampede’ at the New Delhi Railway Station (NDLS) left 18 people dead and several more injured. Initial reports from the ground indicated the incident involved trains headed for the Mela in Prayagraj.

Most recently, on May 3, at least seven people were killed and 80 more were injured during another alleged ‘stampede’, in truth a crowd collapse, in Shirgao village in the Bicholim taluka of Goa, where several lakh people had gathered for an annual religious festibal at the Shree Lairai Devi temple. Chief Minister Pramod Sawant has since called for a magisterial inquiry and sought help from Prime Minister Modi for the State.

Who is responsible?

“Stampede … is a loaded word as it apportions blame to the victims for behaving in an irrational, self-destructive, unthinking and uncaring manner. … It gives the impression that it was a mindless crowd only caring about themselves, and they were prepared to crush people.” Edwin Galea, a fire-safety engineering and evacuation dynamics expert, uttered these words to a journalist after the Seoul crowd crush in October 2022 that killed 150 people.

Events known as ‘stampedes’, ‘crowd crushes’, and ‘crowd collapses’ all start with crowding, when a large group of people gather in a bounded location and their number continues to increase until it crosses an important threshold. Up to five people per sq. m is okay. Between five and eight, the people are pressed against each other, and beyond eight they are packed so tightly that they may develop trouble breathing. A crowd crush occurs when people are constricted by the mass of others around them and can’t move on their own. A crowd collapse happens when one person in this tightly packed mass falls down, with the pressure of others around the individual causing them to fall as well. Seen from above, the crowd seems to progressively implode into a void.

A stampede occurs technically when a crowd of people runs towards or away from something, i.e. when the people are excited or frightened. When people die in a stampede, public conversations concerned with its cause use the word “panic”. But experts like Galea have repeatedly concluded that the people who run in a stampede can’t be to blame for the people who die in the stampede. As Manchester Metropolitan University professor of crowd science Keith Still said in 2015, “People don’t die because they panic. They panic because they are dying.”

The death, or more broadly some tragedy, happens first — due to a crowd crush or collapse, say — before it becomes compounded by the architecture of the space. If the space doesn’t allow people to escape harm quickly, they panic.

More room for some

According to an official estimate, around 66 crore people attended the 2025 Maha Kumbh Mela in Prayagraj, Uttar Pradesh, over 45 days. Say there were 5 crore people at the venue at any point in time. To accommodate them at five people per sq. m, the venue would have had to span 10 sq. km. In reality, the Uttar Pradesh government had created a space 40 sq. km large. This included room for kitchens, parking spaces, bathrooms, roads, hospitals, administrative facilities, and, presumably, some water bodies.

But even accounting for all of them, there should have been more than enough square-footage to admit crowding at a manageable level. Of course, denser crowds gathered at specific locations at various times for different ceremonies, rituals, etc. An important consideration here is that the festival’s administrators also made separate arrangements for VIPs. This “elite” setup allowed for wide spaces between tents, leisurely walkways, and the ability to attend events without having to physically come into contact with others, much less anticipating a crowd crush.

The 2025 Mela is politically significant for the Bharatiya Janata Party at the Centre and in the State, and thus the event witnessed the attendance of many political leaders, including Prime Minister Narendra Modi himself. But the arrangements the State made for these individuals still reflected the larger political economy of public spaces in contemporary India.

Crowding and human health

Consider human access to good health in a typical Indian city. How does the quality of this object vary as we move through different neighbourhoods? Areas that have many trees, wide roads, community spaces like parks and playgrounds, well-ventilated residences, and the ability to reach an affordable healthcare centre within 15 minutes are at a premium and thus have a higher cost of living. At the same time, they collectively represent the living conditions required for any person to ‘get away’ from bad air.

These areas have thus become vouchsafed for society’s upper classes. Even within homes, cleaner air is increasingly only available through the use of expensive air purifiers and likewise on the roads by the use of enclosed private transport like cars. Access to good health thus admits distinct richer and poorer varieties. The latter is typified by poorly ventilated houses, often oppressive heat (especially without air-conditioners), lower water quality, discontinuous power supply, absence of air purifiers, use of public transport options that don’t protect its users from bad air, lower access to open spaces, lower access to affordable as well as good-quality healthcare, and — again — crowding.

In fact, as the COVID-19 pandemic raged around the world amid researchers’ debates about the virus’ changing deadliness, the epidemic in India highlighted in parallel the line between the pathogen’s ability to kill and its human hosts’ ability to lower their risk exposure. In the country, the latter has frequently already been compromised to some degree by subpar living conditions. Medical workers called for people to maintain a distance of six feet between individuals in public spaces, for example, but spaces where crowding remained the norm, like ration shops, vegetable markets, and public water taps, rendered these prescriptions impractical.

Virtually always, a crowded space in India is the product of gaps in the public services provided to people and the quality of public infrastructure supporting these activities. This gap even takes the form of underestimation, as evidenced by the Mughal Garhi crowd crush as well as the 35-hour-long traffic jams outside Prayagraj during the Maha Kumbh Mela. Following the NDLS disaster on February 15, one government official had called the passenger traffic there “unprecedented”. Yet the Railways had not unknowingly sold 2,600 tickets for unreserved seats for the Mela and had to have expected as many people at the station that terrible night as the total number of tickets.

The crowding phenomenon

Studies examining the relationship between socioeconomic status and crowding behaviour have also recorded behavioural adaptations that help individuals tolerate crowding better, although this doesn’t mean they deliberately prefer crowded environments. Yet the factors that encourage disorderly movement are still overlooked when a crowding-related disaster occurs. For example, Deputy Commissioner of Police (Railways) K.P.S. Malhotra revealed a presumption that the people at the station were responsible for the tragedy when he said, “If people don’t rush unnecessarily, there won’t be a catastrophic effect.”

Indeed, in his 2013 paper, Gandhi, the anthropologist, examined how the urban bureaucracy sought to impose the ‘discipline’ of queuing and how it rationalised the behaviour of the people who didn’t fall in line. Excerpt:

“In the 1990s, the municipal government in Delhi responded to dangerous bus accidents. Such incidents were blamed on passengers jostling and elbowing as they entered or exited. … the state finds the unorganised and non-sequential crowd at fault, leading to injuries and fatalities. … This normative effort to shape comportment is echoed by the economist Kaushik Basu, recalling a Kolkata initiative that trained people to queue for buses. His language, like that of the Delhi Police, contrasts the positive queue and negative crowd: ‘there were volunteers with loudspeakers, urging anarchic gatherings of men and women at bus stops to “stand in a line”’. The key term invoked in these examples is discipline: the masses are seen to be unable or unwilling to adopt it.”

Perhaps this is why a day after the NDLS and the Sandhya Theatre crowd crushes, the response was to deploy dozens of paramilitary and police personnel (respectively) to ‘manage’ the crowd. “The crowd,” Gandhi continued, “is understood as pre-rational and overly passionate. Queue jumping is equated with indiscipline. For technocrats and planners, the prefabricated, serial queue moulded by external authority becomes an ideal. It is upheld as the fleeting sign of rational modernity.”

People often rush into buses and trains because they are concerned they may not find a seat to sit on or even a spot to stand in, in turn because there aren’t as many trains or buses as they need. They rush, as The Hindu wrote in its February 18 editorial on the NDLS stampede, because they lack “the money to buy another ticket should they miss a train” or for fear of “suffering injuries due to unsafe pedestrian conditions”. Even if crowds collect everywhere on the planet for many reasons, the Indian one is shaped and motivated by specific local, yet not particularly complicated, constraints.

A not so madding crowd

Scientists are not often able to ‘test’ hypotheses related to crowd safety with real people because they can’t guarantee the participants’ safety. Instead, experts have resorted to models — increasingly fine-turned over the years based on real-world observations as well as findings from studies of human biology and psychology — and monitoring the dynamics of real-world crowds using cameras and sensors. An additional layer of complexity here is that crowds seem to evolve, move or dissipate according to different ‘laws’ depending on how closely packed they are.

Antoine Tordeux, junior professor of traffic safety and reliability at the School of Mechanical Engineering and Safety Engineering, University of Wuppertal, Germany, wrote in Nature earlier this year: “Although pedestrian models have been developed and analysed extensively for the past 30 years, there are few models for dense crowds, and no scientific consensus on how density changes dynamics.”

This said, researchers’ efforts have still yielded some fine-tuned suggestions that have reinforced two ideas. First, there is a science at work: even if human behaviour is unpredictable (although many models have suggested it needn’t be), a combination of (i) behavioural and technological strategies supported by (ii) reliable infrastructure and communication, and (iii) backed up by emergency preparedness can prevent crowd-related accidents. Second, the science offers further opportunities to adjust the ‘landscape’ that people navigate in ways that minimise the risk of crushes or collapses. Just a few examples from the research landscape follow.

A September 2000 study by researchers from Hungary and Germany modelled pedestrian behaviour during an “escape panic” and found a so-called faster-is-slower effect. If people are simultaneously moving towards an exit, some people trying to move faster than others won’t be able to escape first; instead, they will clog the exit because they will reach and try to pass through it even as others are just getting there and yet others are there already. The result is longer evacuation time even though the crowd’s average speed is higher. To counter this, many other studies have found that placing an obstacle a short distance from the exit, in the path of people running (but not walking) towards it, could force them to slow down without stopping and move in a more orderly fashion.

A March 2019 study by researchers from China further this solution by examining the particulars of the space leading up to a 0.8-m-wide exit: “Setting I is a 15-m-wide room and Setting II is a 3-m-wide corridor formed by the same room with lateral constraints. Numerical simulation in Setting I showed that a 1 m-obstacle at 1 m away from the exit can reduce evacuation time by approximately 49%. However, a contrasting increase of evacuation time by approximately 64% is observed in Setting II for the obstacle at the same location.”

In an analysis uploaded two months later, two researchers from the US modelled the behaviour of different sizes and shapes of crowds based on people’s tendency to optimise for two conditions during a panic: “the search for less congested areas and the tendency to follow the stream unconsciously”. As a result, they found spaces with multiple exits could benefit by placing them at certain strategic points rather than close to each other. For example, the duo suggested the exits could be located at opposite ends of a room or close to areas in which people have already been encouraged to congregate at, with the net effect of splitting crowds up and easing individual decision-making. Tokyo’s railway stations make generous use of this simple principle to prevent passenger density from building up.

In a study published in February this year, researchers from France and Spain examined a dense, standing crowd. A particularly popular event in Spain’s annual cultural calendar is the encierro, the running of the bulls, during the San Fermín festival in Pamplona city, held every year from July 6 to 14. In 2019, 2022, 2023, and 2024, the researchers set up cameras on balconies overlooking the plaza where the festival begins, where around 5,000 people gather on July 6 evening for the opening. The crowd density moves on average between four and six people per sq. m in the span of two hours but sometimes reaches as high as nine. The research team collected their videos and used computer-vision techniques to autonomously analyse the movement of people’s heads in the videos.

The researchers found that for about an hour before the festival began, people accumulated in the plaza at a uniform rate, their density increasing linearly from two to six people per sq. m. When the density crossed four, however, “fluctuations” in individuals’ speed started to grow as well, in a non-linear way and in small “activity bursts”. It was a signal that the crowd’s collective behaviour had switched in some fundamental way to a different regime. As the overall density grew further up to nine people per sq. m, the computer’s analysis revealed a curious pattern: smaller groups of hundreds of people packed together over many metres started to rotate in periodic fashion, completing one round every 18 seconds or so.

This newfound movement suggests that even dense crowds are not perfectly static and that it could create opportunities for people to fall, injuring themselves or becoming an ‘obstacle’ to others. The team developed a model to explain the movement and concluded that its period of rotation was related only to the density of the crowd.

This is an opportune finding insofar as it offers planners something to work with until the study of dense crowds matures and the field finds surer footing. Then again, the gaps extant today that lead to unsafe conditions for pedestrians and travellers don’t require planners to wait for the science to firm up: unreliable infrastructure, complacent planning, and exclusive rather than inclusive policymaking.