In a ruling that may reshape how India manages its massive public gatherings, the Allahabad High Court recently observed that crowd management in historic pilgrimage hubs like Mathura and Vrindavan is fundamentally failing because it is not based on science. Hearing a case originally centred on illegal local constructions, Justice Vinod Diwakar expanded the bench’s focus to address recurring, catastrophic stampedes during major religious festivals.
The High Court pointed out a critical administrative blind spot: local authorities routinely reduce crowd management to mere traffic control, reflecting a dangerous misunderstanding of crowd dynamics. The court observed that deploying more police personnel, erecting barricades and issuing standard operating procedures without understanding crowd science would not prevent future disasters. Pointing to established research, the court called for institutionalisation of crowd behaviour as a scientific discipline, recommending dedicated university courses. The court also suggested establishing a Centre of Excellence for Crowd Science and mandatory engagement of crowd behaviour experts for major public gatherings.
This judicial intervention highlights a profound truth that physicists, mathematicians, and crowd scientists have been proving for decades: deadly crowd disasters are rarely the fault of wild, irrational psychological panic. Instead, they are predictable structural failures dictated by the unforgiving laws of Newtonian science, fluid dynamics and particle physics. For decades, news headlines have blamed crowd fatalities on an animalistic, chaotic “stampede” where a crazed public blindly tramples over one another. However, modern studies have thoroughly discredited this narrative. The fatal breakdown is almost always structural, not psychological.
The term “stampede” shifts the blame from poor planning to the victims themselves, hiding the physical reality of the situation. True trampling deaths are exceedingly rare; if a space is open enough for people to actively run over others, it is usually open enough for them to escape. Scientists who study these events have concluded that most fatal incidents are not classic stampedes of people running in blind panic. They are crowd crushes, where extreme density turns a mass of human bodies into a dangerous physical system governed by the same laws that describe fluids, granular materials and even earthquakes.
In this article, we discuss several theories that have been developed to understand how such incidents take place.
Density: The Real Trigger
The single most important factor is local density — how many people stand in each square metre — rather than the total number of people present. At low densities of one or two persons per square metre, people can still move freely, adjust their paths and avoid collisions. Around four persons per square metre, involuntary body contact begins. Between five and six persons per square metre, movement becomes difficult. At six to seven or higher, the crowd starts to behave like a continuous medium. Forces transmit from body to body, and individuals lose the ability to control their own motion.
Keith Still, a leading crowd scientist, has repeatedly pointed out that true human stampedes of people fleeing in panic are rare and seldom cause mass fatalities. People do not die because they panic; they panic because they are already being crushed. Deaths usually result from compressive asphyxia — the chest cannot expand to breathe — either while people remain upright or after they fall and others pile on top. The language of “stampede” often shifts blame onto the victims instead of the organisational failures that allowed density to reach lethal levels.
Careful analysis of video footage from major disasters has revealed clear transitions in crowd behaviour as density rises. The most detailed empirical study was carried out by Dirk Helbing, Anders Johansson and Habib Zein Al-Abideen after the 2006 Hajj disaster in Mina, near Mecca. More than three hundred pilgrims had died in that incident. By tracking thousands of individuals in the footage, the researchers observed two successive sudden transitions.
At moderate densities, the flow remained laminar — smooth and forward-moving, much like orderly traffic. As density increased, stop-and-go waves appeared. People would halt for several seconds and then move forward again, with these waves travelling upstream against the direction of the crowd. When density rose still further, the motion became turbulent. Clusters of people were suddenly pushed in random directions. Pressure built up and then released in eruptions similar to small earthquakes. These sudden displacements caused people to lose balance, fall, and be trampled or crushed by those still being forced forward. The researchers defined a simple measure of local “pressure” as density multiplied by the variance of velocity. When this quantity exceeded a critical value, turbulence set in, and fatalities followed roughly ten minutes later.
This sequence — laminar flow, then stop-and-go waves, then turbulence — has been observed in other disasters, including the 2010 Love Parade in Duisburg. The transitions show that the crowd is not simply panicking; it is undergoing physical phase changes driven by density and force transmission.
The Fundamental Diagram: The Mathematics of Capacity
In civil engineering and architectural design, the relationship between crowd density, walking speed, and overall movement is governed by a mathematical benchmark known as the Fundamental Diagram. Originally pioneered by safety engineers to establish operational “Levels of Service”, this diagram plots pedestrian flow as a function of local density. In low-density environments, individuals walk at their desired free speed, resulting in a linear increase in flow as more people enter.
However, as density approaches a critical threshold of approximately 1.5 to 2.0 people per square metre, physical space constraints force walking speeds to drop. This creates a non-linear parabolic curve where flow reaches an absolute maximum capacity before entering a sharp, congested decline. If density continues to climb toward extreme levels, flow degrades toward zero—a state of complete gridlock. This empirical curve forms the regulatory foundation for global building codes, dictating corridor widths, exit capacities, and stadium designs, serving as the primary quantitative tool for translating macroscopic crowd physics into actionable architectural safety guidelines.
The Microscopic Realm: Newton in the Crowd
To understand the mechanics of crowd collapses, scientists have built highly sophisticated microscopic simulation frameworks. The most influential of these is the Social Force Model (SFM), pioneered by Dirk Helbing, Péter Molnár, and Tamás Vicsek. The Social Force Model takes Newton’s second law of motion as its guiding mathematical principle.
Here, an individual pedestrian is treated as a self-propelled, active particle whose instantaneous acceleration is driven by three main force vectors: a self-propelling driving force that represents the pedestrian’s desire to reach a destination, a social interaction force representing repulsion from other people, and physical forces that arise during direct bodily contact. A separate term representing stochastic behavioural fluctuations, mathematically analogous to thermal fluctuations in molecular dynamics, is added to capture random, non-deterministic micro-decisions.
The self-propelling driving force models the conscious intention of the pedestrian to adapt their actual speed and direction to their desired velocity pointing toward their target destination. This adjustment is not instantaneous but occurs within a characteristic relaxation time, which represents the biomechanical delay of the human body in reacting to changes in the environment.
The core of Helbing’s formulation lies in the interaction force, which separates into a socio-psychological repulsive force and a mechanical, physical contact force. The social force represents the psychological aversion of individuals to get too close to others, reflecting Edward T. Hall’s classical sociological theories of proxemics. Mathematically, it is modelled as an exponential decay function that decreases as the actual physical distance between pedestrians increases.
In this formula, desired interpersonal distance is the target desired interpersonal distance that a pedestrian expects to keep from a neighbouring agent, repulsive strength is the repulsive force strength, interaction range is the characteristic interaction range, and normal unit vector is the normalised vector pointing from the neighbouring agent to the pedestrian. An anisotropic weight factor is often introduced to account for human foresight, meaning individuals are far more sensitive to obstacles directly in their line of sight than behind them.
However, the social force is not a real physical force in the classical Newtonian sense. As crowd dynamics researchers have pointed out, it is an empirical, non-physical mathematical tool that represents mental opinion and psychological motivation. Individuals implement their desire to avoid others by using their muscles to generate foot-ground friction, which is the actual physical force that obeys Newton’s third law.
When the local density of the crowd is low or moderate (fewer than four people per square metre), individuals can navigate comfortably, and the socio-psychological steering drives dominate the system. But as the density rises, the physical reality changes dramatically. When the actual physical distance falls below the sum of the physical body radii, people come into direct physical contact.
At this point, the non-physical social forces are completely overwhelmed by real, Newtonian mechanical forces: a normal body compression force that resists squeezing, and a tangential sliding friction force that opposes relative lateral movement. These forces are governed by high elasticity and friction coefficients. This transition marks a fundamental reinterpretation of the Social Force Model from the perspective of psychological stress. The gap between desired velocity and actual velocity represents time-related stress (time pressure), while the gap between desired distance and physical distance represents space-related stress.
This formulation mirrors the famous Yerkes-Dodson Law of psychology, which states that performance increases with arousal up to an optimal threshold, beyond which performance sharply deteriorates. In a crowd, a moderate amount of time pressure accelerates individuals, leading to a ‘faster-is-faster’ effect. However, if the desire of the crowd to move too fast exceeds the physical capacity of the bottlenecks, the system enters a regime of severe distress. Competitive pushing and spatial compression build up horizontal forces that can bend steel fences, causing stumbles and falls. When someone stumbles, they become a physical obstacle that others cannot bypass, causing a rapid progressive collapse that completely blocks the exit. Under these conditions, trying to move faster paradoxically decreases the outflow rate—a phenomenon widely known as the ‘Faster-is-Slower’ (FIS) effect.
Cognitive Heuristics: The Visual Mind of the Pedestrian
While force-based models treat pedestrians as physical particles reacting to mathematical field forces, modern cognitive theories argue that human navigation is driven by simple visual rules. This visual-routing framework posits that pedestrians do not calculate complex mathematical force fields. Instead, they constantly scan their visual horizon to identify the “first unobstructed gap” in their path and adjust their walking angle to avoid collisions.
By prioritising cognitive vision over physical forces, these models capture how individuals actively seek out open spaces and dynamically adapt to moving obstacles. In normal densities, this visual heuristic approach highly accurately reproduces collective patterns like lane formation in two-way corridors. It elegantly addresses a major limitation of purely physical models by acknowledging that humans are active, decision-making agents with sight and intent, rather than passive, blind molecules. Only when the crowd is compressed to the point of direct body-to-body contact do these cognitive visual decisions fail, handing complete control of the system over to the raw laws of classical Newtonian mechanics.
The Macroscopic Perspective: Crowds as Continuous Fluids
While microscopic models trace individual paths, treating a massive crowd of thousands of people as a continuous medium allows researchers to apply the powerful mathematics of fluid dynamics. When a crowd is sufficiently dense, individual behavioural variations average out at scale, and the crowd’s spatial and temporal evolution can be accurately predicted using continuous density and velocity fields.
The mathematical foundation of macroscopic crowd modelling is anchored in the Mass Continuity Equation, which mathematically expresses the conservation of mass during pedestrian flow. In this continuity equation, density represents pedestrians per square metre, and velocity represents the movement speed and direction of the crowd flow.
To simulate how the momentum of a crowd field changes over time, physicists adapt the classical Navier-Stokes equations from fluid mechanics. However, because human crowds consist of self-propelled, active matter rather than passive, inert fluid molecules, they do not conserve physical momentum or kinetic energy. Pedestrians constantly inject energy into the system as they struggle to reach their targets. To capture this active drive, the standard Navier-Stokes viscosity is augmented with a Rayleigh-like relaxation term.
In this momentum conservation equation, social pressure acts as a density-dependent ‘anticipation’ term that models the psychological tendency of pedestrians to slow down and avoid crowded areas. The desired velocity field represents the velocity individuals would maintain in the absence of other people or physical barriers. The characteristic relaxation time dictates how quickly pedestrians adjust their actual velocity to their desired velocity, and the internal viscosity coefficient represents the social and physical friction generated by lateral interferences and avoidance manoeuvres.
The active term in the equation represents the constant injection of internal motivation energy. At extreme densities of over seven people per square metre, this active drive combined with extreme physical body compression can lead to non-hyperbolic system states, producing severe localised instabilities and violent horizontal pressure waves.
A major milestone in macroscopic theory is the Hughes Pedestrian Flow Model, introduced by Roger L. Hughes in 2002. The Hughes model couples mass conservation with a non-linear Eikonal equation that models human path-planning.
In this elegant framework, a potential function represents the subjective instantaneous travel cost to the target exit, and a monotonically decreasing function of density represents the maximum pedestrian walking speed. At zero density, pedestrians walk at their maximum free speed, but as density reaches a critical limit, walking speed drops to zero.
The direction of movement is determined by the descending gradient of the potential field. The potential field acts as a global pathfinder. The Eikonal equation ensures that pedestrians choose their routes by dynamically balancing physical distance and local crowd density. If a corridor or exit door becomes heavily congested, walking speed drops toward zero, driving the gradient of the potential field to increase rapidly. This mathematical mechanism steers incoming pedestrian flows away from the congested bottleneck toward alternative, less dense exit routes.
Granular Physics, Soft Matter, and Crowd Turbulence
When crowd density crosses a critical threshold of approximately six to seven people per square metre, the crowd undergoes a dramatic physical phase transition. The continuous fluid-like flow breaks down, and the crowd begins to behave like a driven granular medium or soft matter — deformable materials that exhibit both solid and liquid characteristics under extreme pressure.
In a landmark 2007 study, Dirk Helbing, Anders Johansson, and Habib Zein Al-Abideen analysed video recordings of the tragic 2006 Hajj crowd disaster at the Jamarat Bridge in Mina, where 363 pilgrims died. They discovered that at extreme densities, individual human agency is entirely lost, and forces are transmitted directly through body-to-body contact.
Under these quasi-static, granular conditions, applied forces are not distributed evenly throughout the mass. Instead, they propagate along linear networks of touching bodies called ‘force chains’. These force chains branch randomly and terminate at rigid, unyielding architectural boundaries, such as walls, fences, or locked exit gates.
This granular behaviour explains the mechanics of arching and clogging at exits. When an impatient, dense crowd attempts to pass through a narrow doorway, competitive pushing causes individuals to form a curved, load-bearing physical arch across the opening. This physical arch behaves exactly like a stone masonry arch, deflecting forward-pushing forces laterally into the doorframe. Because of strong tangential friction between bodies, the arch remains locked and stable, completely halting the egress of the crowd.
At even higher densities, exceeding eight people per square metre, the stop-and-go waves transition into a chaotic phase called crowd turbulence. Physical interactions generate sudden, multi-directional, seismic-like pressure shockwaves—frequently termed ‘crowd-quakes’—that propagate through the crowd body. These shockwaves can travel up to three metres in lateral directions, instantly lifting people off their feet and carrying them along against their will.
When a crowd-quake occurs, individuals lose their balance, triggering a progressive ‘domino effect’ where people stumble and fall into open voids. Surrounded by an unyielding mass of compressed bodies, those behind cannot stop their forward momentum and are pushed directly over those who have fallen. This leads to the formation of massive stacks of fallen bodies, which in some cases reach heights of up to three metres.
At the bottom of a three-metre pile of bodies, individuals experience staggering vertical loads of 3,600 to 4,000 Newtons, equivalent to 360 to 410 kilograms of force. Even in horizontal crushes without falls, the cumulative pressure from rear ranks pressing forward can generate forces exceeding 450 kgs.
The physical mechanics of death in these disasters is almost always compressive asphyxiation. When a human chest cavity is subjected to forces of this magnitude, the ribcage is compressed, utterly preventing the lungs from expanding to draw breath. Forensic autopsies of victims from major crowd crushes—such as the 1989 Hillsborough disaster in England (where 97 fans died) or the 1991 New York City basketball game crush (which claimed 9 lives)—confirm this mechanism. The vast majority of the deceased show facial and conjunctival petechiae, which pinpoint haemorrhages caused by high venous pressure, with minimal external blunt trauma or bone fractures. They were literally squeezed to death while standing completely upright in the crowd.
The Game-Theoretic Dimension
To understand how individual choices contribute to the build-up of these lethal physical forces, researchers have turned to evolutionary game theory. During a high-stakes emergency egress, conflicts occur when multiple self-interested individuals attempt to occupy the same physical space near an exit. Because individuals cannot coordinate or negotiate binding agreements under stress, they are locked in a dynamic, non-cooperative game.
In a comprehensive study published in the Journal of Artificial Societies and Social Simulation (JASSS), researchers modelled crowd egress as a game where agents can adopt one of four behavioural strategies: Cooperators, who are passive and yield; Defectors, who are aggressive and push; Evaluators, who play strategically depending on their opponent’s physical size; and Retaliators, who start cooperatively but escalate if pushed first.
In this framework, the payoff for an agent attempting to move to a desired location depends on the strategies of neighbouring agents, their physical size, and a conflict time delay. When two defectors compete, the larger and mightier agent dominates the move, while the smaller agent is penalised. Furthermore, when multiple defectors push and jostle simultaneously, the physical conflict introduces a time delay that slows down the entire local flow.
In the above illustration, T = Temptation to Defect — Pushing to escape first, R = Reward for Mutual Cooperation — Orderly queuing, P = Punishment for Mutual Defection — Collective clogging, and S = Sucker’s Payoff — Yielding while others push. The Mathematical Conflict is: T>R>P>S. This inequality is the mathematical engine of emergency egress failures. Because the individual temptation to push (T) is higher than orderly queuing (R), and orderly queuing (R) is safer than mutual pushing (P), which is still better than being pushed over and left behind (S), rational self-preservation drives everyone to defect (push).
To analyse the macroscopic effects of these microscopic strategies, researchers simulated evacuations using heterogeneous populations composed of four distinct behavioural profiles: Risk-Seeking, Risk-Averse, Risk-Neutral, and Best-Response Agents who dynamically update their strategy at each step based on the observed strategies of neighbours in the previous turn.
The simulation results revealed a critical relationship between crowd psychology, local density, and physical pressure. When the proportion of risk-seeking agents in the population increases, the average crowd pressure, local density, and the number of injured agents increase dramatically. Interestingly, the simulations demonstrated that a high crowd density (overcrowding) alone is not sufficient to trigger a disaster. A crowd can be extremely dense near an exit, but if the local crowd pressure remains below a critical threshold, the flow remains stable and safe.
However, when aggressive risk-seeking behaviour dominates, the dynamic conflict time delay increases from half a second to one and a half seconds, causing localised ‘jams’ and spike-like pressure waves. This physical pressure quickly exceeds the physiological limits of the human body, leading to severe injuries and falls.
Crucially, the game-theoretic simulations proved a powerful counter-measure: if the crowd population is composed entirely of risk-averse and risk-neutral individuals, the average crowd pressure remains exceptionally low, and crowd disasters can be prevented entirely, even under potential threat conditions and high physical congestion.
Computational Synthesis: Smoothed Particle Hydrodynamics
To put these interdisciplinary theories into practice, computer scientists and safety engineers use cutting-edge computational frameworks to simulate and predict crowd behaviours in real time. A major breakthrough in modern crowd simulation is the integration of Smoothed Particle Hydrodynamics (SPH) into agent-based models.
Originally developed for astrophysics, SPH is a mesh-free Lagrangian numerical method popular for simulating complex fluid flows. SPH represents a continuous fluid as a set of moving particles, where each particle possesses material properties such as mass, position, and velocity, and moves according to macroscopic hydrodynamic laws.
By treating each individual pedestrian in a crowd as an SPH particle, researchers can simulate high-density, fluid-like behaviours—such as propagating shockwaves and transverse oscillations—without losing track of the individuality, goals, and distinct pathways of single agents.
At each time step of the simulation, the local crowd density at an individual’s position is calculated by interpolating the contributions of all neighbouring pedestrians within a support radius (typically set to one metre) using a smoothing kernel function.
To make SPH suitable for conscious human crowds rather than passive physical fluids, researchers introduced two major innovations. First, each agent is assigned a dynamic personal rest density. Unlike a fluid with fixed physical properties, humans can consciously choose to accept higher or lower densities around them depending on their circumstances. The rest density represents the crowd density that an agent is currently willing to tolerate. It is dynamically updated as a moving average of the actual density perceived by the agent over the last several seconds. The pressure is modelled as proportional to the difference between actual density and this personal comfort threshold. If the local density falls below this personal comfort threshold, the pressure force is set to zero, preventing the unrealistic ‘clustering’ or ‘splashing’ of agents at the boundaries of the crowd.
Second, the simulation utilises density-dependent behaviour blending. At low densities, pedestrians navigate using advanced collision-avoidance algorithms, allowing them to steer smoothly and avoid contact. But as local density increases, individuals have less room to steer, and their behaviour becomes physically constrained.
The model smoothly interpolates from active, agent-based navigation to passive, SPH-derived physical interactions as density climbs. This allows for the realistic, real-time simulation of mixed-density scenarios containing tens of thousands of agents, accurately reproducing propagating shockwaves and fluid-like wave reflections off architectural boundaries.
Spontaneous Oscillations: A Recent Discovery
The introduction of active-matter physics to crowd science has radically shifted our understanding of high-density disasters, moving away from traditional collision-avoidance algorithms and passive fluid-dynamic models. A key breakthrough in this domain is the theory of collective oscillations in confined crowds, developed by physicist Denis Bartolo and his co-workers in a landmark Nature study published in February 2025.
When thousands of self-propelled, active-matter agents are compressed into a confined space, they do not merely scatter or experience linear compression; instead, they undergo a distinct collective phase transition. At a critical density threshold of approximately four pedestrians per square metre, the crowd spontaneously self-organises into macroscopic “chiral oscillators”—massive, rotating vortex structures where hundreds of individuals coordinate their orbital motion in circular patterns without any central leadership, external cues, or conscious co-ordination.
The underlying mechanics of these self-perpetuating vortices lie in the concepts of “odd frictional forces” and non-reciprocal interactions, which are characteristic of chiral active fluids. In standard passive fluids, such as water, molecules experience symmetric, reciprocal friction that obeys Newton’s third law of motion and Onsager reciprocal relations. However, in a tightly packed crowd of self-propelled human beings, these reciprocal symmetries are broken. As pedestrians jostle, slide, and attempt to force their way forward, they exert lateral and rotational forces on their neighbours that are fundamentally non-reciprocal—one person’s action does not trigger an equal and opposite physical reaction due to individual self-propulsion and localised sensory limits. This asymmetrical rubbing generates macroscopic “odd frictional forces” (analogous to odd viscosity) that couple longitudinal compression directly to transverse shearing. Under intense confinement, this non-reciprocal feedback loop triggers a phase transition, causing massive sections of the crowd to coherently rotate in a randomly selected direction (either clockwise or counter-clockwise) with a highly regular orbital period of approximately 20 seconds.
This theory was empirically validated using advanced video analysis of real-world, high-density gatherings. The researchers tracked and analysed the precise trajectories of thousands of individuals during the high-density “running of the bulls” event at the San Fermín festival in Pamplona, Spain, proving that these are stable, physical structures that emerge naturally under confinement. Crucially, the team extended their analysis to archival video footage of the tragic 2010 Love Parade disaster in Duisburg, Germany. They discovered that identical collective chiral oscillations—large-scale orbital swayings and rotational eddies—emerged at the crowded entrance ramp just minutes before the onset of the fatal crowd collapse. This empirical link demonstrates that crowd turbulence and dangerous “crowd-quakes” are not purely random, chaotic events, but rather the catastrophic climax of these predictable, non-reciprocal hydrodynamic states.
The discovery of macroscopic chiral oscillations has profound practical implications for real-time crowd management and predictive safety engineering. Traditionally, predicting a crowd disaster before physical crushing occurred was an elusive task. By proving that dense crowds behave as active chiral fluids with predictable periodicities, Bartolo’s work provides a direct blueprint for automated surveillance. Drone-based or CCTV-mounted video feeds can be processed in real time using optical flow algorithms to map the emergence of these coherent, rotating vector fields. Because these macroscopic chiral oscillations serve as a distinct, measurable precursor to complete crowd collapse, detecting these circular movements offers a non-invasive “early warning” signal. This provides event organisers and emergency services with a vital window of several minutes to execute targeted spatial interventions—such as opening alternative exits or introducing soft, flow-disrupting barriers—to break up the rotating feedback loop and prevent a disaster.
From Blaming the Victim to Scientific Engineering
For decades, the standard administrative response to crowd disasters has been to blame the victims for ‘panicking’ or ‘rushing,’ treating the tragedy as an unavoidable act of human madness. But as crowd science has proven, a crowd of calm, orderly people at six people per square metre is just as physically dangerous as a crowd of frightened people. Once density crosses that critical threshold, physics takes over, and individual human agency is superseded by unyielding mechanical laws.
The solutions, therefore, are structural and operational, not moral or psychological. Crowd disasters must be treated as systemic physical failures. Modern computer simulations based on the Social Force Model and SPH can be utilised by urban planners, police and administration to identify invisible bottlenecks, optimise the placement of architectural barriers, and design exits that actively prevent the formation of load-bearing physical arches.
Furthermore, game-theoretic models show that deploying trained crowd-guidance personnel, who act as calm cooperator or ‘game changer’ agents, can suppress panic contagion, lower overall crowd pressure, and prevent progressive collapses entirely.
By moving away from the inaccurate and misleading rhetoric of ‘stampedes,’ society can finally take structural responsibility for public safety. Ultimately, managing a crowd is not a matter of policing human behaviour—it is a matter of engineering space to respect the laws of physics, ensuring that the right to gather in pilgrimage or celebration is never again paid for in human lives.
