Module 6, lesson 3 · 45 minutes

Faster-is-slower and the obstacle question

A famous simulation says hurrying slows an evacuation and a column by the exit speeds it up. This lesson sets out what human experiments actually found, with the numbers.

You’ll be able to

state what the social force model predicts about desired speed and about obstacles, and why;
summarise the human experiments for and against each prediction, with their numbers;
explain why the evidence points to pushing and contact, not speed itself, and what that means for design.

In 2000, Dirk Helbing, Farkas and Vicsek published a short paper in Nature with a startling result. In their simulated room, people who tried to leave faster got out slower. Above a certain urgency, the crowd jammed at the door in arch-shaped clogs, broke free in bursts, and jammed again. The paper added, almost in passing, that a column placed in front of the exit could improve the outflow.

Both ideas are elegant and counter-intuitive, which is exactly why they need testing. Over the following two decades experimenters put real people, and sheep, through doors to find out. The results are mixed in ways that teach you a great deal about what a model is and is not saying. This lesson presents both sides with their numbers.

What the model predicts

You met the model in module 4’s social force lesson. Each simulated person feels a driving force towards the exit, scaled by a desired speed v⁰, plus social repulsion from others and, when bodies touch, a stiff body force and a sliding friction. That last pair is the key. Once people are packed against each other at an opening, friction between bodies can lock a line of them across it.

The paper’s claims, in its own words where it has them:

Faster-is-slower is the model prediction that raising people’s desired speed beyond a point lowers the rate at which they get through an exit. “Desired velocities higher than 1.5 m/s reduce the efficiency of leaving.”
From v⁰ ≈ 1.5 m/s the outflow becomes irregular, with “an irregular succession of arch-like blockings of the exit and avalanche-like bunches of leaving pedestrians”.
Injuries appear once the simulated pressure exceeds 1,600 N/m, around v⁰ = 5 m/s.
“Improved outflows can be reached by columns placed asymmetrically in front of the exits, which also prevent the build up of fatal pressures.” This is stated in the text; no figure shows it.

For calibration, the paper treats about 0.6 m/s as relaxed, about 1 m/s as normal and up to about 1.5 m/s as nervous. Notice what v⁰ bundles together. In the model, a higher desired speed means pressing harder into the people in front, because the driving force scales with it. There is no separate dial for “hurrying politely”.

The mechanism behind both predictions. An arch (a) forms when people jam shoulder to shoulder across an opening and pressure from behind holds them there; a column in front (b) is meant to break arches up before they form. Schematic.

The social force sandbox in module 4 reproduces the prediction: its “run experiment” button sweeps desired speed and plots leaving time, which dips and then rises. Keep that curve in mind as you read the experiments below.

What human experiments found

Evidence for: when pushing is allowed

Garcimartín, Zuriguel and colleagues ran drills with 85 students and a single 75 cm door. In five drills people were asked not to push; in seven, moderate pushing was allowed. They measured the time between one person passing the door and the next:

Pushing made the door slower. Mean time lapse per person 0.32 ± 0.05 s without pushing and 0.38 ± 0.04 s with it; clogs appeared only in the pushing drills.

A rate of one person every 0.32 s is about 3.1 people per second through the door; one every 0.38 s is about 2.6 (our arithmetic). For the whole group of 85, that is roughly 27 s against 32 s. Clogs, the arches of figure 1, appeared only in the competitive runs, and the distribution of the gaps between people had the heavy “power-law” tail that clogging produces.

This is real evidence of faster-is-slower, with an important qualification. Garcimartín’s team varied competitiveness, whether people were allowed to push, not the speed they wanted to go. Zuriguel and colleagues later showed that the same clogging transition appears for sheep, grains and colloids squeezing through bottlenecks; their pedestrian case, however, was a simulation of 30 agents at a desired speed of 6 m/s, not people.

Evidence against: when nobody shoves

Milad Haghani, Sarvi and Shahhoseini put 114 participants through doors 60 to 120 cm wide at three levels of urgency, all without aggressive physical competition, reaching densities around 9 p/m² at the door. Their finding is in the title of the paper, “When ‘push’ does not come to ‘shove’”: faster “was invariably faster”. Whether explicit, aggressive shoving occurs, they concluded, is what decides the outcome.

Taking stock in 2024, Haghani and Ronchi’s review in Collective Dynamics concluded that empirical research on faster-is-slower has “produced mixed evidence rather than irrefutable support”.

Faster-is-slower: the evidence at a glance
Study	Who or what	What varied	Result
Helbing, Farkas & Vicsek 2000	Simulation, 200 agents, 1 m door	Desired speed	Slower leaving above 1.5 m/s; arches
Garcimartín et al. 2014	85 students, 75 cm door	Pushing allowed or not	0.38 vs 0.32 s per person; clogs only with pushing
Zuriguel et al. 2014	Sheep, grains, colloids; simulated pedestrians	The system squeezing through	Same clogging transition across systems
Haghani, Sarvi & Shahhoseini 2019	114 people, 60–120 cm doors	Urgency without shoving	“Faster was invariably faster”
Haghani & Ronchi 2024	Review	–	“Mixed evidence rather than irrefutable support”

Reconciling the two: contact, not speed

Put the results side by side and they stop contradicting each other. Drills where pushing was allowed showed slower egress and clogs. Rushing without aggressive shoving gave faster egress. The physics of faster-is-slower lives in the contact forces: friction between bodies pressed together across an opening. The model produces it because, in the model, wanting to go faster means pushing harder.

So the model is not simply wrong. It is a model of a crowd that pushes, and its desired-speed dial is really a pushing dial. That is a useful thing to model: crowds do push when they are crushed, competing or desperate, and module 2’s force lesson shows where that leads. But a designer who reads “people who hurry block exits” and concludes that calm, brisk movement is dangerous has drawn the wrong lesson.

Hurrying is not the hazard. Pushing and dense contact at an opening are. Design and management that keep densities at exits low enough for people not to be pressed together remove the mechanism, whatever speed people want to go.

A colleague cites Helbing 2000 to argue that evacuation announcements should tell people to walk slowly, because hurrying makes exits slower. How would you respond?

The model’s faster-is-slower effect comes from contact friction when people push; its desired speed also sets how hard they push. Human drills without aggressive shoving found that faster was faster (Haghani et al. 2019), while drills with pushing were slower (Garcimartín et al. 2014). The useful message is “move quickly, don’t push”, and the useful design goal is to keep density at the exit low enough that people are not pressed together.

The obstacle question

If arches cause clogs, something that stops arches forming should help. That is the idea behind putting a column or panel just in front of an exit. Here too the evidence divides.

An obstacle in front of the exit: the evidence
Study	Method	Finding
Helbing et al. 2000	Simulation	Asymmetric columns improve outflow (text only, no figure)
Yanagisawa et al. 2009	Floor-field model and experiments	“The pedestrian flow increases when we put an obstacle in front of an exit”; off-centre works better
Zuriguel et al. 2016	Sheep	An optimal position exists; too close makes flow worse
Shi et al. 2019	Physica A study	The effect “is sensitive to its size and distance from the exit”
Garcimartín et al. 2018	Three drill series (about 90, 72 and 180 people), doors 69–75 cm, column about 1 m across	“The pedestrians flow rate is not necessarily altered by the presence of an obstacle”, but dangerous transversal rushes fell
Feliciani et al. 2020	Systematic experiments, non-competitive and extremely competitive	“Researchers are split…”; the obstacle helped under urgent but not extreme competition, and can stabilise crowd waves

Read the rows for what they share. Where an obstacle helps, it helps within a window: the right size, the right distance, off-centre rather than central, and at an intermediate level of competition. Too close, it becomes a second bottleneck. And the most careful human drills found a benefit that is not about flow at all: fewer of the sideways rushes the authors judged dangerous.

An obstacle in front of an exit is a permanent change with a narrow window of benefit and a real risk of making things worse. If one is proposed, ask for evidence at the right scale, with people, for the competitiveness the exit will actually see, and ask what problem it solves: throughput, or the lateral surges that Garcimartín’s team saw reduced.

A venue proposes a 1 m column centred 0.5 m in front of its main 1.8 m exit to “speed up evacuation”. Using the table, list three questions you would put to them.

For example: (1) Why central? Yanagisawa found off-centre placement works better. (2) Why 0.5 m? Zuriguel’s sheep and Shi et al. show the effect depends on distance and size, and too close makes flow worse. (3) What level of competition do you expect? Feliciani found benefit under urgent but not extreme competition. (4) Is the aim flow, or fewer lateral surges? Garcimartín 2018 found flow not necessarily changed. (5) What tests with people support this exit’s geometry?

“Experiments proved that hurrying slows evacuation.” Drills with pushing were slower; drills where people hurried without shoving were faster. The evidence is mixed, and the variable that matters is contact.
“So the social force model is wrong.” It is a model of a crowd that pushes. Its desired-speed parameter bundles urgency and pushing together, which real people can separate.
“A column in front of an exit always helps.” Only within a window of size, position and competitiveness; outside it, it can hinder. Its clearest measured benefit in human drills is fewer lateral rushes, not more flow.
“The sheep results apply directly to people.” They show the physics of clogging is general. They do not tell you how people negotiate, cooperate or hold back.