How Crowds Really Behave

AThe vocabulary used to describe large gatherings was settled long before anyone had measured one. Crowds catch fear as bodies catch a virus, lose their reason at the moment they lose their separateness, and turn on whatever stands in front of them: this is the picture bequeathed by nineteenth-century writers on the subject, and it has survived, almost unaltered, into the training manuals of several police forces. Its durability is not hard to explain. The account is vivid, it matches the memory of anyone who has been frightened in a confined space, and it carries the further convenience of locating the cause of a disaster inside the people who died in it rather than in the arrangements that had been made for them. Between 1994 and 2019 the official reports on nine large gatherings that ended badly reached for the word panic 61 times, and in only four of those documents was any attempt made to say what it meant. Definition has never seemed necessary, because the word is felt to describe something everyone has seen. Almost none of it has survived contact with recordings of what people actually do.

BDagmar Reinholt, who has led the crowd research programme at the Kelvinside Centre since 2010, argues that disorder is not a property of a gathering but a relation between a gathering and whoever is policing it. On her account a crowd arrives already divided, most of it indifferent to trouble and a small fraction hoping for it, and what settles the evening is whether those two parts are treated as one. Force applied to everybody converts the indifferent majority into people who now share a grievance with the minority they came in despising. Her team compared 78 football fixtures at Brackenhall in which officers moved among supporters in ordinary uniform with 78 matched fixtures policed by lines of officers in protective equipment, and recorded 23 disorder incidents in the first set against 71 in the second. The gap held once crowd size, weather and the importance of the fixture had been allowed for. Reinholt insists that none of this is peculiar to football, and her group has since reported the same pattern at street demonstrations and at two large festivals.

COwen Marchetti, a public order specialist at Ferrisbridge, accepts the mechanism and doubts the recipe. Graded policing works, he maintains, only where a crowd contains people who can speak for it, because the whole procedure depends on somebody being available to negotiate with, and a growing number of the gatherings that worry planners have no organisers at all. He points to a series of unannounced assemblies coordinated by messaging applications, at which officers who deliberately held back found nobody to hold back for, and where the eventual intervention came later than it would once have done. Planners, he adds, do not get to choose which sort of crowd turns up on a given night. The evidence he offers is slighter than the evidence he attacks. His comparison rested on 11 events in a single country over one season, the classification of police behaviour was supplied by the forces themselves, and no independent measure of what the crowds intended was available at all. Reinholt has since conceded that leaderless gatherings are the harder case, while dismissing the suggestion that they are a different kind of animal.

DA second body of work has grown up almost independently of the first, and it treats a packed gathering as a problem in fluid mechanics rather than in psychology. Above a density of roughly six people to the square metre, individuals can no longer choose where to put their feet, and pressure applied at one edge of the mass travels through it in waves capable of lifting people off the ground. Nobody caught in such a wave is behaving irrationally, and nobody caught in it is behaving at all. The remedies this tradition proposes are architectural: entrances sized for the flow they will actually receive, routes that do not narrow without warning, and announcements specific enough to be acted upon. An audit of 61 venues completed in 2018 found that 19 of them had at least one route whose capacity fell by more than a third partway along it, and that six of those routes were the ones marked for emergency use. The observation most often missed is that the older word does its worst damage after the event, when it supplies an explanation requiring no changes to the building.

EThe two traditions are not rivals, though they are frequently presented as though they were, and the sensible reading is that they describe different stages of the same evening. How a gathering is handled while it can still move determines whether it ever becomes dense enough for the physics to take over. Where organisers exist and are spoken to, Reinholt's evidence is strong and the advice that follows from it is clear. Where they do not, Marchetti is right that the recipe has little to offer, and the weight falls back onto the design of the site. What both traditions reject is the inherited story, which survives less because it explains anything than because it consoles. Neither tradition holds that a gathering is safe merely because nobody within it is frightened, and neither expects one measure to answer for every kind of event. The gatherings that end badly are seldom the ones that were most frightened; they are the ones that were most tightly packed.