Two crowds gather in the same square at the same hour. The first is the crowd that has formed because a fire broke out — strangers from every walk of life, anonymous to each other, momentarily fused by the spectacle. The second is the crowd that has formed because the local synagogue has called its members to a service of mourning — neighbours who know each other by name, who already share belief, profession, and life-trajectory before they walked into the square.
The two crowds will look indistinguishable to a passing photographer. They will be different mechanisms. The first is what Le Bon calls a heterogeneous crowd. The second is what he calls a homogeneous crowd. Each operates by different laws, fails in different ways, and is intervened on through different methods.
The taxonomy is the page.
Le Bon's classification at line 1432:
A. Heterogeneous crowds.
B. Homogeneous crowds.
The structural axis is whether the members share a stable identity prior to the crowd-formation. The street crowd is heterogeneous: members were strangers before the fire and will be strangers again after it. The synagogue congregation is homogeneous: the membership preceded the gathering and will continue after it.
The accountability axis is whether members are anonymous within the crowd. The anonymous street crowd has the highest level of de-individuation — no one is being watched as a particular person. The non-anonymous parliamentary assembly has the lowest — every vote is recorded, every speech is attributed.
Two axes produce four operational types when crossed. Each behaves differently. Each requires different operator moves.
The heterogeneous crowd is the kind almost everything in this corpus has been about until now — the September massacre crowd, the rally, the riot, the mob.
Anonymous heterogeneous crowds run the full Le Bon mechanism at maximum intensity. Anonymity removes responsibility. Suggestibility is high because no prior identity is anchoring the individual. Contagion runs unimpeded because there are no pre-existing social loyalties to slow it. The crowd's behaviour can change direction within seconds.
Non-anonymous heterogeneous crowds — the jury, the parliamentary assembly — run the mechanism at reduced intensity. "The sentiment of responsibility absent from crowds of the first description and developed in those of the second often gives a very different tendency to their respective acts."2 The juror knows their vote is recorded. The deputy knows the speech is on the record. Anonymity is partially restored by the immersion in the body, but only partially — and the partial restoration is enough to introduce friction into the crowd-mechanism.
Le Bon defers full treatment of homogeneous crowds to a later volume, but he sketches the architecture in this chapter.
The sect is the first step in the organisation of a homogeneous crowd. "A sect includes individuals differing greatly as to their education, their professions, and the class of society to which they belong, and with their common beliefs as the connecting link."3 The connector is belief. The sect can recruit broadly across professional and class lines because the shared belief is the only required entry condition.
The caste is the most highly organised form. "The caste is composed of individuals of the same profession, and in consequence similarly educated and of much the same social status."4 The connectors are profession, education, and status — a thicker bundle than belief alone. The military caste, the priestly caste, the legal caste each operate as caste-crowds inside the larger society they nominally serve.
The class is the loosest form. "The class is formed of individuals of diverse origin, linked together not by a community of beliefs, as are the members of a sect, or by common professional occupations, as are the members of a caste, but by certain interests and certain habits of life and education almost identical."5 The connector is shared lifestyle and interest. The middle class and the peasant class are the examples Le Bon names.
Each form of homogeneous crowd is more resistant to the heterogeneous-crowd mechanism than the prior level. The class can still tip into mob-state under the right exciting cause; the sect rarely does because the shared belief mediates the suggestibility; the caste almost never does because the shared profession and status enforce internal accountability that mimics non-anonymous conditions even in physical anonymity. Le Bon's chapter on parliamentary assemblies makes the corollary explicit: "Crowds are open to conviction; castes never are."6
[19TH-C RACIAL ESSENTIALISM] Le Bon at line 1463: "It should be considered as an essential law that the inferior characteristics of crowds are the less accentuated in proportion as the spirit of the race is strong. The crowd state and the domination of crowds is equivalent to the barbarian state, or to a return to it. It is by the acquisition of a solidly constituted collective spirit that the race frees itself to a greater and greater extent from the unreflecting power of crowds, and emerges from the barbarian state."7
Read out the racial framing. Le Bon's race is doing the work of what later thinkers would call culture, tradition, institutional substrate, or thick civic identity. The mechanism he names is real; the substrate framing is wrong.
The non-racial reading: populations with strong, internally coherent tradition-discipline are more resistant to crowd-state than populations with fragmented or hollowed-out tradition. A society with a thick web of pre-existing civic, religious, and family identifications has less suggestibility-surface available for the heterogeneous-crowd mechanism to operate on. A society in which those identifications have been hollowed out — by mass migration, deliberate state policy, or institutional collapse — has high suggestibility-surface and is structurally more vulnerable to crowd-state.
This is the operationally useful version of Le Bon's claim. The crowd-state is the dial; the dial is governed by the strength of pre-existing identifications; the strength of pre-existing identifications is partly cultural inheritance and partly contemporary policy. Hollowing out the identifications increases crowd-state risk; thickening them reduces it.
The English contrast Le Bon draws is the cleanest case. "A Latin crowd, however revolutionary or however conservative it be supposed, will invariably appeal to the intervention of the State to realise its demands... An English or an American crowd, on the contrary, sets no store on the State, and only appeals to private initiative."8 Read non-racially: the English-American crowd reflects centuries of accumulated thick civic identity (parish, common law, jury, voluntary association) that mediates between the individual and the State; the Latin crowd reflects centuries of state-centralised political culture that does not. Same heterogeneous-crowd mechanism. Different mediating substrate. Different operational surface.
Vault page on psychological-crowd-and-mental-unity-law describes the conditions that produce crowd-state in heterogeneous gatherings. The typology page tells you which gatherings are eligible for that mechanism and which are partly insulated from it.
The two-axis classification informs every later page in Le Bon's book: criminal crowds, jury crowds, electoral crowds, parliamentary assemblies are all subtypes within the heterogeneous category, each with specific operational features.
anonymity-contagion-suggestibility-triad describes the mechanism most active in anonymous heterogeneous crowds. The non-anonymous and homogeneous variants partially shut down each leg of the triad.
A riot is heterogeneous-anonymous. The crowd that breaks out of a stadium after a championship loss is composed of strangers, no member responsible to any other, identifiable by no one. Le Bon's full mechanism runs unimpeded. The crowd is capable of arson, looting, and assault that no individual member would have committed alone.
A strike is homogeneous-class. The crowd that gathers outside the factory is composed of workers from the same plant, who know each other's names, who will return to work together (or not return together) on Monday. The pre-existing class identity mediates the crowd-state. The crowd may sing, may chant, may be loud — but it does not arson the factory. The pre-existing identifications enforce friction.
Both look like "a crowd." The two are structurally different mechanisms. The riot can be intervened on by anonymity disruption (cameras, name-tags, dispersal). The strike cannot be intervened on by anonymity disruption — anonymity is not what is producing the crowd-state — and must be intervened on at the level of the underlying class interest. Confusing the two leads to operator failure.
Tuesday morning. You are watching footage of a crowd-event. The temptation is to apply universal crowd-state predictions. Strike that out. Run the typology first.
You ask the first diagnostic. Were these people strangers before this gathering, or did they have a stable identity together? If strangers — heterogeneous. If members — homogeneous.
You ask the second diagnostic. Are members anonymous within the crowd, or are they known to each other and to observers? Anonymous + heterogeneous = full Le Bon mechanism. Non-anonymous + heterogeneous = partial mechanism. Anonymous + homogeneous = limited mechanism. Non-anonymous + homogeneous = the mechanism is largely shut down.
Wednesday morning. You make your operational move accordingly. For the heterogeneous-anonymous crowd, anonymity disruption is the highest-leverage intervention. For the homogeneous crowd, intervention must address the underlying shared identity — and intervening at the level of suggestibility (slogan, prestige, contagion) will fail because the shared identity is what is producing the cohesion, not the crowd-state. Your move depends on which crowd you are looking at.
You are misreading the crowd. The diagnostic:
Your interventions are not producing the expected response. If you are running anonymity-disruption against a homogeneous crowd, the crowd will not disperse — the cohesion is not driven by anonymity. If you are running shared-identity-appeals against a heterogeneous crowd, the appeals will not land — the crowd has no shared identity to appeal to.
The crowd is changing type during the event. A homogeneous crowd that has been infiltrated by sufficient strangers becomes heterogeneous; a heterogeneous crowd that has been organised over time into a stable membership becomes homogeneous. Your reading at hour one may not hold at hour ten.
You are reasoning about "the crowd" without specifying which type. If your operational analysis does not name which of the four cells the crowd occupies, you are running a predictions on a category too large to be useful.
Le Bon's typology has been broadly preserved in twentieth-century crowd research, with the heterogeneous/homogeneous distinction surviving in the sociology of collective action and the anonymous/non-anonymous distinction confirmed by deindividuation studies (Zimbardo, Diener).
The tension Le Bon does not resolve: where does the new digital crowd fit? An online mob is heterogeneous (members are strangers) and anonymous (members are pseudonymous), but the gathering has a persistence and a feedback structure that no nineteenth-century street crowd could have. Subsequent vault pages must extend the typology to cover this case.
Open questions:
Picture Tarde at his desk in 1898, three years after Le Bon's book has appeared. He has read it. He agrees with the typology but resists the strong race claim. Tarde's framework — imitation as the mechanism by which all social phenomena spread — does not require the racial substrate to operate. The Latin/Anglo-Saxon difference Le Bon ascribes to race, Tarde ascribes to differential imitation cascades over time.
Where they converge: both see crowds as differing in kind, not just degree, depending on the prior identifications of members. Both see the heterogeneous-anonymous crowd as the most volatile category. Where they split: Le Bon's mechanism is racial / instinctive; Tarde's mechanism is imitative / historical. Tarde wins on this point in retrospect — the mechanism is historical and cultural, not racial — but Le Bon's two-axis classification is more operationally useful than Tarde's broader imitation-frame for predicting crowd behaviour in the moment. The synthesis is to keep Le Bon's taxonomy and Tarde's substrate explanation.
A new employee walks into the open-plan office. Sixty desks, sixty strangers, sixty conversations she does not yet have entry to. By the end of the first month she knows ten names. By the end of the third month she has been absorbed into the social system — knows who eats with whom, which Slack channels carry which kind of joke, which conversations to stay out of. The anonymous heterogeneous environment has been quietly converted, for her, into a homogeneous-class crowd. The conversion is invisible while it happens. By the time she notices it, the conversion is mostly complete.
That is one of the layers in Group Psychology and Herd Instinct. Trotter's mechanism — written 13 years after Le Bon — describes the human as a herd-animal whose default state is identification with the herd, and whose nervous system reorganises around herd-state with remarkable speed. Le Bon's typology shows where the herd-state is coming from in different crowd types. The two axes complement each other: Trotter explains why the conversion happens at all (the human nervous system is built to seek herd-cohesion); Le Bon explains why some conversions take three months in the open-plan office and other conversions take three minutes in the burning theatre.
Heterogeneous crowds become homogeneous over time when the gathering persists, and the crowd-state mechanism shifts as the conversion progresses. Le Bon's typology is not a static four-cell grid; it is a phase diagram with movement between cells driven by time-and-shared-experience. The street crowd that re-gathers daily for three months has become a sect. The sect that has been operating for three centuries has become a caste. The mechanism each phase runs is different. The operator who is running interventions appropriate to the previous phase will produce no effect at the current phase — and will not see why.
A second handshake to Status Hierarchy and Pecking Order. The page describes how groups produce internal hierarchies that govern member behaviour. Read alongside Le Bon: the homogeneous-crowd cohesion is partly produced by the internal status hierarchy that has stabilised across the membership. The pecking order does work the heterogeneous crowd cannot do — it provides the internal accountability that prevents the runaway suggestibility cascades.
The pecking order is what makes the homogeneous crowd a crowd, not a mob. Le Bon's caste does not riot because the internal hierarchy makes every member visible to every other in their proper rank, with the cost of any rule-breaking concentrated on the breaker rather than dispersed across the body. The riot is what happens when the hierarchy collapses; the riot is the homogeneous crowd reverting to heterogeneous-anonymous state through hierarchy failure. Watching for hierarchy stress is therefore the high-leverage diagnostic for predicting when a homogeneous crowd is about to become a mob.
A third reach to Propaganda and Mass Persuasion Hub — every propaganda strategy must select for crowd type before deploying technique. Slogans engineered for anonymous heterogeneous crowds (mass advertising, viral campaigns) fail when applied to caste-crowds; slogans engineered for caste-crowds (in-group dog-whistles, professional jargon, doctrinal markers) fail when applied to street crowds. Le Bon's typology is the targeting layer that sits underneath the entire propaganda apparatus. The 20th-century apparatus that ran without explicit reference to Le Bon's typology was implicitly running it anyway; the apparatus that ran without it failed in proportion to the mismatch between technique and crowd type.
The Sharpest Implication
You are inside multiple homogeneous crowds right now and one or two heterogeneous ones. Your profession, your political tribe, your aesthetic affiliation — each is a homogeneous crowd of some thickness. Your participation in the open algorithmic feed is a heterogeneous-anonymous-persistent fifth-cell crowd. The four-cell typology lets you see which kind of crowd is operating on you in each moment, and lets you see which interventions on you will work and which will bounce off. The proportion of your behaviour that is yours rather than crowd-driven is smaller than you think; knowing which crowd is which is a partial defence.
Generative Questions