Group Size - Dyad, Triad and the Web of Affiliations

This is Simmel's most rigorous formal argument, and the one that best demonstrates what his method can do. It concerns something so obvious that nobody had thought it worth examining: how many people are in the group.

His claim is that the step from two to three is not a difference of degree but of kind — and that this single structural fact explains an extraordinary range of otherwise unconnected phenomena.

(Lesson 2.2.6 introduced dyads and triads as vocabulary. This lesson gives the full argument and its consequences.)

Three conversations, one evening.

Two friends walking home. They talk about a third friend's marriage — frankly, at length, in a way neither would in company. One admits something he has told nobody. The talk is unguarded, meandering, occasionally silent, and the silences are comfortable.

A third friend joins them at the corner. Within thirty seconds the conversation has changed shape, and everyone can feel it.

It is not that the topic changed, though it did. It is that the structure changed. Now there is an audience. Now every remark is made not only to one person but in front of another. Now two of the three can agree and, in agreeing, form something the third is outside. Now there is a possible majority.

The confidence from ten minutes ago will not be repeated. Not because the newcomer is untrusted — he is a close friend — but because it is no longer that kind of conversation, and cannot be.

And later, an argument. Two of them disagree about something small. The third says: "Well, look, you're both sort of right." And the argument ends.

Ten minutes earlier that sentence could not have been spoken, because there was nobody to speak it. In a pair, a disagreement is between the two of you and must be resolved by the two of you or not at all. There is no outside.

One person, and the entire logic of the situation has changed.

Simmel's question: is group size a merely quantitative variable, or does it produce qualitative differences in the forms of interaction?

The intuitive answer is that it is quantitative. Bigger groups are the same thing, more of it — more relationships, more noise, more coordination cost.

Simmel's answer is that this is wrong at the very first step. Numerical determination produces qualitative transformation, and the sharpest case is the smallest: the difference between two and three is more consequential than the difference between thirty and three hundred.

His reasoning is structural, not psychological. In a dyad, there is one relationship, and it exists only as long as both parties sustain it. In a triad, there are three relationships plus the relationship of each individual to the pair formed by the other two — and the group can survive the departure of any member.

Everything else follows from those two facts.

The dyad

No group above the individuals. Simmel is precise about this: "the dyad is different from all other groups in that it does not attain the super-personal life which the individual feels to be independent of himself." A pair cannot outlive its members, cannot have an existence apart from them, and cannot be experienced as something that would go on without you.

Absolute responsibility, and no delegation. Each member is wholly responsible for the group's continuation. In a group of five you can slacken and it persists; in a pair, your withdrawal ends it. This produces the characteristic dyadic intensity — and the characteristic dyadic burden.

No majority. Disagreement cannot be resolved by outvoting. It must be settled by persuasion, compromise, submission or dissolution. There is no procedure.

No third party. No mediator, no arbitrator, no witness, no audience. Nothing between the two.

Maximum intimacy — and Simmel's finding about it is subtle. The dyad permits a self-disclosure impossible elsewhere, because nothing said passes beyond the two. But he also notes that the dyad is most likely to develop the sense that it is unique and irreplaceable, and that this sense is what makes its ending feel like a death rather than a change.

And the sharpest observation: the dyad is inherently fragile. Every dyad carries the knowledge of its own mortality — that one departure ends it absolutely. Simmel argues this awareness is constitutive of dyadic relationships rather than incidental to them, and that it explains both their intensity and the specific anxiety that attends them.

The triad

Simmel's central case, and where the formal method pays off most spectacularly.

The three triadic roles

The mediator (the impartial third). Stands between the other two and helps them settle. Simmel distinguishes two versions: the third as mediator, who genuinely seeks agreement, and the third as arbitrator, empowered to decide. His observation about mediation is acute: the mediator's usefulness lies partly in translation — each party states its case to the third, and in doing so states it more moderately and more intelligibly than it would to the opponent. The presence of an audience civilises the quarrel.

The tertius gaudens — the rejoicing third. The one who profits from the discord of the other two. Simmel's Latin phrase has stuck. The third does not create the conflict; they exploit an existing one, extracting concessions from both sides who need their support. Two suitors competing; two employers bidding; two factions courting a swing vote; a child whose parents disagree.

Divide et impera — divide and rule. The most active version: the third deliberately creates or maintains the antagonism in order to dominate both. Simmel notes this is the harder and more dangerous strategy, because it requires the third to be strong enough that the other two do not simply combine against them.

Simmel's claim is that these are structural positions, available to anyone who occupies them, regardless of intention or character. The test is whether the same dynamics appear across radically unlike contents — and they do.

A child in a marriage. The child unites the parents into something above themselves — a family that exists independently of their relationship, which the couple alone was not. The child also becomes a mediator, carrying messages and softening conflicts. And the child can become tertius gaudens, playing one parent against the other for permission. Every parent has watched this happen and most have named it without knowing they were quoting Simmel.

The broker. In a market, the party who connects two others who cannot reach each other profits from standing between them — and has a structural interest in not letting them connect directly. Ronald Burt's structural holes theory (1992) formalised exactly this: the advantage accrues to whoever bridges a gap between two unconnected clusters, and it is an advantage of position, not of talent. Burt's work is a direct, acknowledged descendant of this essay.

Great power politics. A third state between two rivals extracts concessions from both. Non-alignment as a strategy is tertius gaudens at the scale of the international system.

Management. A manager between two feuding staff members occupies the mediator position; a manager who quietly benefits from their rivalry occupies the second; a manager who fosters it occupies the third. The same person can move between all three in a week.

And a shift you can watch happen. A two-person startup where the founders resolve everything between them. They hire a third. Within months there are meetings, a decision procedure, and the phrase "let's get input from the team." The company has acquired an existence above its founders. Something has been gained — durability — and something lost, and both are structural consequences of the number three.

Larger groups

Simmel does not stop at three, and the argument scales.

Beyond a certain size, direct interaction becomes impossible and the group must develop representative structures: delegates, offices, written rules, formal procedures. What was managed by acquaintance must now be managed by organisation — which is the seed of bureaucracy (4.3.4).

Individuality expands as the group grows. This is Simmel's counterintuitive and most important claim about size. A small group monitors its members closely, demands conformity, and permits little deviation — everyone knows everyone, and the cost of nonconformity is total. The large group cannot monitor, and therefore permits and even requires individual variation.

His conclusion is one of the most quoted things he wrote: the individual is freer in the metropolis than in the village, not despite the crowd but because of it (4.4.5).

And the group's boundaries become sharper as it grows. Small groups blur into their surroundings; large ones need explicit membership criteria, symbols and rules of entry — the exclusion form from 4.4.1.

The web of group affiliations

Simmel's most sophisticated argument about size, and the one that founded network sociology. The German title — Die Kreuzung sozialer Kreise, "the intersection of social circles" — states it better than the English.

In small traditional societies, group memberships are concentric. You belong to your family, which belongs to your village, which belongs to your region, which belongs to your people. Each circle sits inside the next. Everyone in your family shares almost all your other memberships.

In modern societies, memberships intersect rather than nest. Your workplace, your neighbourhood, your religion, your profession, your hobby, your political party, your online communities — each contains a different set of people, and almost nobody is in all of them.

And here is Simmel's conclusion: individuality is produced by the intersection.

Not by having a unique soul. By occupying a unique position in the web — a combination of memberships that nobody else shares. "The number of different circles in which an individual moves is one of the indices of cultural development."

Why this is a major argument and not a curiosity.

It reverses the standard story about modernity. The conventional account — the conservative one after 1789, and much popular commentary since — says that modern life destroys individuality: mass society, standardisation, the crowd, the anonymous multitude.

Simmel says the opposite, and gives a structural reason. The village knows you completely and permits you no variation. The city knows you in fragments, and the fragments do not communicate. Your colleagues do not know your family; your congregation does not know your politics; your neighbours do not know your work. You are the only person who occupies all your circles at once, and that unique intersection is your individuality.

And there is a cost he names precisely: cross-cutting affiliations produce role conflict (2.2.3). Concentric circles make consistent demands; intersecting ones do not. The obligations of work conflict with those of family, of faith with those of profession. The modern person is freer and more internally divided, and these are the same fact.

Go deeper

How the arguments have fared empirically

Unusually for classical theory, several of these are testable, and they have been tested.

The dyad/triad distinction is robust and has been confirmed repeatedly in small-group research, in network analysis and in experimental economics. Coalition formation in three-person groups is one of the most studied phenomena in social psychology, and Caplow's Two Against One (1968) built an entire theory of coalitions in triads on Simmel's foundation, deriving predictions about which pair will ally given different distributions of power — including the counterintuitive and well-supported finding that in some power configurations, the two weaker members ally against the strongest, so that strength becomes a liability.

Structural holes. Burt's work is the most successful formalisation, and it is well evidenced: individuals whose networks span disconnected clusters receive better performance evaluations, faster promotion and higher compensation, across a range of organisational settings. The mechanism is Simmel's — the advantage of the third who bridges.

The Simmelian tie. Krackhardt's formalisation: a tie embedded in a triad where all three are mutually connected behaves differently from an unembedded one — it is more stable, generates stronger norms, constrains behaviour more, and reduces the parties' bargaining autonomy. This is the network-analytic version of Simmel's observation that the third party changes the dyad's character.

Closure versus brokerage. Contemporary network sociology is substantially organised around a tension Simmel's essays contain: closed networks (everyone connected to everyone) generate trust, norms, sanctions and social capital — Coleman's argument — while open networks with structural holes generate information advantage and innovation — Burt's argument. Both are right, for different outcomes, and choosing between them is a real strategic question for individuals and organisations.

Cross-cutting cleavages. Simmel's affiliation argument has a substantial political science literature behind it. The classic claim — from Lipset and from Truman's work on interest groups — is that societies where divisions cut across one another (some of your co-religionists are your class opponents; some of your class allies are of another ethnicity) are more stable than those where cleavages reinforce one another, because reinforcing cleavages stack everyone into two mutually exclusive camps.

And this is the sharpest contemporary application. Political scientists analysing polarisation describe social sorting: the process by which party, religion, region, education, media consumption, class and even consumer preference become aligned rather than crosscutting — so that knowing one predicts the others. Lilliana Mason's Uncivil Agreement (2018) is the standard account for the US case.

In Simmel's terms, this is a return from intersecting circles to concentric ones — and his framework predicts the consequences precisely: less individuality, more group conformity, sharper boundaries, more intense in-group solidarity, and greater hostility to outsiders. That a 1908 essay generates a live diagnosis of a 2020s phenomenon is a strong argument for the formal method.

The limits

Size is not everything. Simmel sometimes writes as though numbers determined form. They constrain it; they do not determine it. Two three-person groups can be organised entirely differently.

The arguments are about possibilities, not certainties. The triad permits coalitions; it does not require them. Whether the third mediates, exploits or divides depends on interests, culture and personality, which are exactly the "contents" Simmel bracketed (4.4.1).

And the affiliation argument may be too optimistic about modernity. Belonging to many circles produces individuality, and it can also produce thin, instrumental attachments to all of them — which is Durkheim's anomie (4.2.3) reached from the other direction. Simmel was more alert to this than he is given credit for, and The Metropolis (4.4.5) is largely about it.

"Bigger groups are just more of the same." Simmel's core claim is that numerical change produces qualitative transformation, and the two-to-three step is the sharpest instance.

"The triad is more stable than the dyad." More durable — it survives a departure. Not more harmonious: it introduces coalitions, majorities and exclusion, none of which the dyad has.

"The mediator is a good person." These are structural positions, not moral characters. The same individual can mediate on Monday and exploit the same conflict on Friday.

"Divide and rule is the same as tertius gaudens." The tertius gaudens profits from a conflict that exists; divide et impera creates or maintains it. Simmel distinguishes them, and the second is riskier.

"Modern society destroys individuality." Simmel argues it produces individuality, through the unique intersection of multiple non-overlapping memberships — and that the cost is role conflict, not conformity.

"This is just small-group psychology." The same forms operate between states, firms and factions. Scale-independence is the claim.

Why it matters

It is the best demonstration that Simmel's method works. A purely formal variable — how many — generates specific, non-obvious, testable predictions that have survived a century of empirical work. That is a real vindication of formal sociology.

It gives you three roles you will now see everywhere. Mediator, tertius gaudens, divider. Once named, they are visible in families, offices, markets, boards and international politics, and recognising which one someone occupies usually explains their behaviour better than their stated intentions do.

It founded network analysis. Structural holes, Simmelian ties, closure versus brokerage, and the whole enterprise of explaining outcomes by position in a web rather than by attributes of persons — all of it descends from these essays.

And it supplies a structural diagnosis of polarisation. When memberships stop intersecting and start aligning, a society moves from Simmel's modern form back towards his traditional one, with predictable results: less individuality, sharper boundaries, and more hostility across them. That is a serious contribution to understanding the present, made in 1908 by a man who could not get a job.

Numerical change produces qualitative transformation, and the sharpest instance is two to three.

The dyad: one relationship, which dies if either leaves. No group life above the individuals; absolute individual responsibility; no majority; no third party; maximum intimacy; and an inherent, constitutive awareness of its own fragility. Secrets survive in dyads and dissolve in triads, for structural reasons.

The triad adds four things at once: a group life independent of any member; the possibility of majority; the possibility of coalition; and three new roles — the mediator/arbitrator, the tertius gaudens (who profits from an existing conflict), and divide et impera (who creates or maintains it).

Larger groups require representative structures and formal rules — the seed of bureaucracy — and, counterintuitively, expand individuality, because large groups cannot monitor their members as small ones can. The individual is freer in the metropolis than in the village.

The web of group affiliations (Die Kreuzung sozialer Kreise): traditional memberships are concentric, modern ones intersect — and individuality is produced by occupying a unique intersection of circles, with role conflict as the price.

Empirically supported: Caplow on coalitions in triads (including weak-pair alliances against the strong); Burt's structural holes; Krackhardt's Simmelian tie; the closure-versus-brokerage tension organising modern network sociology; and the cross-cutting cleavages literature.

And the contemporary application: political social sorting — the alignment of party, religion, region, education and consumption into a single cleavage — is, in Simmel's terms, a return from intersecting circles to concentric ones, with exactly the consequences he predicted.

Dyad — a two-person group; no life above its members, no majority, no third party, and dissolved by either departure.

Triad — a three-person group; survives any departure and introduces majority, coalition and the three third-party roles.

Mediator / arbitrator — the impartial third who helps settle or is empowered to decide.

Tertius gaudens — "the rejoicing third," who profits from the discord of the other two.

Divide et impera — the third who actively creates or maintains antagonism in order to dominate.

Web of group affiliations (Kreuzung sozialer Kreise) — the intersection of an individual's multiple memberships.

Concentric vs intersecting circles — nested traditional memberships versus overlapping modern ones.

Structural hole — Burt's term for a gap between unconnected clusters, bridging which confers advantage.

Simmelian tie — Krackhardt's term for a relationship embedded in a mutually connected triad.

Closure vs brokerage — dense connected networks producing trust and norms, versus sparse bridging networks producing information advantage.

Cross-cutting vs reinforcing cleavages — social divisions that cut across one another versus divisions that stack.

Social sorting — the contemporary alignment of multiple identities along a single political cleavage.

One — watch a dyad become a triad. Next time a third person joins a two-person conversation, note how quickly the register changes, and what stops being sayable. It takes seconds and it is entirely reliable.

Two — identify the three roles. In a group you know with a persistent conflict, ask who mediates, who benefits from its continuation, and whether anyone maintains it. All three positions are usually occupied, sometimes by the same person at different times.

Three — map your circles. List your group memberships and, for each pair, ask how many people are in both. If the overlaps are large, your circles are concentric; if small, intersecting. Simmel says the second produces individuality and role conflict together.

Four — find a structural hole. In any organisation, identify someone who connects two groups that otherwise do not talk. Ask what advantage the position gives them and whether they have an interest in the two groups not connecting directly.

Five — test the sorting thesis. Take three social identities that matter in your country — party, religion, region, class, education. Ask whether knowing one predicts the others, and whether this has changed in twenty years. That is Simmel's diagnostic, applied to the present.