communicoupling · concept 13 of 26
A virus spreads on a single contact; a risky new behaviour usually needs several. That difference rewrites how each travels through a society — the very network shortcuts that speed a disease across the world can stop a behaviour cold. What spreads on one exposure and what needs social reinforcement obey different laws.
A rumour, a virus, a meme can jump on one touch — this is simple contagion. One exposed friend is enough; the thing is through you and on to the next. Because a single link suffices, the fastest way to spread it is to add shortcuts — long ties that leap across the social world — and this is exactly the "small-world" effect that lets a disease cross continents in a few hops.
But adopting a costly or risky behaviour — joining a protest, quitting a job, trying an unproven practice — usually takes seeing several people you know do it first. That is complex contagion, and it behaves counterintuitively: the long ties that carry a disease are the wrong medium for a behaviour that needs reinforcement, because a lone distant adopter is never enough to convince you. Clustering, not shortcuts, carries the complex kind.
one exposure or several
Set k = 1 and you have simple contagion: one exposed neighbour is enough, so the thing spreads like a fluid, seeping through any link it touches. Set k = 2 — needing two independent exposures before you move — and you have complex contagion. The jump from one to two looks trivial. It is not. It changes which kind of network is good at carrying the thing, and it does so in the opposite direction from what everyone's intuition about "going viral" predicts.
On the clustered ring, both simple and complex contagion sweep the whole network — complex just crawls more slowly, because it must wait for reinforcement to build. Now add shortcuts. For simple contagion the shortcuts are pure gift: the spread leaps across the graph and finishes in a handful of rounds instead of dozens (watch the rounds tile fall). For complex contagion the same shortcuts are poison. A rewired tie drops a lone adopter into a distant neighbourhood where nobody else has adopted — one exposure, never two — so it converts no one, and the reinforcement the behaviour needed has been shredded to make room for reach. The final reach collapses from everyone to almost no one. Same seed, same people, same threshold: only the wiring changed.
What to try
Leave rewiring near the middle and hit Flip k=1 ↔ k=2. The network does not change — only how many exposures a person needs. Simple sweeps it; complex stalls near the seed. The wiring that carries one kills the other.
Hold k = 1 and drag rewiring right: the rounds to spread drop as shortcuts appear — the small-world speed-up. Now hold k = 2 and drag: final reach falls off a cliff the moment the first shortcuts break the clustering.
Load the failed leap: complex contagion on a nearly random graph. It reaches a few percent and dies. The behaviour is no less appealing than on the ring — it simply has no clustered ground to build the reinforcement it needs.
why weak ties fail for behaviour
The famous strength of weak ties is real — for information. A distant acquaintance is your best source of news, precisely because they are plugged into a different world and know things your close circle does not. But information is a simple contagion: hearing it once is enough. Costly behaviour is not. A lone weak tie can tell you that people somewhere are quitting, striking, converting, adopting — and leave you completely unmoved, because one distant example is not social proof, it is an anecdote. What moves you is watching two or three people you actually know, whose judgment you can weigh and whose risk you share, do the thing. That reinforcement lives only where ties overlap — in clusters, not in shortcuts.
So complex contagion crosses from one part of the world to another only over a wide bridge — several ties running in parallel between the same two groups — never over a single long tie. A lone bridge carries the rumour and stops the movement. Below, the identical two clusters are joined first by one tie and then by two; the same k = 2 rule runs on both. Watch which one lets the behaviour cross.
spreading what needs backing
This is why serious organizing looks nothing like a viral marketing campaign. A movement, a union drive, the adoption of a hard practice, a norm that asks people to bear real cost — all are complex contagions, and all of them build density before reach. They saturate a neighbourhood, a workplace, a congregation, a scene, until the local reinforcement is so thick that adoption there is nearly inevitable — and only then reach outward, and only through wide bridges, several committed people carrying the thing into the next cluster together. The instinct to "just go viral" — to broadcast as widely and thinly as possible — is precisely the small-world strategy, and it is exactly wrong for anything that asks something of people. Thin, wide reach spreads a lone example into a thousand neighbourhoods that each dismiss it.
The mirror-image error is just as common: treating a simple contagion as if it were complex. If the thing genuinely spreads on one exposure — an app, a piece of news, a low-cost habit — then hoarding it inside a dense cluster is a waste, and the winning move really is to buy the long ties and broadcast. The practical skill is diagnosis: before you design how to spread something, ask honestly whether it spreads on one exposure or needs several. Get that one number wrong and every downstream choice about network structure is backwards.
Unions and movements win a shop floor or a block before they win a city — building reinforcement dense enough that the next person always sees several before they are asked to risk anything.
New tools, methods, and treatments that cost effort to try diffuse through clustered professional cliques, not through the field's globe-trotting stars, whose single endorsement rarely convinces.
Products with real switching costs seed one campus or city to saturation before expanding — the opposite of a low-cost app, which is simple and should be broadcast as widely as possible.
The mapping
| Mechanism | Social life |
|---|---|
| simple contagion (k=1) | Things that jump on a single exposure — news, gossip, viruses, low-cost imitation. |
| complex contagion (k≥2) | Things that need social reinforcement — risky, costly, or identity-laden behaviour. |
| the threshold k | How many adopting neighbours it takes before a person will move. |
| a shortcut / long tie | A bridge to a distant part of the network — a weak tie into another world. |
| clustering | The redundant local ties, neighbours who share neighbours, that let exposures reinforce. |
| a wide bridge | Multiple parallel ties between two groups — what it takes to carry the complex kind across. |
where it tears
A hard k = 2 — two exposures and not one convert you — is an idealization. Real reinforcement is graded, not a step function, and whose adoption counts is at least as important as how many: two people you trust may outweigh ten you don't, and identity, status, and similarity all reweight the count. The bare integer here is the crudest possible sketch of what actually moves someone.
The same behaviour can be a simple contagion for one person and a complex one for another. A practice that a secure, well-resourced person will try on a single suggestion is a genuine risk for someone with less slack, who needs to see several trusted others survive it first. There is no fixed k for a behaviour — only a distribution of thresholds across people, and the model here flattens that into one number.
The lesson that costly change spreads through density is easily bent into a rationale for never reaching out — for staying inside the cluster and calling it strategy. But some behaviours really do spread best by reaching out, and some clusters are echo chambers that need a shortcut more than another layer of reinforcement. Diagnose the contagion type before you prescribe the network; the model tells you how things spread, not which things are worth spreading.