How did you get your last job? Set the application history aside and think of the person who first mentioned it to you: how often do the two of you actually meet? Many people find the answer surprising — not a close friend they talk to daily, but a former colleague, a classmate's classmate, someone they worked with on one project years ago and barely spoke to since.
This page takes apart three things: why the most useful news tends to arrive from people you are not close to; why the position wedged between two groups is both valuable and uncomfortable; and why "the world is small" and "things spread" are not the same claim at all.
In 1969 Mark Granovetter sent out a survey in Newton, a suburb of Boston, to 282 professional, technical and managerial workers who had changed jobs within the previous five years. The question was plain: how did you first hear about this job?
Among those who got the job through a personal contact, only about 17% said they saw that person often (twice a week or more); roughly 56% saw them occasionally; and nearly 28% rarely — once a year or less.
A word first: sociology calls a single relationship between two people a tie, and its strength is a matter of time spent together, closeness, and how much help passes back and forth.
What needs explaining is not that contacts help, but that less close contacts help more. The mechanism in Granovetter's 1973 paper, "The Strength of Weak Ties", is clean: the people you see daily mostly know each other too — a state where your two closest friends are complete strangers to one another rarely lasts. So strong ties close up into clusters, and inside a cluster the news overlaps heavily: what you have heard, they have probably heard as well.
The value of a weak tie is therefore not about affection but about position: it is more likely to be a bridge joining two clusters that were otherwise unconnected. The point is not the weakness; it is the non-overlap.
State the evidence separately. The 1973 case is a correlation plus an argument: one suburb, one occupational stratum, all male respondents, 1969 data — not a universal law. The causal test had to wait until 2022, when researchers reported in Science a series of randomised experiments run through LinkedIn's "People You May Know" recommendations, covering over 20 million people across five years, during which 2 billion new ties and 600,000 job changes were created. Weak ties did raise the probability of moving jobs, but with a twist — the relationship is an inverted U, and the most useful band is the moderately weak one. Someone who can no longer place who you are transmits nothing, however long the bridge.
One precondition is routinely skipped: having the tie is not the same as the tie being used. The sociologist Sandra Smith's 2005 study, built on in-depth interviews with 105 low-income African Americans, found the bottleneck was often not a lack of contacts holding opportunities but those contacts choosing not to speak up — if the person referred performs badly, the account is charged to the referrer, and some interviewees had indeed lost jobs that way. A referral is not forwarding a message; it is posting your own credit as collateral. As the gift and reciprocity topic argued, what is given freely binds precisely because it enters both parties into an unwritten account; a referral is that kind of account. Evidence: qualitative interviews, mechanism clear, proportions not to be extrapolated.
You get added to a cross-department channel and quickly notice you are the only person who knows people on both sides. So every time the two sides need to align, the message routes through you. That position has a name.
Ronald Burt calls the missing relationship between two groups a structural hole — note that the hole is not a person but an absence of ties — and the person who stands across it and joins the two sides a broker.
His 2004 study worked like this: 673 managers running the supply chain of a large American electronics company in 2001 were asked to write down ideas for improving that supply chain, and the ideas were then rated by senior executives. Managers whose networks spanned more structural holes (technically, low network constraint — their contacts were not all bunched into one pile) had ideas rated more valuable and dismissed out of hand less often; their pay, evaluations and promotions ran higher too.
The mechanism is no mystery: it is vision arbitrage. Talk converges inside a group — everyone reads the same material, explains it to each other a few times, and soon they all think that way. Someone standing between two groups need not be smarter; they simply see both repertoires at once, so "take what works over here and apply it to the problem over there" is obvious to them and to nobody on either side.
Be clear about the causal chain: position cannot be randomly assigned. It is entirely possible that those managers were abler to begin with, and so acquired both the cross-group network and the good ideas. That gap has not been closed. Evidence: a cross-sectional correlation from one firm in one year.
The counterweight matters too. James Coleman stressed something else — closure: in a small circle where every triangle is filled in, everyone learns what everyone did, so norms can actually be enforced and trust has a footing. Bridging and closure do different jobs: new ideas travel over bridges, getting things delivered on time depends on closure.
Nor should the broker's costs be left out: all of the cross-group coordination lands on one person, and neither side quite counts them as their own. As the social stratification topic noted, advantages of position get read as personal traits — same here, and "he always hears things first" is often just "he sits on the hole".
In the late 1960s Stanley Milgram and Jeffrey Travers ran an experiment: people in Nebraska and Boston were asked to get a document to a stockbroker in Boston whom they did not know, with the rule that they could only send it to someone they knew by name, who would pass it on. 296 people were asked to start, 64 chains arrived, and the completed chains passed through 5.2 intermediaries on average. This is where "six degrees of separation" comes from.
The honest reading: fewer than three in ten chains completed, and the broken ones were probably systematically longer — 5.2 is the average over the chains that got through. Researchers later went through the archives of such experiments and found completion rates consistently low, which makes treating "six degrees" as a precise constant an overstatement.
So why are paths in fact short? In 1998 Duncan Watts and Steven Strogatz gave a minimal explanation: take a tidy ring lattice where everyone is tied only to a few near neighbours, so two distant points are many steps apart, then randomly rewire a tiny number of edges to far-off nodes. Average path length collapses immediately, while the degree to which "your friends are also friends with each other" barely changes. A few long-range shortcuts are enough — that is a small world. In 2016 Facebook estimated an average separation of 3.57 across its then 1.59 billion active accounts, though that counts only active accounts on that platform, and reading it as "all of humanity" goes too far.
Here is the turn that matters: short paths mean news can arrive fast; they do not mean behaviour follows. In 2010 Damon Centola reported a study that made network structure itself the experimental variable: 1,528 participants were randomly placed into one of two structures — a clustered one (your neighbours are also each other's neighbours, so ties are redundant) or a random one (the same number of ties, spread wider with more shortcuts) — and a health behaviour was watched as it diffused. The result runs against intuition: the behaviour spread farther and faster in the clustered network.
The reason is that adoption needs repeated confirmation. To know a piece of news, once is enough; to actually change what you do usually waits until several people around you are already doing it. In the random network each person hears it once from one source, and the chain stops there; in the clustered network those apparently "redundant" edges supply the second and third confirmation. Researchers call this kind of diffusion, which only moves under repeated reinforcement, complex contagion — as opposed to simple contagion, which travels on a single contact.
So two logics run on the same network: news takes shortcuts, behaviour needs redundancy. Being in a position where you hear about everything does not put you in a position to make anyone actually change. Evidence: a randomised experiment on network structure, a strong design by social-science standards; but the setting is online and the behaviour is one particular kind, so extrapolate carefully.