Six Degrees of Separation: The Geometry of the Relationships That Bring Us Together and Keep Us Apart

Introduction

In 1990, the playwright John Guare put a sentence on stage that changed how the world talked about connection. "Everybody on this planet is separated by only six other people," his character declares in Six Degrees of Separation — "six degrees of separation between us and everybody else on this planet"[wiki-film]. The 1993 film adaptation, starring Will Smith, Donald Sutherland, and Stockard Channing, turned the line into a cultural fixture; a few years later, a parlor game built around the actor Kevin Bacon ("Six Degrees of Kevin Bacon") made it something you could play at parties[wiki-six-degrees]. What almost nobody realized at the time was that the phrase was not a metaphor. It was a testable mathematical claim about the shape of human relationships — and it had already been tested, twenty years before Guare wrote a word of it.

This guide is about the geometry of that claim. "Degrees of separation" has two very different meanings, and the confusion between them is half the fun. The first is the social graph: the web of friendship and acquaintance in which a stranger three countries away is reachable through a handful of intermediaries. The second is genealogical degree: the literal distance between relatives in a family tree, measured in generations — the number of "removes" between you and your second cousin. Both meanings are rigorous mathematics, and both collapse into the same kind of counting: powers of a branching factor, or repeated halving. Together they explain why the world is simultaneously enormous and intimate.

The tools you need already exist on this site. Every social network can be stored as a table of numbers — an adjacency matrix — and the Matrix Calculator multiplies exactly those matrices, which is how you count the paths of length two, three, or six between people. And the Probability Calculator is what you reach for when the question turns from "are they connected?" to "what are the chances they are?" By the end of this guide, the phrase "six degrees of separation" will not be a curiosity about Hollywood — it will be a small, precise theorem about the universe of people.

A Network Is a Matrix

The mathematician's tool for a collection of people and the relationships between them is a graph (in this context also called a network): a set of nodes — the people — and a set of edges — the connections between them. A friendship graph is undirected: if you are friends with someone, they are friends with you, so the edge is a link without an arrow. A graph becomes directed the moment relationships have a direction, as in a following network on a social platform. All of the classic questions of social physics — "who is closest to whom," "how fast does news travel," "is anyone isolated" — are questions about graphs, and the answers are computed on graphs[wiki-six-degrees].

The single most useful representation of a graph is a table of numbers called the adjacency matrix A, with one row and one column per node:

For a network of n people, A is an n × n matrix of zeros and ones. This is not a picture; it is a data structure, which is why every real social platform stores its "graph" as one. And here is the beautiful part: matrix multiplication already knows how to count paths. The entry in the k-th power of the adjacency matrix counts the walks of length k between two nodes:

Two people are separated by exactly k degrees if the smallest k for which (A^k) has a nonzero entry between them is k. "Six degrees of separation" is therefore a statement about the sixth power of the adjacency matrix of the human network: nearly every pair of nodes is connected by a walk of length six or less. If you have ever wondered what it actually means to "multiply" social networks, this is it — and the Matrix Calculator is built precisely to do this arithmetic, on whatever graph you can type in.

Two other graph quantities will recur throughout this guide. The degree of a node is how many edges it has — how many friends a person has. The path length between two nodes is the number of edges in the shortest route connecting them. "Degrees of separation" is just the everyday word for path length, measured in people rather than kilometers. And just as the All the Ds guide shows that spaces get stranger as their number of dimensions grows, networks get stranger as their number of nodes grows — the geometry of a million-person graph is genuinely different from the geometry of a classroom, and that difference is the subject of the next few sections.

Milgram's Letters: The Experiment That Started It All

The social scientist Stanley Milgram asked the question in 1967, a quarter-century before the film made it famous: how many intermediaries does it really take to connect two Americans chosen at random? His design was elegant and famous. He picked a target person in Boston and a set of starting individuals in Nebraska and Kansas, then handed each starter a letter and a rule: you may pass the letter only to someone you know on a first-name basis, and only if that person is "closer" to the target. The letters were supposed to walk, hand to hand, toward Boston; the experimenters counted the steps[wiki-small-world].

The astonishing result was the chain length. Of the letters that completed the journey, the median number of intermediaries was five — meaning the sender and the target were, on average, six degrees apart. A person in Nebraska, a farmer's community, reached a stranger in Boston through about six handshakes. The experiment had enormous methodological flaws by modern standards — the sample was tiny, most letters never arrived (chains broke when someone refused or lost the envelope), and the people who did participate were self-selected. Milgram's own numbers were reported loosely and the "six" has been disputed by later reanalysis[wiki-small-world]. But the qualitative conclusion survived every critique: the chain was short. Not dozens, not hundreds — a handful. That a person can route a message to an arbitrary stranger through a few acquaintances was so counterintuitive that it became a scientific legend and, eventually, a movie line.

The deeper lesson of Milgram's experiment is what the letters did not need. Nobody in the chain had a map of the whole network. Each participant knew only their own friends and made a local judgment — "she is closer to Boston than I am." Yet the local choices, repeated six times, produced a global connection. This is the first hint of a general truth: short paths in a social network do not require anyone to see the big picture. They emerge from ordinary, self-interested, local behavior. That truth is what the physicists made precise twenty-five years later.

The Film, the Game, and the Phrase

Before the mathematics, it is worth taking the phrase itself seriously, because the cultural history is part of why the concept is so powerful. Guare's play is not about network science; it is about the vertigo of realizing how close strangers are — and how far apart the people right next to you can be. The title is the play's thesis: "six degrees of separation" is the distance that makes a con artist in New York a plausible guest in an Upper East Side apartment, the distance that makes every stranger a potential friend-of-a-friend[wiki-film]. The phrase resonated because it named a feeling people already had: the world had gotten smaller, and we were not sure whether that was thrilling or terrifying.

The Kevin Bacon game turned the sentiment into a hobby. The premise: every actor can be linked to Kevin Bacon through shared movie credits, and "degrees" count the links — Bacon appeared in a film with actor X, who appeared with actor Y, and so on. It was invented as trivia in the mid-1990s and became a full-blown cultural object, complete with a website, a charity, and a growing folklore that "everyone in Hollywood is at most a few films from Bacon"[wiki-six-degrees]. What the game-playing public did not know was that they were playing at graph theory: computing shortest paths in a network of roughly a hundred thousand nodes, exactly the walk-counting that the adjacency matrix automates. The Bacon game is the same mathematics as the airline routing problem, the same mathematics as Milgram's letters — just dressed as a party trick.

The Math of Six: Branching Factors

Here is the arithmetic that makes six degrees plausible, and it is as simple as repeated multiplication. Suppose the average person knows b other people well enough to pass a message or a greeting. From any starting person, the people at distance one number about b. The people at distance two are, roughly, b per person at distance one — so b². Distance three gives b³, and so on. After d hops, the reachable population is approximately:

If b = 100 and the world has 8 billion people, then solving for the number of hops gives:

Five hops. That is the entire mystery of six degrees, reduced to a logarithm: to reach everyone, you do not need a chain of eight billion — you need a branching factor that multiplies. Each hop multiplies the reachable crowd by b, and multiplication is the fastest-growing operation in arithmetic. If instead b were 10 — people know ten people closely — the same world would require about ten hops, and if b were 1000 it would require fewer than four[wiki-six-degrees].

The branching-factor model is an idealized version of the truth. It assumes people's circles are disjoint, when in reality they overlap heavily — the clustering problem, taken up in the next section — and it ignores that a random stranger is far more likely to be reached through a highly connected "hub" than through an average person. But as a first approximation it is exactly right about the mechanism: the network is small because reach grows exponentially while distance grows logarithmically. This is the same trick of scale that the It's Not a Coincidence guide identifies in probability — there, a small chance multiplied by enormous opportunities becomes a certainty; here, a small circle multiplied by a few hops becomes everyone. Multiplication, applied to counts, is the quiet engine behind both the world's coincidences and its connections.

The Watts–Strogatz Model: A Few Shortcuts

If the branching factor explains why six degrees is possible, it does not explain why it is typical. Real social networks have a property that the naive branching model ignores: your friends tend to know each other. Mathematicians call this clustering — the extent to which the neighbors of a node are themselves neighbors. In a tightly clustered world, everyone lives in a small club of mutual acquaintances, and a path from one club to another should be long. High clustering and short paths seem to contradict each other. In 1998, the physicists Duncan Watts and Steven Strogatz showed that they do not — and their model became one of the most cited papers in network science[wiki-watts-strogatz].

Their construction is almost childishly simple. Start with a regular lattice: a ring in which each person is connected to their k nearest neighbors. In this world clustering is maximal — neighborhoods are tightly knit clubs — but paths are long: to cross a ring of 1000 people, a message must hop from club to adjacent club, taking hundreds of steps. Now rewire: take each edge and, with probability p, reconnect it to a random node elsewhere in the ring. The parameter p is a dial between a lattice (p = 0) and a random graph (p = 1). The surprise is how tiny p must be to transform the world.

The table and chart show the signature of a small-world network: path length collapses by p = 0.01 while clustering remains near its lattice value. Only about one edge in a hundred needs to be a long-range shortcut for the whole ring to become six-degrees-like. The intuition is that a single shortcut is a teleport: it lets a message jump from one side of the ring to the other in one step, bypassing the hundreds of local hops the lattice would demand. A handful of such teleports, sprinkled randomly, turn a vast clustered world into a small one while every neighborhood still looks like a tight-knit club[wiki-watts-strogatz].

This is why the small world is not a paradox but a design: real networks are mostly local, with a thin layer of long-range ties. Watts's book, Six Degrees: The Science of a Connected Age, argues that this thin layer is one of the most consequential facts about human societies — the difference between a world in which an idea reaches the village next door and a world in which it reaches the planet[watts-book].

The Small World, Measured at Scale

The Wattss–Strogatz model explained how a small world could exist; the data revolution of the 2010s measured how small the real world actually is. When researchers at Facebook computed the average path length over the entire friendship graph, the folklore number "six" turned out to be an overestimate. In 2011, a study of 721 million users found an average of 4.74 degrees; in 2016, with the network grown to 1.59 billion people — more than a fifth of humanity, all connected in a single component — the average had fallen to 3.57 degrees[arxiv-facebook]. Not six. Between any two people on the planet's largest social platform, there are, on average, fewer than four intermediaries.

Two facts from the 2016 study are worth sitting with. First, the network's giant component is essentially everyone: 99.9% of the 1.59 billion users belonged to a single connected web, a scale at which "isolation" is statistically negligible. Second, the path length keeps falling as the network grows. This is the Watts–Strogatz signature in reverse: adding people and the shortcuts they bring does not stretch the world — it shrinks it. Every new user and every new friendship is, on average, another teleport thrown into the ring[arxiv-facebook]. The world's separation is not a fixed number; it is a property of the network's wiring, and we are all still busy rewiring it.

The Other Six Degrees: Family and Genes

There is a second, older meaning of "degrees of separation," and it predates social networks by centuries: the degree of consanguinity — how close two relatives are in a family tree. Genealogists and geneticists count this precisely. The coefficient of relatedness r between two individuals is the probability that a randomly selected gene in one is identical to a gene in the other because of shared ancestry. It is computed by a simple law: each step up or down the family tree — each meiotic link, in the geneticist's phrase — halves the contribution, and paths through every shared ancestor are added[cornell-relatedness].

For a parent and child, the path has length one, so r = 1/2: a child carries half of each parent's genes. For a grandparent and grandchild the path has length two, giving 1/4. Siblings share two ancestors (both parents), so each path of length two contributes 1/4 and they sum to 1/2. First cousins share two grandparents through paths of length four, contributing 1/16 each, for a total of 1/8. The pattern is exact, and the "degrees" of kinship are literally the exponent in the halving[wiki-coefficient].

This is the same phrase — "degrees" — doing different work: the social version counts connections (each hop is a person), the genealogical version counts distance in generations (each hop is a meiotic link). But both are exponents of a half or a base, and both compress enormous genealogies into a single digit. A genealogist will happily tell you that you and a stranger on another continent may share a common ancestor from eight generations back — a mere eight degrees in the family-tree sense — which is one reason the social and the genetic "six degrees" are not as far apart as they look. The relatedness formula has real applications beyond family trees: it underlies genetic counseling, pedigree analysis, and the evolutionary theory of kin selection, where an organism's behavior toward a relative is famously governed by r times the benefit minus the cost[cornell-relatedness].

Why Short Paths Matter

The small world is not a curiosity; it is the mechanism behind some of the most consequential dynamics of modern life. An epidemic spreads along the edges of a social network, and the short average path length is exactly why a novel virus can cross a continent in weeks rather than years — each hop is an exposure, and few hops separate any two cities. The same topology that makes the network feel intimate makes it dangerously conductive; this is why modelers of disease and rumor and financial panic all study the same graphs[wiki-six-degrees].

Information rides the same edges, and here the small world is an engine of coordination. Ideas, job offers, and opportunities flow through acquaintances rather than close friends — the sociologist Mark Granovetter's famous "strength of weak ties" — precisely because the long-range shortcuts of the Watts–Strogatz model are usually weak, low-frequency ties. Your close friends form a tight cluster (they mostly know each other); your acquaintances are the teleports that connect your cluster to everyone else's. Marketing exploits the same structure: "going viral" is nothing more than a message that finds its way across the small-world shortcuts before it dies out, and the probability calculations for who-sees-what are the bread and butter of the Probability Calculator. Once you can see the graph, you can see why a rumor with a tiny starting audience can reach millions — the path is short, the branches multiply, and the world cooperates by being small.

Common Misconceptions

• "Everyone is exactly six hops from everyone else." No. Six is the median of Milgram's completed chains; the 2016 Facebook data puts the average at 3.57. It is a statistical summary, not a universal law, and any particular pair can be closer or farther.
• "Small world means everyone knows everyone." The opposite. A small-world network is locally dense but globally sparse: your neighbors know each other, yet few long-range links span the world. Short paths and high clustering coexist; that is precisely the Watts–Strogatz discovery.
• "The six degrees number is fixed." It shrinks as networks grow and densify. The human graph measured 4.74 degrees in 2011 and 3.57 in 2016 — the number is a property of the wiring, not of human nature.
• "Six degrees of separation and the coefficient of relatedness are the same thing." They share the word "degrees" but measure different distances: one counts social hops (people), the other counts meiotic links (generations). Confusing them produces claims like "we're all cousins" that are only loosely true.
• "If everyone has 100 friends, six hops reach everyone, so everyone is six steps away." The branching model is an idealization that ignores clustering and assumes non-overlapping circles. It explains why small worlds are plausible; the Watts–Strogatz model explains how they arise despite clustering.
• "A random network is the best model of society." A purely random graph has short paths but almost no clustering — it feels nothing like a neighborhood. Real social networks sit between order and randomness: clustered like a lattice, connected like a random graph.

Frequently Asked Questions

❓ What exactly does 'six degrees of separation' mean?

✅ It is the claim that any two people on Earth can be linked through a chain of about six social connections — each person knowing the next on a first-name basis. It entered popular culture through John Guare's play and its 1993 film, but the underlying claim was tested by Stanley Milgram in 1967.

❓ Did Milgram really prove six degrees?

✅ He showed that letters routed through acquaintances reached a target stranger in a median of about five intermediaries — six degrees — but the experiment had a small, self-selected sample and many chains broke. Later reanalysis disputes the exact number while confirming the qualitative result: chains are short, not long.

❓ Is six degrees of separation actually true today?

✅ Modern measurements on massive platforms put it lower. A 2011 study of 721 million Facebook users found an average of 4.74 degrees; a 2016 study of 1.59 billion users found 3.57. So the folklore number is a slight overestimate — the human network is even smaller than six.

❓ What is the Watts–Strogatz model?

✅ A 1998 model by Duncan Watts and Steven Strogatz that generates a small-world network by taking a highly clustered lattice and rewiring a tiny fraction of edges as long-range shortcuts. Just a few percent of shortcuts collapse average path length while preserving clustering, explaining how real networks can be both dense and connected.

❓ What is an adjacency matrix and why does it matter?

✅ It is a table of zeros and ones recording which nodes in a network are connected. Matrix multiplication on it counts paths: the entry (A^k) of the k-th power counts walks of length k between two nodes. 'Degrees of separation' is therefore a statement about the sixth power of the human network's adjacency matrix.

❓ What is the coefficient of relatedness?

✅ The probability that a gene shared by two relatives is identical due to shared ancestry, computed as one-half raised to the number of meiotic links, summed over all shared ancestors. Parent–child and full siblings are 1/2, grandparent–grandchild 1/4, first cousins 1/8, and second cousins 1/32.

❓ How is the Kevin Bacon game related to the mathematics?

✅ It is shortest-path computation on the network of actors connected by shared film credits. Each 'degree' is an edge, and the game asks how many edges separate an actor from Kevin Bacon. It is the same graph theory as airline routing or Milgram's letters, expressed as trivia.

❓ Why are weak ties and long-range shortcuts so important?

✅ Strong ties form tight clusters of mutual acquaintances, which are poor at connecting different parts of the network. Weak ties — acquaintances rather than close friends — are usually the long-range shortcuts that collapse path lengths, letting information and opportunities travel between otherwise separate communities.

❓ Does a short path length mean everyone is close to everyone?

✅ In the topological sense, yes: any two people are only a few handshakes apart. But that is a statement about paths, not about trust, intimacy, or influence. A short path is not the same as a strong relationship, and most of the paths are usable only in principle.

❓ Why does the small world keep getting smaller?

✅ Because networks grow by adding nodes and edges, and every new long-range connection is another shortcut. As more of humanity joins the same giant component, average path length falls — the 2016 Facebook average of 3.57 is likely lower today.