Explore the Watts-Strogatz small-world model: start from a ring lattice where every node only knows its nearest neighbors, then randomly rewire a fraction p of the edges and watch the average path length between any two nodes collapse — while local clustering barely changes.
A handful of randomly rewired long-range "shortcut" edges act as bridges across the whole network. Even a tiny rewiring probability p dramatically shrinks the average number of hops between any two nodes, while the local neighborhood structure — and therefore the clustering coefficient — stays almost unchanged.
Drag the Rewiring p slider from 0 toward 1 and watch cyan shortcut edges appear and the average path length (L) plunge. Adjust Nodes (N) and Neighbors (k) to change network size and density, then click Rewire to generate a fresh random rewiring at the current settings.
Stanley Milgram's 1967 "small-world" letter-passing experiment found that most chains between strangers in the US took about six intermediaries — the origin of "six degrees of separation," later formalized mathematically by Duncan Watts and Steven Strogatz in 1998.
"Six degrees of separation" refers to the idea that any two people on Earth are connected through a surprisingly short chain of acquaintances. The phrase traces back to social psychologist Stanley Milgram's 1967 "small-world" experiment, in which he asked people in Nebraska and Kansas to route a letter to a stranger in Boston using only personal contacts. Letters that arrived typically passed through around five to six intermediaries. The result suggested that social networks, despite their enormous size, are tightly interconnected in ways ordinary intuition does not predict.
In 1998, physicists Duncan Watts and Steven Strogatz gave this intuition a rigorous mathematical foundation in their Nature paper "Collective dynamics of 'small-world' networks." Starting from a ring lattice — where every node connects only to its nearest neighbors and clustering is high but paths are long — they showed that rewiring even a small fraction of edges at random creates long-range shortcuts that collapse the average path length dramatically, while the local clustering coefficient barely drops. This "small-world" regime, sitting between an ordered lattice and a fully random graph, turns out to describe an enormous range of real networks: friendship and collaboration graphs, neural connections in the brain, electrical power grids, and the "Six Degrees of Kevin Bacon" game linking actors through shared film credits. This simulation lets you explore that transition directly by adjusting the rewiring probability p.
A small-world network is a graph that combines two properties usually thought to be in tension: high local clustering (your friends tend to know each other) and short average path lengths between any two nodes (a few hops connect almost everyone). The Watts-Strogatz model shows that adding just a handful of random long-range edges to a highly clustered lattice is enough to produce both properties simultaneously.
In the Watts-Strogatz algorithm, p is the probability that each edge in the original ring lattice gets detached from one endpoint and reattached to a randomly chosen node instead. At p = 0 the network is a pure regular lattice with long paths and high clustering. At p = 1 it becomes essentially a random graph with short paths but low clustering. Small-world behavior — short paths and high clustering together — emerges in the narrow intermediate range, often for p as low as 0.01 to 0.1.
The clustering coefficient of a node measures how many of its neighbors are also connected to each other, expressed as a fraction of all possible connections among them. Averaged over the whole network, it captures the tendency of a graph to form tightly knit local groups, such as friend circles or research collaboration clusters, rather than being uniformly sparse.
Average path length is the mean number of edges on the shortest route between all pairs of nodes in a network. It matters because it determines how quickly information, disease, or influence can spread across a system. In a pure ring lattice of a few hundred nodes, average path length grows roughly linearly with network size, making the network "big" in a functional sense. Adding just a few random shortcuts causes average path length to grow only logarithmically with size instead, which is why huge social networks can still be traversed in a handful of steps.
Stanley Milgram was an American social psychologist at Harvard best known for his controversial obedience experiments, but his 1967 "small-world problem" study is equally influential. He gave several hundred people in Nebraska and Kansas a folder addressed to a specific stockbroker in Boston and asked them to forward it only to someone they knew personally who might be closer to the target. Completed chains averaged around five to six intermediate links, giving rise to the popular phrase "six degrees of separation," although the original study had significant methodological limitations, including a low completion rate.
Duncan Watts, then a graduate student, and his advisor Steven Strogatz, an applied mathematician at Cornell, published "Collective dynamics of 'small-world' networks" in Nature in 1998. They introduced a simple algorithmic model — start with a regular ring lattice and rewire each edge with probability p — and showed computationally that a narrow range of p produces networks with both short average path length and high clustering, matching properties observed in real networks like the electrical power grid of the western United States and the neural network of the nematode worm C. elegans.
Small-world structure has been documented in an enormous range of systems: social and friendship networks, scientific collaboration and citation graphs, the neural wiring of brains including C. elegans, large power transmission grids, the internet's router-level topology, and film-actor collaboration networks — the basis of the popular "Six Degrees of Kevin Bacon" trivia game, where almost any actor can be linked to Kevin Bacon through a short chain of shared movie credits.
No. "Six degrees of separation" is a memorable approximation from one specific historical experiment, not a universal constant. The actual average path length in any network depends heavily on its size, density, and degree distribution. Large modern online social networks have been measured with average distances closer to four steps, while sparser or more geographically constrained networks can have longer typical paths. The Watts-Strogatz model explains why such short paths are possible at all, not that every network has exactly six degrees.
Average path length is estimated by running a breadth-first search from a sample of up to 40 source nodes and averaging the shortest-path distances to every other reachable node, which keeps the computation fast even for larger networks. The clustering coefficient is computed exactly for every node as the fraction of possible edges among its neighbors that actually exist, then averaged across all nodes. Both statistics recompute automatically whenever you change the rewiring probability, node count, or neighbor count.
Sigma is a commonly used summary statistic that compares a network's clustering and path length to those expected in comparable random and regular networks. A sigma value noticeably greater than 1 indicates genuine small-world structure — high clustering relative to a random graph combined with short paths relative to a regular lattice. The estimate shown here uses simplified analytical approximations for the baseline comparisons rather than generating full comparison graphs, so it should be read as an illustrative indicator rather than a precise research-grade measurement.