A generous choice can alter an encounter that the original giver never sees.
That was the central result of a 2010 analysis of laboratory public-goods games. People who observed higher contributions tended to contribute more when they were later placed with new strangers. Those strangers then carried some of that behavioural change into still later groups. Across the constructed chains of interaction, cooperative behaviour spread through as many as three degrees of separation.
This was not a study of friends passing favours along a real community network. It was a tightly controlled experiment involving anonymous players, monetary decisions and random reassignment. That design makes the causal evidence unusually strong within the task, while limiting what can safely be claimed about generosity outside the laboratory.
The public-goods game made generosity measurable
James Fowler and Nicholas Christakis reanalysed data from experiments originally designed to study cooperation and punishment. The cooperation-cascade paper included 240 participants.
In each round, players entered groups of four and decided how many monetary units to keep and how many to place in a public project. Keeping a unit maximised private earnings. Contributing imposed a private cost but increased the amount available to the group. Everyone benefited from the shared fund, including anyone who contributed nothing.
This is the public-goods dilemma. Collective welfare is highest when people cooperate, but each individual has an incentive to free-ride on everyone else. The underlying experiment was initially used to ask whether costly punishment could sustain contributions.
Participants completed six rounds. After each one, they saw what the other members of their group had contributed. They were then randomly assigned to an entirely new group, and no pair of participants met twice. Identities remained anonymous.
Reshuffling turned six rounds into a network
The repeated reshuffling created a traceable path of observation. Imagine that Alex played with Blair in round one. Blair saw Alex contribute generously. In round two, Blair joined Casey and contributed more. Casey then moved into a later group with Devon and raised their own contribution.
Blair was one degree from Alex because Blair directly observed Alex. Casey was two degrees away because Alex’s possible influence reached Casey through Blair. Devon was three degrees away through Blair and Casey. Devon had never met Alex and had never seen Alex’s decision.
These “degrees” were not friendship ties. They were directed links defined by who could observe whose earlier contribution before making a later choice. The network existed because the experimental schedule connected decisions across rounds.
That distinction is essential to the headline. The original giver did influence people they never met, but not by transmitting a known gift through a queue of recipients. The proposed mechanism was behavioural carryover: one person’s contribution changed a partner’s later contribution, which became evidence for someone else.
The effect weakened but remained detectable
At one degree of separation, each additional monetary unit contributed by one player was associated with about 0.19 additional units from an observing partner in the next round. In the punishment version, the estimate was about 0.18.
At two degrees, the estimated increase was smaller: about 0.07 units in the basic game and 0.05 in the punishment condition. One analysis detected an effect of about 0.06 units at three degrees in the punishment game. The authors’ preprint describes the general result as spread from person to person to person to person.
The initial effect also persisted across later periods. When the researchers added the statistically significant direct and indirect pathways, one extra unit contributed in the first round generated an estimated total of three additional units from other players over the experiment. The 95 per cent interval was 2.4 to 3.6 units in both game versions.
That “tripling” depended on this particular network structure, group size, sequence and model. It is not a universal multiplier for kindness. Uncooperative behaviour also cascaded: lower contributions induced lower contributions downstream.
The estimates describe average changes across many experimental paths. They cannot identify a dramatic individual chain in which one known generous player caused one particular distant player to give a specified amount. The causal claim concerns the pattern produced by random assignment and repeated decisions.
Random assignment answered a difficult causal question
People connected in real life often resemble one another, but resemblance does not establish influence. Generous people may choose generous friends, a process called homophily. Neighbours may respond to the same local conditions. Friends can also coordinate or select similar situations.
Random reassignment removed the choice of partner. Anonymity and the rule against repeat pairings prevented direct repayment, reputation building and durable personal relationships from explaining the chain. If Alex’s contribution predicted Devon’s later behaviour through Blair and Casey, the planned sequence supplied a credible path of influence.
This is why the experiment is stronger than merely finding that friends of friends give similar amounts. A later methodological paper on inference across multiple degrees of separation distinguishes association, contagion and direct causal effects. Randomised experiments can address some of those problems more cleanly than observational network data.
They do so by simplifying the world. A laboratory contribution is not the same as helping a colleague, donating blood or caring for a relative. The participants were strangers, the stakes were bounded and the researchers controlled when information arrived.
Institutions can lengthen or shorten the cascade
The 2010 analysis compared a basic game with a version in which participants could pay to reduce a low contributor’s earnings. Cooperative spread appeared in both. Punishment was not necessary to create the basic carryover, and the original paper found minimal overall differences between the two versions.
Later work showed that details matter. A 2019 analysis of different punishment arrangements found that low-cost, high-impact punishment allowed influence to persist farther than weaker or more expensive versions. The network did not transmit behaviour independently of its rules.
Other controlled research has likewise found that allowing people to update their social ties can stabilise cooperation. Who can leave, who can observe, whether people meet again and how rewards are divided all change the incentives surrounding an apparently personal moral choice.
Three degrees is a finding, not a law
The experiment demonstrates that cooperative influence can cross three links under controlled conditions. It does not show that every generous act travels exactly three degrees, that influence always ends at the fourth, or that any one act will transform dozens of real-world strangers.
Statistical detection also depends on the strength of the remaining effect and the amount of data. A cascade may become too weak to measure before it literally reaches zero, while different institutional rules may preserve it for longer.
The authors explicitly left open whether pay-it-forward behaviour spreads through multiple degrees in natural human networks. Relationships outside the laboratory can amplify influence through trust and repetition, but they can also block it through competing norms, unequal power or selective attention.
A related ScienceBlog report on generosity among four-year-olds shows another form of paying kindness forward. It is complementary evidence, not a replication of the three-degree monetary cascade.
The cautious conclusion is still consequential. People carry information about how others behaved into their next encounter. A cooperative act can modify what one observer treats as normal, and that observer can display the modified norm to someone new. Generosity need not stop with its first recipient, because the behaviour it evokes can become the next person’s example.
