Using decades of data from the Framingham Heart Study, they found that loneliness was not scattered randomly through a community. It clustered among connected people, reached as far as a friend’s friend’s friend, and appeared disproportionately around the network’s edges. A person’s loneliness was associated most strongly with that of direct contacts, but the relationship remained above chance at two and three degrees of separation. They controlled for a person’s earlier loneliness and examined whether one contact’s loneliness predicted another person’s later response. A 2025 model recreated the pattern through different routesA 2025 study in Scientific Reports returned directly to the loneliness question with an agent-based computational model.