You become friends with whoever you run into

Leon Festinger, Stanley Schachter and Kurt Back published in 1950 a study of a married student residence at the Massachusetts Institute of Technology, where housing allocation did not depend on preferences. That detail makes it almost a natural experiment.

They measured who declared friendship with whom and found a strong relationship with physical distance between doors, stronger still when obligatory paths, stairs and mailboxes, were taken into account. Residents living at the foot of a staircase were thus named more often than their neighbours on the same floor.

The authors distinguish distance measured in metres from functional distance, defined by the probability of running into each other. Two neighbouring flats separated by a wall with no passage are functionally far apart, whereas two doors opening onto the same landing are close even at an equal distance.

That distinction is what made the result's fortune: it explains why the layout of a place produces social effects, without invoking individual preference. It was taken up by urban planning and architecture, often with more confidence than the data warranted.

A related result was obtained by Mady Segal in 1974 at a police academy where seats were assigned in alphabetical order: trainees disproportionately named as friends those whose surnames placed them next to them.

Repeated exposure to a person raises the positive evaluation made of them, an effect documented by Robert Zajonc in 1968 as the mere exposure effect and found across many protocols, including with stimuli the participants do not remember.

An experiment by Richard Moreland and Scott Beach in 1992, where people attended a class without ever speaking, found that the number of sessions attended was enough to raise the liking reported by other students, most of whom nonetheless said they barely recognised them.

Contexts producing repeated and unchosen encounters are concentrated in youth: school, study, service, first jobs. They become scarce afterwards, as days come to be organised around more closed places and more individualised schedules.

Surveys do find that the share of friendships formed after thirty is small in most personal networks, a result compatible with the gradual disappearance of those contexts. It describes a structure of opportunities, not an aptitude that fades with age.

Proximity raises the probability of meeting; it does not determine what follows. The same studies find that perceived similarity then predicts how long the relationship lasts, while proximity no longer contributes. Proximity can also produce enmities, since it multiplies the occasions for friction.

The result is therefore about entry into a network, and its transposition to what makes a tie last is an extension the data do not support.