What a child calls a friend

Interview work, notably by Robert Selman in the 1970s and 1980s, describes an evolution: a friend is first an available playmate, then someone who helps, then someone with whom inner states are shared.

That progression is found in later work, with a variable timetable. Descriptions in terms of trust and loyalty rarely appear before preadolescence, whereas physical proximity and shared activities dominate the answers of the youngest children.

These stages describe ways of talking about friendship, collected through interviews. They do not say that a young child feels nothing lasting for a classmate: they show that the vocabulary available to the child for saying so is not yet that of an adult.

Children's friendships are measured by reciprocal nomination in class, a technique inherited from Jacob Moreno's sociometry in the 1930s. Their stability over a school year is moderate, and it rises with age.

That instability is not a flaw in the relationship but a property of the age: criteria of choice change quickly, and group composition is largely imposed by school organisation. A friendship that comes apart during the year says nothing, on its own, about the children involved.

Nomination surveys find that a substantial share of declared friendships is not reciprocal, including among adolescents and adults. An unreturned nomination may stem from a difference in definition, a capped list or a simple omission, as much as from a gap in feeling.

That asymmetry is a robust and generally unwelcome result. It is more frequent among younger children and decreases with age without ever disappearing.

The composition of the peer group is largely determined by school allocation. Studies exploiting class changes show that friendships follow the structure of opportunities more than declared preferences, as in adults.

That finding echoes work on proximity in adults, and it explains why a substantial share of childhood friendships does not survive a change of school. The seat assigned in the classroom, the composition of work groups or the canteen timetable are variables that research takes seriously.

Asking the same question of a six-year-old and a fifteen-year-old does not measure the same thing, since the word does not designate the same relationship. An apparent rise in the number of friends between two ages may thus reflect a change of definition rather than a change of network.

Serious surveys adapt their instruments by age band, which makes it impossible to read a single curve of the number of friends across childhood.