Optimal Distinctiveness Theory
People are pulled by two needs at once, to belong and to stand out, and no single group size satisfies both. Marilynn Brewer's 1991 theory argues the relationship between a group's inclusiveness and how strongly people identify with it is curvilinear, not linear: identification peaks in the middle, in groups small enough to be distinctive and large enough to include you. A whole company is too broad to answer "who am I here". A pair is too narrow to function as an identity at all.
What it is
Most everyday thinking about belonging assumes more inclusion is simply better, the bigger and more welcoming the group, the stronger people should feel about it. Brewer's argument, set out in her 1991 paper "The social self: On being the same and different at the same time", is that this assumption is wrong, because belonging is only half of what a group has to supply.
The other half is differentiation. People also need to feel distinct, to have some part of their identity that is not shared with everyone else, that marks them out rather than dissolving them into an undifferentiated mass. These two needs, inclusion and differentiation, pull in opposite directions. A very large, very open group satisfies the need to belong almost automatically, anyone can join, but does nothing for the need to stand out, being one member among thousands offers no distinctive identity at all. A very small, very exclusive group does the opposite, membership is highly distinctive, but the group may be too narrow and fragile to supply a real sense of inclusion, and it disappears if a single member leaves.
Because the two needs point in opposite directions, Brewer's model predicts that identification does not rise steadily as a group becomes more inclusive. It rises, peaks, and then falls again, a curvilinear rather than a linear relationship. The peak sits at whatever group size manages to satisfy both needs at once, distinctive enough to answer who you are, inclusive enough that the answer holds up.
This has a direct bearing on why cliques, teams and clubs tend to converge on a similar range of sizes rather than trending toward either extreme. A group of five or six is small enough that membership means something specific and excludes almost everyone, and large enough that it does not collapse the moment one relationship in it goes wrong. Somewhere in that middle range, not at either extreme, identification is predicted to be strongest, which is roughly where many everyday social units, cliques among them, actually sit.