This is some follow-on from my musing about too many protagonists.
When we sit down to play a role-playing game, there are three to seven of us (more or less) who want to make a narrative about a bunch of characters. Great. Depending on the game, one of us might play a more general role, wrangling all the walk-ons and antagonists and such, freeing everyone else to really get into the protagonists’ roles.
This doesn’t actually say anything about the nature of the protagonists’ relationships with each other.
A lot of games do, however, come with expectations about the protagonists’ relationships with each other; very often, that expectation is that they will form a group of some sort, working towards common goals, maybe not always on the best of terms, but pulling together when the going gets tough. This is an assumption that I think is inherited from D&D. It’s not a bad one, either, as long as games are clear about it.
You’ve got to manage expectations here, just as in any other part of a game. In the absence of instructions, people will often act in whatever way they’re most used to. For gamers, that’s often towards the D&D model: a motley bunch of misfits who all want something in common, and work together. And that’s not right for all games. Consider, for example, Smallville: in that game, you really need a group who are thematically aligned (not so much a motley group of misfits, unless that is the theme), but whose interests are all at cross-purposes. And once you know that, the game clicks.
Now, as a final thought, you don’t have to make that management of expectations explicit, necessarily. Certain things in a game will encourage certain behaviors. For example, a two-way divide between the participants, with one “GM” and many players, will align the players like a magnet aligns iron filings. It will encourage the players to act as a team, whereas the absence of this divide (as, say, in Fiasco) helps players all see themselves as peers who can have goals at cross-purposes.