Yeah, the truth is somewhere between that and the other theory that the overhead for internal communication goes at least quadratic with the number of group mem

Yeah, the truth is somewhere between that and the other theory that the overhead for internal communication goes at least quadratic with the number of group members, so a group of four looses 16/9 more of their productivity to internal friction than a group of three. And all this is assuming you actually have a quantitative measure of programmer productivity that is accurate enough to measure a 30% change. But let's no go down that drain again.