One is a value function, which is best described by Tilly-Odlyzko: v = n * log(n)
(Tilly would be @tilly on HN, BTW.)
The "log(n)" factor means that as you add additional elements, the positive benefit decreases as n increases.
The other is a cost function, which is probabalistic between any two nodes. It's actually closer to the old (and incaccurate) Metcalfe rule: c = k * n^2
Where k is some cost constant. The larger k, the smaller the productive size of the network, the smaller k, the larger the network can grow. Hence: network size is constrained by the cost constant.
What you're observing is the cost constant of team size as a factor of interruptions. Since any member might interrupt you at any time, and those interruptions all have roughly equal cost, but the positive contribution a team member can make is limited by their ability to provide useful inputs, and the k factor is relatively large, few technical teams seem to scale much beyond about 6-12 members.
(Larger teams are almost always subdivided into smaller effective working groups.)