One of the principal impediments to participation is the time and effort necessary to compose, edit and refine a contribution for internal "publication" so that it can be clear to all potential readers. If you post a note about your specialty (e.g. your practice) for your own use, it can be short and cryptic. If you know that only trusted senior members of your practice group will see it, you may be a little more thorough, but you still have confidence that others of equal experience will understand it.
People don't like to contribute (participate) when they don't know who might be listening in on the conversation. This bears out in the suggestion above. The further your audience is from your inner circle, the more work you have to do in sharing what you know. At some point, the effort of sharing becomes too high for the perceived value of doing so. This lame drawing gives you a better feel: at some point it's too much (perceived) trouble to share, even though the potential value of the shared information could be higher too.
I didn't get it at the time, but the value of the Caterpillar communities project is in line with this (described here). Each community has a limited membership, and members can see very clearly who has access to their community discussion. New community members request access through the community moderator.
This also relates nicely to the idea of transaction costs, which are normally missing from social network maps because they represent only one kind of flow. In network maps of transportation or supply chain networks, the cost of moving along any one of the links must be explicit because it is not uniform across the network (the shortest "path" isn't always the cheapest). In the environment where I am asked to share my knowledge directly with a wide body of people, the cost is inversely proportional to the strength of the farthest relationships.