Here is an article from the McKinsey Quartely on a topic of great personal interest to me: how to boost performance of R&D workers.  I believe knowledge barriers are increasingly important in increasingly complex and increasingly virtual R&D teams and have written about it quite often.  Clearly, the idea of forming focused virtual communities is new to many R&D executives:

Nonetheless, many executives have a hazy understanding of what it takes to bolster productivity for knowledge workers. This lack of clarity is partly because knowledge work involves more diverse and amorphous tasks than do production or clerical positions, where the relatively clear-cut, predictable activities make jobs easier to automate or streamline. Likewise, performance metrics are hard to come by in knowledge work, making it challenging to manage improvement efforts (which often lack a clear owner in the first place).

The article points out five significant barriers to effective R&D community building and what to do about them:

1. Physical: As I have discussed in the past, even R&D workers separated by just one floor become more or less virtually connected.  The article suggest building communities of practice (COP).  However, COP need help to overcome barriers 2, 4 and 5 below.
2. Technical: Much R&D collaboration involves different technical disciplines.  Each technology has its own specific jargon.  It is quite difficult to have effective communications across different different disciplines.  The article suggests role rotation as a possible solution.  However, these rotations are expensive and provide mixed results because of 4 and 5 below.
3. Social / Cultural: Pretty obvious for multi-national teams.  However, as I have pointed out, significant cultural differences exist between two different organizations (such as marketing and manufacturing) in the same office.  The article again focuses on COP.  In my experience, COP are a good start, but not the ultimate answer to addressing social barriers.
4. Contextual: This to me is the most important barrier.  Most R&D communication occurs through documents (reports, papers, etc.).  These communications describe the results, but rarely provide the context of why particular decisions were made.  Some of it is between different technologies or organizations (2 and 3 above), but the barrier is even more critical to team members in the same discipline… The article focuses on forums and meetings as a possible solution.  I do not think you can understand the context of another discipline through meetings.  
5. Time: Given enough time, it is possible to overcome the barriers above.  However, there is never enough time.  Most people need the right information, at the right level of detail – instantaneously. The article suggests a central database of learnings.  I think that is a great idea.  However, it is critical to structure the database so that the information is easily accessible.  That is easier said than done.