We seemed to uncover our own version of dunbar's number with groupsize in trying to hold a group discussion with more than about 11 people very quickly we break off into two smaller groups.

This is a well-known result from organizational psychology where meeting with more than 12 people just don't work. I first read about this in Christopher Alexander's 'A Pattern Language: Towns, Buildings, Construction'. A quick google returns e.g. this: http://www.iwritewordsgood.com/apl/patterns/apl151.htm