Dunbar is the last I've looked at, honestly.

I know in my work that there's a "phase change" from seminar to audience around.... 16-20 people.

Most studies are probably qualitative. Factors that influence group size and their performance at that size. It depends on the ability to partition tasks across domains. von Hippel wrote several papers on this subject:
1. Von Hippel E. Task Partitioning: An Innovation Process Variable. Research Policy. 1990;19:407-418. Web: http://stuff.mit.edu/people/evhippel/papers/Task%20Partitioning.pdf

The ability to partition tasks is aided by modular design, clear interfaces, etc.

I would hypothesize (and expect) that with simple tasks, usually a bigger group is less efficient per person but can yield bigger returns (vele handen maken weinig werk!)
I also expect that with complex tasks, e.g. innovation tasks, this reduced performance is much less pronounced and the performance per participant goes up because of the distinct knowledge that everyone brings to the table. This probably is hard to measure since these projects are more effective (solve a problem in a better way) rather than more efficient (solving more problems).
"When users’ needs are heterogeneous and when the information drawn on by innovators is sticky, it is likely that product-development activities will be widely distributed among users, rather than produced by just a few prolific user-innovators."
"... most of the important innovations attributed to users in these studies were done by different users. In other words, user innovation does tend to be widely distributed in a world characterized by users with heterogeneous needs and heterogeneous stocks of sticky information."

1. von Hippel E. Innovation Communities. In: Democratizing Innovation. Cambridge: MIT Press; 2007:93. At: http://web.mit.edu/evhippel/www/democ1.htm Choose Chapter 7. Although the whole book is probably relevant for you.

I'm interested in this because I'm comparing innovation by an organization with limted R&D team (Stratasys) and innovation by an open innovation community (RepRap). Group size and innovation output seem to be related.

Trust and familiarity of group members could also help. The literature on this is from psychology and is probably more extensive than the innovation and economical literature that I'm more familiar with. (I only know some neuro-psychology from AI research that I did)

Ton,

Vanmorgen hadden we een discussie dat je eigenlijk weinig liet zien over je echte werk...ik zie nu je blog van afgelopen zondag. Had je toch al over groepsdynamiek geblogd....dus ik heb voor mijn beurt gepraat vanmorgen.

Was een inspirerende lunch vandaag. Dank.

@Richard Dank je wel! Je algemene punt is ondanks deze post hierboven nog steeds geldig volgens mij: meer bloggen over werk-thema's. Ga ik aan werken ;)

Ton,
My experience is very similar to yours... re groups and sizes. The job of the facilitator is to keep everyone engaged. Often I think it is just how one sets the scene... and works the discussion and reporting. My rule of thumb is 2 people take 5 minutes to draw a quick conclusion, 3 people 10, 4 people 20, 5 people 40... which is why you like a max of five in your groups.

So frankly I usually manage... in pairs, in threes, or in groups based on the number of people in the total group, --- but recognize that once you get over 24... even groups report out with four reports... at 15 min each...

A corollary to this is the total time that the group will work together. Eg a two day workshop with 24 is much easier where you want innovation or strategy than a one day workshop... for one is 8 hours and the other is at least 24 even if you are only in session for16.