Evidence cited in support of open offices ... but is it true?

I just today read a University of Michigan Study from 2000 that concludes that "radical collocation" type open offices show significant productivity benefits over more private workspaces.

Maybe, but I have some questions about this study that may require another look at the results:

1. There was no control group. There were six groups studied, but none of the study groups were using "closed" offices. The study groups' results were scored against an existing corporate baseline of productivity.
2. The productivity gain, while impressive, is within range of the 10 to 1 difference that's been measured in the literature repeatedly. This means that the 4x improvement could easily have been due to the particular programmers chosen rather than the office environment's effects as cited.
3. Finally, there's the Hawthorne effect! All groups knew they were being studied, which has been shown repeatedly to increase productivity no matter what details of changes in the environment are made!

Does anyone know of a controlled study of both kinds of offices compared against each other with no change other than the office space styles between groups? Or of a similar study of 'closed' offices against baseline data collected from 'open' offices?

Maybe 'radical collaboration' does work better, but the Michigan study does not prove the point to my satisfaction.