Yeah, but where are the *interesting* comparisons?
The study by Williams et. al. concludes that two programmers working together can accomplish something with slightly less than twice the effort as one individual programmer. I have a hard time getting very excited about that conclusion. What I'd like to see is a comparison between pair programming and inspections, or 3 pairs vs. 1 well-run 6-person team. Until I see that kind of data, I'm going to continue to believe that most programmers who pick up XP do so as a way of legitimizing their existing poor practices.