I just got this from an old colleague:

Despite all of our many successes, our architect has started making noises to the product management team that pair programming is a waste of money.

Why should he let mountains of academic research and our own project metrics get in the way of his own personal opinions!?!?!

Sometimes I just want to pack up and go home

Now first off, when people start talking in favour of Pair Programming, I tell them two things:

# From what I’ve seen, it’s the most efficient way to transfer skills, reduce “truck factor”:http://www.agileadvice.com/archives/2005/05/truck_factor.html, ensure quality and (say it quietly) stop slacking. It’s also damn hard to get some companies to try it for a number of reasons.
# Read “this Hacknot article(Hacknot – A Critical Look At Some Pair Programming Research)”:http://www.hacknot.info/hacknot/action/showEntry?eid=50 before you start quoting studies.

I don’t happen to agree with the editorial bais of the Hacknot site – my own (and my company’s own) experience has shown that we produce better software more efficiently when pairing. That said the study quoted is clearly flawed for the reasons mentioned. There have been more studies since – one day I should really get round to pulling them all together. The major problem in producing such a study is that I don’t know anyone who would spend the money creating a real world experiment over a long enough period of time to produce conclusive results.

I am however fairly sure that “pairing with a cat”:http://www.hacknot.info/hacknot/action/showEntry?eid=21 yields no benefit over pairing with an alternate programmer.