I've always wonder why Norman, Dvorak etc. preform so much better than qwerty by most metrics, but real-world studies showing that much of a difference are rather lacking.

I can imagine it must be hard to design a good study comparing them when almost everyone uses qwerty.