I looked on your site but didn't see a description of the design process for Colemak. You mention in the FAQ that in contrast to Dvorak it was designed with the aid of computers. You also mention some design principles for Colemak, and there's the comparator applet that calculates stats like distance moved.
So what was the design process? Did you randomly generate layouts (within some constraints) and run them all through an evaluation program like the comparator applet? Did you use genetic algorithm techniques, modifying successful keyboards randomly and resubmitting them to the "pool"? Or did you design most of the potential layouts "manually", and just use computers to evaluate them? (Oh... looking at Capewell's web site it sounds like you used genetic algorithms.)
What was the list of evaluation criteria that were actually used for selection, and how were they weighted/prioritized?
Have you written up an academic paper on it?

Just curious. :-) Seems like you've done enough work that it would be worth publishing.
Thanks.