Interesting. Just typing "password" yields "pa,s,s,w,or,d", so it seems you register groups by keypress delay. It actually opens up multiple possibilities to factor typing style in to the password. The same keys one fingered grandma-style would not necessarily register as the same password when "properly" typed in.<p>Idea itself is interesting, though password validation brings its own can of worms. Where do you do key groupings - client or server side? How do you capture keypress timing data? What amount of timing variation is allowed? How do you factor in changing typing swiftness? How do you handle keyboard layouts?<p>Essentially, the idea is different keys pressed at tightly controlled intervals, which sounds awful lot like music tune played on a keyboard (pun intended). Are tunes more or less secure than strictly textual passwords we use now?