Does usability research suffer from historical bias? I've seen papers suggest that old-fashioned buttons with a bezel are better than the modern flat style. That seems reasonable, but I wonder if the effect would be smaller if early GUIs were flat.