Here's an actual eye tracking study that comes to the opposite result of the linked article: http://www.lukew.com/ff/entry.asp?571

This isn't really surprising. The article claimed he just knew from experience and didn't need to do a test, so the whole article was just bullshit.

I was laughing out loud when he said users read all the options before picking one anyway, since I've seen that not happen in countless tests. Users don't generally give a crap about your interface, won't read instructions or screens fully, and will just scan and grab on to something that matches their current goal.