Via Robert Scoble:

"Poynter does scientific study on how people read Web sites

Everytime Poynter Institute does EyeTrak research they change how people design media. This time they use their fancy eye tracking machines to look at how people read Web pages. Steve Rubel has the links, including an interview with the guy who ran the study."

In my view, this sort of study should be required reading for anyone involved in online experience design. Some of the findings confirm current design theory some not...

Text Rules:

"Photographs, contrary to what you might expect (and contrary to findings of 1990 Poynter eyetracking research on print newspapers), aren't typically the entry point to a homepage. Text rules on the PC screen -- both in order viewed and in overall time spent looking at it."

Navigation at top performed best:

"Navigation placed at the top of a homepage performed best -- that is, it was seen by the highest percentage of test subjects and looked at for the longest duration. In a survey of 25 top news sites, we found 11 that used top position navigation."

Is it Hot or Not:

"We observed that with news homepages, readers' instincts are to first look at the flag/logo and top headlines in the upper left. The graphic below shows the zones of importance we formulated from the Eyetrack data. While each site is different, you might look at your own website and see what content you have in which zones."

The wandering eye:

"While testing our participants' eye movements across several news homepage designs, Eyetrack III researchers noticed a common pattern: The eyes most often fixated first in the upper left of the page, then hovered in that area before going left to right. Only after perusing the top portion of the page for some time did their eyes explore further down the page."

Full article is here, and plenty more Eyetrack stuff:

Eyetrack history
FAQ
How people read
The Eyetrack III team
Eyetrack is not a solution
How accurate is eyetracking?