I finally got around to reading the eye tracking blog posts. Thanks for those.
While a limited study to be sure, they make a lot of sense. It's widely established that faces are interesting to us, and that we have huge amounts of brain devoted to studying them. If there's a face in-frame, it's practically a no-brainer that's what we're going to look at most. Human figures, animal figures? Sure, that also makes sense.
If you think about it, it's ridiculous to suppose that we're going to slavishly be directed by geometry, by tonal values. Our eyes exist to examine and evaluate the environment, not to blindly "follow lines" or to "identify high contrast points".
In the absence of other clues, probably tonal and color values are going to serve some sort of function.
Slobodan is fond of citing Molly Bang's book "How Pictures Work" and if memory serves a great deal of the supplied example can be interpreted primarily as our ability to identify fairly abstract shapes as "figures and faces" with some minimal cues, and the rest is really just supporting those basic identifications with color and tone.
The eye tracking study was very limited, and I'm certainly not saying that it's Truth. It does seem to be consistent with other things I know to be true, however. There's definitely room for a much more detailed study, with a lot more abstract pictures examined, and a bunch of work to determine if whatever we learn about abstracts applies usefully to realistic pictures, and so on and so forth.