[hughprime]

There is, as I understand it, some evidence for the idea that people learn what's attractive in a member of the opposite sex by essentially averaging out all the faces which they see while growing up. This makes a helluva lot of sense as an evolutionary strategy -- you learn what your species looks like by observing as many other members as you can, and then you try to mate with the least unusual-looking person you can. Aside from protecting your offspring from the weird disability-causing genes which are sometimes signalled by unusual-looking features, it's also a great strategy if you assume everyone else is using it too: if you mate with a non-unusual looking partner you'll get non-unusual-looking kids, who will then be sought-after mating partners.

Anyway, if our brains really do use this strategy I think it goes a long way towards explaining some of this data.

edit: I don't mean to imply that this is the only criterion our brains use -- we probably find most attractive people who are slightly taller, more muscular and/or larger-breasted than the average we see while growing up. However in other things like facial structure where no particular face is objectively fitter than any other, heading towards the average is a pretty good plan.