when you're presented with some 48 figures, you simply overload your brain and it fails to evaluate aesthetics of each rectangle, sort and pick the highest score. so in my opinion that experiment is flawed.
you brain can only operate on a limited set of objects especially when evaluating and comparing qualities. (7+-3 or thereabouts). pick any book about brain/psychology - it's there.
http://en.wikipedia.org/wiki/The_Magical_Number_Seven,_Plus_...
http://www.ted.com/talks/barry_schwartz_on_the_paradox_of_ch...
I'll check the PDF, thanks. Also see:
http://www.lhup.edu/~dsimanek/pseudo/fibonacc.htm
http://plus.maths.org/content/os/issue22/features/golden/ind...
People do seem to like ratios of small integers, though. Which is why the supposed aesthetic elegance of 1.618 ((1+sqrt(5))/2) is most probably within statistical significance of 1.5 (3/2) or 1.666 (5/3), it's impossible to test whether people don't just enjoy ratios of small integers instead of the irrational number Phi.