My brain must be nonstandard — I see two groups of circles in all three figures, regardless of color.

The change of color in the second set of figures is powerful, though, and is an excellent illustration of (as you say) an inexpensive change that better conveys the groupings.

What I like most about this post is your point that the brain “sees” things automatically based on certain visual attributes. I find it a persuasive argument against my (soon-to-be former?) position that no interface feature is by its very nature intuitive, that anything we call “intuitive” is really just “familiar”. You are arguing, I think, that the brain is hard wired to interpret visual attributes in a certain way; is that correct? To me this means that design features based on those attributes do not in fact draw on learning but on something innate.

I must ask, however — do you know of studies of variations in the Gestalt principles? Do some cultures see the circles & triangles and others not? Do some types of people see smaller/lighter/grayer/etc. objects as being farther away, while others do not? I still wonder if familiarity might not have something to do with it…