Motion and physical dynamics captivate the human brain. I’ve always found the iTunes visualizer hypnotic; the video above takes it to the next level. Have a look and see what you think.
Informally, I conjecture that our perceptual and cognitive systems intuit how much information a scene contains, and then allocate attention in proportion to the quantity of information. I can’t offer a precise mechanism, but FINST theory comes to mind, as does pre-attention. It’s tempting to think of perception and cognition as compression routines—recall that a classifier with no inductive bias can do no better than rote memorization of all examples—that are saturated, or hypnotized, by visual scenes that are rich in complex information. Information is complex to the brain if and when it cannot be easily compressed (cf. Chaitin complexity).
(Before I forget, kudos to my former colleagues at NVIDIA for the awesome CUDA, a real-time hypnagogue :D)