On the right you can see an arrangement of lozenges or elongated diamonds.
Question: Where are the diamonds darker, at the top or at the bottom?
Answer: All diamonds are identical, but it sure doesn’t look like it.
Pick any diamond with the mouse. Now you can move this diamond around and compare it to all the others.
The left slider controls the luminance difference between the top and bottom of each diamond. The right slider controls colour saturation (not relevant for the main effect).
There is a “luminance gradient” along the diamonds: darker at top, lighter at bottom. The slider allows to adjust the strength of this gradient.
Shallow changes in lightness are already suppressed by neural processing in the retina. What is transmitted from the retina to the visual centers of the brain is the contrast step from the dark bottom end of each diamond to the lighter top end of the diamond in the next lower row. So the brain receives information on a lightness step, not of the slow return of lightness, and “concludes” the every row is lighter then the one above it. I interprete this effect as a variant of the Craik-O’Brien-Cornsweet illusion
I am thankful to Akiyoshi Kitaoka for digging up the original source for me. Not only did he do that, he also promptly produced this beautiful variant.
Watanabe I, Cavanagh P, Anstis S, Shrira I (1995) Shaded diamonds give an illusion of brightness. ARVO Annual Meeting, Fort Lauderdale, Florida. [According to him, the illusion originates from Pat Cavanagh.]