What to see
Above, the tiles are moving left and right in alternating rows. In the ‘half-shifted condition’, the ‘mortar lines’ (the horizontal lines between the tiles) appear to slope alternately upward and downward. This gives the impression that the tiles are wedge-shaped. As seen when the tiles align or make up a chequerboard, the lines are actually parallel, and all tiles are perfectly square and of the same size. So during the movement, the illusion “comes and goes”.
What to do
There's lots to play around here. You can stop the movement at any time, or –for starters– stop it at maximal illusion offset by pressing “1/2”. You can also experiment with the width of the “mortar lines” between the checks, preset to 2 px. If you reduce the line width to zero, the illusion is nearly gone, this condition is the “Münsterberg Figure”. There is not just an optimal width for the mortar lines, but also an optimal luminance for maximal illusion, namely half way between the bright and dark checks. Color is certainly not important here, if you press the “bw” button the same illusion occurs in black/white.
The Café Wall Illusion was first reported by Richard L. Gregory and Priscilla Heard in 1979. A member of Gregory’s lab had noticed that the front of a café (St Michael's Hill, Bristol, England) had been adorned with black and white ceramic tiles. The mortar between adjacent rows of tiles was visually apparent, and the black/white pattern was offset by half a tile width in alternating rows.
Ten years ago I wrote here “This illusion demonstrates the effect of some simple image processing occurring at the retina combined with some complex processing in the cortical cells of the striate cortex. The incoming image is first filtered by the centre-surround operator of the retina. The apparent tilt of the mortar lines is caused by orientation-sensitive simple cells in the striate cortex. The cells interact with one another to interpret the diagonal bands produced by the retina as a single continuous line, tilted in the direction of the diagonal bands.”
Now I would write: I do not understand how the apparent tilt comes about… ;-)
Lulich DP & Stevens KA (1989) Differential contributions of circular and elongated spatial filters to the Café Wall illusion. Biol Cybernetics 61:427–435
Nice interactive Java Applet at <www.cs.ubc.ca/nest/imager/contributions/flinn/Illusions/CW/cw.html>
For an explanatory text see Brain Connection