Cool visual illusions (with animations!), and an effort to explain why they occur

The cafe wall illusion has the dramatic effect of making a straight line appear slanted:


i-eca0cf2af9fc3ac4445c7dff7d8aab70-research.gifThat's right, the line is precisely horizontal. It was created by Akiyoshi Kitaoka, one of the world's foremost authorities on visual illusion, who is also a wonderful artist. In addition to the hundreds of other illusions he's created, he's posted an entire page of illusions all based on the cafe wall effect.

But why does the line appear to be slanted? It must have something to do with the juxtaposition of light and dark patches, but what exactly? Take a look at this more elaborate illusion which relies on the cafe wall phenomenon:


This illusion was the subject of a study by Rob van Lier and Ãrpád Csathó. In the background between the rectangles (yes, they really are all rectangles) is a gradient which smoothly changes from white to black. The brightness of the background gradient, combined with the color of the rectangle's border, determines whether we see a rectangle or some sort of trapezoid. Now watch what happens when the picture is put in motion:

The rectangles appear to change shape, as if they are rippling in the wind. In fact, all that is happening is that the gradient behind the rectangles is moving horizontally across the picture. The juxtaposition of the different shades of gray with a light or dark line appears to be causing the illusion. To make the phenomenon even simpler, take a look at this animation:

As before, the only thing that's changing is the background. The two rectangles on the left appear to change size as the color changes, but the rectangles in the center and on the right do not. The interior color of the size-changing rectangles is the same, but the border color is different. Notice that the two middle rectangles are the same color as the two on the left, but they don't seem to change size. The two on the right have the same border color as the two on the left, but the color inside changes with the background, and they don't change size.

Focusing in on the rectangles that seem to change, when the rectangle with the dark border has a dark background, it seems bigger. The rectangle with the light border seems bigger on a light background.

Returning to the original animation, when the shading of the background varies constantly, then one side of a rectangle can have a dark background, while the other has a light background. Depending on whether the border of the rectangle is light or dark relative to its inside color, this will make one side of the rectangle appear bigger than the other -- so now the rectangle appears to be a trapezoid.

Van Lier and Csathó's study broke the illusion down further, and in separate experiments they varied the shading of the inside, outside, and border of rectangles, showing how the relative brightness of each of these components affects the perception of the illusion.

What makes this phenomenon really interesting is when rectangles are combined in creative ways. In the cafe wall illusion, for example, the placement of the blocks next to a line makes the same line appear "darker" or "lighter" depending on what it's next to:


In the hands of an artist like Kitaoka, there's no limit to what can be done with such a phenomenon:


van Lier, R. and Csathó, Ã. (2006). Dancing shapes: A comparison of luminance-induced distortions. Perception (35), 775-798.

More like this

I strongly suspect the origin of the illusion is in depth cueing. In fact, the grayscale/moving image originally did not work for me; the rectangles appeared to have 90 degree angles but I also got an mild depth illusion. I had to concentrate on the effect a bit before I could see the trapezoidal interpretation.

It appears that the outline of the rectangle juxtaposed against the background is the key. When the background is white, the outline disappears and is no longer part of the rectangle. We judge the size of the rectangle by the size of the inner color only. As the background color darkens, the outline becomes more and more prominent and becomes part of the overall size of the rectangle thus, it becomes bigger visually. With a gradient from black to white for the background, on the white end of the gradient, the outline disappears, on the black side of the gradient, the outline is most prominent creating the illusion of a change in size of the rectangle - in fact the presence or absence of the outline does indeed visually change the size of the rectangle because the outline adds size to the rectangle. This changing in visibility of the outline is what causes the apparent slanting of the lines, the apparent trapazoidal shape of the rectangels. It's not entirely an illusion as the size of the rectangles actually does change depending of whether the outline is visible or not.

That's my guess at it at any rate. Correct me if I am wrong.

One of my lecturers at Uni (Prof Dave Earl) did alot of research into this and taught me alot about it. I love this illusion! There is a nice site that looks into it in some details and has an interactive applet here .

Jeff - That's what I thought at first too. But then I watched it again, and it seemed to me that the exact opposite was happening. When the rectangle with the light border had a darker background, it seemed smaller.

So I think that when the outline is more prominent, it makes the inner color seem smaller, and when the border is not visible, the inner rectangle doesn't seem to be bounded as much.

Your explanation made much more sense to me, and I don't quite understand why it might be the opposite (unless it's just me, and nobody else is seeing it grow bigger against the same background).

Sorry, but the cafe wall "illusion" is really something of a cheat! The so-called 'straight line' that is supposed to appear slanted is actually a dashed line of dark segments each of length equal to the width of the squares. The squares have been offset from the dashes by half a length, so that each black square now genuinely has a stepped (ie slanted) edge. The whole secret of this effect is that the dashed line is not completely black but grey. This means that your eye also sees the white squares as stepped. Thus you are faced with the seemingly impossible conclusion that the inside edges of every square are stepped upwards. If the dashed line were black, the illusion wouldn't work. Very clever, but still a cheat! :)


No, it's not a cheat. It is composed of a single, solid gray line, and alternating black boxes. If you don't believe me, take the image into photoshop and crop the image so that only the line remains.

I just did it myself, and I can assure you, the line was uniformly the same color.

The line can appear to be dashed, because it seems relatively light next to dark squares, and relatively dark next to light squares.

Much as I _absolutely_ value Akiyohi Kitaokas contributions to the field of illusions, this one was not "created" by him. In my webpage referenced in the URL I give the origin, a movie variant and some literature pointers.

It is NOT a cheat, and IMHO still not well understood.

Best, Michael

PS: "Best", I always thought that's rather nice, but from the recent signoff debate... well on the other side of the Atlantic and as a non-native speaker I can't get it right anyway ;-).

Michael -- you're absolutely right. I meant "created" in the sense that Kitaoka created the image, not in the sense that he discovered the illusion.

(and if any of our readers haven't visited Michael Bach's fantastic illusion web site, you should definitely check it out. Be prepared to spend at least an hour or two, though!)

My very first thought was to do my bedroom wall with this. However, i probably wouldn't be able to stand up with my eyes open in such a room.

Call me weird, but I had another, different thing happen when I looked at the first picture. The line did look as if it slanted up from left to right, and the two ends stayed still, but when I looked at the left end, the line seemed to have a "concave" curve going up, and when I looked at the right end, the line became "convex."

BTW, AFAIK I don't have astigmatism.

By anomalous4 (not verified) on 08 Dec 2006 #permalink