The checker shadow illusion redux

August 16, 2011 • 4:58 am

I’ve posted on this before—it’s the most striking optical illusion I’ve seen—but this video is a far better demonstration than the graphics I used before.  It’s called the “checker shadow illusion” and was invented by Edward Adelson at MIT.  The link in the first sentence explains it: it’s based on how our visual system compensates for shadows, making the checker in shade look lighter than it really is.

I can already hear the theologians crowing (as they often do) about how our senses don’t really perceive an external reality—that our idea that we can trust the world through our senses, as in science, is simply another form of faith. (Forgive me, for I just read this very argument yesterday.)

h/t: Matthew Cobb

62 thoughts on “The checker shadow illusion redux

  1. Saw that illusion in one of Steven Pinker’s books. It’s fantastic. I showed it to everyone and nobody believed me.

    The guy who made the illusion, Adelson, pointed out that optical illusions work because the human visual system is an adaptation that works very hard to see them (the designed optical illusions) as coming from a real world based on past experiences of visual processing: some fifty regions of the brain organise raw pixels into meaningful components to correspond with aspects of reality that were relevant to the survival and reproduction of our ancestors. Pinker writes, “if we were in a world of ordinary 3-D objects that had projected those images [in the Shadow-Checker illusion] onto our retinas, our perceptual experience would be accurate”.

    Our perceptual systems evolved to tell the difference between reality and fiction, and cognitive scientists believe that the ability to entertain prepositions without necessarily believing in them is a fundamental ability of human cognition (an ability that, to a large extent, does not exist in schizophrenics.)

    Although external reality is perceived through the brain, this does not imply that reality is an arbitrary construct created by expectations or the social context. An image can be both a product of the mind and a genuinely existing entity.

    Adelson has carefully manipulation some of these assumptions (as in the Shadow-Checker illusion above) to lead an observer to an incorrect conclusion; Adelson explains, “As with many so-called illusions, this effect demonstrates the success rather than the failure of the visual system”.

  2. I see these illusions as the very reason we need science. We know how easily our senses can be fooled, so we desperately need an objective tool to discern was is real verses illusionary. Empiricism cuts the heart out of theology.

    1. Exactly. Science can tell us THAT we have been fooled, and HOW we have been fooled. Theologians have no idea that they have been fooled, much less how.

      1. “our idea that we can trust the world through our senses, as in science, is simply another form of faith. ”
        Which is exactly why scientist do experiments, somewhat like the movement of the paper in the video.

      2. Matt, I think it’s worse than that. Theologians *embrace* the tricks our brains play on us as the “highest” form of reality. Mystical visions, hyperactive agent detection, confirmation bias, “ghost” sightings, and on and on are celebrated by religious and spiritual belief systems as true and deeply meaningful, and believers refuse to even consider the possibility that they may have been fooled.

  3. What is really amazing about this illusion is that even when you know the trick your eye can’t see reality. Same as the inability of creationists to see reality. If were not for deductive reasoning we would be taken by the illusion.

    1. I’ve seen this illusion a few times, and in previous times I could never get my brain to see that the colors were the same without cutting out or covering over parts of the board. However, I noticed this time that I was able to make myself see it. If you look at the two dark squares at the bottom of the screen and at the three light squares in the shadow, you can focus on them as two parellel rows and allow your eyes to ignore everything else going on with the shadow. At that point, you can actually see that they share the same shade of gray.

  4. Jerry,

    You said, “I can already hear the theologians crowing (as they often do) about how our senses don’t really perceive an external reality, that our idea that we can trust the world through our senses, as in science, is simply another form of faith.”

    Actually, the neo-Kantian position is well-respected in epistemology. Sense impressions, even when taken as indubitable, do not justify matters of fact outside of sense impressions.

  5. I’m calling shenanigans on this one — though, I must say, it’s quite well done.

    The whole point of the Adelson illusion is that two objects with different albedos can reflect the same amount of light if the light source is adjusted appropriately. With the cylinder casting that shadow, our brains know that the square in the center must have a higher albedo and therefore be lighter in absolute terms. It’s just that the light levels (and choice of square coloration) have been matched such that the a spot meter pointed at the two squares would read the same value — but it would read different values in uniform illumination.

    So, when you create a real-world representation of this scene, you most emphatically cannot swap a tile between light and dark and have it remain exactly the same — any more than you can have a real-world Escher-style perpetual staircase. You can build a real-world model that, from a particular perspective, looks like a perpetual staircase…but, take even one step away from the proper vantage point, and the illusion is shattered.

    There are two possible explanations for the video. Either they’ve done the same thing as in the perpetual staircase example, or there’s digital fakery going on. The former doesn’t seem to apply, unless they swapped out multiple sets for the multiple camera angles. However…if you look closely and pause the video at just the right moments, you can see that there are times when the moveable square is partially overlapping the outer square and the shades aren’t quite the perfect match that they are a few frames later. Therefore, I’m going with digital fakery.



    1. So, when you create a real-world representation of this scene…

      Ah, but did they? Or did they mock it up to look like one? You’re thinking too hard.

      1. Unless the entire thing is an award-winning CGI rendering, they most certainly did create a real-world model. Not only do they have the camera flying through the scene, you can even see the cylinder casting a shadow on the girl’s hand and arm.

        And that’s actually another point where you can see that there’s digital fakery going on. The relative values of the moveable square and the girl’s skin change during the swap. You would expect her skin to darken the same amount as you would expect the square to darken…but it doesn’t.

        As I wrote, it’s very well done — but we’re seeing a manipulated version of the video. The raw footage straight out of the camera would not have shown the illusion.



        1. Not necessarily. Look at the floor around the illusion. It’s all fairly equally illuminated. There are two light sources. One shining on the pillar on the left, and another shining directly down onto the checker board.

          I did see a fair bit of macroblocking that disappears at times, but this is the nature of video compression, not necessarily trickery. Distortions like this are most common in the very types of textures present in the video — nearly uniform or smooth gradients.

        2. I dunno. I’m looking, but to my eye, her arm is not darkening when it’s in the “shadow”.

          My guess: Notice that there are TWO lights in the scene.

          If you look at all of the true shadows in the scene – e.g., from the girl’s arm, or the edge of the platform – they actually fall toward the left, created by the overhead light. I can’t see any double shadows.

          So I expect that the top light is bright enough as to wash out almost all of the side light’s shadows (the side light is also angled upward), and the “shadow” on the platform is actually just painted or printed on. I think the light to the side is just to complete the illusion by creating the expected highlight on the column.

          1. Still don’t see it.

            Note that it’s the top part of her arm where we’d want to see a clear shadow even under the overhead. Like the tiles appear to show. But there doesn’t seem to be one.

            There’s some shadowing on the LEFT (from the viewer’s perspective) underside of her arm at certain angles — when it is outside the cone of both lights — and this tends to sell the illusion because it flashes a bit of “dark” at the right time as her arm moves into the shadow of the cylinder. But if you look closely, there’s nothing perceptible on the TOP part of her arm, the part illuminated by the overhead light, as the tiles are.

    2. Nope, there’s a third explanation, and it’s sneaky.

      You’re making the assumption that the squares really are the same albedo/reflectance, and that the apparent difference is produced by the light and shadow. But if the shadow cast by the “key” (stronger) light is an illusion, then it works just fine.

      In other words, the light is not as strong as you think, and the “shadow” it casts is augmented by the tiles actually being darker.

      There are some subtle indications of this. If you look at the shadow in the overhead shots, you’ll see it is rigidly defined near the base of the cylinder, then becomes more obscure as it gets further away. This is as it should be, but this should have made it extremely obscure (barely making any difference at all) as it got to the far side of the checkerboard, where the light falloff from the key light got overwhelmed by the secondary softbox light.

      Another clue is if you look at the sides of the raised checkerboard early on, where the edges closest to the key light are very much brighter than the edges farthest from it, even though the softbox does not illuminate them – this difference is caused only by the distance from the key light, yet it appears almost the same where the softbox is illuminating the surface. Pause at :43 or so and notice that the side (vertical surface) of the rightmost square, most distant from the key light, is brighter than the top surface, even though it is receiving the softbox illumination obliquely and should be darker.

      Full credit to the people that set this up – getting the light levels and the fake shadow to look right, especially from different angles, took a lot of care. Very, very well done.

      I do a lot of photo editing, and pride myself on spotting color casts and being able to match, for instance, different portions of sky that many people think show no difference at all. The original checkerboard illusion completely fooled me, and this one was subtle enough for the indicators to slip past the first go around. Love it!

      1. Wait — you’re saying that the “shadow” is painted into the tiles, that the lighting is uniform, and the small round light behind the cylinder isn’t casting a (significant) shadow?

        Hmmm…could be. I’ll have to watch this a few more times, methinks….


        1. The lighting isn’t quite uniform, but by all rights the shadow should have faded faster than it does. It’s a combination of both, but basically, where the tile sits it is evenly matched between the key (cylinder) light and the softbox.

        2. Ben, thanks for this thread – you were absolutely right to point this discrepancy out. I believe the other poster(s) is correct that the lighting is part of the trickery here.

      2. I was trying to figure it out and noticed that there is a ‘fake shadow in the coloration of the central diagonal tiles behing the cylinder, it is confusd by the real shadow which of course has diverging outer borders. This fake shadow is due to all the central lighter tiles being darker than the lighter tiles outside of shadow, and gradation from light to dark within the darker tiles immediately next to this diagonal line of lighter tiles….this ‘fake shadow’ has parallel borders that align w/ the edges of the cylinder footprint accentuating the illusion that it is a real shadow and also blend in with the diminishing real shadow giving the illusion that it is part of the shadow.
        Therefore, there are medium gray and white tiles in the periphery outside the shadow; within the true shadow there are medium gray tiles in the central diagonal line that *appear* white because they are all adjacent to gradient dark to medium gray tiles which appear to be like the medium gray tiles on the periphery and which appear monochromal due to the confusin w/ the real shadow.

      1. Thanks for that.

        I’d still like to know how it was done…the painted-on shadow with uniform lighting still seems like the best bet…but I’d absolutely love to get unrestricted access to that set.



    3. see my entry below which gives a forth explanation..that of deliberate shading of the squares, with a resulting fake shadow (barely visible w/in the real shadow, but w/ parallel borders) which is cleverly hidden due to confusion with the actual shadow.

    4. see my post below…I think it desrcribes the how correctly on the shading…false shadow mixed w/ real…also..a clue…watch her when she moves squares and her hand fails to get as dark as we would expect it should in the shadow, confirming that some of the ‘darkness’ of the shadow is not real. Also at exactly 1:00 when the moved paper overlaps the dark squares in the edge of the shadow, you can see that the bottom half of the underlying square is darker compared to the paper than the top part.

  6. Keep this in mind when you read Plantinga’s evolutionary argument against naturalism.

    I quote from Wikipedia:
    EAAN argues that the combination of evolutionary theory and naturalism is self-defeating on the basis of the claim that if both evolution and naturalism are true, then the probability of having reliable cognitive facilities is low.

    Our sense perceptions and our cognition are far from perfect. In fact, they are reliably imperfect. We are subject to a great many illusions which illustrate that our senses and our cognition evolved to be good enough under many common conditions.

  7. For heaven’s sake, this is something that can be replicated. If they are using trickery, it can be found out. Would they really set themselves up to be shown to be liars?

    1. I’ve got to head off to an appointment in a little bit, but I think a series of screen captures coupled with the eyedropper tool in Photoshop should be most revealing. Somebody feel free to have at it….



  8. Interestingly, Buddhist philosophy (in its advanced forms) completely rejects the evidence of the senses.

    During the Monlam festival at the Tibetan New Year in pre-occupation Tibet, the climax was the expulsion from Lhasa of a human scapegoat. The scapegoat would turn at the last moment and address the assembled high lamas: “All that you teach is false; the evidence of the senses is true.” [I think I’ve quoted that accurately, but I’m writing from memory and may have garbled it slightly.]

    Another expression of the disbelief in the evidence of the senses by Tibetan Buddhists is found in the difficult visualization exercises more advanced practitioners went through. These would, for example, involve the progressive visualization of one’s tutelary goddess in greater and greater detail until your visualization not only conformed in all points to the canonical description (ornaments, colors, posture, gestures, etc) but was indistinguishable from reality. Alexandra David-Neel in her book “The Secret Oral Tradition of the Tibetan Buddhists” makes the point that some practitioners at that point would fall in love with their visualized goddess and make no further progress. The real point of the exercise was, however, that we can jigger our brains to produce at will hallucinations indistinguishable from reality – and therefore we should dismiss all sensory evidence as nothing more than a side effect of mental activity.

    I’ve always been sympathetic to the Tibetans, but reading the account of the scapegoat’s parting shot at the Monlam festival convinced me that I could not be a Buddhist for the simple reason that I accept sensory evidence as being based in an objective reality. It may be flawed at times (or even very often), but it’s a good starting point. And science provides many ways of overcoming any distortions in raw sensory evidence.

    1. The real point of the exercise was, however, that we can jigger our brains to produce at will hallucinations indistinguishable from reality – and therefore we should dismiss all sensory evidence as nothing more than a side effect of mental activity.

      The Tibetans are not the only culture that discovered this.

      Carlos Castenada jiggered a half dozen best selling books and a position as a tenured professor of anthropology out of the same idea.

      After about 3 books, I decided that I could not be a Toltec. Dismissing all sensory evidence as nothing more than a side effect of mental activity just seems like a bad idea to me.

  9. Dear cretins who don’t think there are two lights:

    Then what exactly is that second stand for (the one in front of the scene… the one that has the giant fucking LIGHT coming out of it at about 0:27.)

  10. I don’t think there is any claim made that we are seeing anything beyond very carefully adjusted light levels, and the idea that our visual system is making assumptions about the checkerboard pattern.

    Interesting are the things that are not done in the “experiment”.

    After moving the square and replacing it, she might have moved the square next to her that she covered into position over the square in shadow.

    Move the square diagonally to the right of the trick square into the light. I would bet that that square is white. It reinforces our idea of the thousands of checkerboards that we have seen in the past, and gives us a shade or tone that our brain can match to.

    What would this video look like to someone that had never seen a checkerboard pattern?

    After all the square shuffling, I bet that the Amazing Randi would have simply walked over and picked up the cylinder and moved it out of the light.

  11. It looks suspicious when the woman’s face morphs at 0:50, so I confirmed the validity of the illusion by hitting my Print Screen key and pasting a screen capture into Paint, then selecting and copying a section of the dark square and moving it over top of the light square. It’s real!

  12. Unfortunately, this video was digitally faked (or enhanced to exaggerate the effect). Using a color sampler the tile in the center is measurably about 8% lighter than the darker-looking tile on the edge. When she moves the tile its brightness does, measurably, change.

    In the still frame with the arrows claiming that they’re the same color, they are not in fact the same color. They’re still off by about 8%.

    If you have a Mac you can verify this by opening “DigitalColor Meter” from the Utilities folder inside Applications. You can also use that tool to verify that drawing of this effect at MIT’s web site is the real thing.

    This fakery can be done using a real model and a professional color correction tool (the kind used in movie post-production). Those allow you to draw believable shadows, adjust brightness, and change color on real-world objects.

    The real illustration at MIT does have to be one of the coolest optical illusions I’ve seen, however — and there the tiles really are the same brightness and color.

    1. According to the Youtube uploader of the video:

      “the trick is ..that when you move the square you are moving it WITH the shadow still printed on the square.”

      1. Mtran: That would explain how they “moved” the tile while keeping it the same shade, but also means he’s admitting to doing some digital manipulation: the shadow was printed on digitally.

        The fact is, the square is 8% lighter when it’s on the center than it is moved to the side, and in the still frame claiming they’re the same color they were not actually the same color. The one that looked lighter really was lighter, at least in the image being shown.

        1. Yeah, at first I was puzzled by the stability of the color and thought it was just a nice trick, though not interesting enough for me to think about it further. Then the questions and explanations here turned out to be more fun than the video!

          The person who uploaded the video wasn’t very clear with his explanation but it seems that quite a few people on this thread (such as yourself) were able to figure it out on their own. Good work!

    2. In the film industry, the phrase “practical effect” means an effect produced on-set physically (using props, lighting, etc.), without computer generated imagery. The video is a “practical” with no effects applied after recording. If that leads to the 8% you measured, OK, but the point is if you stood on-set where the camera stood, you would see what the video shows.

    3. see my notes above…I believe it is the board which is altered, but appears unaltered due to the shadow. The squares are the same w/ that explanation.

  13. Is there a connection between the YouTube legend that is brusspup and Edward Adelson? Are they the same person?

  14. John Sadowski shows how to make “color afterimage” videos at His most famous one is “Spanish Castle” but I’ll post an example that’s even more vivid.

    The first half of the video is false color (from 0:04 to 0:31). The second half of the video is physically black-and-white (from 0:32 to the end at 0:59). But if you’ve been staring at the dot, and you continue to stare at the dot, you’ll see the second half of the video in color (with a blue sky, etc.).

    h/t: 4chan

  15. Aren’t illusions like this proof that we were not Intelligently Designed? Wouldn’t the omnipotent Creator of the Universe leave bugs like this out of the system, just by simple fiat, and we would see things as they really are?

    1. A person suffering from “subjective tinnitus” (perceived ringing in the ears) may be treated by tinnitus masking (constant application of another sound, like filtered noise, sometimes by in-ear devices). I’m not very familiar with it, but I think that in some cases: 1) The masker can block the ability to perceive the tinnitus, and 2) Our perception may be able to process away the constantly-applied masker. But I’m sure the subject is more complicated than what I just wrote.

      1. Also, I haven’t read this journal paper firsthand, but in the 1980s, researchers programmed a Yamaha TX816 synthesizer to sound like a concert piano, then they had classical pianists play it with a weighted keyboard (where “weighted” means the keys have mass to feel like piano keys). When the researchers changed the amplitude of attack in the sound, the pianists felt the action of the keyboard changed physically, even though the keyboard was not able to change physically.

        In the McGurk video, vision altered hearing, and in the TX816 experiment, hearing altered sensation. We could look for more examples between the five senses.

  16. I will try to describe why this should not work.

    The checker board photo illusion works because, although the two squares *would* be different shades under *equal* lightning, with the shadow in place, they become the same shade. When we compare the shades, side by side, exactly as they appear in the original photo (ie, a dark square in the light, and a light square in the dark), we can see they are in fact the same colour.

    Fast forward to the video. Those shades may very well be identical, but when the square in the shadow is moved into the light, it should not be the same colour. What I mean to say is, they cannot be equivalent shades under equal light AND when one is behind a shadow – that is the paradox.

    So, while it’s really neat that they are the same in the lighting, it only proves that it is the “wrong” shade in the shadow. It makes me wonder what colour/shade the adjacent squares actually are.

    1. It’s starting to become clear that the video is genuine, and that it’s done with a series of layers of trickery.

      At its heart is the original, static, two-dimension Adelson checkerboard. That is, the checkerboard you see in the video has a painted-on shadow. The cylinder on the left is not casting (much of) a shadow, because the light behind it is aimed in such a way that it only illuminates the top part of the cylinder. Instead, the scene is evenly lit by the softbox to the right.

      See here:

      for what the checkerboard would look like without the cylinder, in even lighting, etc.



  17. Recently, by coincidence maybe, I have seen a lot of ‘arguments’ (well…) along the lines of ‘sensorial relativity’ or how our senses (or science) do not tell us anything about the ‘real’ world, but that it’s all imagination, models in our head. Or that we need ‘faith’ in our senses/science as much as faith in god because it’s all in our interpretation.

    To people who espouse those views I always want to say, if you think science isn’t real, climb to the top floor of the tallest building near you. Open a door to the outside. Science will tell you you will fall to your death if you step off the ledge. Is science really wrong? Then take the plunge, prove science wrong.

    Or, do your senses tell you you see a hammer lying there? Do you feel the heft when you pick it up? Do you believe this is just a model of reality interpreted by you? Then go ahead, beat yourself over the head until you pass out. It’s not really pain you’re feeling and you’re not really unconscious – it’s just a model of reality.

    Stupid arguments these are.

  18. I tried masking the image with two 3×5 cards, leaving a “strip” with parallel sides just wide enough to see both of the squares in question. It seemed to take a few seconds, but the two squares then did seem to be identical in color. I’m convinced the demonstrated illusion is real, but it is a bit disturbing. Luckily, as far as I know, there are no circumstances where this effect could put a person in danger. I suppose that there must be some survival advantage in this, but I’m clueless as to what it is.

Leave a Comment

Your email address will not be published. Required fields are marked *