Kanizsa's Triangle: Illusory Contours and the Inferring of the Perceptual World

Since I said a little bit about the title of the blog, I figure I'd say a bit about the image in the banner as well. It's a "Rubin Vase," a "bi-stable" image (sort of like a necker cube, a duck rabbit, or the "Old Woman-Young Woman") in which our brains alternate which part is the figure, and which part is the ground, so that at times it appears as a vace, and at times it appears as two faces, (usually) without us seeing both at the same time. Now, since I'm lazy, and since it's unlikely anyone's reading this anyway, I'm not going to talk about the Rubin's Vase, because to explain it would require a detailed explanation of figure-ground processing, as well as facial recognition (our brains have special areas just for faces, and it turns out the Rubin Vase activates those areas). That's just too much work. But I included the image in the banner because I dig visual illusions, so instead I'll talk about a (comparatively) simpler one, and maybe someday get to Rubin's.

My thing, my groove you might say, in the study of the mind is all about the ways in which our brains build our world, and how little “we”, by which I mean our conscious selves, have any say in it. At higher, cognitive levels, heuristics and biases, unconscious processing, and “hot” reasoning – that is, affect or emotion-driven reasoning – do the bulk of the work in ways that we're all familiar with (if you don't know what I'm talking about, get into an argument on the internet and watch these biases, “hot” processes, and all the other unconscious ways in which we build the world play out in virtual time). At a lower level, though, visual illusions reveal even more interesting, and perhaps frightening, ways in which our brains build our perceptual world. They show that we don't simply take in reality and represent it as it is out there, independent of us. Instead, we meet it with a bunch of built in assumptions and rules of inference that, when conditions are right, cause us to mis-see things out there.

This mis-seeing happens at all levels of visual processing, but just as I'm more fascinated by visual illusions than I am by cognitive ones, I'm even more fascinated by visual illusions that happen early in visual processing than I am by those that happen later. One of the more striking examples of lower, which is to say earlier, level visual illusion is the famous Kanizsa triangle, first reported by Gaetano Kanizsa back in 19551 (you can read a short biography of the interesting man here). The image at the top of this post shows the classic version of this illusion in which 3 packman-like semi-circles cause the appearance of a non-existent triangle. It doesn't really have to be a triangle, and you don't have to use packmen (you can use offset lines) as these stimuli from a study by Mendola et al² show:

You can even make 3D versions. I created this one myself (based on one first demonstrated by the folks at the Human Technology Lab):

Cool, right? The illusory lines of the triangle, combined with the different shades of gray, create the perception of depth, even though there's no triangle there. I created this one myself so that I could motionify it (you might say "animate it," but I believe the technical term is "motionify"; I've heard it both ways) and show you that all this image contains is 3 diferent shades of gray with 3 black packmen on top. There is no 3D triangle:

Now you see it, now you don't.


These phenomena are called "illusory countours," and they basically show that our brain really, really wants to see objects where it looks like they should be, and is willing to put them there if they're not there on their own. And when there are a set of packmen that seem like they could all be occluded by the same object, our brain thinks there should be an object there. Bam! Triangle.

Even though this illusion was first described almost 60 years ago, vision researchers are only now starting to figure out exactly how, when, and where the brain produces it. The visual cortex, where our brains processes the information it gets from our retinas, makes up a substantial portion of the brain, and uses as much as half of its resources. Researchers divide it into several sections numbered based on how "early" they start processing retinal information. These days, researchers are debating whether illusory contours like this in Kanizsa's Triangle are created in V1 (earliest), V2 (still pretty damn early), which process basic visual information like edges, or perhaps a bit later in the lateral-occipital cortices (LOC), which respond to (and help us track) whole objects. A study just last year by Knebel and Murray suggests that the LOC which is doing the bulk of the work in producing the illusory contours³, though it's likely that the earlier parts of the system, V1 and V2, are sending it information to suggest there are edges or lines there, and the LOC is putting them together to produce a non-existent triangle. Of course, even higher parts of the visual system, such as the inferior temporal cortex which is responsible for object recognition are going to be involved as well⁴. Basically the higher parts of the visual system, those that allow you to recognize that the object in front of you is a computer, and not a fish, are taking the info from the LOC and saying, "Oh look, a triangle!"

So there you have it, your visual system is fooled from bottom to top: it sees a figure that looks like it should have lines in certain places, creates a whole object out of those non-existent lines, and then says, "Triangle!" This reveals more than a little about what the brain is doing when it gets input from the world out there: instead of just looking at what's there, it looks for clues to what should, based on its experience (both phylogenetically and ontogenetically, which is a fancy way of saying both in its evolutionary history and its lived experience), and using those clues it infers that things are there, even if they aren't. Of course, it's pretty easy for the cognitive parts of our brain to say, "Silly visual system, there's no triangle there" (though I admit the first time I saw the illusion in a version similar to the one at the top of the post, it took me a whie to convince myself that, no, there is no triangle there), but our brains are making these sorts of inferences constantly, and most of the time we're not going to catch them. The great things about these demonstrations of visual illusions is that they make this fact apparent.

1 Kanizsa G. (1955). Quasi-perceptional margins in homogenously stimulated fields. Rivista di Psicologia, 49, 7-30.
2 Mendola, J., Dale, A., Fischl, B., Liu, A., & Tootell, R. (1999). The representation of illusory and real contous in human cortical visual areas revealed by functional magnetic resonance imaging. Jounral of Neuroscience, 19(19), 8560-8572.
3 Knebel, J., Murrah, M. (2012). Towards a resolution of conflicting models of illusory contour processsing in humans. Neuroimage, 59(3), 2808-2817.
4 Sary, Gy, Koteles, K., Kaposvari, P., Lenti, L., Csifsak, G., Franko, E., Benedek, G., & Tompa, T. (2008). The representation of Kanizsa illusory contours in the monkey inferior temporal cortex. European Journal of Neuroscience, 28(10), 2137-2146.