Culture and Perception, part II: The Muller-Lyer illusion

Another post from our holiday collection of oldies but goodies.

The first post in the series dealt with Nisbett's findings on different patterns of attention in Asian and Western cultures, and I talked a bit about how certain differences are more likely a priori than others. I mentioned that we cannot expect people to differ too much in being able to perceive, e.g., orientation, because it's difficult to imagine a functional visual system with orientation sensitivity. There are no visual environments without orientation. On the other hand, there is some variation between visual environments along other lines, and it would not be completely surprising to find that it causes differences in certain aspects of people's visual perception. An obvious example is in the perception of faces: in some Western environments people relatively rarely encounter Asian faces and in some Asian environments it's the opposite. There is a well-documented handicap in Europeans in the identification of Asian faces, and vice-versa (it's called the "other race" effect, holds for other populations, and is possibly the single greatest source of racist jokes). It's an interesting topic, but I won't discuss in today's post, saving it for some other time.  Instead, I will deal with less obvious sources of variation: depth clues.

Most readers have probably seen the Müller-Lyer illusion. It's a Psych 101 staple that dates back to 1889. Michael Bach has a page devoted to it on his (fantastic) website, here. Here the illusion is in its standard version:

I'm counting on the reader perceiving the figure with the outward-pointing arrow as longer. I probably won't kill the suspense by revealing that the two segments are actually of the same length, that's what makes it an illusion.



The classical explanation for the Müller-Lyer illusion is that the outward-pointing figure is perceived as longer because it is also perceived as further away (Gregory, 1968). The figure below attempts to explain why, in vintage Wolfenstein 3D graphics for the nostalgic:



On the left-hand side they walls appear to form a convex corner, on they right-hand side the corner appears concave.  The contours outlined in black are the same figures that appear in the Muller-Lyer illusion. Just like in the Muller-Lyer illusion, the segment where the two walls meet appears longer on the right (again, they are actually the same length). The walls on the left seem to pop up towards us, while those to the right recede. Our visual system appears to conclude that the segment on the right must be longer, because it is the same size as the other on the retina, yet is further away in space.

This is the basis for the seeing-in-depth explanation for the Muller-Lyer illusion, first tested experimentally by Richard Gregory (1968), who in turns attributes the theory to Thiéry (1896). That theory explains why we are subject to the Muller-Lyer illusion, but not why we should see depth where there isn't any.

A classic study by Segall et al. (1963) deals precisely with that point. Their "carpentered world" hypothesis is that we see depth in the M-L illusion because we live in worlds full of right angles (in buildings, furniture, etc.). When right angles project on the flat surface of the retina they give rise to M-L patterns: lines and arrows. A heuristic is at work, but that heuristic only makes sense in environments with many right angles. In cultures with "non-carpentered" environments, the heuristic is pointless and the illusion should vanish.

The authors set out to test the M-L illusion on a wide variety of populations (17 groups in total), with the help of a team of anthropologists. They had people adjust the length of a M-L segment until it matched – perceptually – the length of the other one. The so-called Point of Subjective Equality is the adjustment needed to make the two segments look of equal length, and summarises the strength of the illusion.

The data seem to support the "carpentered world" hypothesis. Europeans and Americans were the most susceptible to the illusion, and Kalahari hunter-gatherers among the least susceptible. They also point to wide variation in susceptibility to the illusion, across populations and age groups.

The data can be interpreted as proof of strong cultural influences on perception. However, the causal link is indirect: material culture influences the visual environment, which in turn impacts the visual system. The M-L illusion is no evidence for a link between visual perception and the symbolic aspects of culture, its particular contents or structures. It is strong evidence for something far less controversial – that what you see influences what you perceive.

[I glossed over a lot of problems with the seeing-in-depth explanation and the carpentered world hypothesis. More on that some other time].


Gregory RL (1968) Perceptual illusions and brain models. Proceedings Royal Society 171:24279–296. Available here.

Segall, M., Campbell, D. and Herskovits, M. J., (1966). The Influence of Culture on Visual Perception. New York: The Bobbs-Merrill Company.

There's a delightful article by Deregowksi that discusses the Segall data and a lot of other studies:

Deregowski, J.B. Real space and represented space: Cross-cultural perspectives.  Behavioral and Brain Sciences 1989 12 (1): 51-74.



  • Olivier Morin
    Olivier Morin 9 February 2009 (11:49)

    Thank you Simon, nice post! Researchers working on cultural influences on perception, like Richard Nisbett are actually quite ready to admit that cutural influences on perception may be due to differences in the physical environment. In a paper published in 2005 in TICS, Nisbett and Miyamoto show that their \”holistic perception\” effect can be replicated on non-asian subjects simply by exposing them with pictures of asian streets. I agree with you that it weakens considerably their strong claim, namely, that culturally transmitted ideas and social practices influence low-level perception in a non-trivial way. But something bothers me. I have tried to make that point repeatedly with many supporters of relativistic views of cognition, and they are quite unimpressed. They do get the point, they do admit that \”cultural\” effects may just be caused by the shape of the external environment; they just don\’t see the problem. The distinction between direct and indirect effects of culture is deemed artificial – and my objections are seen as typical of a philosopher (that is, immaterial). Did you have similar experiences? Why do you think cultural relativists are unconvinced by this kind of argument? and what could we do?

  • Simon Barthelme 9 February 2009 (20:32)

    To be honest the last time I had to deal with strong cultural relativism was some years ago when I took a course on the sociology of media. The person who taught the course spent the first session discussing the relativity of all knowledge and the perils of scientism, and in the next session he claimed to have written “the definitive work” on a certain problem. Oh well. I’m surprised you find that some people don’t care how cultural differences arise. As far as I can tell most people seem to care very much about the origins of differences between genders, for example. Are they all representational in nature? Or do differences arise because our societies build effectively different environments for men and women? I’d point to that and to other debates (such as the one on intelligence, if you want one that’s really messy), they’re more or less guaranteed to get your the attention of your audience.

  • Ilkka Pyysiäinen 13 February 2009 (10:30)

    As many of you may know, Jerry Fodor used the M-L illusion as an example of the informational encapsulation of modular input systems, which supposedly testified to the fact that not all perception is theory-laden. Bob McCauley and Joe Henrich argue against this view in the following paper: McCauley, R. N. and Henrich, J. (2006). “Susceptibility to the Müller-Lyer Illusion, Theory Neutral Observation, and the Diachronic Cognitive Penetrability of the Visual Input System,” Philosophical Psychology 19, 79-101. There is a link to this paper on McCauley’s website:

  • Olivier Morin
    Olivier Morin 13 February 2009 (10:46)

    Most proponents of modularity today admit that modules may be acquired or modified through education; reading is an obvious example. Innateness has been dropped from the defining features of modularity since Fodor. A reading faculty, not being innate, wasn’t modular from the start; it was, and to a certain extent remains, informationally open. Still, it makes sense to say that reading is to a certain extent automatic, encapsulated, etc. For the same reason, I am not convinced by McCauley and Henrich’s contention that cross-cultural data argue against modularity of the ML illusion. It argues against an outdated notion of modularity that few people today endorse.

  • Ilkka Pyysiäinen 13 February 2009 (17:56)

    Well, if there are modules, it is only obvious that modularization can also be the result of learning. For example, being a grand master in Chess does not seem to correlate with general intelligence etc.; yet it is not of course “innate.” The very concept of ‘innateness’ seems to be preferred by some evolutionary psychologists but rarely by biologists. It is notoriously vague. The moment of birth is not a magical boundary line and genes are expressed through a developmental process that begins in utero (the “evo-devo”). I do know the more recent arguments about modularity a la Carruthers (2006) and others but I am still confused about the fact that the (massive) modularity of mind is regarded as somehow self-evident, while it is extremely unclear what it means for the mind to be “modular.” Saying that the mind is modular only to a “certain extent” is not very helpful. The question is to *what* extent, or rather, in what sense? One option is the so-called micro-modularity (sorry, I don’t have the reference now): it may be that cognitive processes are modular (in a basically Fodorian sense) only at some very low level which does not correspond to the concepts of folk-psychology.

  • thinking difference 16 January 2010 (03:28)

    Very interesting discussion. It could be a very good way of introducing social construction of reality theory to students. It’s a neat way to make us aware of how we learn to see things in relation to our contexts. And it seems related to an experiment in a computer science department that tested our recognition of perspective in virtual environments. There was a long discussion about how people who grew up playing video and computer games see perspective in virtual environments in quite a different way than non-gamers.