Thursday, 29 April 2010

A Note on Holt on Visual Illusions (Heft, 2001)

This is a visual illusion. When you look at it, there appears to be a great deal of motion as things rotate in various directions. But there's no actual motion; this isn't an animated display and you can verify this by fixating the centre or touching your computer screen - no motion. I just fooled your visual system.

But Andrew, you say, you've been talking all this time about how great perception is! It's direct, functional, reflects experience, and lawful: if perception is all these things, how can it be fooled so easily by some pixels? Direct realism surely cannot handle this: perception is flawed, and readily shown not to correspond to the way things really are. Clearly this James Gibson is a mad man and we can go back to 'knowing' that perception is clearly mediated by pretty good but fallible representations.

You're in plentiful (if not especially good) company. This is a common way to dismiss ecological psychology and it does need to be addressed. Fortunately, it has been, extensively, so it's time to remind the representational camp of exactly why they are wrong, again. Holt has a few choice words on this, in a chapter from The New Realism entitled 'The Place of Illusory Experience in a Realistic World'. Gibson will want a word as well, but I'll leave that to the ongoing Gibson (1979) reading group. I'll focus on Holt because a) he went after this obvious objection to a realist ontology early and hard, and b) it was discussed in the chapter from Heft's book on Holt - I left it from there out because it's worth a post by itself.

Remember, James, Holt, Gibson et al are all engaged in a very specific game, namely redefining the metaphysics underpinning psychology. When you're working at this level, the first step in any counter argument is to examine the metaphysical assumptions of the opposing claim: it is often that case that you can simply define the problem away. This is essentially what Holt does here.

The excitement about visual illusions comes from a simple dualistic assumption: there is an objective true state of affairs, and our perception of this, and that in the case of illusion, these do not correspond. The perception is therefore in error and illusions reveal the assumptions mental representations are making in order to be of any use. But James and Holt and the rest of the New Realists have been making that case that experience does not reside 'in the head'; it is a functional relation that we enter into. Correspondence ceases to be a problem because there are no longer two things (world and represented-world) to be placed in alignment; there is only the relation.

Visual illusions, then, don't reside in the head: they live in the relation between the perceiver and perceived. To quote Heft's summary:
Perceptual errors and illusions are real parts of experience. Illusions arise under particular specifiable, objective conditions. The problem is not one of correspondence, or a lack thereof, but one in which circumstances lead to a lack of coherence or consistency within perceptual experience (Heft, 2001, pp 80).
James, Holt and Gibson all consider experience to arise from a dynamic process of selection, extended over both time and space. From this perspective, the illusory status of the motion in the above image is revealed by inconsistencies in the perceptual experience of it: the motion vanishes if we fix our eyes, we cannot feel the motion if we touch the image, we can continue to explore the computer display showing the image to confirm that nothing is changing, etc.

But surely this implies that the perception of visual motion is still an error, as revealed by further exploration? Not so, says Holt. Exploration has simply revealed the limits of the scope of the true visual experience, which, for a realist, is always occurring in a broad context. Take the commonly cited case of the straight stick that appears bent when placed in water; this appearance is surely an error. For Holt, the error is with the analysis: psychologists fail to notice that the context of the experience has changed, which has naturally led to a different visual experience. One case (when the stick is out of the water) leads to the true visual experience that the stick, when viewed under these circumstances, appears straight; the other case (when the stick is half in water, half in the air) leads to the equally true but different visual experience that the stick, when viewed under these circumstances, appears bent (a consequence of the behaviour of light as it moves across boundaries between media). Recall the account of the man walking past the window; context is everything.

All visual illusions depend on a trick to work. The Ames Room, for example, is a trick of static perspective depth information that only works from one location; moving from there breaks the illusion, although it is compelling when you're in the right place. Visual illusions are getting more compelling: there's a regular competition at the Vision Sciences Society meeting to show of new illusions, with the emphasis on making them more and more stable over wider and wider viewing conditions. But because they all depend on a trick, they don't actually tell us about vision in the wild; they do reveal certain boundary conditions, but no more. Given the opportunity (i.e. the space and the time to explore) they are revealed to be illusions when they lead to inconsistencies in our experience. This is a necessary condition for perception to succeed, and Gibson will insist on allowing motion and exploration to be fundamental parts of the senses considered as perceptual systems.Illusions arise when a necessary condition has been violated, and are therefore not as interesting as people think for the study of perception. Indeed, Gibson will dismiss them as entirely useless because they depend on breaking the rules under which perception actually occurs.

Summary: illusions do pose an important question to direct realist accounts of perception that must be answered, but the challenge has been met; illusions create true visual experiences under artificially constrained circumstances and are only problems for a perceiver when the constraint is successfully concealed. Illusions reveal only the role a given constraint plays in allowing perception to succeed normally, but do not reveal the contents of the hypothesised mental representations that supposedly mediate vision.

Holt, E.B. (1912). The place of illusory experience in a realistic world. In E. B. Holt et al. (Eds.), The New Realism: Cooperative studies in philosophy (pp. 303-377). New York: MacMillan. Download


  1. All illusions are revealed as such with enough space and time to explore them? The rotating snakes illusion above is pretty resistant to damn near anything (although, if you get drunk enough they stop rotating, so it's likely to do with eye movements). Also, the size weight illusion is pretty much impossible to reveal as an illusion until you place the stimuli on the scales - there is no way to 'accurately' perceive the (identical) weights of the stimuli. All this being said though, visual illusions have never seemed all that damning to ecological psychology - I assumed they are treated as leftovers from nice tricks our visual system has developed to deal with more common stimuli in an easier way (like the cool direct area measurement device from several posts ago). So, rather than only representing boundary conditions (which some certainly do), plenty of illusions seem to give real insight into the cheats that the system can employ to get around excessive computation. Whether this direct perception is all that is ever employed... Well, I look forward to being convinced by further posts.

  2. Eduardo Garza29 April 2010 14:23

    Wow, the site looks nicer. I haven't been here for a while.
    I was just going to ask you about the Illusions, and it's great that you mention them.
    Did Gibsonians ever talk about Auditory Illusions as well?

  3. @Gav
    The rotating snakes aren't that resistant: touch your monitor. The visual experience is persistent because it arises from specific reasons it is hard to alter; but you know there's something weird by a) the motion elicited being a bit weird and b) exploring the apparent motion in more detail.

    The size-weight illusion is a better example of what you mean. I need to think some more, but some work by Geoff and Arthur Zhu has suggested the size-weight 'illusion' is actually you correctly perceiving the affordance of throw-a-bility. I need to get into that in more detail though - iirc it's more complicated than that.

    Auditory illusions don't really get talked about only because when Gibson was writing, visual illusions were where it was at and ecological types who came later aren't interested in illusions anyway. The principle is the same: vision and hearing are perceptual systems operating according to the same types of rules just with different energy (light vs sound).

  4. Speaking of illusions:

    Note the careful use of exploration to reveal the illusory nature of the set up :)

  5. Holt's explanation - that perceiving the stick bending in water is a true visual experience - makes me think of interpreting statistical interactions. The way we see objects always depends on the medium, but we don't notice this relationship until we compare our visual experience of the same object across mediums. Underpinning this experience is the interaction between light and medium, such that seeing something via two mediums creates a discontinuity in how we see the object. Confusion arises because it feels as though we're seeing things the way they really are under normal conditions. But, each of us spent a considerable period of time in our lives learning to reach for things using the visual information available in the environment. As humans, we perceive visual information through the medium of air. But, if that visual information was somehow made to act as if the medium was water (which, in terms of the illusion is the "distorted image"), we would have learned how to use it to guide our actions every bit as well. The illusion reveals a dependency we don't usually encounter because we see things via one medium at a time. But, the impression that there is something wrong with the bent stick is due to our forgetting that both air and water interact with light - we don't see things the "right way" via air, we see them the usual way.

    Very cool fire illusion, by the way.

  6. "All visual illusions depend on a trick to work. The Ames Room, for example, is a trick of static perspective depth information that only works from one location; moving from there breaks the illusion, although it is compelling when you're in the right place."

    Ok. I get the "trick" for the Ames Room. What is the trick for the snakes?

  7. It's apparently an example of a peripheral drift motion illusion (fixate the centre and that stops rotating, for example). Under certain viewing conditions (i.e. peripheral vision) it apparently provides information from contrast and luminance which specifies motion. This paper also suggests that part of it comes from eye movements themselves.

    Note that it's clearly a trick and readily revealed as such, because the snakes are not actually moving which can be readily established by further examination via touch.

  8. Speaking of sticks in the water: with respect to sabethg's comment about how if our normal medium was 'water+air' we might not be fazed by the behaviour of the stick half in the water: it turns out archer fish live that life and aren't fazed at all (ignore the awful non-Attenborough commentary)

  9. What about saliency of categories that differ through experience?

    I can't recall the paper right now, but I remember reading that the differing length illusion <—> vs >—< and some other stuff about parallel lines did not work with people from a nomadic tribe. Hence the assumption that it is categorical perception as shaped by living in a modern environment of geometrical symmetry in buildings etc that makes the mind susceptible to these kinds of geometrical illusions.

  10. Learning is part of the information story, so that's not a problem. The particular illusions you're susceptible to may easily be a function of experience.