Tuesday, March 03, 2009

What Is an Illusion, Exactly?

I'm confused again. Publicly expressing confusion seems to have become a (perhaps tiresome) professional habit of mine these days.

A classic example of an illusion -- classic in the sense of dating back to ancient Greece, not classic in the sense of central to 20th century perceptual psychology -- is the oar partly submerged in water. Due to (what we now know as) the laws of the refraction of light, the oar typically, in some sense, "looks bent" as it angles into the water. But one might argue (does John Austin argue this?) that that bent appearance is not really an illusion: If one knows enough about the world, one should know that an oar partly submerged in water (seen from a particular viewing angle) should look bent just like that. If it looked straight, I suppose, a longtime oarsman or a person very familiar with the laws of refraction might think the oar looked strange, might even think that it looked like an oar that is actually bent (bent in such a way as to exactly compensate for the bend a straight oar would seem to have at that angle). Perhaps part of what it is for an oar to look straight is for it to also (in a different sense) "look bent" when it is partly submerged in water. (This formulation is indebted to Alva Noe's "dual aspect" view of perspectival appearance.)

So is the skilled oarsman experiencing a visual illusion as he looks at the oar? If we say no, then I'm worried we're off onto a slippery slope to entirely denying the possibility of illusions that are known to be such. When I press gently on the side of one eye, I seem to see double. Is this also no illusion, since I know that's how things are supposed to look when I press on one eye? When I look at the Poggendorff illusion (below) I know that if the upper segment of the line going behind the rectangle really is aligned with the lower segment emerging near the bottom, they should look (in some sense of "look") offset. I've seen this illusion so much that now I think that if I saw a figure of this sort where the lines didn't look (in the relevant sense) offset I would probably infer that they really were offset. Parallel to the oar case, perhaps, what it is for the line segments to look aligned (for me) is for them to look (in a different sense) like they don't align.

(image reproduced from Titchener 1901-1905)
But if the Poggendorff illusion is no illusion for me, then is any illusion an illusion to someone who knows how the illusion works?

So maybe we should say the oarsman does experience illusion. But then there's the risk of a slippery slope on the other side, toward saying that much more is illusory than we ordinarily think. If the bent oar is illusory, it seems that looking at things through a curved glass of water must be, since the refraction similarly distorts things. And then it seems like a magnifying glass held at arm's length similarly creates illusion. But then also does a magnifying glass held near the eye? A telescope? Ordinary corrective lenses?


D said...

I would answer that telescopes, magnifying glasses and cups of water do create optical illusions, making things appear bigger, closer, or distorted. Ordinary corrective lenses are an exception because they are very carefully crafted to cancel out an illusion caused by a defective cornea.
That's my take on it as a computer vision researcher.

Pete Mandik said...

How about "systematic divergence between appearance and reality" for a working definition of illusion? I'm not impressed by the "stick *should* look bent half out the water" consideration against considering that an illusion. If it *is* straight and *looks* not straight in a systematic way, then that's an illusion.

(By the way, one way to tell that the lines are the same length in the Muller-Lyer Illusion is that the caption usually says "Muller-Lyer Illusion". But its still an illusion.)

I'm not super happy with my use of the word "systematic" here, but I'm aiming to rule out cases where for just a brief moment I, for example, mistake you for Elvis. If you aren't usually taken for Elvis by me or anyone else, then we don't have a Schwitzgebel-Presley illusion on your hands even though appearance and reality diverged for a moment.

Anyway, regarding your dilemma of the two sliper slopes, I'd be comfortable sliding down the latter one: appearance and reality diverge systematically much more frequently than naive observers realize.

Anonymous said...

These are interesting questions. Here is Austin from Sense and Sensibilia:

"Refraction again--the stick that looks bent in water--is far too familiar a case to be properly called a case of illusion. We may perhaps be prepared to agree that the stick looks bent; but then we can see that it's partly submerged in water, so that is exactly how we should expect it to look.

"It is important to realize here how familiarity, so to speak, takes the edge off illusion. Is the cinema a case of illusion? Well, just possibly the first man who ever saw moving pictures may have felt inclined to say that here was a case of illusion. But in fact it's pretty unlikely that he, even momentarily, was actually taken in; and by now the whole thing is so ordinary a part of our lives that it never occurs to us even to raise the question. One might as well ask whether producing a photograph is producing an illusion--which would plainly be just silly" (pp. 26-7).

Some thoughts:

(1) One way to read "familiarity takes the edge off illusion" is this: familiarity destroys illusions. Yet there's clearly no absurdity in speaking of the Mueller-Lyer (to take a familiar example) as a "familiar illusion." The fact that I (like many philosophers and psychologists) have become familiar with the M-L to the point I'm not taken in by it should not lead us to say that the M-L is no longer an illusion for me. Rather, it should lead us to say that I'm no longer taken in by this particular illusion. It seems, then, that if there's any sort of familiarity that destroys illusions, it must be something like cultural familiarity. Perhaps: x ceases to be an illusion when "we" are no longer misled by x?

(2) On the other hand, I'm inclined to think that the "robustness" of illusions such as the M-L is such that even if we as a culture were no longer misled by the M-L (as a result of repeated exposure), it would still count as an illusion. I say this because, even though I'm not misled by the M-L, the lines still look to me as if they are of unequal lengths. Moreover, I want to say that they not only look this way to me, but that they look this way, full stop, and so, would look this way to "us" even if we were thoroughly familiar with it. (By contrast, the stick in water doesn't look bent to me, or, I think, to anyone else. It simply doesn't look bent.) This connects with Pete's point about "systematic divergence between appearance and reality." Such systematic divergence is presumably to be explained by structural features of the illusions in question: e.g., the way the wedges at the ends of the M-L lines produce effects similar to those exploited in perspective drawing. (In fact, perspective drawing might be a good example of a kind of illusion with which we are all familiar.)

(3) Of course, as Pete notes, the notion of "systematic divergence of appearance and reality" is hardly crystal-clear. But so what? Why shouldn't there be borderline cases? And why shouldn't different things count as "systematically divergent" in the relevant sense in different contexts? Why insist on a context-independent criterion for illusoriness?

Anonymous said...

One more thought:

(4) Austin writes: "One might as well ask whether producing a photograph is producing an illusion--which would plainly be just silly." Agreed: as such, no photograph is an illusion. But we can imagine placing a life-sized photographic image in such a way as to create the illusion that a certain room is much larger than it is. The image itself would not be an illusion; the illusion would consist in the image's being placed in such a way as to give the room a look that systematically diverges from reality. (Similar considerations apply to video and film.)

Anonymous said...

Excellent question. Another nice case is that of mirrors. Does looking at a mirror create an illusion that twin-you is in front of you? When one doesn't know it's a mirror, the "yes" answer seems plausible, but this answer rings less true for the familiar case when one knows it's a mirror. But if one says "no" in this case, then the question is whether there's a principled distinction between this case and the bent stick case, or any other case of a "known illusion".

I suppose that for there to be a principled distinction, one will have to make the case that in some cases (e.g. the mirror case), knowledge of the relevant effect penetrates the experience and changes its content, while in other cases (e.g. the stick case) knowledge does not penetrate the experience in this way. Perhaps there is some intuitive plausibility to the claim that the relevant spatial phenomenology is changed by mirror knowledge in a way that it's not changed by bent-stick knowledge. But it's not clear what the source of this difference is.

Matthew J. Brown said...

I'm inclined to think that the problems you raise show that "illusion" shouldn't have much weight in philosophical arguments, to be relegated to a rather circumscribed area of psychological/psycho-physical research on perception. Our thinking about "illusion" seems to be tied up with hopelessly vague ideas about viewing under "normal conditions" (perhaps this goes back to Aristotle?). Something like what Pete and Jason (hi Jason!) suggest will probably do for our ordinary concept of "illusion," but then what's so useful about the concept.

(I am inclined to go just the opposite way of Pete and say that appearance and reality diverge not at all, when one properly understands "appearance" and "reality," and what diverges are ones expectations about what the appearances "should" be under certain conditions.)

Anonymous said...

The mirror example is an interesting one, because it seems importantly different from how we think of many of the standard cases of illusions. In the bent stick case, we say there is a stick, and it is being experienced as if bent. In the Muller-Lyers case, we say there are lines, and they are experienced as if unequal. In the mirror case, however, we do not think that there is a person, a person experienced as if a twin. Rather, it is more like a hallucination--the subject, which is being represented as a certain way, does not exist. When we learn about mirrors, it is the putative subject of the experience that changes, not the properties we represent the subject as having. Does this make sense?

Anibal Monasterio Astobiza said...

How can be explained an illusion or the difference between how thinks look like and how really are from an enactive (sensorimotor) approach to perception?

Anonymous said...

Eric--is there some theoretical purpose you need this concept to do? Is a sharp line required?

Something about error seems important to illusion. Perhaps illusions are misleading perceptual experiences. Taking the oar in and out of the water changes what we see, but there is reason to think the oar itself does not change. This is potentially misleading--one might think the oar is bending back and forth--but it need not actually mislead. Further, the misleading information is perceptually driven--things perceptually appear in misleading ways.

I agree with Pete about embracing the slippery slope towards illusion. But, again, it depends on what your purpose or worry is. Avoiding the "argument from illusion" might drive you to the other slope.

Eric Schwitzgebel said...

Thanks for all the thoughtful comments, folks!

D: I feel the appeal of that perspective. I wonder what you'd make of djc's example of mirror images, or -- perhaps somewhere between ordinary mirror images and refraction cases -- reflections on water (e.g., streaming lights).

Pete: I've been more often compared to Al Franken than Elvis, but I'll take it as a compliment if you mean *early* Elvis! More seriously, I can see going this way -- but as with D, I worry that that slippery slope goes unappealingly far, as in djc's example of mirror images or the case of light reflected off water. Are those, too, illusory?

Jason: Thanks for the Austin quote. That was indeed the passage I was thinking of. On (1) and (2): (2) seems to be the early consensus here. But I still wonder whether we might get some "illusions", like djc's mirror images, by this criterion, that don't intuitively seem to be illusions at all. Is it a distortion that the person seen in the mirror is presented as being... well, where is he presented as being? On (3): I agree that there's nothing in principle wrong with having vague cases. On (4): That's a nice analysis.

djc: Nice suggestions. As you can see, I make liberal use of your mirror case in the replies above. It's going to be hard, though, I think, to settle the extent to which knowledge penetrates the experience in the relevant way. This seems like a good candidate for the kind of dispute about phenomenology that is frustratingly intractible.

Matt: You could be right that some background assumption of "normal conditions" is playing a toxic role here. My own inclination would be, rather than discard the idea of illusion from philosophical discussion, to evaluate it on a case-by-case basis: How much is turning on a problematic haziness about these kinds of cases? In some discussions, perhaps, not much.

Martin: Nice point. In some ways that makes it similar to the mirage case. Yes?

Anibal: That's a good question! It seems that Noe or somebody must have discussed this, but nothing is leaping immediately to mind. The old Gibsonian view, of which the enactive view is a descendant, was notoriously dismissive of illusion.

Josh: "Misleading experiences" seems like a potentially helpful idea here. I came to this concern not due to the "argument from illusion" but rather in thinking about Noe's "dual aspect" view of perspectival experience, where I think similar issues arise. (Maybe even the bending of the oar can be seen as a perspectival change, or the "elliptical coin" as an illusion.) Maybe untangling my thoughts about illusion will help me in untangling my thoughts about the experience of object constancy through perspectival change....

Anonymous said...
This comment has been removed by the author.
Anonymous said...
This comment has been removed by the author.
Anonymous said...

[Sorry for the deleted comments. I was trying to collapse two into one.]

The case of the mirror-image seems to parallel that of the photograph. Echoing Austin: it would be "just silly" to say that producing a mirror-image (of, say, oneself) is ipso facto producing an illusion. Even so, we can easily imagine setting things up in such a way as to produce mirror-images that are legitimately illusory. Primary here are the cases djc refers to, in which, for whatever reason, we don't know it's a mirror we're looking at. But knowing it's a mirror doesn't necessarily destroy the illusion. Imagine a series of mirrors that produce a surprising effect: e.g., you're standing in the doorway, but hidden mirrors make it such that you appear (your twin appears) to be sitting behind a desk across the room. Something like this plausibly counts as an illusion, even if the observer knows it's achieved by means of mirrors.

These considerations suggest that whether something counts as an illusion is less a matter of our knowledge than of whether the effect (the divergence between appearance and reality) is appropriately surprising. And now we seem to be back to something like Austin's point about familiarity.

The question is then how to take this point so as to do justice to the phenomenon of "known illusions" such as the M-L. After all, I'm very familiar with the M-L, but I still find it impressive. I'd even say that I'm repeatedly surprised by its robustness. Why? What exactly makes the M-L surprising, even in its familiarity?

(Hi Matt!)

Eric Schwitzgebel said...

I agree with you, Jason, that it's not very appealing to say that mirror images in ordinary conditions are illusory, but also it seems hard to deny that there are conditions in which mirror images can be illusory. But whether the relevant difference is familiarity or surprisingness I'm not so sure. As you point out, the Mueller-Lyer illusion is very familiar and unsurprising (to some of us). There must be some kind of robustness, it seems, but how exactly to capture that...?

Richard Y Chappell said...

This is a very interesting discussion! I've added my two cents here, suggesting that a certain kind of practical competence or fluency (over and above mere propositional knowledge) seems to be key in dispelling what might otherwise qualify as an illusion.

Eric Schwitzgebel said...

Thanks, Richard! Interesting suggestion, over on your blog.

Anonymous said...

Using the word "illusion" suggests that the perception of X does not match up with the way that X actually is in the physical world. Implicit in the suggestion however is the idea that X (in this case the oar) exists as X in both perceptual "reality" and physical reality, but I would argue that the oar does not exist (as the oar) in the physical world. The "oar" is a product of your perceptual experience, and your perceptual experience would be very different if you had 4 cones instead of 3 or if you were more sensitive to very low spatial frequencies or if you had been raised in an environment devoid of vertical lines. The "oar" is constructed by your perceptual system, from the physical stimuli to which it is sensitive. In this sense, all of perceptual reality is an illusion. There is an ontological nuance that is being overlooked... there isn't one way that things are, for things (as things) don't exist in the physical world.

Eric Schwitzgebel said...

Thanks for the comment, Michael. I'm not sure I fully understand your suggestion. What does, on your view (if anything), exist in the physical world? And does it follow from your view that there are no cases of illusion whatsover?

Anonymous said...

It's something I am still trying to work out myself, but I guess I would say that the physical world consists of energy and empty space (and perhaps even the empty space isn't really empty); that's it. All of the "things" that we encounter in the perceptual realm are manifesting out of that cosmic soup of energy, but they manifest in a way that is dependent upon the perceptual system of an observer, the environmental context in which that perceptual system developed, and in the case of vision, the type of light used to view said thing. Thus, it seems odd to talk about an illusion in the sense of something appearing different than it actually is, because my argument is that there isn't one way that a thing is, ever.

If the range of frequencies that the sun emits suddenly shrank to a range of 510-520nm light, the oar that used to appear white (let's say it was made out of birch) would now suddenly appear yellow! Is that an illusion? We tend to think that the way it is now is the way it is on some fundamental level, but it only appeared white before because of the nature of the light shining on it. Which is the "real" oar by which we can judge the illusory oar? I would say that they are equally illusory in the sense that neither one of them appears as raw energy and empty space.

I'm a grad student doing work in visual perception, so it's actually quite helpful to think through this. What am I overlooking from a philosophical perspective?

Anonymous said...

I think djc nailed it. And I don't think it will be hard for psychologists to settle the extent to which knowledge penetrates the experience in (what philosophers will recognize to be) the relevant way. I think it will be surprisingly easy. Indeed, Fodor's book on the Modularity of Mind contains plenty of the data.

Of course, there may be borderline cases. As Jason points out though, that needn't be a problem.

Eric Schwitzgebel said...

Thanks for the comments, Michael and Paul!

Michael: I find these sorts of metaphysical positions very difficult to evaluate. What does it mean to say that "all there really is" is X? What are the rules or standards that govern the intellectual inquiry that yields answers to that sort of question? I'm not even sure how to start thinking about that. Colors seem easier to lend themselves more naturally to a secondary qualities type view (on which there's no mind-independent fact about yellow vs. white) than do shapes (bent vs. straight).

Paul: When it involves consciousness, in my experience, it's never easy!

Anonymous said...

I'm very sympathetic with Michael's first post (but not his second). We shouldn't confuse what we see as a straight oar with the reality of the matter. The straight oar is merely a representational model that serves us well. Seeing the oar as bent is confusing, because it prompts us to update our model, though it's not clear how exactly. Then we discover that the original model is not actually undermined by seeing the oar as bent, so we start calling the phenomena an illusion. We keep the model, and disregard the anomalous perception. Or go for the more sophisticated response: We attribute refraction to the water, and update our model of water, or even transparent substances in general.

In other words, "illusion" only arises in local relation to a particular model. In a more global perspective of reality-modelling, there are no illusions, only data that can be anomalous to the theories or models that we happen to employ. We call it "illusions" when the anomaly is irrelevant or too obscure to be bothered with.

Alternatively, the act of modelling itself can be interpreted as the real illusionism. In which case everything is illusion, like Michael argues.

As to what really exists beyond our models: How could we possibly do anything but guess? We have some models that work impressively well, but no model can explain everything, and even if a theory could do that, how would we go about comparing it with model-independent reality? The way out of this mess is not a very satisfying one: Radical agnosticism, at least as a backdrop for scenery dependent on differing degrees of faith (philosophy aims for the minimally faith-dependent, science for the most explicit and comprehensive theories we can find to believe in, religion for the most personally or socially satisfying).

Jim Hamlyn said...

Hi Eric,

Apologies, I'm very late to this discussion but I hope my contribution may shed some light.

Illusion, as you know, is an ancient concept and as such it is a tool with many uses. It can be applied skilfully or it can be wielded, like Maslov's hammer, as if it were the solution to all our philosophical ills. Like you, I'm skeptical of the explanatory efficacy of the concept of illusion in dealing with the questions of appearances.

My suggestion is this. Every instance of what we would be prepared to attribute to illusion — whether it be the stick in water, the tilted circular disk or even the conjurer's sleight-of-hand is in fact a case of sensory discrimination failure.

Our perceptual system has obviously evolved to overcome many of the systematic weakness to which it is necessarily subject, yet many of the underlying susceptibilities remain. For example, it is vital that we perceive distant things as distant things, not as things of diminishing size. However, as a species, we have gradually discovered that our sensory susceptibilities are exploitable in representational terms.

This is emphatically not to analogise seeing with picturing. Whilst we perceive an obliquely viewed penny as a tilted penny we can nonetheless efficaciously represent it with a flatly presented ellipse. It is a matter of historical fact that the numerous pictorial techniques of this kind were not obvious to our ancestors for the precise reason that evolution has been at pains to overcome the sensory weaknesses that they exploit. Perspective wasn't discovered until 1452 so it would be a serious error of reasoning to assume that vision is perspectival.

Anyway, if this brief response doesn't strike you as incoherent then you might be interested to read some more thoughts on the subject here. or here.