Friday, August 25, 2006

What is illusion?

Here's an interesting (new-to-me) proposal implicit in a recent talk from Alex Byrne: illusions are when perceptual experience itself misrepresents reality (as opposed to when we draw false conclusions based on our experiences). This then gives rise to what we might call the illusion test: if misjudgments about X are not illusions, then X-judgments are not built into the content of experience itself.

Since all kinds of standard perceptual misjudgments -- say, mistakenly thinking that a facade in Fake Barn Country is a real barn -- are not really illusions, it follows that the contents of experience are rather more "thin" than sometimes supposed.* They involve basic things like colour, shape and length, but not such concepts as "barns" or the like. Your experience does not represent that there is a barn in front of you. It just represents the more basic visual qualities, and it gets all them quite right. The error comes later, when on the basis of this experience, you infer that there is a barn in front of you.
* = Byrne hence argues that the contents of experience have little to offer epistemologists. It's an interesting suggestion, but I won't discuss it further here.

Compare this to the Muller-Lyer illusion, where your experience (mis)represents that the two lines differ in length. The error is in the experience itself, rather than any further judgment. Indeed, the illusion persists even when you know the truth. Perceptual representations are hence informationally encapsulated from our reflective judgments or beliefs. The latter fail to correct the former. This might provide another argument for the "thinness" of perceptual content, since you probably will not continue to see the facade "as a barn" after learning the truth. Correcting the belief is enough. There's no enduring perceptual misrepresentation here.

These claims spell trouble for all sorts of proposed "thick" perceptions. Consider those who think that we "see" mental properties (e.g. another's happiness), or moral properties, etc. The illusion test flatly contradicts such proposals, since there is clearly no illusion involved in the relevant misjudgments (and nor do they persist after belief-updating). They go beyond the basic visual qualities presented in experience.



  1. Illusions are when perceptual experience itself misrepresents reality (as opposed to when we draw false conclusions based on our experiences). This then gives rise to what we might call the illusion test: if misjudgments about X are not illusions, then X-judgments are not built into the content of experience itself.

    Isn't this circular, or am I missing something? Are there indendent criteria for distinguishing illusions from non-illusory errors that don't alreayd appeal to whether it's built into the content of experience itself?

  2. >Illusions are when perceptual experience itself misrepresents reality (as opposed to when we draw false conclusions based on our experiences). This then gives rise to what we might call the illusion test: if misjudgments about X are not illusions, then X-judgments are not built into the content of experience itself.

    that requires a objective definition of what set of information is sufficient to consider it to "represents/misrepresents reality". I dont think there is such a determination so the debate is unstructured and the conclusion arbitrary.

  3. Isn't Byrne's proposal similar to Descartes' theory of the material falsity of some ideas?

  4. hmm; i have a similar worry to brandon for this test. If we rely simply on our pre-theoretical judgements about what is illusory there may be some cases we pre-theoretically judge as illusions that would favour the rich content guy; on the other hand, if we're just introducing a technical notion of 'illusion', it's open for the proponent of rich content to simply claim that the facade really is illusory (and hence the circularity worry).

    But suppose that our pre-theoretical judgements go the way you want (and I'm inclined to agree you've put some pressure on the rich content guy here). With respect to the consequences for what we can see, there I'm not so sure. It would spell trouble for that if we could see just what is represented by our experience, which seems reasonable. But that would conflict with our pre-theoretical judgements about what we see, which suggests some tension in the argumentative strategy. Put bluntly, why should pre-theoretical judgements about what is illusory carry more weight in determining experiential content then those about what we can see?

  5. Brandon - as Russ suggests, I'm appealing to "our pre-theoretical judgements about what is illusory" here.

    Russ - do you really think there are intuitive cases of rich-content illusions? I'd be very interested to hear one.

    "why should pre-theoretical judgements about what is illusory carry more weight in determining experiential content then those about what we can see?"

    That's a nice point. Though I personally don't have the intuition that I literally see rich content stuff. The word "see" has a range of uses -- at the most liberal extreme, we sometimes use it just to mean anything we know or understand ("oh, I see"). Strict "seeing" is restricted to the contents of perceptual experience, which I take to involve only appearance qualities (shape, colour, etc). In between, perhaps, we have the idea of judgments that we make on the basis of visual perception. When I say, "I see that Bob is happy", I really mean something more like, "I see the appearance as of a happy Bob, and thereupon infer that Bob is happy."

    (Some cases might also be explainable as mere de re reports, e.g. the ancient astronomers saw other planets, but they didn't see them as such, i.e. they didn't see that certain celestial objects were planets. The "seeing planets" report merely indicates some non-representational fact, e.g. the true cause of their visual experience.)

    Those explanations sound a lot more intuitive to me than the suggestion that facades cause me to undergo barn-illusions. But your mileage may vary.

  6. Would our tendency to see things as three dimensional on a flat screen (e.g., a movie) be counted as an illusion (it certainly doesn't appear to be a mere misjudgment)? If so, illusions still cover a lot of ground. If not, however, we seem just to be artificially restricting what counts as an illusion.

    Seeing the appearance of a happy Bob is also still rich content. I suspect, in fact, that this is one of the clear cases where we can't thin down the content very much -- we do seem to perceive faces, as such. (And the illusion criterion itself would seem to suggest this, since there seem to be instances, like the hollow mask illusion, where the illusory effect for faces is harder to shake than it is for non-faces.)

    Likewise, I'm not convinced that the illusion test tells us that we can't strictly see emotions, because there are cases -- like fear inducement as a result of artificial stimuli that are recognized as artificial -- that could be regarded as emotion-based illusions. So either the illusion test lets these past, or there are illusion-like non-illusions, and we need to have independent criteria rather than pre-theoretical intuitions.

    The illusion test becomes especially interesting when we stop thinking of standard optical illusions and start thinking of multimodal illusions. The McGurk illusion is a good example: if we see someone pronounce /ga/ but the sound produced is /ba/, we hear it as a /da/, which is intermediate between the two (and so on with other sounds) The illusion is as stable and persistent as any other illusion; and this would suggest that we can't assume that the illusion test will get us a strict empiricism, i.e., that it will give us the result that we only directly sense appearance qualities. The connection between phonemes heard spoken and phonemes seen spoken is not an appearance quality, but it passes the illusion test with flying colors: according to the illusion test, we directly sense the connection between phonemes heard spoken and phonemes seen spoken.

  7. Yeah, I think flat images can give rise to illusions of depth. I'm not sure about the relevance of hollow-mask illusions -- surely they involve misjudgments of shape, concavity or 3-D structure. That they're caused by face-like shapes doesn't mean that we misjudge them as faces. The idea of a "face-illusion" seems no less odd to me than a "barn-illusion".

    (Note that by "appearance of a happy Bob" I meant merely to denote (de re, not de dicto) the appearance qualities that Bob has when happy. I didn't mean that the concepts of 'Bob' or 'happiness' [or 'face'] themselves featured in the representation.)

    Induced emotions are presumably something we feel, not see, so I'm not too sure how they fit in here. Do you mean that they might cause us to misjudge that a situation is dangerous (even when we consciously know better)? Perhaps something similar occurs in irrational phobias. I guess there could be "illusions of danger" in this sense, which does seem a relatively rich sort of representation? Curious.

    I don't follow your McGurk illusion argument. Doesn't it merely involve misjudging the heard sound? (On the visual side, I don't think there are any "phonemes seen spoken", but only various (facial) movements and whatnot.) I take it you want to claim that we judge that there is a "connection between phonemes heard spoken and phonemes seen spoken". But it's not clear to me that I would make any such judgment. The assumption of such a connection might drive some of our cognitive processing, but that doesn't mean it's ever explicitly in a representation.

  8. Clearly the hollow-mask illusion does involve shape; and you can create illusions with other hollow-shapes. However, this is not, it should be pointed out, inconsistent with the illusion being partially independent on the fact that it is the shape of a face; and from what I understand (it's always possible that I misunderstand, of course) there are quirks of the illusion (e.g., in how we estimate distance for motor functioning) that are persistent regardless of our judgment that occur when it's a face-shape rather than some other shape. In any case, even if I am misunderstanding, my point (related to the overall moral of this comment, found at the end) is that the illusion criterion itself doesn't obviously rule out such possibilities. For all we know a priori, there may be such illusions; we can only determine this by studying perceptual experience.

    I may have explained myself poorly in the fear contagion example. The question is whether we perceive emotions; the feeling of fear is merely a sort of marker. The real illusion is the persisting inability to shake the perception (taking that in a neutral sense that doesn't prejudge the issue) of fear in others, so that regardless of one's actual judgment about their fear, one begins to feel fear in sympathy.

    The only way I can see around this is to say that the fear contagion from afraid faces recognized as only artificially afraid is not due to any direct sensing of fear but due to 'appearance-qualities' that are directly sensed. But to make this distinction we have to have (ex hypothesi) already used the illusion criterion to put fear on the not-directly-sensed side of the distinction, i.e., before we can apply the illusion criterion properly to perception of fear, we have to have ruled out direct perception of fear, which we do by applying the illusion criterion properly. So either we are going in a circle or we are smuggling in other assumptions about what is directly sensed (or perhaps we are smuggling in assumptions that make 'illusion' more precise than the pretheoretical notion is).

    On the McGurk illusion, I'm not claiming anything about judgment. The phoneme seen / phoneme heard illusion has been shown to persist regardless of our judgments of the connection between the two or of what phoneme was spoken. Which makes it a genuine illusion. Thus the conclusion seems to be that we directly sense, independently of any judgment, a link between seen-phoneme and heard-phoneme or (perhaps?) that spoken phonemes are a genuine part of direct visual experience. (And note that if it's the latter, it's difficult to see how we could do so if we our direct visual experience did not include mouths to speak them.) As an example the McGurk illusion also has the advantage of being a surprise -- who would have supposed it beforehand? Certainly not you and certainly not I. It's a case where the illusion criterion gives us a surprising result.

    The overall point I am getting at, in a very roundabout way, is that we can't assume that the illusion criterion gives us the result that only appearance-qualities (shape, color, etc.) make up the content of direct experience. It can only do so if no misjudgments about X are illusions, whenever X is anything other than an appearance-quality. But the McGurk illusion, in particular, seems clearly to be a case where a non-appearance-quality has misjudgments that are perceptual illusions; and the face (recognition of a face, not recognition of specific faces) and fear cases are at least borderline as it is, and may have genuine cases that would pass the illusion test. Such cases are either the content of direct experience; or the illusion test, using a pretheoretical notion of illusion, is not adequate for ruling out non-appearance-qualities.

  9. Three points that I intended to make in the previous comment, but didn't (apologies for having to tack them onto the end here):

    (1) The real issue with the hollow mask is that our unshakable tendency to see certain shapes as (convex, normal) faces even when we know that we are just viewing a hollow mask seems well-supported; and this certainly looks like what we would call an illusion.

    (2) When we recognize what is going on in a movie, it's not enough to say that we have an illusion of depth; we have an illusion of discrete objects as well. This was why I said that accepting something like that as an illusion actually gives us more than we might expect.

    (3) Berkeley in his works on vision had already pointed out that when we have learned a language the suggestion of the meaning is irresistible. This means that there can be illusions about the meaning of lines or sounds that (1) are genuine illusions, because they persist whatever our judgments about the lines or sounds; but (2) are nevertheless learned. If this is so, then the illusion test might well rule out very little at all (and, as Berkeley notes, seeing shame in a face is very similar to seeing meaning in the marks on the page).

  10. I think my statement of the illusion test may have been unclear. What's crucial is that our (de dicto) misjudgment that X is the case constitutes an illusion. It is not enough to be "about" X in the looser sense that the illusion is distinctively caused by Xs. (For the latter may be so without X ever figuring as such in the conceptual content of our mental representations.) So the hollow mask illusion is the wrong kind of illusion for your purposes; what you need is strictly a face-illusion, i.e. where one's misjudgment that they are seeing a face constitutes an illusion.

    Your elaboration of the fear contagion case sounds interesting -- I hadn't heard of it before. It does sound like there are encapsulated mechanisms drawing mistaken "conclusions" in some sense. But are they really involve explicit experienced representations? That is, do the subjects persist in saying that the others look fearful? Or is it merely a kind of (exclusively) sub-personal registration? If the latter, it does not seem to meet the intuitive criteria for illusionhood.

    Similarly with the McGurk case, I'm still not clear on what exactly is being misrepresented. We might "sense" (or register) some fact about X in a de re sense, without the concept of X entering into our conscious representations. It's the latter that matters here.

  11. Hmm, I guess it's not really the misjudgments themselves that constitute illusions, but rather whatever external stimulus is causing the misjudgment. But hopefully my intended meaning was clear enough...

  12. Thanks, Richard; that clarifies a lot; you're right that the hollow mask illusion wouldn't pass this test. It seems to me that it becomes less clear that the illusion test is workable -- how would one distinguish between having a direct and persistent experience of faces and having a direct experience of shapes that persistently are faces? I.e., how would we distinguish between a misrepresentation that X, which is clearly Y, exists, and a misrepresentation that X is Y, without begging the question? I'm also not convinced that we can set aside the causes so easily if they also enter into conscious representation (since in such cases they seem to be just more rigorous descriptions of the experience -- if I have the colored Muller-Lyer, and I find that the illusion is affected by visible lightness contrast, then it would seem that a full account of what I'm actually experiencing directly has to include lightness contrast; if I see convex, normal faces whenever there is a face-shape, even a hollow one, and its being a visible face-shape makes a difference, then it isn't clear how I can give an accurate account of what I directly experience that does not appeal to the fact that I see a face-shape). But that seems to be a secondary issue.

    The McGurk case, though, I think, still stands, even if we were to interpret it as a plain auditory illusion. I see someone pronounce /ba/, but the sound that reaches my ears is /ga/; thus I hear /da/ even if I know that what I'm seeing is /ba/ and what I'm hearing has to be /ga/. (And so on with other phonemes: we hear phonemes intermediate between the two when there is a disparity.) So (1) There is something consciously represented (/da/); (2) it misrepresents both the phoneme seen (the viseme /ba/) and the sounded phoneme (/ga/), whichever one (if any) you take to be the key one; (3) it persists regardless of our judgments about the actual phonemes. However precisely we may cash it out, phonemes of some sort can't be ruled out as the content of experience by the illusion test. (If this doesn't qualify as an illusion, it at least shows that our pretheoretical notion of illusion is not adequate for the illusion test, and that we need a set of precise, independent conditions.) The only way I can see of getting around this is to show that there is nothing special about phonemes in the illusion; as far as I am aware this is not true, but I don't know how well it has been studied.

  13. question is: what is X really?do we have a way of really knowing what X is?i think the reason why illusion occurs is because there are certain things that influences what is perceived/experienced.and influences are something that unfortunately, could not be totally stripped away even how hard oen tries to.

  14. Brandon, I'm inclined to agree that the McGurk case involves illusory phonemes. They don't seem to be an especially "rich" form of content, do they?

  15. They're certainly not rich in the sense that say, the meaning of words is rich content. But the serious question here is whether (as seems at first glance to be so) the McGurk illusion is really phoneme-specific. And if so, then it seems to me that phonemes would be rich content in the sense that phonemes are not what we'd usually consider basic auditory qualities. And the question, of course, is how many other illusions are there that would introduce contents that are not what we would usually consider basic sensory qualities.

    When people say our experience has rich content, after all, they usually aren't saying that everything in our understanding of the world is found immediately in our experience, even if they are a bit vague about where the line is. What they are doing (I take it) is arguing that the distance between the contents of direct experience and our robust understanding of the world is much smaller than we usually assume empiricist principles can allow it to be.

    To use the McGurk case as an example, if phonemes, distinctively, are contents of direct experience, then the contents of direct experience nudge up much closer to spoken language than one might have thought originally. Something still has to be learned -- patterns and rules of usage. But that would be just about it; so (in only a slightly looser sense) we could say that spoken language as such is directly experienced, since we directly experience its matter (phonemes) and can build up an acquaintance from direct experience, by habituation and memory, with its form (syntax and usages). Contrast that with what follows from the supposition that phonemes are not distinctively parts of our direct experience; one would have to work much harder to connect language to what we directly experience.


Visitors: check my comments policy first.
Non-Blogger users: If the comment form isn't working for you, email me your comment and I can post it on your behalf. (If your comment is too long, first try breaking it into two parts.)

Note: only a member of this blog may post a comment.