Wednesday, September 8, 2010

Gibson on Aperture Vision

In The Ecological Approach to Visual Perception, Gibson speaks rather disparagingly of what he calls "aperture vision":
The eye is easily deceived, and our faith in the reality of what we see is therefore precarious. For two millenniums we have been told so.
     The purveyors of this doctrine disregard certain facts. The deception is possible only for a single eye at a fixed point of observation with a constricted field of view, for what I called aperture vision. This not genuine vision, not as conceived in this book. Only the eye considered as a fixed camera can be deceived. The actual binocular visual  system cannot. (Gibson, 1979, p. 281)

Ok.  So, grant that aperture vision is not genuine vision as Gibson conceives it.  Concede, just for the sake of argument, that only the eye considered as a fixed camera can be deceived and that the actual binocular visual system cannot.  Still, something is going on during episodes of so-called "aperture vision" and a scientist might well want to know what.  So, why not have a scientific study of aperture vision along with a scientific study of "the actual binocular vision system"?

Just be sure, I guess we can add, not to simply assume that what one learns from aperture vision applies to "genuine vision".  What's wrong with that?

27 comments:

  1. Gibson wasn't saying aperture vision is just a special case; he was saying it doesn't exist and the entire concept is meaningless.

    Modern eye-tracking research has pretty conclusively shown that the eye is never, ever, ever fixed. New 1000Hz eye trackers reveal micro-saccades even when you're trying to fixate something, and their purpose is to create flow.

    So actually there isn't anything going on during so-called episodes of aperture vision.

    ReplyDelete
  2. Hi, Andrew. The eye movements angle is a very interesting take on the non-existence of aperture vision. I hadn't connected them. Thanks.

    So, what do you say about experiments with retinally stabilized images? Is this a legitimate technique?

    ReplyDelete
  3. It's legitimate in that, to the best of my knowledge, it reveals the absolute necessity of motion. Place a contact lens with a dot on it on the eye, and because it moves with the eye and thus creates no flow it simply fades.

    I have recently come across the idea of glueing the eye to a small stick and holding it in place that way; I'm a little too weirded out by the idea to have gone looking for more details but as far as I know you get similar sorts of effects.

    ReplyDelete
  4. Yes, it did occur to me that a lot more needs to be said about temporal integration of information. But, I'm not sure how that favors a Gibsonian. After all, I would think you have to save the information over time in order to integrate it. And saving information doesn't sound like the kind of thing Gibson would go for.

    ReplyDelete
  5. Oh, I think that one has to distinguish the putative "absolute necessity of motion" from a putative "absolute necessity of non-constancy" of stimulation. In truth, I think that a close look at the retinal stabilization literature suggests the latter, rather than the former.

    ReplyDelete
  6. re suggesting non-constancy over motion - how so?

    For a Gibsonian, the 'integration' is done by having access to information defined over both space and time. Actually, the information is never dis-integrated, on the grounds that it is an intractable problem to put it back together again given only the pieces.

    (the internet ate another attempt to post this, so if you see two, that's why)

    ReplyDelete
  7. My understanding of the retinal stabilization stuff is that what it shows is not that the retina or eye has to move, but that there must be changing retinal stimulus.

    But, how does a person have access to what is past, if it is past? This seems to me to require some method of preserving the impact of past events. But, that sounds like recording information.

    ReplyDelete
  8. I think birds are proof positive that the *default* case of perception in the animal world (information pickup) depends on motion to some extent. They are always bobbing their heads because this allows them to extract or "sample" ecological information from the environment. Gibson famously claimed (like the phenomenological tradition of Husserl/Heidegger), that ecological information is extended in time because it is an awareness or discrimination of invariant molar properties from micro properties. This could be likened to a categorial intuition of sorts that is attuned to resonate to certain properties which are invariant overtime. The environment is always in flux. In order to use the environment to regulate and time its behavior, the animal needs to latch onto properties that are invariant over time. For example, if a bird is flying, it can extract certain invariant patterns from the rate of change on the optic flow field which are nomothetically related to the distance from the bird to the ground. Detecting this information is crucial for the success of the bird's flight. Through an awareness of the invariant properties of the changing optic flow field, the bird can navigate effectively. This awareness is both exterospecific (invariant in respect to what the animal does) and propriospecific (variant in specific ways as a function of the animals behavior). Another example is how the spontaneous, self-organized mobility of unicellular organisms allows them to be aware of differential flow fields in nutrient gradients, a primitive kind of affordance-based regulative process.

    Gibson's point is that ecological information (opportunity for action) is temporally extended in time. Coming back to the birds, they bob their heads because the extraction of meaningful information only happens during the retinal transformations (because the relevant information being sought is exactly the differential rate of change over retinal transformation, since that is when you can detect the relevant affordance properties which are invariant over time or space).

    ReplyDelete
  9. With that said, I want to respond to something Ken said in the OP. He said:

    "Still, something is going on during episodes of so-called "aperture vision" and a scientist might well want to know what."

    Definitely. Gibson never denied this (See his chapter on perceptual illusions in "The Senses Considered as Perceptual Systems"). Something is still going on in aperture-looking but for Gibson, perception is not taking place because, for him, perception is a technical term which means "the pickup of meaningful information from the environment". When you look through the aperture, you are not picking up meaningful information because the information available in the stimulus is ambiguous (basically, you wouldn't know how to operate on the stimulus effectively).

    This point is important and often overlooked in discussions of Noe and Gibson's theory. Noe follows Gibson on the point and says in Action In Perception (p. 17) that "Mere sensation, mere stimulation, falls short of perceptual awareness".

    Now, I think Noe chose the wrong word there. What he meant was that mere sensory stimulation falls shorts of *meaningful* perceptual awareness i.e. an awareness of an affordance. It isn't like when you look through an aperture your vision fades to black or anything, *it just doesn't mean anything*. Gibson liked to talk about the Ganzfeld experiments wherein subjects looked at an undifferentiated visual field (you can try this with cut-open pingpong balls; it's fun). When you perceive an undifferentiated visual field (such as when looking at the pure sky), there is *sensory stimulation* but no available *stimulus information* and thus no perception an affordance, which is defined in terms of stimulus information (the information discrimination of which can help effectively regulate your changing relationship with the environment).

    So Gibson would respond to Ken by saying that in the aperture experiment, there is mere sensation going on, but no perception, because perception is defined as the discrimination of meaningful ecological information and there is none available in the ambiguity of the aperture experiment.

    I hope that clarifies things.

    ReplyDelete
  10. But, how does a person have access to what is past, if it is past?
    By accessing information defined over time. (Spoiler alert: the answer is always information :)

    I've blogged a little about this topic as I was working my way through Heft's great book. The key idea is that time is the wrong way to carve up the size of the present; information and dynamics are the key.

    A (slightly spherical cow :) example: push a toy car along a wooden floor then let it go. While you're pushing it reflects being in perceptual contact with something which is maintaining a behaviour; after letting it go, that behaviour persists for a little while due to the dynamics of the system in question. The whole event has structure which provides you with access to the whole thing; this structure fades, eventually of course, so it's not permanent, nor is it perfect. There is no specific size to the 'present'; it's task specific.

    A less dodgy example is the calibration of reaching behaviour (but I didn't want to jump right away to a perception-action example you might not know about). Reach for your coffee mug with your eyes open; make successful contact with it, etc. Then close your eyes and repeat the reach a few times; you are still able to do the reach, although the quality of the reach fades over time as you get further from the point when you had visual contact. For ecological types this isn't memory, it's perception (which of course weirds a lot of people out). I think this is the paper I have in mind.

    It's certainly a hard question, though, and I'm not going to pretend we have all the answers on this one. But we have a plan: investigating the temporally extended perceptual information used to control task specific devices (like a reaching device) that have specific dynamic properties.

    ReplyDelete
  11. @Gary #1:
    I agree that perception is a temporally extended process. And, I am happy, at least for the nonce, the idea that optic flow is important and is what birds use. The catching fly ball cases probably work in terms of flow too.

    The problem, I see, comes with the extrapolation to things like the handling of objects, as in the little video I made or in Noe's account of how children identify unseen objects in a cloth sack. Take the exploration of a flat head screw driver in the sack. At time t1, you feel the flat end, then at time t2, five seconds later, you feel the handle. It seems to me that the ability to perceive a flat head screw driver requires some sort of preservation of the information gathered at t1 until one can integrate this with the information gathered at t2.

    This is why I thought that temporal integration can be a problem for Gibsonians. It looks like it requires information storage from t1 to t2, then integration of the information gathered at t1 with the information gathered at 2.

    ReplyDelete
  12. @Gary #2
    "So Gibson would respond to Ken by saying that in the aperture experiment, there is mere sensation going on, but no perception, because perception is defined as the discrimination of meaningful ecological information and there is none available in the ambiguity of the aperture experiment."

    Ok. But, this just transforms the basic question. Before I asked, "Why should scientists not study aperture vision?" Now, I just have to ask, "Why should scientists not study mere sensation?"

    Gibson seems to want to discourage the study of what he calls aperture vision and what he calls sensation, but why should anyone be discouraged? There is still something going on in aperture vision and mere sensation, so why not study it scientifically?

    ReplyDelete
  13. @Andrew,
    I'm thinking that my reply to Gary's first comment might clarify what I'm up to vis a vis your comment.

    ReplyDelete
  14. On Gary's example of the Ganzfeld experiments: if you've ever been in proper white-out conditions up a mountain or on the road, you'll know what happens: everything fails utterly, fairly quickly. People fall over, get lost immediately, are very disoriented in general and it is incredibly dangerous.

    The effect is exactly the same as in total darkness (and I mean total; we are really sensitive to even a little bit of light). My only experience of this was in a WWII era underground tunnel on Waiheke Island in NZ; we turned our flash lights off for a second. The experience was incredible; we were all instantly on edge, reaching for the walls, entirely unsure about what was going on and getting more so every second. We had to flick the lights back after 10s or so, it was so disturbing.

    The reason someone interested in perception doesn't need to study sensation like this is that it isn't perception. Light energy with no information = no perception, just light.

    I'll think about the temporal extent thing a little more too.

    ReplyDelete
  15. OK, temporal integration.

    Perceptual information is extended in time and space. This means the process of information pickup is extended in time and space. With the screwdriver, there's a lot of information about the fact that what happens and t1 and what happens in t2 are about the same object, not least of which what happens at time t1<=t<=t2. Even if you are only pinching the flat head of the screwdriver, picking it up will lead to information (via dynamic touch) about the size and shape of the object (ie that it has an extent, etc).

    The problem with your suggestion is taking t1 and t2 too seriously. These things are continuous processes. If you did just present the child with the flat head, then 5s later with the handle, there wouldn't be any reason to connect them and the child presumably wouldn't. But if the child engages in the continuous (relative to the time scale of the dynamics of perception for this task) exploration of the object they'll generate a lot of information about the fact that it's a single object.

    Does that make sense?

    ReplyDelete
  16. "The reason someone interested in perception doesn't need to study sensation like this is that it isn't perception. Light energy with no information = no perception, just light."

    Fair enough. But, what I am getting at is why shouldn't someone else, e.g. folks who don't care about Gibson's kind of perception be interested in sensation.

    Or, put the matter another way. Why shouldn't there be a scientific study of mere sensation that would look at things like aperture vision? Gibson seems to want to eliminate such a study.

    ReplyDelete
  17. Let me tweak the details of the screwdriver example a bit. So, let's talk about time intervals. So, from t1-t2 a subject spends pinching the flat end of the screwdriver, then from t2 to t3, the subject slides her hand along the shaft up to the handle, then from t3 to t4 the subject grips the handle, then at t4 the subject recognizes the object as a flathead screwdriver. Let each interval be, say, three seconds. So, now the example seems to me to involve the kind of continuous (or uninterrupted)contact with and exploration of the screwdriver that I think you are alluding to.

    But, it seems to me that the subject has to have some means of saving some information about what happened in the t1-t2 interval in order to integrate it with the information gathered in the interval t3-t4. In particular, this seems to be crucial to identifying the things as a flat head screwdriver, rather than a philips head or robertson drive.

    ReplyDelete
  18. People who study sensation as you suggest are called physiologists. I'm happy for them to keep their jobs so long as they don't start calling themselves psychologists :)

    One suggestion for the screwdriver: by t2 you have 'flat end', by t3 you have 'flat end at one end of shaft', by t4 you have 'shaft with one flat end and one handle end'. Each bit in quotes is a unit in and of itself; by t4 you don't have flat head + shaft + handle, you have flat-head-shaft-handle. We happen to have a word for that: screwdriver.

    ReplyDelete
  19. Ok. So, there is a legitimate discipline that will investigate aperture vision, this non-perceptual stuff. So, we are down to what to call it, right.

    "People who study sensation as you suggest are called physiologists"

    Now, as a descriptive claim about how things go in the profession, I think you are wrong. Many of the folks who have worked on the Ames room are called psychologists and vision scientists. Maybe what you want is a normative claim that they *should* be called physiologists.

    ReplyDelete
  20. "by t2 you have 'flat end', by t3 you have 'flat end at one end of shaft', by t4 you have 'shaft with one flat end and one handle end'."

    Ok. But, how is it that in the interval t2-t2 you have "flat end at one end of shaft", rather than just "shaft", without storing information about what happened earlier? It sounds like what one is doing is building up a representation. But, I know that that is not the Gibsonian line.

    ReplyDelete
  21. I don't think it is a matter of what people should be called professionally. I think Andrew's point was that cognitivism (and classic behaviorism) is wedded to a physiological (rather than biological) description of the animal's perceptual system.

    They start off with a very close up model of the retina wherein light particles come in and bounce off the retina, which registers various wavelengths, etc. The problem then is how do you go from detection of wavelengths (something a mere instrument could do) to perceptual experience.

    For ecological psychology, the problem here is that the question is set up wrong because on the biological level (the level relevant to survival) animals aren't interested or concerned with photons and wavelengths. They are interested in the so-called molar level, which is extended over time. According to Gibson, the perceptual system is precisely attuned (for evolutionary reasons) to the spatially and temporally invariant properties which can be utilized for specific functional purposes (affordances).

    This simple shift from a physiological view to an ecological view has radical implications for the basic question-set of psychology. If we accept the physiological set-up, we are stuck with the question of how the brain makes meaning out of meaningless input. As far as I am aware, this is an intractable problem. If we accept ecological psychology, the question is not how does the brain generate meaning, but rather, *how does the brain find the meaning in the environment and then use it to regulate its behavior*. For a bacterium, meaningless sucrose molecules are intrinsically meaningful because they afford nutrition.

    This allows us to construct a scientific and natural explanation of meaning. Meaning is not in the head (as Noe provocatively puts it), but rather, external. But this doesn't mean that meaning is something completely subjective. The sucrose is perfectly objective. But it is inherently meaningful to a bacterium with the proper metabolic system.

    ReplyDelete
  22. Sensation is merely the physical process of energy transduction; that's why it's physiology that studies it. People studying the Ames Room aren't actually studying that, hence they can call themselves psychologists if they want - sensation is not the basis of perception, though.

    I don't know exactly how the persistence I'm talking about works. Building a representation is one way to describe it; perceiving the information generated by the extended event of exploring the object is another. This is certainly where it gets down to your metaphysics; but given that the world is it's own best model and we have a theory about how we might be able to perceive that adequately there's no need to assume you have to represent anything.

    ReplyDelete
  23. But, Gary, Andrea, bear in mind a comment Gary made earlier.

    "Something is still going on in aperture-looking but for Gibson, perception is not taking place because, for him, perception is a technical term which means "the pickup of meaningful information from the environment"."

    When you say they are studying sensation and not perception, these are technical Gibsonian terms. That are not the pedestrian notions. So, to be more explicit about what you are claiming, it would be that folks who study aperture vision are studying sensation-G, rather than perception-G, where the "-G" is a marker that draws attention to the point that these are technical terms. This is not the same as saying that folks show study aperture vision are studying sensation-P, rather than perception-P, where the "-P" is a marker that draws attention to the point that this is the pedestrian sense of the terms.

    Ok. So here is what I did next. I went to EBSCOhost and did a search on "Ames Room"

    Emmert's law in the Ames room

    Jonathan Dwyer, Roderick Ashton, Jack Broerse

    Abstract. The Ames distorted room illusion, in which the perceived sizes of objects placed within the room differ from their objective sizes, has been used to support arguments for indirect perception. A study is reported in which Emmert's law of the apparent size of after images was examined in relation to the Ames room's illusory alteration of apparent and actual distances. Size judgments of afterimages projected into the Ames room were compared with control conditions in which both actual and apparent afterimage projection distances were reproduced. Results indicate that Emmert's law may not provide a simple geometrical relationship between proximal image size and actual viewing distance, and that the processes involved in making afterimage size judgments are similar to those processes involved in making size judgments of 'real world' objects.

    Dwyer J, Ashton R, Broerse J, 1990, "Emmert's law in the Ames room" Perception 19(1) 35 – 41

    So, this looks to be a study of the Ames room, hence a study of aperture vision. Now, it seems to me that, while this is a study of sensation-G and not perception-G, it is not a study of sensation-P, but a study of perception-P.

    So, here is what I am driving at. Gibson appears to be disparaging studies such as this by Dwyer, Ashton, Broerse as not even perceptual-P psychology, because it is not perceptual-G psychology, simply on the basis of the target of the study, namely, aperture vision. (There could be other problems with the study, but Gibson seems to be willing to dismiss this work out of handle, simply because it is aperture vision.) But, as I've claiming, there seems to be something going on in aperture vision and someone might legitimately study it and those someone's might be perceptual-P psychologists, even if they are not perceptual-G psychologists.

    ReplyDelete
  24. Gibson gets to rule this out as being uninteresting because he has a theory of information in which the Ames Room becomes a mere limited case of static perspective structure and thus not interesting. Ruling things out as uninteresting is what theories are for - you may be wrong, but that's not the point.

    These people can study judgments of the size of after images all they like, but they'll need to work hard to convince me that their study tells me anything interesting about vision, and I don't see that that's a bad thing. The bar's been set too low in psychology for too long because of it's lack of theory; I'm all for making people work a bit harder.

    ReplyDelete
  25. Earlier I wrote, "Ok. So, there is a legitimate discipline that will investigate aperture vision, this non-perceptual stuff. So, we are down to what to call it, right." But, now, you seem to be taking that back. Now this discipline is uninteresting, right?

    So be it. Just to be clear, then, the view is that what is going on in the Ames room case is uninteresting because it is a special case?

    ReplyDelete
  26. Sensation = the transduction of light energy to neural firing. This is a perfectly sensible thing to study if you are a physiologist but it is not the basis for perception.

    Aperture vision is Gibson's term for studying vision with no motion; the suggestion that by examining the static perspective structure in the 2D retinal image projected to a certain point in space you are studying the basis of vision. This is not identical to studying transduction, although they are connected to one another.

    The Ames Room is boring because even though it's compelling at that one point in space it's only compelling at that one point, in that 'snapshot' of the room, as evidenced by the fact that the compelling 'illusion' is shattered by moving three feet to the left. Perception is the time extended process in which the Ames Room can be broken, and it is not merely the sum of a series of snapshots. Thus the Ames Room doesn't tell us anything about the act of perception.

    ReplyDelete
  27. Let me take stock about what you guys say that Gibsonians say about the Dwyer, Ashton, & Broerse study.

    Gary suggests that they are studying mere sensation. By contrast, Andrew suggests that they are not studying sensation, because sensation = he transduction of light energy to neural firing.

    So, whatever they are doing, maybe we can agree to call it "Old Fashioned Perception."

    Andrew notes, "Aperture vision is Gibson's term for studying vision with no motion; the suggestion that by examining the static perspective structure in the 2D retinal image projected to a certain point in space you are studying the basis of vision."

    Ok. Don't suppose that aperture vision is a basis for extrapolation to other cases of vision. (I added this proviso in the initial post.) Still, it looks to be pretty interesting in its own right.

    And, what might make it interesting is, as Gary suggests, the Gibsonian approach just does not apply: "So Gibson would respond to Ken by saying that in the aperture experiment, there is mere sensation going on, but no perception, because perception is defined as the discrimination of meaningful ecological information and there is none available in the ambiguity of the aperture experiment."

    ReplyDelete