Monday, September 13, 2010

I had forgotten about this study

In recent discussions of aperture vision, Gary appears to have been articulating a view according to which studying aperture vision is not doing ecological psychology.

He writes,"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."

But, there is this study:
The Distorted Room Illusion, Equivalent Configurations,
and the Specificity of Static Optic Arrays

Sverker Runeson
Uppsala University, Uppsala, Sweden

The distorted room illusion (DRI) and the attendant argument for perceptual ambiguity is critically analyzed from a Gibsonian/ecological point of view. The notions of multiple specification, conflicting information, and perceptual skill are invoked in showing how the ecological approach can accommodate illusion effects that may remain under mobile binocular viewing conditions. Static optic arrays are shown not to be ambiguous. So-called equivalent configurations
are found to be analytic artifacts, appearing when the problem of information is treated in geometrical terms without regard for constraints due to physical and ecological regularities. The relative importance of motion-based and motion-independent information is discussed.
Journal of Experimental Psychology: Human Perception and Performance 1988, Vol. 14, No. 2, 295-304


  1. This is all occurring in the context of a specific fight, I should add; specifically, if the scene is intrinsically ambiguous then Gibson can't ever be right about anything. You kill that idea by redefining what information means:

    "So-called equivalent configurations
    are found to be analytic artifacts, appearing when the problem of information is treated in geometrical terms without regard for constraints due to physical and ecological regularities."

    So what the Ames Room is supposed to demonstrate (that there's something excitingly ambiguous that needs to be accounted for) is not what it actually shows, on an ecological analysis.

  2. "This is all occurring in the context of a specific fight". Absolutely, which might partly explain why I had forgotten about this study.

    And, indeed, it is one things to say that the physical and geometrical information is ambiguous, but another to say that the "ecological information" is ambiguous. The Gibsonian line is apparently that, while there is physical and geometrical ambiguity, there is no "ecological ambiguity".

    So, the next step, for me, is to get a handle on the "ecological information". This was what I found elusive in the discussion of the occluded pac-man. The Gibsonian analysis is that the "ecological information" is not ambiguous, but univocally speaks to there being a circle. But, I need to be able to get a handle on "ecological information" apart from what one perceives. It won't do just to say that I perceive a circle because I get the ecological information that there is a circle. That's the path to saying I perceive a shoe because I get the ecological information that there is a shoe.

  3. Well, I've said it before and I'll say it again: that (shoe) argument gets dealt with by Turvey et al (1981) in the 'ecological laws' paper, the reply to Fodor & Pylyshyn. If you want to get a handle on ecological information, that's your next stop :)

  4. Ken,

    I think the best way to get a grip on ecological information is to forget about the complexities of higher mammalian perception and look at the bacterium as the exemplar of detecting ecological information.

    If you put a bacterium in a sucrose solution, the bacterium is able to detect or discriminate the differential sucrose gradients and orient itself such that its propulsion will maximize the exposure to sucrose. For ecological psychology, *this* is the paradigm case of successful perception. The ecological information in this example is the nutritional properties of the sucrose. There is nothing intrinsic to the sucrose molecule that makes it "ecological information" or "nutrition". Rather, it is ecological information only in relation to the nature of the bacterium's specific metabolism. But this doesn't make it "subjective".

    For Gibson, the bacterium's perception is best explained in terms of a reaction to the available information in the environment. From a physiological perspective, this information is sucrose molecules. From an ecological (psychological) perspective, this information is nutrition.

    Now, in the case of visual perception, ecological information is defined exactly the same way. We are surrounded by ambient energy fields that specify affordances in virtue of being lawfully related to the actual structure of the environment. And since it is the actual (i.e. objective) structure of the environment which is inherently meaningful to an evolved organism (does this thing afford nutrition? locomotion? cover? sex?), perception is a matter of detecting information that specifies the affordance-opportunities of the actual world. By moving through the world, we can literally "bump into" or "come into contact with" the energy fields and the complex spatial-temporal transformations that such encounters trigger in us specify ecological information under normal conditions (in virtue of learning about the affordances through experience).

    In other words, the ecological information is simply the thing-in-itself in relation to the internal needs of the organism encountering it. Many Gibsonians make the mistake of saying that ecological information only exists when animals are perceiving the environment. This is a grave mistake. The ecological information exists independently of whether we are able to successfully find it and use it. The ground affords support whether or not I am around to use it. But the utilization of the information is entirely relative to the user.

    Accordingly, the affordance property is an *objective* fact about the things-in-themselves. The process of finding and detecting it can go awry in all sorts of ways (as demonstrated by the ample experiments on visual illusions). We can fail to detect all the available information or localized brain damage can interrupt and deform the detection-utilization process, leading to ineffective action.

  5. Gary,
    My problem is not so much ignorance of the Gibsonian apparatus as it is seeing how it does, or does not apply, to static viewing in the Ames Room. And, more generally, it is linking up this Gibsonian to cases that seem to me relatively simple and comprehensible.

    The nub of your claim seems to me to be that in static viewing of the Ames Room there is no meaningful Gibsonian information regarding the shape of the room. If you "No information, hence no perception," then it looks like your take on the Gibsonian analysis leads to the prediction that in static viewing of the Ames room there is no perception of the shape of the room. But, we do perceive a rectangular shape in static viewing.

    Runeson's analysis, by contrast, spells out how he thinks we get unambiguous information about rectangularity. So, his analysis is there is Gibsonian information, your analysis is that there is no Gibsonian information. It seems to me Runeson is more likely to be correct.

  6. Oh, I should add. Your analysis is that there is no perception-G, which is fine. Since that's true by the definition of perception-G. But, we do perceive-p a rectangular shape. So, there is getting past the Gibsonian definitions to get to the point. That point is that, by your lights, Gibson's theory does not apply to the case of static viewing of the Ames Room. That seems bad.

  7. Watching my Philosophy TV episode with Mark Rowlands, I had forgotten his contention that the Gibsonian view is supposed to say that no information is contained in any particular optic array. (See about 10:05.) So, that at least to me sounds very close to the view that there is no information contained in a static display.

    So, it sounds like Mark's take on Gibson is like Gary's take on Gibson, which is not like Runeson's take.

  8. As I prepared a post on Runeson and the Ames Room, I saw this again; I haven't watched your Philosophy TV appearance (yet :) but I think you're partially right here. There is information in a static optic array, which is a problem Gibson was grappling with in the '79 book in the section you took the aperture vision quote from. Gibson did not have a definitive answer, which Runeson talks about too; his analysis of the Ames Room allows that there is indeed motion-independent information (and it's this that a) allows you to see the room at all and b) leads you to fall for the illusion when it's all you have available).

    But Mark and Gary are right in that, in ecological visual perception, static arrays might be around but are merely a fraction of the actual information around and it's this invariant-over-transformation information that is going to be typically used, because it's so much more robust.

    Even in the Ames Room, while Runeson is right that there is static structure specifying a square, it quickly loses out as all the other (consistent) information about the actual shape is allowed in.

  9. My sense was that Gibson wanted to avoid the topic of illusions, since he had no account. But, Runeson at least provides an account. And, he probably should have an account.

  10. Er, Gibson had an account of illusions: they aren't misperceptions, they are perceptions of inconsistent or incorrect information (this comes from Holt's influence). This is Runeson's take too.

  11. You call that an account? I thought that this is the kind of thing psychologists call a redescription of the phenomenon.

  12. It's firmly based in his theories about information, so yes, it's more than mere redescription.

  13. What's Gibson's account of Mach bands?

  14. I've no idea. Aren't they just a feature of the organisation of the retina that enhances contrast? The kind of thing you might build into a visual system interested in differences? That seems uncomplicated for everyone.

  15. The standard account of Mach bands does appeal to lateral inhibition in the retina, but the EPists seem to like to complain about views in which vision begins with the retina. What gives?

  16. Perception requires a detection device, which will have it's own characteristics and dynamics. It's not like we don't think the retina is involved at some point in the proceedings; it's just not the first thing to structure light.

  17. Can you provide me some references where Gibson says something at all supportive regarding what the retina does? Everything I know of is pretty dismissive, although it is not entirely clear to me what basis there is for this dismissal.

  18. "The registering of differences in intensity in different directions is necessary for visual perception; the formation of a retinal image is not."
    Gibson, 1979, p.62

    i.e. of course the retina is doing some work (it's just not image formation or processing).

  19. Ok. But, this is not committal regarding the retina (or its parts). For all I can tell, Gibson is still free to say that it is not the retina (or its parts)that registers differences in intensity in different directions. It is the eye or the visual system or the organism as a whole.

  20. That sentence is Chapter 4 of Gibson '79 (a useful summary is here). In it, Gibson talks about what does happen on the retina (stimulation by light, transduction, etc) and simply pointing out that this, while part of the eyes-head-body perceptual system, is not the basis for perception.

    He's not down on the retina; he's just down on the flawed concept of the retinal image.

  21. I agree that he rejects the concept of the retinal image.

    How about this? Does Gibson ever talk about lateral inhibition in the eye?

  22. Ok. You write on your blog post that "Sensation (e.g., stimulation of photoreceptors) doesn’t tell us anything about the world. Sensation cannot be the basis of perception."

    How do you square these claims with the idea that events in the retina apparently explain Mach bands?

  23. I don't know off the top of my head if lateral inhibition ever came up; he was focused on ecological optics by '79. Maybe in the 1950 book; I'll have a look sometime.

    Then, two things:
    1. I didn't write that post, sabethg did.

    2. Mach bands aren't a phenomenon of perception, as defined by Gibson. Remember Gary's point that perception is a technical term with specific meaning for Gibson, and perception defined this way is not based in mere sensation. The detection and transduction of light happens, but that's not the starting point for perception, simply part of the chain of events in detecting visual information.

  24. Or, here are Wilcox & Katz,
    "Sometimes it seems as if the advocates of J. J. Gibson's ideas are in danger of being drowned out by a chorus repeating the same refrain:
    "Gibson's theory is incomplete". The most frequent allegation is that Gibson, although providing insight into the environmental support for perception, fails to account for the internal processes which make it up (e.g., Johansson, 1970; Hamlyn, 1977; Neisser, 1977; Epstein, 1977; Haber, 1980). It is of course, true that Gibson never gave an account of these alleged processes, but that is because, as Reed and Jones (1978) recently pointed out, there are none for Gibson." (p. 313)

    Wilcox, S., & Katz, S. (1981). What Gibson isn't missing after all: A reply to Heil. Journal for the Theory of Social Behaviour. Volume 11, Issue 3, pages 313–317, October 1981

  25. Specifically, "Gibson’s theory does not require internal processes to bridge the gap between sensation and perception, because for Gibson no such gap exists." (Wilcox & Katz, pp 313). He's rejecting very specific internal processes (constructivist, enrichment processes), not the entire inside of the organism.

  26. Can we agree, then, that Wilcox and Katz don't clearly state Gibson's view? They overstate the scope of Gibson's rejection?

    The more I read the more I find that really one has to read Gibson as responding to psychological theorizing circa mid-Twentieth Century, rather than to late Twentieth Century. So, that maybe Gibson would not resist things as they are now. Or, would he?

  27. W&K seem pretty spot on to me - even handed too. What makes you think they're over stating anything?

    How late 20th century? Gibson, let's face it, died in 1979 and most of his empirical work is from the 50s and 60s. I agree that his specific arguments are indeed about earlier theorising (but one of the things I really respect about Gibson is how clearly he articulates what it is he is arguing against at any given moment, so it's usually obvious what's going on).

    I wish things had changed, but I'm not convinced things have moved on all that much. Cognitive science nods to affordances but ignores ecological optics, clear evidence that they've missed the point. Many of Gibson's intuitions have indeed snuck into modern thought, but little of his conceptual rigour made it through. All the mainstream progress has been in articulating more and more sophisticated representational models in which more and more sophisticated maths is brought to bear to solve a problem Gibson explained doesn't exist (enrichment of an impoverished stimulus). Dynamical systems has certainly been a healthy move, and it's no accident eco-psych people have all gravitated towards it; it's a language designed to express what we want to say. Dynamics gets misused as a theory of behaviour, though, which it isn't, but that's a side note.

    So my view is that Gibson would still have plenty to resist.

  28. Well, W&K seem to think Gibson rejects *all* internal processes, where you think he rejects only constructivist and enrichment processes.

    Well, I am most familiar with vision stuff that has been done since, say, 1980 with Fodor, Marr, and Pylyshyn. And, I find Gibson just baffling a lot of the time. I disagree a lot, but I also find that there are too many times where I have very little idea what is going on. Part of it has to be that he's responding to things I'm not aware of.

    And, I don't see the appeal of dynamical systems.

    I guess that EPers do think that, e.g. Fodor and Pylyshyn are wrong, but it also seems to me that they are deploying arguments that don't really apply to F&P.

  29. Reread that quote I noted: "Gibson’s theory does not require internal processes to bridge the gap between sensation and perception, because for Gibson no such gap exists." This is all they say on the matter and it's what I'm saying.

    Have you read Gibson 1979 cover to cover? It's a great book that builds an argument nicely and it's best taken as a whole. sabethg was working on a reading group and I'd like to re-read it anyway, maybe this could be an excuse to get that part of the blog going again. She has a new job taking her time but I know she's got things in the pipeline.

    Dynamics is a very useful language to express systems level thinking in a disciplined way, but like embodied cognition there are many people who think they do dynamics who aren't. Plus many people (eg Scott Kelso, Pier Zanone, etc) who treat it as a theory of behaviour. The coordination posts I need to get on with will get into this in detail because it's an area I do a lot of research in.

  30. Ok. Sensation is retinal stimulation, but perception is not retinal stimulation, right?

    So, what happens after retinal stimulation that eventuates in perception? I'm thinking its this business of lateral inhibition. That looks to be an internal process that would bridge the gap between sensation and perception.

    I've not read all of Gibson, 1979. I've tried several times, but after a while I feel I'm just not getting anything out of it. But, there is a good chance I'm going to do a conference symposium on EP this in the spring, so I'll probably be at it again. And, I'm working on posts on Gibson, 1979. So, maybe it would be good to have a joint blog, so we could both post images, etc.

  31. So, what happens after retinal stimulation that eventuates in perception?
    Nothing. It's what comes before that matters, because information is created by the interaction of energy with surfaces and surface layout. Read Gibson '79, laying this out carefully is what that book is entirely about. sabethg's posted summaries of Chaps 1-5 on our blog.

  32. Gibson argues that retinal stimulation is not the level at which perception occurs. This is not the same thing as saying that events in the retina don't explain anything. But, events in the retina don't explain perception. As Andrew said, Mach bands aren't a phenomena of perception (as defined by Gibson), so this is consistent.

    It is clear that the presence or absence of retinal stimulation does not result in perception. Total blackouts are the absence of stimulation (at least, stimulation caused by photons) and whiteouts are the presence of considerable stimulation. However, we are entirely unable to organise our behaviour with respect to either of these situations because their is no structure, no information, available to us. If we focus research on the retina, this is apparently all vision has to go by and this does certainly seem to require some additional processing to extract any meaning. But, this is the wrong level of analysis.

    Structure in the ambient optical array already contains meaning because the relationships between visual solid angles (caused by light reflected from objects) has a lawful relationship to things in the world. We perceive information and it already has meaning. This is why Gibson says there is no gap. And, this is why it's unnecessary to invoke an event "after retinal stimulation that eventuates in perception." I think it's fair to say that perception (defined by Gibson) occurs concurrently with sensation (as is commonly construed in cog sci), but at another level of analysis.

    At this ecological level of analysis, vision isn't so mysterious. Yes, things are still happening in the retina and if you ask questions about retinal events, then you'll need to talk about the retina to provide adequate explanations. But you're not going to explain perceptual events by focusing on the retina.

  33. As Andrew said, I'm going to pick back up on the Gibson '79 posts, so I'd be interested to hear what you have planned.

  34. @Andrew:
    "So, what happens after retinal stimulation that eventuates in perception?
    Nothing. It's what comes before that matters, because information is created by the interaction of energy with surfaces and surface layout."

    Nothing? That can't be right. So, I can accept that the physical environment structures an ambient optic array. But, perception also requires more than just an ambient optic array. Having the optic array strike a rock and having it strike a human visual system makes a difference. The rock doesn't perceive anything, but the normal human does. What comes after retinal stimulation may not be the whole of the story, but how can it not be part of the story?

  35. Ok. So, what again is Gibson's definition of perception?

    I can see that the white out cases do not lead to perception-G, but the physical layout that induces the Mach bands are not instances of a physical white out.

  36. @Sabrina,

    I am haltingly reading through Gibson, 1979. I started Chapter 2, but got bored. Chapter 3 was essentially a non-starter. Chapter 4 was better and I have prepared some posts on it.

    But, if you wanted to do a reading group kind of thing I would sort of like to be able to post some images. I find them helpful. It might be useful to create a separate joint blog for that special purpose. That was the idea I had.

    I'll (probably) be part of an EP symposium next semester, so I'll probably read at least some good chunks of Gibson, 1979. I'm not yet sure what I am going to say, but I figure something will come to me.

  37. @Andrew

    Let me try another approach. For Gibson, perception requires light structured by the environment. (This is the point of the white out discussion.) Ok. So, structured light is necessary for perception. What, then, is sufficient for perception? It cannot be *just* the ambient optic array. There have to be some organisms.

    Ok. Then, is the ambient optic array plus sensation (i.e., retinal stimulation) sufficient for perception?

    I'm assuming that the answer is "no", so that one needs more. What more does one need?

  38. OK, yes, I was being a little extravagant. But you're asking what to a Gibsonian is the wrong question - nothing happens after the the retina that 'eventuates in perception'. You don't take the retinal input and process it. That's not how it goes. What more does one need? A perceptual system (Chapt. 14 of Gibson '79 is where a lot of this is going to happen - I was going to and summarise it but I don't really have time today).

    Honestly, you need to read Gibson cover to cover. I know the early chapters are about the environment, but given that this is what we want to perceive what Gibson does in those chapters is pretty important and will answer a lot of these questions that keep coming up.

    Where's the symposium?

  39. Ok. I'll read chapter 14. But, remember, it's a 300+ page book and a really long slog for someone who's not really buying it. And it's always a matter of catering to Gibson's jargon and reading more.

    I'm betting that there's more to the story than just *having* a perceptual system. The system needs to do something in order for there to be perception. So, I'm betting that it's going to be something that system does that will eventuate in perception.

    Or, as I've tried to put it in other words, even an ambient optic array + retinal stimulation + having a perceptual system is not sufficient for perception. Among the sufficiency conditions I assume there will have to be something about the activity of the system.

    The Southern Society for Philosophy and Psychology is having a session in New Orleans. Michael Riley and Tony Chemero and Zoe Drayson are lined up now, but we are looking for another psychologist from Cincinnati to maybe show. (The invite is still up in the air.)

    You interested in hopping the pond to give a talk? There's no money, but there is New Orleans, which is a great American city.

  40. Ok. So, the SSPP line up is set. It will be
    Kevin Shockley
    Michael Riley
    Tony Chemero
    Zoe Drayson
    and (probably) me.

    I might drop out for time reasons, but then again I would be the only one voice of resistance.

  41. Ok. I'll read chapter 14. But, remember, it's a 300+ page book and a really long slog for someone who's not really buying it. And it's always a matter of catering to Gibson's jargon and reading more.
    And yet, still the best way to find out what it is that Gibson is actually claiming :)

    (I'm surprised you're finding it a slog, though - in my opinion it's one of the most clearly written science books I've ever read.)

  42. Well, take Chapter 2. Sentences 3 and 4 are these bold and dismissive claims regarding past science, but there seems to be nothing at all in the chapter to substantiate it. There is, instead, this long list of rather pedestrian observations and giving of definitions.

    Or take the summary: "We do not live in 'space'". To me, there is a clear sense in which we obviously do live in space, but some sense in which Gibson is right. So, here is this annoying ambiguity. But, is it worth pondering out how to articulate that ambiguity.

    Also, there is just very little in the way of reporting scientific results. So, there are large tracts where I'm not learning new facts, just memorizing the Gibsonian system.

    Compare Gibson's 1979 with Pylyshyn's *Seeing and Visualizing*. Pylyshyn's book is a much harder read. It's a very challenging read, but I feel like I am learning something. There are studies discussed and interpreted. I read parts of it in a grad seminar I saw it on and I could see that it was hard for the students, so I know it's hard. But, there is a grappling with evidence. There is also not the same sweeping dismissal of vast parts of the profession. Pylyshyn's views are out there, but he seems to me to work through a lot of experimental detail.

    But, a last thing is that I don't have intellectual background that makes this kind of thing palatable. I have much more background in mainstream vision science that Gibson rejects, so I sense a lot of ambiguity.

    And I think it's not just the relative age of Gibson's book. I get Skinner's Verbal Behavior despite the vast theoretical differences between him and me.

    Incidentally, I am working through Chapter 14 (which I've done part of before), but I'm not feeling it. I don't see in it what you see in it as an account of how there is no gap between sensation and perception. I'm sure Gibson has some such view, but I'm still thinking one ought to be worried about the tenability of that view.

  43. Also, there is just very little in the way of reporting scientific results. So, there are large tracts where I'm not learning new facts, just memorizing the Gibsonian system.
    This is certainly a fair comment. Part of the problem was that this reflected the state-of-the-art in 1979; the experiments hadn't been run.

    The other issue is that Gibson has reached a point in his own thinking and empirical work where he's realised the conceptual tools of psychology were entirely inappropriate to handle the questions he was addressing, so he's had to throw it all out and start again. This book is laying out the form of that do-over, so it is very much a manifesto, a programme, rather than the sum of an experimental career.

    Think of it this way: Pylyshyn is summarising his normal science, Gibson's having to handle a paradigm shift, and the result for Gibson is that he doesn't get to rest on specific empirical results. That's not to say the book isn't grounded in data - but his empirical work has all led him to the conclusion that he had to rethink how he was handling the data.

    The 'long list of rather pedestrian observations and giving of definitions' is part of this, and one of the things I personally find fascinating about the book. Gibson is the only person studying visual perception to spend any time talking about what there is to be perceived! He has to spend three chapters laying out the environment because no-one ever thought to do this before. This, to me, has always marked the book out immediately as an entirely different sort of science book, and I think it's precisely the correct place to begin.

    It's interesting to hear your reaction to it, though. Finding ways to bridge these sorts of gaps is an important part of communicating the ecological approach, which is something I'm obviously very interested in doing.

  44. Hi, Andrew,
    Vis a vis the very last bit of your comment, I just now put up a post that I have had idling in the queue. I think it's timely and relevant.

    I'll try to return to your other points in a bit.