Tuesday, January 18, 2011

Gibson, 1979, Chapter 8

Andrew just suggested that the exploding box example is not a problem for EP.  But, it's the kind of case that appears to be relevant in evaluating empirical claims suggested in the following:
I have described the environment as the surfaces that separate substances from the medium in which the animals live. But I have also described what the environment affords animals, mentioning the terrain, shelters, water, fire, objects, tools, other animals, and human displays. How do we go from surfaces to affordances? And if there information in light for the perception of surfaces, is there information for the perception of what they afford?  (Gibson, 1979, p. 127).
So, consider the last question: "If there information in light for the perception of surfaces, is there information for the perception of what they afford?"

Now, there is, to my mind, an ambiguity in this.  Notice that, strictly speaking, Gibson is writing about surfaces, not objects.  To see this, replace the pronoun with its antecedent:
(A) If there information in light for the perception of surfaces, is there information for the perception of what surfaces afford?
Contrast that with the question: 
(B) If there information in light for the perception of surfaces of an object, is there information for the perception of what the object affords?
Now, (A) is a little odd, since it is perhaps not the surface of the fish, say, that affords edibility, but the "meat" of the fish that affords edibility.  Or, to use a TSRM example, it is not the surface of a thing that affords potability.


Now, contrast these questions with some more restrictive questions:
(A') If there information in light for the perception of surfaces, is there information for the perception of what surfaces afford prehension?
Contrast that with the question: 

(B') If there information in light for the perception of surfaces of an object, is there information for the perception of what the object affords prehension?
Maybe this too reads a bit awkwardly, but one gets the idea (I hope) that it is not the information for all affordances that is provided by information in light for the perception of surfaces; instead, it is only the information for prehension affordances that is provided by information in the light for perception of surfaces.

Finally, there are questions about even weaker matters, like the one's Noë hints at,
(A'') If there information in light for the perception of surfaces, is there information for the perception of some of the properties that surfaces afford?
Contrast that with the question: 
(B'') If there information in light for the perception of surfaces of an object, is there information for the perception of some of the properties that the object affords prehension?

24 comments:

  1. I'm not sure I see the problem. Gibson is indeed talking about surfaces, not objects (Chapter 2 is called 'Medium, Substances and Surfaces'; surfaces start on pg 22).

    The reason is that surfaces are where a lot of the action is. Surfaces reflect light, for instance. Surfaces are what animals touch. So information is information about surfaces. It might be the meat of the fish that you eat, but it's the outside of the fish that you detect.

    Are you trying to say that if all you can see is the outside of something, you can only get to what's inside via inference? And if the affordance is on the inside (eg the meat of the fish) then you can't perceive edibility by detecting the outside of a fish?

    ReplyDelete
  2. The problem is that it sounds so crazy that I am very unsure that I am getting him right.

    But, one way of thinking about what I'm driving at with the boxes is that surfaces don't constitute affordances. The surface gives you data for perception, but the surface does not give you pick-up-abilty.

    Or, think about sit-on-ability of a chair. You don't get that just from the surfaces of the chair. A chair has to be sufficiently strong and typically the strength is due, in part, to stuff beneath the surface.

    So, here's an exercise: Think of an affordance of a thing that is not determined by the surface of the thing.

    ReplyDelete
  3. But, one way of thinking about what I'm driving at with the boxes is that surfaces don't constitute affordances. The surface gives you data for perception, but the surface does not give you pick-up-abilty.
    Close. The surface indeed gives you data for perception, which is information for the affordance (e.g. pick-up-ability).

    What ever you (perceptually) know about the world has to come from the surfaces. Gibson was quite clear on this in Chapter 2.

    So, here's an exercise: Think of an affordance of a thing that is not determined by the surface of the thing.
    The walk-on-ability of a surface.

    ReplyDelete
  4. So, here's an exercise: Think of an affordance of a thing that is not determined by the surface of the thing.
    The walk-on-ability of a surface.


    Ok. Good. So, now you have offered a counterexample to Gibson's claim about the connection between surfaces and affordances. This is what I have been driving at with the boxes example.

    ReplyDelete
  5. Look at the sentence after the ones quoted in the main post:
    "Perhaps the composition and layout of surfaces constitutes what they afford".

    What I have been driving at, and your counterexample of walk-n-ability of a surface, are challenges to this.

    ReplyDelete
  6. In the Noe-on-Gibson-on-Affordances comments, Ken offered me both a sucker bet and an easy-win bet, and I took the former by trying to redefine affordances. So, I want to renege and go back to the easy-win:

    "Charles will be all over the distinction between looking pick-up-able and being pick-up-able."

    Let me rephrase that to "between an affordance being only apparent and being real". If that's Ken's major concern, I think the appropriate response is simply to note that affordances, being a perceptual concept, are necessarily always merely apparent. That's the difference between a perception and knowledge - the former may require additional work to become the latter. A stool affords the appearance of being sit-on-able but collapses when you sit on it; a box affords the appearance of being pick-up-able but explodes when you touch it; a floor affords the appearance of being walk-on-able but you slip and fall when you step on it because it was undetectably slippery. So, now you know. Surely no one assumes Gibson was proposing an approach to perception that provides a free ride from perception to knowledge.

    "surfaces don't constitute affordances"

    Depends on what "constitute" means. The structure of light reflected from surfaces may - or may not, depending on the info in the reflected light - offer affordances to a perceiver. But any affordances offered are only apparent.

    "Perhaps the composition and layout of surfaces constitutes what they afford".

    This admittedly seems poorly worded. So, Gibson wasn't the best tech writer ever - we established that with my debut comments about system engineering.

    ReplyDelete
  7. Walk on ability of a surface isn't determined by the surface of the thing; it's determined by the anchoring properties which lead to that surface. You can then perceive the surface.

    ReplyDelete
  8. Surely no one assumes Gibson was proposing an approach to perception that provides a free ride from perception to knowledge.
    Hardly a free ride; perceptual learning is hard. But he certainly doesn't want the extra work you propose is required to go from perception to knowledge. Avoiding that work is what affordances are for, in his theory.

    The error cases are solved with information. Prior to sitting on the chair that is made of, say, styrofoam, if the information specifies a sit-on-able chair then you may indeed try to effect that affordance. You didn't pick up an apparent affordance; you perceived the affordance that was specified, which wasn't the right one because the full thing was not specified.

    The distinction between what the observer can possibly know and what a third party analysis of the situation knows must be maintained.

    ReplyDelete
  9. I think we are now down to misunderstandings of specific words. That's progress!

    I understand that Gibson's claim is that there is lots of information available in the complex structure of reflected light and that one can extract "knowledge" from it. I'm just adding that even after you've extracted all that's available in the light, there may be things you still don't "know" - eg, that the box will explode when touched. That's what requires "additional work" to get at.

    As I tried to make clear, in calling an affordance "apparent" now I'm not distinguishing between that and "real" (although before, I mistakenly was), I'm just observing that affordances are necessarily apparent because they are products of perception, and perceptions can be misleading.

    Perhaps I'm still missing Ken's point, but all I see is that the two boxes are perceived as offering a certain affordance, and trying to effect that affordance for one box will definitely fail because of a stated (ie, known by us but not by the perceiver) but imperceptible feature of that box. I'm just noting that trying to effect the affordance may fail in the other case as well for a different (or even the same) unstated reason (known by neither the perceiver nor us). And in that sense the perceived affordance is apparent in both cases. And in that sense was "apparent" pre-actualization even if the affordance does get effected for the non-exploding box.

    I actually prefer what I said at the beginning of the other thread, viz, affordances have nothing to do with unforeseeable consequences of attempts to effect them. But that didn't seem to fly, so I'm just trying to express it in the EP vocabulary, augmenting a bit where it seems necessary. Either way, I don't see a problem.

    ReplyDelete
  10. Walk on ability of a surface isn't determined by the surface of the thing; it's determined by the anchoring properties which lead to that surface. You can then perceive the surface.

    Precisely. But, the problem is that the anchoring properties are making the affordance and the surface is making the perception. Absolutely.

    But, the problem is making the link between the surfaces and the anchoring properties/ affordances. Gibson conjectures that surfaces constitutues affordances. But that's not right. The exploding boxes case is meant to drive a wedge between these.

    ReplyDelete
  11. Charles,

    I don't think this is correct:
    affordances are necessarily apparent because they are products of perception, and perceptions can be misleading.

    Affordances are supposed to be something in the world to which an organism is related. They are not like, say, mental images.

    ReplyDelete
  12. Charles, here's the first sentence from TSRM:
    "This paper is both a reply to Fodor and Pylyshyn 1981) and a systematic explication of one of Gibson's (1979) basic claims, namely, that there are ecological laws relations organisms to the affordances of the environment."

    (italics added)

    ReplyDelete
  13. Typo:

    "Precisely. But, the problem is that the anchoring properties are making the affordance and the surface is making the perception. Absolutely."

    This should be:
    "Precisely. The anchoring properties are making the affordance and the surface is making the perception. Absolutely."

    ReplyDelete
  14. I think my three states of an affordance might be helpful here.

    Dispositional affordances are persistent and therefore in some sense "of the environment". So, I labeled the state of an affordance prior to being perceived its "persistence" state. One might want to call these "real" or "natural" or something else, but I don't immediately see any benefit to doing so.

    Once perceived but before being actualized, an affordance is in what I called its "perceived" state. And it's affordances in this state that I'm describing as "apparent" because all perceptions are apparent in the sense that they don't - and can't - capture "reality".

    I see the message of the exploding box as being that an affordance in its "perceived" state can be misleading. If it is claimed that affordances are properties of entities in the environment that are always perceived "as they really are" (whatever that might mean), then the exploding box seems a counterexample. I don't see that claim as necessary and am suggesting a way of avoiding it.

    ReplyDelete
  15. But, the problem is making the link between the surfaces and the anchoring properties/ affordances. Gibson conjectures that surfaces constitutues affordances. But that's not right. The exploding boxes case is meant to drive a wedge between these.
    Well, in fairness to Gibson he says "Perhaps the composition and layout of surfaces constitutes what they afford", so I'm still not sure I see a wedge. Note that not only are the anchoring properties making the affordance, they are also making the surface (making these meaningfully connected).

    Charles:
    The information for affordances can be 'apparent' in the way you intend, I think. But Ken's right, affordances are real properties (for TSRM) and don't actually depend on being perceived to be there. However, no-one claims they are always perceived 'as they are'; perception is measurement and this introduces thresholds, noise, learning, etc.

    I'm just adding that even after you've extracted all that's available in the light, there may be things you still don't "know" - eg, that the box will explode when touched. That's what requires "additional work" to get at.
    But is that additional work inference? Or is it further exploration of the environment and light?

    ReplyDelete
  16. Note that not only are the anchoring properties making the affordance, they are also making the surface.

    No. That's precisely what I am arguing is not correct! I'm not going to just note that.

    ReplyDelete
  17. "Note that not only are the anchoring properties making the affordance, they are also making the surface."

    Here's another example. Think of the frame of a carbon fiber bicycle that has been painted. What gives the frame its rigidity (which is an anchoring property for the ride-on-ability of the bike)? It's mostly not the surface, i.e. the paint. It's mostly the carbon fiber matrix and the resin.

    I'm picking the example of rigidity, because TSRM use rigidity as an anchoring property in their example of the marsh periwinkle.

    ReplyDelete
  18. So reading TSRM on periwinkles:

    The analysis of an affordance as a natural kind for an animal is two stage. The first stage is very much a matter of ecological physics: to isolate occurrent physical properties that are invariant over the extension and the perspectives and which are non-accidentally related to the extension of the affordance. The tricky term is ‘non-accidental’. The case of the marsh periwinkle, for example, requires the determination of a lawful relation between an occurrent property o of plant stems and the affordance c (climb-upable) of plant stems so that (for the ecological world of the marsh periwinkle) the concommitance of the intensions o and c is nomically necessary rather than accidental.
    So TSRM seem to be making the argument I suggested, yes? (I'm just checking my reading of this, it's dense in the way only Turvey can manage :)

    So in the case of the paint job: the question would seem to be, does the paint obscure information about the ride-on-ability of the bike? Well, being able to ride a bike entails being able to sit on it (so it's size and shape relative to you matter) - paint doesn't interfere with that. It also entails being strong enough to hold you - not much in the way of visual information for that but a quick explore once you're in contact would reveal it was ok, and again, the paint job doesn't interfere with that.

    So my first pass on this is that your paint job has not obscured any relevant properties of the bike, so we're good. This is why it's safe to paint bikes.

    ReplyDelete
  19. What they are proposing here is that affordances have anchoring properties (that are occurrent physical properties that are invariant over the extension of the affordance).

    That's fine. We agree on this. Affordances, I'm allowing here, have anchoring properties.

    We also agree that surface properties structure light.

    So in the case of the paint job: the question would seem to be, does the paint obscure information about the ride-on-ability of the bike?

    Close. I've been limiting the discussion to vision. So, the question is does the paint job keep the anchoring properties from imparting information to the light? Of course, it does. The paint is opaque.

    Sure, you can detect ride-ability by riding the bike (or maybe some other forms of contact). But, you cannot detect ride-ability through vision. TSRM and Gibson wanted you to be able to perceive ride-on-ability through vision. That's what I'm challenging.

    ReplyDelete
  20. In TSRM, here is a sentence that comes close to what I am challenging:

    "the assumption is of a nomically necessary relation between a property e of the optic array and the occurrent property o, that is a law of the form 'e-ness -> o-ness'" (pp. 265-6).

    What they are proposing here--kind of awkwardly by my lights--is that the anchoring property (which they here call an occurrent property o)structures light.

    But, I am arguing that against the view (assumption) that anchoring properties structure light. The rigidity of the bike frame is an anchoring property for ride-on-ability, but the ridigity cannot structure light because there is a coat of opaque paint over it.

    Similarly, anchoring properties of the pick-up-ability of the normal box (that it is not wired with a touch-sensitive explosive mechanism) does not structure the light; the outer surface of the box structure the light.

    This is a crucial assumption in defending the ecological laws view against Fodor and Pylyshyn. It's important to TSRM. I'm challenging it.

    ReplyDelete
  21. Andrew -

    It occurs to me that no one has detailed how a perceiver identifies an affordance. We have been working under the assumption that it is obvious that both boxes afford "pick-up-able", and Ken has correctly observed that one "really" doesn't. Perhaps the resolution is that neither does owing to insufficient information.

    That actually makes sense to me because there are innumerable reasons that a box might not afford "pick-up-able", so to presume that it does is, well, presumptuous. Why should a perceiver do that based only on the box's size (and perhaps shape)?

    Or put another way, you have emphasized that the available information is determinative. So, isn't the consistent conclusion that if there is insufficient information to determine that one box isn't "pick-up-able", there is insufficient information to determine that either is?

    This occurred to me while reading about the bike and "ride-able". To determine that a contraption like a bike is "ride-able" would seem to require either a considerable knowledge base and a fertile mind adept at inference or prior knowledge adequate to identify a bike on sight.

    ReplyDelete
  22. One box does afford pick-up-ability. You can pick it up. You just can't perceive it.

    This occurred to me while reading about the bike and "ride-able". To determine that a contraption like a bike is "ride-able" would seem to require either a considerable knowledge base and a fertile mind adept at inference or prior knowledge adequate to identify a bike on sight.

    Absolutely. And one uses that information stored in memory, along with one's computational apparatus, and one's visual stimulation to "infer" that there is something rideable. That's what cognitivism says.

    Here's the story told for the boxes. The information in the light does not enable one to perceive pick-up-ability. Now, suppose you do think that subjects perceive pick-up-ability. In that case, maybe it's some sort of computational processing that gets to the perception of pick-up-ability.

    The point is that Gibson and TSRM have to be careful here, else they will be on the road to the much despised cognitivism.

    ReplyDelete
  23. Charles: I missed this comment but spontaneously came to this conclusion too on the other, more recent post.

    Ken: I think restricting the discussion to vision is to sell the perceiver short. I also think (as I say in the other post) that your analysis is effectively leading me to think we just have the wrong affordance in our task analysis. Gibson was very clear that discovering affordances and the information for them was an empirical affair - there's no obvious basis other than that for figuring out what an organism is actually using.

    Gibson (1979) and TSRM may have focused on vision; but Gibson (1966) is much broader and even in '79 he's constantly on about perception more broadly conceived.

    ReplyDelete
  24. So, Andrew, is this your way of admitting that one does not visually perceive pick-up-ability?

    ReplyDelete