Friday 14 January 2011

Reading Group - Gibson (1979) Chapter 6 Part 1

Chapter 6: Events and the information for perceiving events

Thus far, Gibson has been talking about how we perceive objects. But, things in the world often move around, so we'd better be able to perceive things that are extended in time as well. First off, Gibson describes the properties of surfaces that are relevant to events. He then identifies a number of things that can happen to surfaces during events. The nature of these changes will determine what types of optical information might specify particular events (this will be discussed in Chapter 6 Part 2).

Surfaces differ in their resistance to deformation and substances differ in their persistence in a given form. Surfaces can be large (e.g., table) or small (e.g., salt granule) and substances can be solid (e.g., rock), liquid (e.g., water), viscous, viscoelastic, and particulate. Gases aren’t substances, they are components of the medium. Some objects are attached to the ground and some are not. These properties are all relevant to events on an ecological (rather than microscopic or astronomical scale).

A classification of terrestrial events

For the moment we’re assuming that these are terrestrial events and that neither the motion of the sun across the sky nor changes in shadows and light is relevant. What remains are three main types of events that can happen to surfaces: changes in layouts, changes in colour or texture, and change in existence.  

Change of layout due to complex forces

When the layout of a surface changes, it is due to some force acting on it to change its shape. Some familiar changes include moving detached objects from one place to the next. This is a translation of an object’s spatial position (this includes falling and spinning objects, as well). Another important type of change involves deformations of surfaces, such as pouring water, molding clay, or a change in the surface of a moving animal (see p. 95 for a list of changes and examples). It doesn’t make sense to reduce these events to interactions between elementary particles of matter. Dropping a rock is an event, itself and it is not understood any better by invoking atoms or molecules.

These events can combine in various ways. Throwing a ball against a wall involves a collision between two objects of different elasticity. Some events cause others (think of Rube Goldberg machines). And, sometimes, lots of things happen at once (think of the workings of a clock).

When an event causes a break in the continuity of a surface it is usually a big deal. The ball hits a window instead of the wall and the glass shatters. The surface of an animal is torn by a predator. At the extreme, a surface completely disintegrates. This is a change in state, rather than layout.

Only some events are reversible. We can walk backward and forward, but we can’t re-make the window pane.

Change of colour and texture due to change in composition

The idea that colour, texture, and shape are independent is kind of misleading. Colour and texture are both consequences of a substances composition, and texture is really a kind of small-scale shape. Chemical reactions change composition, so they also influence colour (chromatic and achromatic) and texture. These changes can indicate changes in affordances. For example, an apple that changes from green to red goes from sour to sweet – it changes in edibility. Chemical changes with animals can lead to physical transformations – adult birds have different feathers than juveniles. 

Waxing and waning of a surface due to change in the state of matter

When a substance becomes gaseous it becomes part of the medium – it is desubstantialised. Critically, it isn’t specified in the optic array, so for all intents and purposes, it no longer exists for us. Surfaces going out of or coming into existence are important because they indicate a radical change in affordance. If a pond evaporates (changes from liquid to gas) or freezes (change from liquid to solid) the terrain might now afford crossing by an animal.

Summary: What shall we take as an event?

Thinking about what constitutes an event can be difficult if we get too hung up on the role of time. Regardless of what time is, it is events that we perceive, not the passing of time. Similarly, objects do not populate an otherwise empty space. There is no such thing as empty space and never was. Gibson argues for a distinction between sequential and adjacent events and objects. Events are ordered sequentially – things happen one after another in a particular order that is often non-reversible. Objects are adjacent – they occur next to one another and quite often they can be rearranged.

Events may recur (stepping, sunrises), but they do not recur identically. One sunrise is always a bit different from all others, and not in a silly poetic way.  Only some events are reversible. Generally, you can reverse a change in position, but not a change in state. This contrasts with the formal physical laws that (except for thermodynamics) are directionless. But, on an ecological scale this isn’t true – a broken egg isn’t going to be re-made.

Events are nested. An episode of interest can consist of many sub-events and can, itself, be nested within a super-event. Gibson says,

“Some of the best examples of a nested hierarchy of sequential events are found in the human production of events such as speech, music, and the theatre. If we can understand these nested sequences, it may be possible to understand how it could be that in some cases the outcome of an event sequence is implicit at the outset  - how the end is present at the beginning – so that it is possible to forsee the end when an observer sees the beginning” (p. 102).

Events have affordances. For example, a fire affords warmth, a looming ball affords collision, and an approaching deer might afford being eaten. Of course, we don't perceive the world, we perceive information about the world. So, in order for events to have meaning we have to be able to perceive specifying information. In the next part of this chapter, Gibson talks about what information might be available in the optic array to specify various types of events.


  1. There was a special issue of Ecological Psychology (vol 12, issue 2) about whether people can perceive events or just affordances; Tom Stoffregan had a target article about this, and argues ecological psychology doesn't predict events are primary objects of perception. Bingham has a rebuttal, though, that's very much drawn from this Gibson text, that always made good sense to me. (Just another example of how we like scrapping over the details :)

    Only some events are reversible. We can walk backward and forward, but we can’t re-make the window pane.
    Bingham had a great paper on this; he and a former student showed babies point light displays of events that only run one way (e.g. a ball rolling after being hit, so it accelerates then decelerates under friction). You can run the display forwards or backwards, and the latter looks entirely incorrect to adults. Babies can tell the difference between the displays but don't prefer either, showing sensitivity to the information but that they have yet to learn about the underlying dynamics this information specifies. Love it :)

  2. Hi Sabrina -

    With Andrew's patient tutoring, I've been reading your summaries. Thanks for doing them - they've been a real help in trying to tackle Gibson.

    I find one part of the first Bingham paper cited in Andrew's comment confusing. Towards the bottom of page 34 he lists some affordances offered by a knife and others offered by a fly ball. But I see the two lists as qualitatively different.

    The knife offers various affordances each of which can be actualized by an effector that takes some action - hammering, cutting, chiseling, et al. - with the knife. This is consistent with my understanding of affordances. But the items in the second list - affordances supposedly offered by a "fly ball event" - seem qualitatively different: effect an out, achieve a home run, harm a spectator, smash a window. (I infer that the "fly ball event" is a "change of place", one of the things Gibson includes in the definition on p. 30 of Bingham's paper.) But I have problems with each of these.

    First, consider the affordance "effect an out". I'm guessing that Andrew's objection to my description of this case in a BofC comment was due to my injecting "catchable" as a separate, intervening affordance, and I now agree. As Bingham notes, the significance of catching a fly ball is context-dependent. So, in order to take account of the rather epistemically rich context necessary to make "catch a fly ball" become "effect an out", we should consider the affordance to be the latter, which presumably includes (in some sense) that context. But then it would seem that the "fly ball" event must also include (in some sense) that context.

    In which case, isn't it misleading to say - as Bingham does - that "both affordances ["effect an out" and "achieve a home run"] are exhibited by the single fly ball event"? Eg, if the trajectory of the fly ball takes it out of the ballpark, then the "fly ball event" can only offer one "affordance" (or less - see below) - a home run. If the trajectory remains in the ballpark, then there are numerous other affordances, "achieve a home run" being perhaps the least likely to be actualized.

    (As an aside, is the combination of "catch the ball" and "effect an out" an example of nested events?)

    Additionally, in the case in which the fly ball's trajectory is an out-of-the-park home run, I don't see in what sense any affordance is offered. The fly ball will be a home run no matter what (within reasonable expectations) any would-be effector can take. It appears that the environment is determinative of the denouement of the "fly ball event" independent of any action by anyone. And that objection seems to apply to the supposed affordance "smash a window". (I tend to think in terms of the effector being a person, so perhaps that's clouding my understanding, at least in the case of the window.)

    ========== continued ============

  3. "Harm a spectator" is trickier. I can envision three scenarios in which a spectator could be harmed. First, the spectator tries to catch the ball and gets hurt in the process. But then, the affordance is "catchable" and the harm seems merely an unintended consequence of attempting to actualize that affordance (like picking up/touching the exploding box). Second, the spectator is masochistic and intentionally moves so as to be hit. But then, the affordance would seem better described as something like "collidable with" (one of Sabrina's examples). The harm is now an intended consequence dependent upon the spectator executing specific movements, ie, actualizing the pre-actualized affordance. Finally, a spectator could be hit totally by accident. But then the spectator seems not to be an effector, at least as I understand the term. By the symmetry of the relation, I suppose one could view the spectator as affording "hittable", making the fly ball the effector, but since the role of the ball is merely to passively participate in the "fly ball" event, that seems to stretch the concept of "effector" beyond what I (mis?)understand it to be.

    To reiterate an earlier point, my quibbling over nits like this isn't an attempt to put anyone down, it's just my way of exploring the nuances of the concepts.

  4. Hi Charles,

    There are a few things going on here which make the fly ball examples seem qualitatively different than the knife examples. I'm going to argue that they are, in fact, the same, but we'll see what you think.

    I know you know this, but just to be clear, affordances "can exist independently of their actually being in the relevant dispositional relationship" (Bingham, 2000, p. 30). It's easy to keep this in mind for inanimate cases like salt affording dissolution in water and it's also fairly simple for objects that afford some action by a person (e.g., the hammer affords grasping even when there's no one around). Bottom line, affordances are there regardless of context, but the ability to effect the affordance is not.

    Bingham argues that events are things with properties (rather than being, themselves, properties). This means that, just like objects, events have affordances. But, there's a critical difference between objects and events. Objects seem to persist (more or less) over time, but events come in and out of existence.

    If I leave my coffee mug sitting on the table while I leave the room, I expect it to be there when I come back. An event, on the other hand, doesn't exist when it's not happening. As a consequence, each instance of an event is different. I think this difference leads to difficulty in thinking about the properties of events, but, I think this difficulty is overcome by focusing on the types of things events are.

    Bingham makes the point that events are the same to the extent that they exhibit the same spatiotemporal dynamic, and he shows that people can discriminate between many events on the basis of this information only.

    The thing that convinces me that the knife example is of the same type as the fly ball example is to remember that for events their thingness is their spatiotemporal dynamic. It's this dynamic that persists in instance after instance of the same event.

    In Bingham's example we have two different things - a knife and a fly ball. One of the things is an object and one is an event. Things have affordances, which exist whether or not there is currently the capacity to effect them. Knives afford cutting even if there's no one around to use them. Fly balls afford harming a spectator, even if there's not one in the stands. Fly balls also afford home runs, although a particular instantiation of this event might not make it out of the park.

    You're right that a particular fly ball, once in action, either is or is not going to be a home run. This is because the event was in the proper dispositional relationship with the environment and hitter. Similarly, if a knife is in the proper dispositional relationship with a person and the environment it's always going to cut, say, an apple. There are lots of reasons why the fly ball event might not make it out of the park, just as there are lots of reasons a knife might not successfully cut an apple.

    In terms of harming a spectator, I don't think that the intention matters. In each of the alternatives you propose, the fly ball event effects harming a spectator. And, whether the fly ball hits someone or not, it affords harming a spectator. I take your point about whether the spectator is effecting something when he tries to get in the way, but I don't think this changes how we think about the affordances of the fly ball event.

    Again, I think this is easier to think about with object examples. A knife affords cutting flesh. So, this means that knives can cut people. There are lots of ways that this affordance can be effected. I might cut myself by accident while slicing apples. I could cut myself on purpose. I could drop the knife and cut my foot. These examples contain various degrees of intention but each of these cases depends on the same affordance.

  5. And, yes, I do think that catching the ball and getting an out are examples of nested events :)

  6. Sabrina -

    Thanks for the detailed reply. I'm actually with you on most of that, but I still have a problem with the "achieve a home "run affordance.

    It appears that we need a somewhat more precise definition of the "fly ball" event. Implicit in your comment is that that phrase represents a specifiable entity (ie, one can say what characteristics a generic event of that type must have) that is instantiated at the time of impact between ball and bat. And as you suggest, once the event "fly ball" is instantiated, the total environment dictates whether or not the event will end up "achieving a home run". So, we agree up to that point.

    But here are my essential questions: when does an affordance begin and end, and what action must a prospective effector take - if any - to actualize the affordance? I can easily see that a "pitched ball" event (subject to some obvious conditions) affords "achieve an out-of-the-park home run" up to the point of impact with the effector (bat or hitter) at which point it is either actualized or not. Likewise, I can easily see that an in-the-park "fly ball" event can afford "effect an out" (and again, other affordances) until some effector (player, glove, etc) actualizes one of the affordances. And I can - a little less easily - see that one of the affordances of the in-the-park "fly ball" event might be "become injured/damaged by", actualized upon impact with someone or something (only, however, if mere existence is sufficient for being an effector - as in the case of the window). But I can't answer those questions for the "fly ball" event and the affordance "achieve an out-of-the-park home run". So, I continue to question whether an instantiated "fly ball" event can afford both "achieve an out-of-the-park home run" and "effect an out". And I would argue that the affordance "achieve an out-of-the-park home run" has been actualized before instantiation of the "fly ball" event and thus is not even available to that event.

    And as a trivial point of interest, although not a big sports fan and therefore not sure, I think that an in-the-park hit can become a home run (that's why why I keep referring to "out-of-the-park" home runs) so that, strangely enough, an "in-the-park fly ball" event can afford "achieve a home run".

  7. Hi Chris,

    Yes, I see a bit more clearly where you're coming from now.

    We agree that an object always has a whole complement of affordances. But, objects always occur in a particular context which limits the affordances that can be effected at a given time. We can think of these limiting contexts as instantiations of the object. When my coffee mug is sitting on my desk it is not droppable. The context rules out effecting that affordance. But, the affordance still exists.

    Similarly, an event always has a whole complement of affordances, but a given instantiation of an event might only permit effecting some of those. From this perspective, the affordances are always there and don't have a beginning and an end. As a consequence, a given fly ball continues to afford both effecting an out and getting an out-of-the-park home run, even though only one of those is possible in a particular instantiation.

    Perceiving affordances is often about perceiving possibilities. A child perceives that she would be able to get the cookies on the counter if only she were a bit taller. And, a hitter perceives that he would have been able to hit the ball out of the park if only he'd swung a bit harder. I also don't need to see a spectator get injured by baseball to know that getting hit by one would hurt. Being able to perceive these possibilities is only an option if affordances persist when they are not able to be actualised.

    "And I would argue that the affordance "achieve an out-of-the-park home run" has been actualized before instantiation of the "fly ball" event and thus is not even available to that event."

    It seems to me that effecting the affordance is coincident with the beginning of the event. The shape of the event is dependent on the pitcher, the hitter, the bat, maybe the weather and these all come together at the moment the ball hits that bat. It is the shape of the event that determines whether or not that particular affordance is actualised. I don't see that this would make the affordance unavailable (I'm taking unavailable to mean 'not able to effect the affordance'). Am I misinterpreting your point?

  8. As a consequence, a given fly ball continues to afford both effecting an out and getting an out-of-the-park home run, even though only one of those is possible in a particular instantiation.
    I think I'm with Charles on this one. The event class 'fly ball' is capable of affording either, but a specific fly ball can only afford one or the other (with some interesting close calls possible, of course).

    In fact, this is one of the important things you have to learn when learning to field. I have a vivid memory of playing softball and the day I realised I could perceive 'catchable by me' almost immediately - some balls just weren't going to land anywhere near me or sail over my head. Learning this distinction is actually quite critical in become an expert fielder.

  9. Andrew,

    My take on this is that "catchable by me" is realising that you can effect a particular affordance, not perceiving that the ball is, in principle, catchable.

    To repeat from the earlier comment, we all agree that affordances "exist independently of their actually being in the relevant dispositional relationship" (Bingham, 2000, p. 30). Whether or not you can catch the ball, it is the type of thing that could be catchable if you were in the right relationship to it. That's the affordance, and that persists even if you are no where near being able to actualise it in a particular instance.

    I agree that the event case, particularly in reference to an out-of-the-park home run, seems strange. But as far as I can tell I'm drawing a straightforward analogy to the object case.

    We all seem to be on the same page with saying that an object continues to have affordances even if some of those can't currently be effected. No matter how weird it seems, this is identical to saying that an event continues to have many affordances even if a particular instantiation rules out the possibility of actualising some of them.

    I gave the example of not being able to drop my coffee mug when it's sitting on my desk. I can't actualise that affordance unless I change the instantiation of the coffee mug, say, by picking it up. Do you agree with this?

    For the fly ball, I can't actualise the out-of-the-park home run affordance if the given instantiation of the fly ball is in the right relation for someone to effect an out. But, the affordance hasn't disappeared. If you think the the coffee mug example was correct, then it seems like this would be correct as well.

    If you disagree, can you reply in terms of how these two cases differ?

  10. @Charles - I apologise for randomly calling you Chris. I have no idea where that came from!!

  11. Not to worry - compared to what my wife calls me (but lovingly!), "Chris" is a welcome alternative.

    Something more substantive (hopefully) to follow after I have time to digest your last comment. Believe it or not, affordances have turned out to be a serendipitous find for me, seemingly relevant to something else I'm doing and are keeping me busy.

  12. Sabrina -

    Are you following the Gibson-related threads at Bounds of Cognition? The exchanges here and there are sufficiently intertwined to keep me a bit confused as to what to comment on where. I don't mind cross posting when it's clear that something is relevant to both, but I don't want to needlessly clutter up multiple blogs.

    Eg, my first (and so far only) comment on the "S does not perceive ..." thread seems relevant to a response to your last comments here - with which I am struggling because although there seems a lot to be said, it's not clear that I'm capable of saying it!

  13. That last sentence reminds me of a description of one of the prominent US practioners of punditry (our principle growth industry at present), which unfortunately applies to most of them:

    Nothing to say but the ability to say it well.

  14. I think there's an asymmetry between catching the ball and it being an out-of-the-park home run.

    a) Catching: any ball affords catching, because if you can get to the right place at the right time then you will intercept it (the ball must fall back to earth).

    b) Out-of-the-park: only certain balls are actually on a trajectory that will take them out of the park.

    This occurs to me out of the discussion of 'anchoring properties' on Ken's blog. These are the physical properties that produce a disposition and 'anchor' it to the real world. Salt is disposed to dissolve in water; it's anchoring properties which lead to this is that it is an ionic salt and water has the right electrical characteristics.

    So: all fly balls are projectile motions and will fall to earth. They therefore all share the key properties which make them all afford catching. Only some fly balls are projectile motions which will carry them far enough; only these afford going all the way.

  15. Hi Charles,

    I will keep an eye out on Bounds of Cognition for ways to consolidate these discussions! Your comments on that post did home in a bit on what you were getting at in terms of the persistence of affordances.

    There is some awkwardness in the language I have available here and I'm not sure I'm entirely consistent in what follows so I just want to begin by clarifying a few things:

    I'm taking a cue from Andrew in thinking about affordances as the sub-set of dispositions for which perceptual information exists. An object may be disposed to something, but if no information exists about this then I can't use it to guide my behaviour. Following from this distinction, the persistence of an affordance should hinge on the potential to perceive information about it. If there is no information, then it's just a disposition.

    My background is in categorisation, so I have exactly the same concerns as you do about claiming that affordances have some Platonic truth about them. I haven't exactly decided how to deal with this yet, but I have taken a preliminary stab at the problem below.

    We all seem to agree that affordances persist if there is no one around to actualise them. But, that doesn't necessarily mean that affordances persist if some change occurs in the object that makes the affordance unactualisable. There are two ways that an affordance might become unactualisable. One is reversible and one is not. The irreversible case refers to a destruction of or permanent change in an object. The reversible case refers to some immediate contextual factor which makes it impossible to directly actualise the affordance.

    The irreversible case: If I break my coffee mug then it doesn't afford holding liquid anymore (and it doesn't have that disposition either).

    The reversible case: If my coffee mug is full of coffee then it doesn't immediately afford my pouring anything else into it (although it still possesses anchoring properties that permit the object to potentially have more liquid poured into it). As a perceiver I know that I could in principle pour something into this coffee mug if I just get rid of the stuff that's in it right now. The liquid in the mug affords actions like being poured out or drunk (if it's coffee). Actualising one of these affordances leads to an empty mug. This leads to my being able to effect the affordance of being able to fill the mug with liquid. The process works because I can perceive a chain of affordances that led to being able to fill the mug with liquid.

    I can also imagine a different type of reversible case. Imagine a situation where the mug is physically intact (so, in principle, is fill-able), but I cannot perceive a chain of affordances that leads me to my goal. Say, the mug is buried underground and no one knows it's there. There is no information available to any perceiver that could lead to an intentional chain of behaviours culminating in filling the mug with liquid. It seems like the underground mug no longer affords being filled with liquid. I suppose it still has this disposition, but since it cannot be perceived it is not an affordance.

    This isn't a permanent state of affairs since it is possible that someone will accidentally unearth the mug. This creates a weird possibility where an affordance can come in and out of existence depending on whether there is any perceptual information available about the affordance.

  16. To summarise this hodgepodge of ideas:

    1a) An affordance exists if there is perceptual information available about an action-relevant disposition of an object

    1b) An affordance exists if there is perceptual information about other affordances, the effecting of which lead to the perception of the affordance of interest, regardless of whether there is anyone around to perceive it (this is the full coffee mug case).

    2) An affordance does not persist if the object is destroyed or changes state such that its anchoring properties no longer support the behaviour related to the affordance. Obviously, as a result of such a transformation there will be no perceptual information available about the disposition and hence no affordance.

    3a) An affordance does not persist if the context means that there is no perceptual information available about that property or any other property that might play a role in an intentional chain of behaviours that culminates in the perception of affordance x.

    3b) If the context changes in a way that makes perceptual information about the property available again, then the affordance exists again.

    The thing that troubles me is the status of 1b - an afforance that cannot be effected because some temporary contextual factor means that it's not actualisable. I think Andrew would argue that the "fill-able" affordance doesn't exist for the full coffee mug (tell me if I'm wrong).

    Clearly we are not totally stumped by how to handle this situation. We know that pouring out the liquid will permit the action we want to perform. But, I'm struggling to come up with an explanation for how we know things like this that neither requires that we perceive the affordance when the mug is full nor use some type of inference to solve the problem.

  17. Sabrina -

    This helps. Something that had been bothering me about the exchange at BofC was that an affordance - supposedly a key feature of an approach to analyzing perception - had seemingly become devoid of any perceptual content (eg, various imperceptible dangers). And since I wasn't clear on how to think about dispositions, I just took "disposition" and "affordance" to be essentially synonymous. Clearly distinguishing them and levying perceptual information requirements on the latter (as Andrew also has done) substantially clarifies the picture for me.

    I have been rethinking the idea of actualizing an affordance in a sequence of steps and now question whether that's a meaningful idea. If I'm correct in understanding that Gibson views perception as a continuous process, viewing affordances as discrete entities that are dealt with in discrete steps doesn't seem to fit.

    We've been taking about events in terms of objects in motion, and implicit in the concept of effecting affordances is that the effector is also in motion. So, our perceptual system comprises a prospective effector perceiving affordances and responding to those, and in that system both the position of the prospective effector and the affordances are continuously time varying. How do discrete events get into the picture?

    Additionally, the nature of the source of affordances - ie, an object - would seem to enter in only in that recognizing the source as a particular kind of object will determine the specific affordances available. But at least at first, it would seem wise to simplify the scenario as much as possible, and eliminating (or at least minimizing) epistemic aspects will help. If you don't recognize the source of perceived affordances as a specific object called a "full coffee cup" but merely as a surface, then neither "pour-into-able" nor "pour-from-able" is an available affordance - and the concern about dealing with that object evaporates.

    Which is meant not to beg the question but just to simplify the problem. So, I'll also take a stab at a direct answer in cases where you do recognize the affordances as being offered by a specific object like a full coffee cup. In order to evade the requirement of immediate actualization, I have been distinguishing affordances that are immediately actualizable from those that aren't in order to deal with affordances that can only be actualized via intermediate steps. I now doubt the wisdom of that view. Rather than a problem, immediate actualization now seems a possible solution.

    Andrew has tried to lead me down this path a time or two, but I didn't even see that there was a path. Now perhaps I do. Recalling that the perception process is continuous rather than discrete, the immediate actualization requirement on affordances would seem to mean that an affordance that is immediately actualizable at time T-must have been actualized at T+. You introduced the idea of "intentional" actions, which suggests distinguishing affordances - things that require immediate actualization due to the continuous nature of the perception-response system - and intentions - things that are, as you suggest, involved more with inference and long-term planning. Yes, a full coffee cup can be refilled by approaching it, grasping it, upending it, uprighting it, etc, etc. But is that complete complex motion (continuous notwithstanding being described as a sequence of discrete steps) really something that qualifies as "actualizing an affordance" and needs to be - or even can be - captured in Gibson's math model?

    Thankfully, these ideas seem to offer hope of killing off imminent disaster scenarios. Eg, a box that is disposed to explode on contact isn't disposed to be picked up or touched, and in any avent can't be offering those as affordances since they can never be immediately actualizable. Similarly for objects that explode due to proximity detectors. Etc, etc.

  18. In that last comment, I obviously redefined affordances, though not intentionally - (what I hope was) the logic just "made me do it". I think I now see more clearly why.

    There appears to be an inescapable tension between the concept of an affordance as being something that can, so to speak, be "perceived at a distance" as actualizable - where "distance can be spatial, temporal, or both - and levying a requirement of immediate actualization on affordances that can be so seen. Adding a constraint that certain conditions must be in place before actualization doesn't appear to help - there is still a difference between the nature of the affordance before those conditions are in place and once they are in place. And that's what I was trying (although I didn't know it at the time) to capture by calling the former "apparent". And all of Ken's examples are, I think, cases in which there is an interceding event that precludes the conditions being met.

    If I have that right, something has to give. In that last comment I gave up on "apparent" affordances. Maybe the wrong choice, but a choice seems mandatory.

  19. Sorry for the multiple fragments, but I'm thinking about this during breaks from doing something else, so it's unfolding in real time.

    Let's return to Sabrina's fly ball. At the moment it's hit into the air it acquires a variety of dispositions: to be lofted by a gust of wind, to hit a window or a spectator, to be caught by a fielder resulting in an out, to land on the ground outside the ballpark resulting in a home run (or a foul ball), et al. And a person may be able to perceive some of those depending on their knowledge: eg, one can't perceive an out disposition unless one knows what "out" means.

    A perceiver of those dispositions may also perceive conditions under which a specific disposition may afford the opportunity to effect a result, and may also perceive actions that can be taken in order possibly to be in a position to take advantage of that opportunity, ie, to become an "effector". One could then say that at that point the combination of the disposition and the conditions necessary for effecting the result become a "conditional affordance" offered to the perceiver. The perceiver then activates the continuous perception-action control loop that is intended to result in being positioned so as to be able to effect the result. If successful, the conditional affordance becomes an "immediately actualizable affordance" and actualization occurs.

    During this real time event, the perceiver's assessment of the conditions required for allowing effecting of the result may change (eg, Andrew's skillful outfielder). The perceiver may even conclude that conditions required for effecting the desired result are no longer possible, at which point the conditional affordance ends. Consistent with Sabrina's sense that a fly ball event retains something throughout its duration, the dispositions of the fly ball event don't end when the conditional affordance ends.

    Seems to hang together, although it obviously diverges in terminology from the way we've been describing things. Problems?

  20. As I understand what you're saying, you're trying to give affordances a dynamic quality; you want the affordance to have a time-varying quality.

    I'm not sure I'd put the time-varying aspects on the affordance (although some affordances may actually be dynamic, I don't think they all need to be). It still strikes me that this work is being done by the perceiver; affordances hove in and out of 'actualisability' as a function of what the perceiver is up to and what information they have access to.

    I think giving affordances a certain stability, a certain invariance, is quite important. Learning requires a stable target: higher order invariant information variables have the required stability, but this stability depends on there actually being something out in the world that is interacting lawfully with (for instance) light.

    The time extended nature of the perception-action process is great. But I think it's this process, and not what the process is trying to find out about, that is the home of the particular dynamic quality you're looking for. Obviously events and affordances in the world can be dynamic, but I don't think that's the flex you want.

  21. My hypothesized scenario comprises an object, possibly in motion; a perceiver, also possibly in motion; and the rest of an environment, possibly changing with time. If (as I assume) any reasonable concept of an affordance is a function of the contents of such a scenario, it is necessarily time-varying. For me to be "put[ting] the time-varying aspects on the affordance" would seem to require that some aspects of that scenario are substantially wrong. But which ones?

    I've explicitly assumed that dispositions are long-term persistent. And since the process I envision is continuous (and presumably "smooth", ie, continuous first derivative), the conditional affordances are short-term persistent. I have hypothesized that the perceiver continuously reassesses the conditions for actualization based on changes in perception, ie, "learns". What sort of additional learning is precluded?

    Re your last paragraph, I don't understand the phrase "what the process is trying to find out about". The process I envision is a continuous perception-reaction feedback loop trying to bring the perceiver into a position in which the conditions for actualization of the conditional affordance (one of which is the perceiver's position relative to the object) are met so that it becomes an immediately actualizable affordance. The quoted phrase suggests the presence of something other than that process and its components, but I don't see what that "something" might be.

    In any event, I see the key observation in my last few comments as being the inescapable tension between affordances perceived by a prospective effector but not immediately actualizable and the requirement of immediate actualization in the presence of an effector. I'm inferring from Ken's persistence in attacking specific presumed "affordances" that they are used as examples in the books he's critiquing. If they are, he may well be right - they don't appear to meet the requirements for "affordances" as I understand those to have been defined. If so, it seems that the definition needs some tweaking, the examples need to be dropped, or some of both. I'm just proposing a possible starting point that hopefully highlights some underlying assumptions and focuses the discussion.

  22. I'm on board with Andrew about wanting to keep affordances relatively stable, but I like the idea of paying attention to the time-varying aspects of what actions are available to us. I'm also interested in this problem of achieving some seemingly goal-directed complex series of actions to actualise some affordance.

    I'm going to return to the idea that an affordance is a disposition that we have some perceptual information about. The anchoring properties of the object will mean that its dispositions are relatively stable, although things can always happen to change this.

    The relationship between the disposition and the perceptual information available about it will also be stable because that's what information is all about. However, our access to this information might not be stable depending on particular contextual factors (e.g., is the information is visual and the object is occluded).

    Now, let's also go with the notion that affordances must be actualised in the presence of an effector. If we take this idea seriously, then we always actualise any affordance for which we are an effector.

    To the extent that we don't actually freak out and try to do a million things at one time this suggests a couple of possibilities: 1) the requirement to immediately actualise an affordance that we can effect is incorrect or 2) being able to effect an affordance is actually a pretty high bar.

    I suggest sticking with the second possibility for the moment.

    Part of the tension that Charles identifies is that there is quite a leap between the simple scenario (immediate actualisation) and the complex one (actualisation via a series of actions). There is something non-ecological seeming about enacting complex sequences of actions in the pursuit of some goal. One thing that creeps in here is language about accomplishing a series of discrete steps when, as Charles mentions, this is problematic for a continuously perceiving/acting system. So, how do we acount for complex or nested actions that involve eventually actualising some affordance of interest...

  23. Here's what I'm thinking. Perceptual information tells us about affordances, so having access to this information is equivalent to perceiving affordances. For any given situation, we can probably perceive multiple affordances. However, the amount of attention we pay to these could depend on a number of things (their perceptual salience, for instance). I would guess that the amount of attention that we pay to particular affordances also depends on what type of device we are at the moment. If we are a walking device, then we will pay more attention to walking-relevant affordances and if we are a catching device then we will pay more attention to catching-relevant affordances.

    For the set of affordances that we perceive we will only be able to effect one at a time (I don't think this is strictly true, but true enough for now).

    Whatever actions we take might change our relationship to the environment to permit us to actualise a different one of the set of affordances that we perceive. They might also change how much attention we pay to particular affordances, for instance if we move closer to something and can see it more clearly.

    Now, without any explicit goal in mind, I'm guessing that an organism that operated like this would produce at least some behaviour that looks intentional and planned. In cases where our behviour changes both the salience and the perceivability of affordances in the environment it will be the case that actions at tn depend on having gone through tn-1,tn-2,tn-3...So, for instance, our third party interpretation of an organism acting like this would probably invoke causal language - it did x to reveal y to get to z. But, we don't need to invoke this language to explain the organising principles of the organism's behaviour. All it was doing was effecting whatever affordances were available at a given moment (also taking into account that it might not be paying attention to all possibly acutlisable affordances at any given moment).

    I realise that I'm still using the language of discrete actions and affordances. I do think this is potentially an issue, but I need to mull it over for a while before writing anything about it.

  24. Sabrina -

    All of that sounds good. I have just a couple of questions (for either you or Andrew):

    "an affordance is a disposition that we have some perceptual information about. The anchoring properties of the object will mean that its dispositions are relatively stable, although things can always happen to change this."

    This description suggests that persistence is really a property of dispositions rather than affordances. This makes sense to me since being functions of perception, it would seem that they wax and wane with the changing geometry of perceiver and entities in the environment.

    Affordances seem to exist in multiple "states", so to speak: when they are first perceived, when the perceiver has identified them as actualizable as the end of a multi-step process, when they become immediately actualizable, et al. It seems natural to add a modifier to indicate which state an affordance is in, but I have sensed some resistance to doing so (or at least to my specific choices of modifier). If there is no such resistance, it seems desirable to explicitly define and name the distinguishable states in order to clarify discussion.

    I agree with your assessment of the handling of perceived affordances that require intermediate steps before actualization. It appears that the problems Ken is raising stem to some extent from focusing on affordances identified from an initial perception of a surface but ignoring changes as the intermediate steps unfold and new information is acquired (changing the relative "salience" of the affordances - a term I like!). Eg, a surface that appears approachable and sit-on-able at a distance may appear no longer sit-on-able once approached.

    Using the language of discrete steps to make discussion more intuitive seems OK to me, although there is the danger of forgetting that the underlying process is actually continuous - which I did for a while.

  25. Part of my reluctance is simply that the states aren't inherent to the affordances. Affordances are dispositions, not relational. Organisms, however, stand in a dynamic (time-varying) relation to that affordance, which is reflected in the state of the information at each time.

    So it's not that the affordances come in different states: it's that the affordance-information-organism system has different states.

    Sabrina, I like the SR type formulation (yay behaviourism! :). That said, SR chains of this type failed to explain behaviour because they aren't goal directed, and this certainly seems to be a feature of behaviour to be accounted for somehow. One solution might be learning to effect the higher order affordance 'getting-up-to-locomote-to-the-pickupable-object'. Then, of course, the question becomes what is the informational basis? Is there information for the higher order affordance, or are we actually effecting the bits?

  26. As I suggested before, searching for just the right phraseology strikes me as great progress. And capturing the distinction as different states of the whole system seems an improvement. So I'm happy on that point.

    "Affordances are dispositions ..."

    But the converse, I take it, is not so. Hence, ...

    "Organisms, however, stand in a dynamic (time-varying) relation to that affordance, which is reflected in the state of the information at each time."

    ... seems problematic since at times that dynamic relation involves the affordance "disappearing" in the sense that the flow of perceptual information ceases (eg, occlusion). As with Ken's booby traps, it seems inconsistent to insist that something is both persistent and temporary.

    The disposition is persistent - why isn't that enough? Eg, the inertia in the system will keep it going during brief absences of new perceptual information from an affordance. There seem to be benefits to viewing it that way. What are the costs?

  27. The disposition is persistent - why isn't that enough? Eg, the inertia in the system will keep it going during brief absences of new perceptual information from an affordance. There seem to be benefits to viewing it that way. What are the costs?
    Actually there is exactly this kind of perceptual persistence; the 'present' isn't a razor edge, but a tunable, calibrate-able amount of time. But this is because the affordance does, indeed, swing in and out of view. It's just a fact.

    So to capture the whole process, you have to acknowledge this fact and then investigate how the system deals with it; that's why you need to full systems account.

    Gibson's initial swing on this, by the way, is 'progressive occlusion', the event of one thing passing in front of another. Sabrina covered it some here.