Tuesday 7 June 2011

Perceiving long distances in action scaled units

I have so many things I need to write up just now, but it's been a struggle finding the time. I hope to post on Chemero's last chapter, task-specific devices, calibration and some new coordination data soon. In the meantime, I thought I'd take advantage of the fact that I'm reading some new articles on an interesting topic, and I wanted to organise some thoughts and see if anyone had any comments!

Perception is action-scaled

Traditional theories of perception claim that we perceive the world in generic terms, and must transform that perception into a task relevant variable after picking the information up. The ecological suggestion is that the act of perception itself is directly scaled in action-relevant units, and that this perception will therefore be task-specific. In order to directly perceive action relevant properties (i.e. affordances) perception must be smart (think of the analogy of the polar planimeter).

We are capable of perceiving the distance of things in the world; but we don't perceive them as being '6m away'. Instead, the system is interested in how to reach for an object, so you need to calibrate your perception of distance in terms of, say, arm length units. Calibration is the process of placing a measurement on a scale, and the ecological approach has been interested in action relevant scales such as arm lengths (for reaching; Mon-Williams & Bingham, 2007) and leg lengths (for stair climbing; e.g. Warren, 1984). One of Chemero's points is relevant here; body scale is probably only a proxy measure for ability to perform the action and the real action scale the system is using (the effectivity) will be more complicated. But body scale is mostly where the field is at right now.

Perceiving distances beyond reach space
 
I'm currently interested in distance perception and the scaling it uses because I've begun collaborating with Arthur Zhu and Geoff Bingham on their throwing research (one paper of which I blogged here). We ran some studies in parallel this year on learning to throw to hit a target at 10m, and one of the interesting questions lurking in here is, how are people perceiving the distance to the target? Learning to throw accurately to a specific distance entails learning to produce a useful combination of release angle and release velocity; but these release parameters are the last chance you have to affect the ball's trajectory and so you must select these parameters on the basis of the perceived distance to the target (amongst other things). At 10m, your eyes are parallel and hence there is little information from this geometry about where things are in space. You need to visually perceive the location of the target along the floor from you, and you need to perceive this in throwing relevant units.

There are two basic suggestions for action relevant scaling over long distances: eye height units (Sedgwick, 1986) and effort (e.g. Proffitt, 2008). Eye height is potentially quite useful, because it's always available in optic flow. The suggestion here is that you treat one eye-height as a unit, and perceive distance on a scale using that unit; 10m would then be some number of eye heights away. This is fine, and there is some evidence that eye-height matters, but it's not clear what is action relevant about it. If you wish to perceive the affordances of a spatial layout related to distance, what is it about eye-height that affects your ability to effect that affordance?

Effort seems to be more in line with Chemero's suggestion that 'ability to perform the action' is the relevant dimension to go looking. However, it's terrible vague and hand-wavy. Essentially, the suggestion from Proffitt's group is that when viewing a space, your perception of those distances is scaled in terms of the amount of effort required to perform the task you're about to do. If you're about to walk the distance, your perception is scaled to locomotor effort; if you're about to throw something the distance, your perception is scaled to throwing effort. Perceived distance is therefore directly scaled to action relevant units at the time of perception by virtue of you being one task-specific device rather than another. 

When and how are spatial perceptions scaled?

To test this idea, Witt, Proffitt & Epstein (2010) ran a simple study to tease apart when your visual perception of distance is scaled (at the time, or by a post-perceptual cognitive process) and to demonstrate task-specific effects on this scaling.

There were two groups; the first was told they were going to walk to a target 6m or 8m away, the second were told they were going to throw to that target. The two groups then walked on a treadmill; this recalibrates the relationship between walking and optic flow such that walking at a set speed produces minimal optic flow, instead of flow at the set speed, because you aren't actually translating through the environment. This makes people feel that they need to expend more effort to walk a given distance (because effort now isn't producing much flow, you have to increase effort to get the appropriate flow to suggest you've moved far enough). Both groups were then asked to walk blindfolded to the target distance (so the throwing group were now doing a different action).

The hypothesis was this: if you viewed the target at 6m with the intention of walking to it, your perception of the distance was scaled according to the effort required to locomote that far. If you recalibrate this scale with the treadmill so that more effort is required to cover that distance, and you are then asked to walk blindfolded until you think you have covered the distance, you will expend more effort and end up walking further than you would have. If you viewed the distance with the intention of throwing an object that far, your perception would be scaled in terms of the effort required to throw that far; this effort is not recalibrated by the treadmill experience and hence, when asked to instead walk the distance, you won't expend extra effort. If perception is not action scaled at the time of perception (pre-treadmill) but rather transformed in a post-perception cognitive process, then both groups should show the effect of the treadmill.

The results showed that for both distances, walkers always blindwalked further than throwers-turned-walkers if there was a treadmill manipulation. This demonstrates that distance perception is action-scaled at the time of perception, relates to effort, and is task-specific.

Thoughts

There are some devilish details to worry about:
  • Blind walking is designed to be an action measure of your visual distance perception at time t, in the absence of the ability to refine and update that perception with ongoing visual perception of the scene. It gets used quite a bit, and seems to work ok. It is, however, a little noisy and over longer distances people begin to get conservative and undershoot distances. This doesn't mean they thought the distance was shorter than it was, it can simply mean that, under uncertainty, they are playing it safe. In this study, at 8m, both groups undershot in the control condition and the throwers-turned-walkers did so in the treadmill study. So blindwalking is a complicated measure of what you thought the distance was.
  • They tried this in reverse (trying to get walkers-turned-throwers to overshoot because of the treadmill experience) but blind-throwing is even more variable than blind-walking, so nothing came out. 
  • I'm interested in other changes in people's locomotion: does their stride length change? Stride frequency? The speed and hence the time taken to walk a given distance? And how do these relate to the treadmill activity? What about other measures of effort (e.g. VO2 max?)
Other thoughts:
  • If throwers-turned-walkers perceived the distance to the target in terms of throwing effort, why were they then able to use this information to walk quite accurately? 
  • What does effort mean, anyway? What is the informational basis for this? I can kind of see a story about locomotion, where over the course of the day and all your experience walking, you would develop a highly stable calibration based on the effort you needed to expend to traverse a distance which you could then use to prospectively control later actions. But what about throwing? When my participants are looking at the target and preparing their throw, what type of effort scale do they have information for? Does it take time to develop? Is this why you need to warm up a little? Is this why the human perception-action system seems to contain biases that help structure and inform throwing - it was an important enough skill that wiring in some help to overcome the informational limitations was warranted?
Effort seems to be me to be the kind of scaling we're going to need, but it is as yet entirely unclear how to cash it out with any useful precision. I'll keep blogging on this topic as I read because it's helping organise my thoughts, and frankly, any bright ideas from any quarter are welcome :)

References
Mon-Williams, M. & Bingham, G.P. (2007). Calibrating reach distance to visual targets. Journal of Experimental Psychology: Human Perception and Performance, 33(3), 645-656.  Download

Proffitt, D. R. (2008). An action-specific approach to spatial perception. In R. L. Klatzky, M. Behrmann, & B. MacWhinney (Eds.), Embodiment,ego-space, and action (pp. 179–202). Mahwah, NJ: Erlbaum.

Witt JK, Proffitt DR, & Epstein W (2010). When and how are spatial perceptions scaled? Journal of experimental psychology. Human perception and performance, 36 (5), 1153-60 PMID: 20731519 Download

Sedgwick, H. (1986). Space perception. In K. R. Boff, L. Kaufman, & J. P. Thomas (Eds.), Handbook of perception and human performance (Vol. 1, pp. 1–57). New York: Wiley.

Warren, W. H. (1984). Perceiving affordances: Visual guidance of stair climbing. Journal of Experimental Psychology: Human Perception and Performance, 10, 683 703. Download

9 comments:

  1. Random thoughts:
    "body scale is probably only a proxy measure for ability to perform the action"
    I'm increasingly suspicious of this approach. On the one hand, it is clearly true. On the other hand, I'm worried that the ecological approach loses some of its cleanness. 'Ability' is a messy, messy word. You seem to admit this when you call Chemero's suggestion 'hand wavy'.
    ----

    Effort also seems problematic. If you think people are basing perception on effort, why not manipulate effort in more obvious ways -- have them walk the same distance up hill vs. down hill, or over obstacles, or on an oiled floor, or while wearing a heavy back-pack, etc.

    ---

    The cleanest stories you can tell in ecological psychology are about how the shape of the body makes certain actions possible, and organisms are attuned to information specifying the layout of the environment in terms relevant to their particular bodies. Anything more than that requires more care in talking about than we like to think.

    ReplyDelete
  2. Although I can't recall the specific references, Proffit and co have done almost all the effort manipulations you can think of (hill steepness, wearing backpacks, glucose depletion etc etc)

    ReplyDelete
  3. Eric, Gavin is right; Proffitt's group specialise in manipulating effort in exactly the ways you suggest. This paper uses the treadmill manipulation because it has been shown to have task specific results.

    I'm increasingly suspicious of this approach. On the one hand, it is clearly true. On the other hand, I'm worried that the ecological approach loses some of its cleanness. 'Ability' is a messy, messy word. You seem to admit this when you call Chemero's suggestion 'hand wavy'.
    Actually I was calling out Proffitt and 'effort', but I agree with this entirely. It certainly seems to be the case that 'ability' is the way to go, and it really is a worry.

    But I don't accept that we can't get into it. Especially in the throwing case, which is what I'm actually interested in right now; there are no clear body scale parameters accounting for affordance judgements in this task, and as I was writing this post I realised that eye height units aren't functional in any obvious way - they aren't about performance of the task, mostly. So it's interestingly messy, and it needs a solution, because people can throw objects to targets at set distances.

    ReplyDelete
  4. well... but lots of body scale parameters are relevant to the affordance of throwing. It is certainly difficult, for example, for a quadriplegic. I suspect that most of the questions you are asking only make sense if we all already agree that our subjects can, in fact, perform the task.* So, the behavior IS afforded by the match between organism and environment: We then go on to study whether or not the afforded action is realized, and test hypotheses as to why certain afforded things happen while others do not. Such investigations are really very straightforward to describe if we stick to body size, shape, etc.

    If we try to talk about "ability" in more sophisticated ways, I think it gets awfully hard to talk sensibly about what we are doing.


    *Unless you are specifically interested in threshold judgements, which are not as clean as the statistics make them seem either.

    ReplyDelete
  5. @ Gavin,
    Yeah... embarrassing. Right after sending that message I thought "Hey, haven't I seen talks in which lots of those manipulations were done. That might explain why I thought of them so quickly." I was hoping no one would notice.

    ReplyDelete
  6. well... but lots of body scale parameters are relevant to the affordance of throwing.
    Such as? Geoff and Arthur have already tried a bunch related to hand span, etc, and none predicted judgements of the affordance. Effecting a throw is more about coordination between limb segments that simple sizes.

    I suspect that most of the questions you are asking only make sense if we all already agree that our subjects can, in fact, perform the task.*
    Ruling out people who cannot possibly effect the affordance for some straight forward reason, like quadriplegics, is ok by me. The ball does not afford throwing to them; their perception of the distance to the target is not scaled in locomotion or throwing scaled units.

    Such investigations are really very straightforward to describe if we stick to body size, shape, etc.

    If we try to talk about "ability" in more sophisticated ways, I think it gets awfully hard to talk sensibly about what we are doing.

    Straightforward does not equal true. If body scale variables aren't explaining performance, they clearly aren't the dimension along which performance is organised. Warren clearly got a bit lucky with stair climbing - leg length is a pretty good proxy for ability-to-raise-your-leg. There are no such obvious proxies for throwing. (I'll get more into this as I blog more on throwing, I think, a lot of this comes up in the papers by Arthur and Geoff.)

    ReplyDelete
  7. "If body scale variables aren't explaining performance, they clearly aren't the dimension along which performance is organised."

    I know Arther and Geoff's work a little, but obviously not as well as you do. In any case, this seems like an exaggeration. My four year old daughter cannot throw a football 60 yards, your average large, muscular NFL quarterback can. The physical differences between the quarterback and my daughter seem non-trivially related to throwing ability.

    So, within a limited range of physical differences, and given people who - based on body size and shape are afforded the task - it is difficult to predict who can perform and who cannot. Looking for good predictors in such a case is a perfectly good for an empirical research program. But if we don't frame the problem carefully I think it can lead to big confusion between (not within) research programs, and creates false problems in understanding how different projects fit together in a bigger whole (i.e., a coherent research field).

    ReplyDelete
  8. Sorry, bad punctuation and a wrong word:

    So, within a limited range of physical differences, and given people who - based on body size and shape - are afforded the task, it is difficult to predict who DOES perform and who DOES not.

    ReplyDelete
  9. Strength is obviously part of the deal for executing a throw. But think of John McEnroe; little skinny guy with one hell of a serve. A big chunk of maximising release velocity is timing, not strength, and QBs are mainly big for other reasons.

    But this isn't about the act of throwing - it's about perception of affordances. People can perceive the affordance of an object's 'throwability to a maximum distance' if they have some level of throwing ability. Variation in that ability is not explained by any obvious body scale metric; therefore body scale (in terms of things like lengths of bits of you) isn't the action relevant variable used to scale perception. Body scale things like eye height units also probably won't do the trick for scaling the perception of distance, because they aren't obviously action related. However, we haven't run those studies yet.

    ReplyDelete