Tuesday 1 February 2011

The Size-Weight Illusion is Functional, and It's About Throwing

My colleagues, Geoff Bingham and Qin Zhu, have recently published some fascinating data which has emerged from their work on the uniquely human skill, long-distance throwing. This is a novel and rich perception-action task which Bingham and Zhu (and recently, me) have been investigating for some time, with many interesting results. I'll get onto blogging about this project once I've caught up with the coordination studies and have had some time to get my head around the data I'm helping generate.

I wanted to blog about this new paper, though, because it's an exciting result which deserves all the attention it gets. The result is about the size-weight illusion, one of the most robust illusions around. As I've talked about before, illusions are a concern to ecological psychologists only in that they suggest the task has been incorrectly characterised. This paper presents data that suggests the size-weight illusion is actually functional, and that it reflects the readiness of the human perception-action system to throw objects long distances.

This paper has seen some activity in the popular press already (e.g. here and here): Geoff's hoping for the NYT Science section too! 

UPDATE: Geoff being interview on NPR

The size-weight illusion
The size-weight illusion is, as I said, extraordinarily robust. It is typically described as a mis-perception of weight. Two objects of the same weight but different sizes are misperceived as being of different weight, with the larger one rated as lighter. People seem unable to perceive weight without taking size into account. 

There are two basic types of theories about the cause of the illusion (references from Zhu & Bingham, 2010)
  1. Top down: You expect the larger item to weight more, so you prepare to lift a heavier weight and the illusion is caused by it feeling lighter than you expected (e.g. Ross, 1966). This is 'top'down' in the sense it is being caused by an expectation, a cognitive state;
  2. Bottom up: People actually perceive a complex variable composed of both size and weight and use this in their judgements (e.g. the inertia tensor, Amazeen & Turvey, 1996; density, Huang, 1945). This is 'bottom-up' in that it is about perception.
None of these types of theory have ever quite been able to fully explain all the data (indeed, the inertia tensor hypothesis was recently explicitly tested and rejected in a paper by Bingham's group that is under review, although it's still an area of active research). Both approaches entail some learning, but children as young as 18 months reliably show the illusion (Kloos & Amazeen, 2002). The expectation account fails because the illusions persists when people are fully informed about the weight of the objects, and the illusion persists even when people are actually performing lifts appropriately for the correct mass (Mon-Williams & Murray, 2000); there's no evidence in their lifting behaviour that the person was initially surprised.

Zhu & Bingham then suggest that, given the illusion seems to be about perceptual experience, rather than online action control, perhaps it is about an affordance. Gibson (1979) suggested that these are what animals perceive in order to control their behaviour; they are real properties of the environment, but not necessarily properties as described by physics. The illusion suggests that the affordance in question should require the perception of both the size and weight of an object; such an affordance is the throw-ability of an object.

The dynamics of throwing
Throwing an object to a maximum distance means optimising the parameters of projectile motion. Some of these parameters are constant or out of the control of the thrower (e.g. gravity, air resistance). Some of the parameters are about the action of throwing (release angle, release velocity) - these are what the thrower must control to produce a throw. 

Critically, some of the parameters are about the object: for a given set of other parameters, maximising the distance of a projectile motion requires a specific combination of object size and weight. For throwing, this combination must fall within the range that is graspable and liftable by a human. By hypothesis, this relationship needs to be perceived when selecting an object to throw if you wish to maximise the distance of your throw, and empirically humans are very good at selecting the size/weight combination they can, actually, throw the furthest (e.g. Zhu & Bingham, 2008). Imagine the common game of throwing stones into a lake; we have very clear preferences about which stones we think we can throw the furthest, and these preferences are very accurate. Stable, accurate performance implies informational control (although the specific information variable being perceived has, as yet, failed to reveal itself; e.g. Zhu & Bingham, 2010).Information implies an affordance, and the critical object property for throwing is a combination of size and weight. Perceiving the affordance of 'throwable-to-a-maximum-distance' would require perceiving this combination. How does this variable, whatever it's precise form is, relate to the size-weight illusion?

The experiment
The study was simple. 12 participants (all capable of throwing a tennis ball 20m) were asked to make two separate sets of judgements about a set of objects. These were 48 balls which had been custom made; they came in 6 sizes (all graspable) and each size came in 8 weights (all liftable; see Table 1 in the paper for details). 

Participants then made two judgements about the objects; 

First, they were presented with a row of objects all the same size but different weights, and asked to choose and rank the 3 objects they thought was best suited for maximum distance throwing. Participants were allowed to hold and heft the objects (palm up, moving at the wrist).

Second, the participants were asked to heft a comparison object, and then asked to choose  and rank the 3 objects from each size set which felt about the same heaviness. The comparison object was either the object selected for throwing from the smallest set, or from the largest set.

In each case, the three choices were combined into a weighted mean, with the first, second and third choices weighted as .5, .33 and .17 respectively. This weighted mean allows for the fact that the object set contains discrete weights and sizes, and the actual 'equivalent weight' may not exist in the set although be on the continuous function the set samples; the weighted mean can then be a legitimate value that just doesn't exist in the set. The results are in Figure 1.
Figure 1. Mean selected object weights for throwing judgments and heaviness
judgments as a function of object sizes (from Zhu & Bingham, 2010)
The dark line shows the object weights judged as optimal for throwing to a maximum distance as a function of size; larger objects need to be heavier to be optimal. The dotted line shows the object weights judged as equal heaviness to an object of a different size but that had been selected as optimal for throwing. At each size, an object of different weight was required before people thought it was equally heavy to the small or large comparison object; this is the size-weight illusion. People don't just pick any different weight, however; they pick specific objects. In fact, the weight chosen was the weight identified by those people as the optimal weight to throw to a maximum distance. In other words, objects at different sizes which afford throwing to a maximum distance all feel equally heavy.

Summary
The implication is clear: people do not misperceive weight as a function of size, they correctly perceive the optimal size-weight (heaviness) value for throwing to a maximum distance. The size-weight isn't an illusion, it is a side effect of asking a perceptual system to judge weight when what that system perceives is heaviness; throwability. The size-weight illusion therefore reflects a functional capacity of the human perception-action system.

This capacity seems to be unique to humans (apes and monkeys famously throw faeces, but not far and not accurately), and it is possibly an intrinsic capacity of humans. Zhu & Bingham speculate that the nervous system is 'ready' to be able to throw in the way it is 'ready' to learn language; what's innate isn't a module (a la Chomsky's universal grammar), but a perceptual bias - here, the tendency to perceive not size, nor weight, but heaviness. Throwing has long been thought a critical skill that helped us hunt animals otherwise too strong for us (e.g. mammoths) and it has been suggested the skill enabled to us to keep hunting successfully in the face of climate change during the last Ice Age. It is therefore exactly the type of skill you might expect to be favoured strongly by evolution; these data suggest we are indeed 'ready' to throw in just this way.

References

Zhu, Q., & Bingham, G. (2011). Human readiness to throw: the size–weight illusion is not an illusion when picking the best objects to throw. Evolution and Human Behavior DOI: 10.1016/j.evolhumbehav.2010.11.005

22 comments:

  1. This paper has been getting more press than any other work on weight illusions over the last decade, so I feel that I’d like to have my say on it in a more in depth way than the usual commentaries on articles allow. I hope it can start a good debate on some of this stuff with the regular contributors to this blog, and hash some things out for an upcoming pub that may have something more direct to say on direct and indirect contributions to this illusion.
    For the past few years, I've been doing research on controlling the fingertip forces, and have found the size-weight illusion (SWI) is a nice tool to use. Subjects behave in pretty stereotyped ways when lifting SWI stimuli, experiencing a whopping great big perceptual illusion that does not get smaller with repeated experiences, while making force application 'errors' in the opposite direction to the illusion (lifting the larger object with more force than the smaller object, since they think it will weigh more). Unlike the perceptual illusion, these lifting errors get smaller with repeated lifts, and pretty quickly (~4 lifts of each) you are lifting all the stimuli with roughly equal forces. But still experiencing the same perceptual illusion. So, one conclusion you could draw from this (as many ecological types have) is that your (initially) incorrect expectations do not cause the illusion, as they are obviously no longer incorrect after a bit of practice with the stimuli. This leaves the door open for saying that lifters are just experiencing some other variable, incorrectly tagged as heaviness.
    The other possibility, championed by Randy Flanagan and co, is more representational, positing independent representations for action and perception/cognition/whatever. In other words, your motor system has expectations (priors) that are independent from the expectations of your cognitive system. The motor system’s expectations alter rapidly, to accommodate the fact that individual items change in weight (i.e., the prediction about the upcoming lift gives a high weighting to recent information). In the context of the SWI, this rapid adaptation of the expectations is why people can lift these identically-weighted things appropriately. The cognitive system’s expectations, on the other hand, alter very slowly to accommodate the fact that, in general, big things outweigh small things and metal outweighs wood etc (high weighting on long term, average, information). In this framework, our experiential percept of these illusorily different heaviness’s is formed by (glossing over the unresolved details) the fact that the system is always trying to reconcile the fact that one object feels lighter than it should while the other feels heavier than it should. I’ve got some work (Buckingham & Goodale, 2010) that strongly supports this view, making a single object feel as if it weighed different amounts as a function of what the lifter expected to be lifting (lifting without vision, following a short preview period, controlling for haptic differences in the stimuli. A bit like a priming study). Of course, this doesn’t rule of directly-perceived variables also contributing to the percept (and I’d be astonished if they didn’t), but it does show pretty conclusively (I would say) that your expectations (i.e. representation of the weight of the object you are about to lift) can alter your perception of heaviness. Based on various other cobbled together studies of mine, I’d venture to say expectations account for approx 40% of the illusory difference in weight.

    Continued in next comment…

    ReplyDelete
  2. Now where this paper goes off the rails for me is claiming that the SWI in all context is just the perception of throwability, based solely on the fact that the 2 (between-subject averaged) lines on the graph overlay one another (unless I’ve misunderstood the crux of the analysis). I also really don’t get why these lines have hugely different variability to one another (why wouldn’t the SD also be similar?) – if they are the same variable, albeit arrived at by different psychophysical techniques, shouldn’t they have similar SD? Also, no variables are independently manipulated (this is nothing but correlation here, albeit a very high one). Also, one of these lines (but, to my reading, not the other) is log-transformed, presumably how the high correlation comes about. Also, the sample were selected as people who can throw a pretty impressive distance (obviously this had to be the case for the perceptual expertise aspect of the judgement, but how does this generalise?). And why was this weighted mean necessary at all? The strong conclusions of the paper should still work just fine with subjects picking the best single object to throw – the appropriate object should exist for both tasks. Why does the weight illusion manifest in those who do not throw, and have presumably not had the same perceptual learning experience of throwers? Why is the illusion experienced at all under contexts where throwing is not afforded (all of my work is done by lifting with a precision grip on a handle mounted on top of the stimuli)?
    Basically, this paper tells me that people are pretty good at judging what they can throw farthest. When this idea is tested using (a pretty limited set of) stimuli that would induce the SWI, a nice correlation emerges. The conclusions that are drawn far outstretch the data, and, while the evolutionary story is nice enough, it is no better or worse than the face preference/mate choice/good genes stores that Sabrina has been bashing. Just to reiterate, the study is nice enough and I think it tells us something about throwing. But I do not think, however, that it has solved the SWI.
    Over and out!

    ReplyDelete
  3. Observations from a passing non expert - treat as possibly hopelessly naieve:

    My abiding question is: is one handed lifting a special case, or does the percept apply equally to two handed lifting ?

    While I've thrown a lot of objects in my life, I've spent much more time usefully employing two handed lifting, typically of objects of unequal size/weight/density - wood and stone. These objects would typically be highly irregular in shape, thus where the weight involved is toward the upper limit of my capacity it is essential that I'm able to make a judgement of conceptually apportioning the potential load, between to hands/arms.

    Is two handed/armed lifting subordinate to the skill required for throwing, or indeed vice versa, or are these two wholly separate cognitive cases ?

    That throwing is an important evolutionary gain for humans seems without doubt, however given the allocation of roles in the hunter gather societies of modern humans, were throwing to be exclusively driven by hunting enskillment, one might expect there to be notable gender differences in the SWI. Throwing behaviour is evident in many simians as a protective practice involving both genders and all ages. Throwing also plays a role in human gathering of fruit,again as a gender non specific activity.

    Two handed lifting is important as a gathering skill - exposing invertebrates and other smaller animals. It is also a key skill in creating defensible spaces, and ultimately in the building of structures. Of course in early technological homomin development, one handed manipulation of a heavy object - the hand ax, might be considered to be a defining skill beyond any other.

    IVI

    IVI

    ReplyDelete
  4. Hi IVI, thanks for stopping by!

    SWI studies usually just involve one handed lifting. That said, there's no reason it wouldn't show up in two handed lifting. The important thing is that heaviness must be defined; so there must be a size/weight relation available to perception. But I don't know of any studies that do two handed; Gav may know.

    I know there are lots of gender differences showing up in the throwing kinematics Geoff and Arthur have measured in other studies. However, a lot of this is sociological; girls aren't encouraged to throw the way boys are. The perception/action system may be 'ready' for long distance throwing, but it still needs to spend time practicing.

    Some interesting questions though, and certainly answerable. I'll add them to the list of things to do over the next few years :)

    ReplyDelete
  5. Gav: some thoughts, not quite in the order you had:

    Now where this paper goes off the rails for me is claiming that the SWI in all context is just the perception of throwability, based solely on the fact that the 2 (between-subject averaged) lines on the graph overlay one another (unless I’ve misunderstood the crux of the analysis)..
    You have slightly. First, the whole thing was within subject; the same people made all the judgements. Second, the idea is that people didn't pick any old heavier bigger object as being the same heaviness: they picked the bigger object that was the weight they'd picked as optimal to throw. So this is a stronger result than just a correlation between two lines.

    I also really don’t get why these lines have hugely different variability to one another (why wouldn’t the SD also be similar?) – if they are the same variable, albeit arrived at by different psychophysical techniques, shouldn’t they have similar SD?
    This is a good question and I don't know the answer; I'll ask Geoff. My guess is that the larger objects start being less graspable (7.62cm is the span of my palm, for example) - the variability may reflect between person differences in hand size. But I'll check.

    Also, no variables are independently manipulated (this is nothing but correlation here, albeit a very high one).
    The size and weight are independently manipulated; they had this set of objects custom made so that they could control these. The relationship is between the measures from two tasks using those objects.

    Also, one of these lines (but, to my reading, not the other) is log-transformed, presumably how the high correlation comes about.
    Both are log transformed.

    Also, the sample were selected as people who can throw a pretty impressive distance (obviously this had to be the case for the perceptual expertise aspect of the judgement, but how does this generalise?).
    Generalise to whom; people who can't throw? As I said above, readiness to throw still requires some learning. The issue was first, how do people who can throw perceive size/weight vs throwability?

    And why was this weighted mean necessary at all?
    I tried to explain this in the post. The objects are a set of discrete size/weight combinations; a sample from the continuous size/weight functions you can make. If, by bad luck, the objects made weren't the right ones, and I asked you to just pick your top choice in each task, they might look very different just because you picked two different objects.

    This way, you 'smooth out' the discreteness using the top three choices, and you weight it to reflect the order of choices. This then gives you an estimate of the value on the continuous function, which may not be a value that exists in the set of objects.

    So:
    The strong conclusions of the paper should still work just fine with subjects picking the best single object to throw – the appropriate object should exist for both tasks.
    But it may not exist in the limited set of objects they have. This approximation works because the sizes and weights were carefully generated from a continuous function of the two variables.

    Why does the weight illusion manifest in those who do not throw, and have presumably not had the same perceptual learning experience of throwers?
    This is a good question. The suggestion is that this reflects the readiness. A strong test of that would be to examine the variability of judgements of novice vs. expert throwers: you'd expect the same basic pattern but with larger variability.

    Why is the illusion experienced at all under contexts where throwing is not afforded (all of my work is done by lifting with a precision grip on a handle mounted on top of the stimuli)?
    Size is visually available, though?

    ReplyDelete
  6. More:
    The conclusions that are drawn far outstretch the data, and, while the evolutionary story is nice enough, it is no better or worse than the face preference/mate choice/good genes stores that Sabrina has been bashing.
    I don't agree that the specific conclusions they draw do outstrip the data, given the specific weights people chose.

    The evolutionary story: yes, it's a risky move. Part of the problem with most of evolutionary psychology is that it suffers from WEIRDness: a lot of the things they claim as universally true (like waist to hip ratios) just aren't. There's still plenty of work to do here, and I certainly haven't captured the care of the arguments in the paper.

    That said, there is a fairly strong evolutionary biology literature on throwing and it's role in our evolutionary success. The paper cites a bunch of this. It is also apparently the case that this kind of throwing is unique to humans; the throwing apes do just isn't the same. (Do apes show the SWI? I have no idea.) The precise timing required to produce a throw is supported by things like the cerebellum, brain areas H. sapiens was developing faster than other species. So the idea is certainly only suggestive, but there's a lot of lines of converging evidence.

    More on the two-visual system stuff later on, I have meetings now :)


    WEIRD - Western, Educated, Industrialised, Rich and Democratic, from Henrich, Joseph, Heine, Steven J and Norenzayan, Ara (2010) The weirdest people in the world? Behavioral and Brain Sciences, volume 33, issue 2-3, pages 61-83.

    ReplyDelete
  7. I'll await your next slew of comments to rebut.. RE 2 handed lifting, there's not much. Some stuff on weight perception is out there (should the same item feel half as heavy when lifted with 2 hands? - that kind of thing), but not much illusion stuff. Interestingly though, I'd suspect that 2-handed lifting rarely affords throwing, so this situation might evoke a dissociation between perceiving throwability and judging weight.

    ReplyDelete
  8. Gav: so, Bayesian things and two-visual system stuff. Some separate but related comments:

    Bayesian analysis: I think Bayes is a great framework for data analysis (vs GLM). It isn't a theory of behaviour, though, because it's purely descriptive; like dynamical systems, it's a toolbox for describing but not explaining the specific behaviour of the perception/action systems. So I worry about it's ability to actually provide an answer to the question of mechanism underlying the results.

    Also, from my point of view, it's a sophisticated solution to a problem I don't think quite exists, namely making reliable and robust inferences on the basis of impoverished stimulation. If you are going to assume poverty of stimulus and the need for enrichment from prior knowledge, this is indeed the right framework - but given I think that's the wrong starting point, I don't think it's the right way to do the science.

    Two visual system stuff: yes, anatomically the brain does have a dorsal and ventral stream of visual processing. There are two issues, though:

    1. In the intact brain, the idea that these 'streams' are independent is simply wrong. One of the things I actively dislike about fMRI and the way it's analysed is that it's about spatially localising function, and I think this has mislead us into the naive assumption that only certain bits of the brain light up when you do something. The problem is that the entire brain is on the go all the time; the idea that a given function is performed in one place seems flawed (even the neuropsychological literature is incredibly mixed: DF (is that the right patient?) didn't simply lose the ability to do object recognition, for instance. Mark tells stories about being down the pub and her identifying some weird object with no problem.

    Actually, the neuro-psych literature is all filtered through a narrative. The craziest video I ever saw was a split-brain patient having an object shown to his right hemisphere only and generating a verbal name for it. He got it wrong, but his right hemisphere came up with a word! Blew my mind; I need to find that video. But the moral is these results are never, ever as clear as they get laid out in the textbooks.

    2. I'd be more impressed with the two visual system as a theory if it could decide what the two visual streams did. I've heard at least five theories (where vs what, etc etc) and they don't map onto each other clearly: they just reflect people following their data.

    I think it's a weird way to think about perception/action systems, for the reason I worry about illusions: they are generally the result of either a trick or a mis-description of the task space. I firmly believe perception has no time to waste being as crap as this approach allows.

    Feel free to pile back on on this - it's an interesting topic and it's been a while since I've thought about it in detail. It'll be good to bang some of these ideas around :)

    ReplyDelete
  9. Ah, but an object's weight is a perfect example of how we have to deal with a poverty of stimulus on an almost daily basis because (1) there is no adequate visual cue to an object's weight (the best it gets is watching someone else's dynamics while they lift it or something). Object size or material gets a bit of the job done, but we somehow have to fill in the gap because (2) our fingertip forces scale to the predicted weight of the object, and the vast majority of these variables are pre-liftoff. Behaviorally, we make a guess before the lift, and this guess is impovererished until we get some feedback on how successful our initial lift was. Direct information about weight can only come from experience, and we cannot have the appropriate experience until after we have finished the lift.

    now, on to the sticky 2-visual-stream issues

    Object lifting in the context of the SWI says little to the 2 visual streams idea, for the simple reason that it's not an ostensibly visual task. While you do have a 'dissociation between perception and action', the illusion is 'perceived' with haptics, not vision. So you have a dissociation between haptic perception and the guess you have made about the upcoming lift (even with practice, the forces are still the product of a guess, albeit a much more refined guess). This makes the dissociation somewhat less surprising, and really just lets you draw conclusions about what function perception has in the world (to point out oddities and possible dangers with a big blaring siren), while action just has to act in the most efficient way possible (minimizing comedy falling over with inappropriate force applications and such).

    What people do like to talk about object lifting for is how it is an example of how these streams for visually-guided action and visual perception must be talking to one another (identify a material and associate it to a memory is ventral, while retrieving a sensorimotor memory is dorsal blah blah blah), at this point its really all metaphor, since little is known about either process, let alone how they would interact with one another.

    But no serious object lifting researcher will tell you this is a 2-visual stream story (although it certainly has parallels with the grasping visual illusions stuff).

    You should do a larger 2 visual streams post at some point though - i doubt we'll manage to come up with all that much more than the reams that have already been written back and forth on this topic, but it's not a debate i've seen in an ecological psych context (I often think that the 2-vis streams story smacks of ecological psychology anyway, which is why it is so appealing to the lay person(

    ReplyDelete
  10. Ah, but an object's weight is a perfect example of how we have to deal with a poverty of stimulus on an almost daily basis because (1) there is no adequate visual cue to an object's weight (the best it gets is watching someone else's dynamics while they lift it or something).
    This isn't quite poverty of stimulus; this is the situation of no visual specification of mass unless the object is moving, which is different. The fact that there's no visual information for mass simply means that there's no ability to prospectively control fingertip forces; if fingertip forces are to be set ahead of time, it must be done via prediction, which is not very ecological. So the question is, is there prediction?

    Which leads to
    Object size or material gets a bit of the job done, but we somehow have to fill in the gap because (2) our fingertip forces scale to the predicted weight of the object, and the vast majority of these variables are pre-liftoff.
    Really? I know appropriate fingertip forces are generated quickly; but you don't start generating force pre-contact, and as soon as contact is made there's information available about the mass and whether you're generating enough force.

    I know the standard line in force control is that it must be done via prediction because it happens too fast for perceptual control. I'm not an expert, I've never played with force control in lifting, but it occurs to me that there isn't force production until contact with the object. Of course it's pre-lift off: lift off doesn't happen until you generate the right amount of force, which takes time.

    Behaviorally, we make a guess before the lift, and this guess is impovererished until we get some feedback on how successful our initial lift was. Direct information about weight can only come from experience, and we cannot have the appropriate experience until after we have finished the lift.
    I would argue you start getting information about mass as soon as you attempt to move the object. Well, actually I would say you immediately start getting information about the object's move-ability and whether you need to ramp up the force. What happens force wise pre-contact?

    I'd actually love to play with this, but I don't have any finger tip force transducers. We should chat sometime about some studies :)

    ReplyDelete
  11. Two visual systems; yes, of course, you're right about this SWI. It's funny, the 'dissociation between perception and action' just triggered all my 'two visual systems' rage :)

    I will disagree with this, though:
    ...and really just lets you draw conclusions about what function perception has in the world (to point out oddities and possible dangers with a big blaring siren), while action just has to act in the most efficient way possible (minimizing comedy falling over with inappropriate force applications and such).
    Action can't do what you suggest without accurate perception, and my problem with the dissociation literature is precisely that it claims perception is full of disastrous errors but that it's ok, because perception-for-action is doing the business. But the idea that there's lots of brain real estate dedicated to a system that seems to do nothing but get things entirely wrong all the time seems insane to me.

    Lots of people think the two-visual-system stuff is ecological. This has always confused me. Part of it is that the two-visual-system stuff does have built into it the idea of perception-for-action, which is on the right track (hell, any conference running a 'perception and action' session usually means 'two visual system' content, which bugs me too). But it's fundamentally not Gibsonian, for the simple reason that the idea of perception that isn't about action isn't part of ecological psychology at all and it's not clear why you'd want such a system.

    ReplyDelete
  12. re: the log transform. I'm just reading another of their papers and they note they log transform the weight data to linearise it for the correlation, because the weights were generated by a geometric progression (Wn+1 = Wn * 1.55). That's why they had to do it, and to both sets of data.

    ReplyDelete
  13. Commenting on this blog post has become a nice part of my morning routine. I shall be sad when it's over!

    "I would argue you start getting information about mass as soon as you attempt to move the object. Well, actually I would say you immediately start getting information about the object's move-ability and whether you need to ramp up the force. What happens force wise pre-contact?"

    Probably the best evidence for this would be that the rates of change of the forces scale to the expected weight of an object (so in the swi, peak grip force rate and load force rate are higher for the large object than they are for the smaller object), and these rates peak well before liftoff - sometimes only 50ms after the initial application of force. It's been pretty well mapped out that online correction of fingertip forces requires at best 90ms for error detection and spinal cord mediated correction, and most of the error signal should follow (the expected) liftoff, depending on whether the lifter has overestimated or underestimated (once gravity has had its wicked way with the ratios between grip and load force).

    On this note, there's a nice paper that came out in j Neurosci a few months back by a guy called Marco Davare which seems to indicate that the information you use to make your prediction changes very rapidly as you get closer to your lift. If you tms motor cortex to elicit a motor potential in the lifting hand, the MEP will scale to the last thing you just lifted (your next lift is based on/corrupted by your last lift). If you have visual information suggesting that a new force is necessary (like a differently sized block), your motor plan is dynamically weighted to the new information over the next few hundred milliseconds, becoming less and less corrupted by your past (and now redundant) sensorimotor memory.

    The idea that large chunks of the brain are devoted to the conscious processing of things that do not subserve action. I'd hardly say that the errors that our conscious perception (or ventral stream, if I may) are useless. In the SWI, the errors are quite useful, since they provide a warning flag that your initial guess about the weight may be wrong. But hey, even though our perceptual systems did initially evolve to deal with action (and hence the unified perception action overview), that does not rule out a perceptual system that is not ostensibly involved in the control of action, which, given skills such as recognizing faces and other things that do not obviously afford action is surely the case. I find the solution of having multiple perceptual streams evolved for multiple purposes (action and non-action, which seems a sensible dividing line giving the importance of action to us) a rather elegant solution - certainly moreso than just shoehorning all of our perceptual tasks (be it action or otherwise) into the same processes (be it ecological or representational.

    re the log transform stuff, it was only an issue for me if the log transform was only done on one of the lines (I obviously misread that point).

    ReplyDelete
  14. So fingertip forces at contact are already roughly scaled as a function of object size, but then begin scaling to the actual object's mass once you actually begin to move it. Is that right?

    And with the TMS stuff: am I getting this right? The potential in the arm you evoke with the stimulation scales to what you just did (sensible) but then gets quickly retuned as a function of what you're about to do, as per what I said in the first paragraph?

    ReplyDelete
  15. Terribly sorry, the obfuscatory nature of your comment hid the message you were trying to convey :)

    ReplyDelete
  16. I'm asking: with the TMS, the MEP scales to what you just did (reflecting calibration) but this MEP quickly gets over ridden by the task at hand?

    Then I'm asking: once you start scaling fingertip forces to the specific lift, you do this first as a function of object size then quickly as a function of actual object mass?

    I'm trying to get a feel for the timeline of events.

    ReplyDelete
  17. Yeh, with the TMS that's about right (all happening within the time course of a single 'trial')

    With the kinetics, I was referring to things on a trial by trial basis. At the start of trial 1, your guess is based purely on size/material/whatever other cues you have. On trial 2, it is those things combined with the sensorimotor memory from trial 1. So, still not perfect information, but better than trial 1, and your force rates scale more accurately.

    This also holds true within the timecourse of a lift - your forces proper change (presumably for the better, getting more refined), and if you have the time in your task, do things like reducing the ratio between grip and load force.

    ReplyDelete
  18. Update for those not following the Twitter feed: Geoff being interview on NPR

    ReplyDelete
  19. I'm not going to solve everything.

    Another miserly concession!

    ReplyDelete
  20. Wrong place, I think Ken, plus a bit weird anyway.

    ReplyDelete
  21. Hi, I am not sure I can return to read a response to my comment, but as a lay person-non professional scientist- Something happened here with me today. I noticed you referenced visual connection with a weight object and throwing quite alot in the original blog above. ....What happened with me today had nothing to do with my actual visual/ocular connection with the weight object so much as having to do with my minds eye & experiential connection..AND...COMPRESSION. Not the throwing; not lifting shocked me, but the compression of the objects near the item (items I was unaware of and not looking at them nor looking at the item which I was trying to lift). I was very shocked when my light, plastic, jar was unliftable. I pray blessings upon your study...illusions are a key thing to now about in many aspects of life. If for no other reason then to discern truths and act with better efficiency. In short this very light item became heavy due the compression of the objects surrounding it. I bet Larry, Curly, or Mo have a scene where they tried to lift something and couldn't. Not likely due to compression of the object though...it was probably just nailed or glued down and they didn't know it. Nice day to you, sirs.

    ReplyDelete
  22. Gavin, I finally have an answer re the variability (having just remembered to ask yesterday :)

    The high variability is caused by the fact that, while a given person will choose the same object in the two tasks, different people choose different objects. So everyone is doing the SWI/throwing thing reported in the paper, but the specific ball chosen reflects, presumably, their ability to effect the throwing affordance.

    Exactly how that effectivity is composed is unclear. Geoff and Arthur have looked at hand span and a bunch of body scaled measures, but nothing predicts it. It's going to be complicated, because it will be about 'ability to throw' and that's more about coordination between limb segments than about body scale.

    ReplyDelete