Wednesday 8 September 2010

Tools and Brains and Embodied Cognition

Arguing about embodiment with Ken Aizawa over the last few days has opened up a lot of topics that I hope to cover over the next little while. But it also primed me to notice this article at Scientific American by Patrick Haggard and Matthew Longo, summarising a recent paper adding to the growing literature on the neuroscience of tool use. I like this work, and it got me thinking how this relates to the embodied cognition literature; Ken, I'd be interested to hear your thoughts on this on.

I actually really do like this work, although as you may have guessed, not for exactly the reasons everyone else likes it. The basic result is this: if you pick up an object (say a pair of tongs, or a stick), then the cortical representation of your arm very very quickly reorganises itself to include the tool; as far as the brain is concerned, your arm is now longer than it 'actually' is.

This is (of course) grounded in perception. Dynamic touch is the name given to the way in which we can non-visually perceive characteristics such as the length of an object. Get a ruler, close your eyes, hold it at one end in your hand and wobble it round a bit. You'll quite quickly get a fairly good sense of where the tip of the ruler is, and if I asked you to reach using it you'd be ok at landing that tip somewhere without vision. Now hold the ruler in the middle and twiddle it; same basic thing. You aren't perceiving length, per se; the evidence suggests you are perceiving information about the distribution of mass in the object by the act of rotating it, which reveals the inertia (resistance to rotation) of the object. Mathematically, this is an inertia tensor.

The suggestion is that this is how you maintain 'knowledge' of where your limbs are in space. The cortical representation that everyone is excited about is being actively maintained by perception of the inertia tensor resulting from how your limbs are moving around; this is how it can change so fast, because it isn't a static thing, it simply reflects the brain maintaining perceptual contact with your arms. 

Yes, the cortex does indeed seem to have some real estate that knows things about the arm; isn't this a representation and if so, is this blog now over? Well, no, I don't think so: the reason I like this tool research is that it shows that the cortex involved isn't anything like what cognitive people talk about when they talk about representations. It's fluid, in constant flux, resonating with the world and changes in it; this sounds a lot more Gibsonian than most neuroscience, and this is why I like it. We ecological types get a bad rap for dissing the brain - this is mostly just because we think there's work to do before getting to excited about fMRI. This work shows, I think, the kind of behaviour ecological types should be looking for in the brain, which, let's face it, is clearly involved in perception, action and cognition. 

I wonder how people like Ken, who don't think cognition is ever really embodied, think about these cases - what the brain knows about the arm has changed in a manner driven by perception, and future behaviour with the arm is affected accordingly. An object from the world has literally crossed the boundary from the external world into the seat of bounded cognition - what does this imply? I think it implies the boundary is actually a pretty low fence; but I would be interested to hear from Ken on this when he has time :)

References
Carlson et al (2010) Rapid assimilation of external objects into the body schema. Psychological Science, 21(7), 1000-1005.
Turvey, MT (1996). Dynamic touch. American Psychologist, 51(11), 1134-1152.

17 comments:

  1. Andrew,
    I don't see anything in this kind of embodied cognition that would bother a cognitivist. As you noted in another of your posts, a lot of different things have been described as "embodied cognition".

    The one thing that jumped out as me as premature is your claim that "An object from the world has literally crossed the boundary from the external world into the seat of bounded cognition". Yes, using a tool causes the brain to change the way it dedicates neural hardware. That, alone, however, does not mean that the tool use constitutes part of one's cognitive processing. To suppose otherwise is what Fred Adams and I have notoriously labeled "the coupling-constitution fallacy". If you are going to commit that fallacy, then a cognitivist such as myself will object.

    ReplyDelete
  2. You can get the basics of the "Adams-Aizawa line" (of which the C-C fallacy is only mentioned in passing) by a google scholar search on "The Bounds of Cognition". The first item will take you to a free pdf.

    A draff of a longer paper (now published) on just the C-C fallacy is at http://www.udel.edu/Philosophy/papers/adams2007.pdf

    ReplyDelete
  3. In regards to the question of cortical maps being "representations", I think Edward Reed says some nice things about it in his book Encountering the World: Toward an Ecological Psychology. He says:

    "The somatotopic 'map' of a finger is perhaps not really spatial at all but temporal -- a dynamic correlation of sets of overlapping neural processes resulting from a unified behavioral process...On this view, the cortex is not a map of the periphery, nor is it any kind of fixed network receiving signals and issuing commands; instead, cortical sites are places where specific patterns of covariation are selected out of larger rangers of variation within a behavioral time frame" (p. 75). This seems to dovetail nicely with what you said in the OP.

    I think this is a good way to find solid ground between the cognitive idea of a "cortical map" and the Gibsonian idea of behavioral regulation through action selection (driven mainly by environmental constraints e.g. affordances). Cortical maps turn out to not be representations in the classic symbolic sense, but rather, "vehicles for weaving behaviorally significant aspects of the environment into a system that has the capacity to preserve some varieties of behavioral units rather than others" (p. 76).

    ReplyDelete
  4. Ken

    Yeah, I wasn't really thinking this was going to cause you trouble; I just wasn't sure how you might think about it, because to me it's so clear that the tool has simply been dynamically worked into the system and it seems so odd to think it's not now a part of the cognition going on.

    So while this could be mere coupling, do you have a test? A way to tell the difference between coupling and constitution? If we applied such a test to this case, what would we find and why?

    Gary: as usual Reed summarises what I mean with much greater clarity than I can rattle off :) As much as I think current neuroscience is a bit rubbish, I do firmly believe that we need to include the brain in our investigations because it's clearly pretty important. Job 1 is to break the assumption of mental (symbolic) representation, though, so we can get freed up to think about the brain the way Reed describes.

    Olaf Sporns is the only neuroscientist I've ever heard talk this kind of way, and he's not even particularly ecological; just a very smart person who takes the systems approach very seriously.

    ReplyDelete
  5. @Gary:
    "Cortical maps turn out to not be representations in the classic symbolic sense, but rather, 'vehicles for weaving behaviorally significant aspects of the environment into a system that has the capacity to preserve some varieties of behavioral units rather than others" (p. 76)."

    But, being a cortical map and being a vehicle for weaving ... do not seem to me to be inconsistent. They do not seem to me to be rival hypotheses. A cortical map could be a vehicle for weaving ...

    So, why reject cortical maps?

    ReplyDelete
  6. Andrew,
    The path away from the C-C fallacy, it seems to me is to have a theory that distinguishes between cognitive processes and non-cognitive processes. This is what Adams and Aizawa refer to as having a "mark of the cognitive". I just posted at my blog some text that I hope will give some illustrative examples of the distinction. We described them as "intuitive", but alas that word is apparently to Ross and Ladyman what a red cape is to a bull.

    ReplyDelete
  7. The path away from the C-C fallacy, it seems to me is to have a theory that distinguishes between cognitive processes and non-cognitive processes.
    Agreed. Frankly conversations about the lack of theory in psychology is what motivated us to start this blog in the first place.

    Do you have a mark of the cognitive? I flicked through 'Why the Mind is Still in the Head' but all I saw were critiques of other people's suggestions. You presumably need a theory too, in order to be able to rule out various suggested forms of extended cognition?

    ReplyDelete
  8. We do not have a mark of the cognitive, but what we take to be a necessary condition: Cognition is restricted kind of manipulation of representations bearing non-derived content.

    Technically, we don't need a complete theory of what characterizes cognition processes in order to block EC. We only need necessary conditions that the examples fail to meet.

    ReplyDelete
  9. OK, you're going to have to explain what 'manipulation of representations bearing non-derived content' means.I saw it in the paper and didn't quite get it there either.

    ReplyDelete
  10. It's a weak version of cognitivism. So, the representations are data structures. They have non-derived content, so the content they have is not determined by social conventions, as is the content of, say, stop signs and traffic lights. Not just any kind of manipulation will count, however. So, a chess computer program will not manipulate data in the way that humans do, so the chess computer program will not be a cognitive processor for that reason.

    It's pretty familiar 1980's kind of philosophy of mind, so we don't go into it much. It's in our paper "BoC" and the BoC book.

    ReplyDelete
  11. Ken, have you heard of William Ramsey's job description challenge?

    The question he poses for "weak cognitivism" goes as follows: Can you give an explanation for how these "data structures" function that does not ultimately fall back on a purely causal story? That is, can you give an account of representations that actually tells a story about how the representations function *as* representations, and not, say, as merely causal mediators between input and output or as dispositional "indicators" (something a thermostat can do)?

    ReplyDelete
  12. Technically, we don't need a complete theory of what characterizes cognition processes in order to block EC.
    Hmm. Given that you want to rule non-brain things out of the picture I would say you certainly do need a theory of what counts as cognition, otherwise you surely run the risk of cherry picking. On what basis do you rule things in and out?

    This blog was inspired by the conversations Sabrina and I have been having in which we inevitably return to the lack of a theory as the single root cause of most of the trouble psychology faces. Representations aren't a theory, as evidenced by the endless definitions of them in different papers and the ability of every cognitive person I know to tell me 'that's not what I mean by representation' when I call them on something.

    ReplyDelete
  13. Ken, you asked:
    So, why reject cortical maps?
    Because they aren't maps! The reason I like this tool use data is that it clearly points to the fact that the brain isn't mapping a world onto a model of the body; it's just not that static. When you wield a tool it becomes as much a part of you as your arm (well, up to the limits of the information that's available - but that's true of your arm as well). The brain hasn't mapped arm+tool onto arm: it's simply recalibrated so that arm now means arm+tool.

    Cortical maps are not mapping anything onto anything else, nor are they representing (standing in for) anything. And this is what cognitive scientists mean when they say maps.

    ReplyDelete
  14. I know that there are anti-representationalist arguments, e.g. Ramsey's. My point is that I don't see further support for anti-representationalism in the paper discussed in the post.

    Why not say that the cortical maps are not static cortical maps; they are dynamical maps?

    Say they are "Vehicles for weaving behaviorally significant aspects of the environment into a system that has the capacity to preserve some varieties of behavioral units rather than others" Again, no worry to me. Cortical maps can be this, right?

    ReplyDelete
  15. To you maybe, but not to any cognitive psychologist worth their salt.

    ReplyDelete
  16. "The brain hasn't mapped arm+tool onto arm: it's simply recalibrated so that arm now means arm+tool."

    The crux of this point is whether or not the information about the new 'arm' has been diluted by its apparent extension. Do we wield and understand the dynamics of the stick-arm as well as just our arm alone. I doubt it (unless you are a pro athlete with a racquet). Any remapping must come at a pretty steep cost for behaviour..

    ReplyDelete
  17. There may be less information about some things (the stick tip doesn't have nerve endings, etc) but for anything there is information about (length, etc via dynamic touch) the calibration seems pretty fast and effective.

    Although that said, presumably this is something you can get better at, the way a tennis pro does. I don't think there's anything like 'dilution' though, this isn't an averaging process.

    ReplyDelete