What is “the grain problem”?

Bechtel and Mundale, (1999), propose that the putative
multiple realization of psychological functions is the product of a kind of
trick.  They suggest that defenders of MR
type psychological properties at a coarse grain and potential realizer properties
at a fine grain.  By doing this, it looks
as though the one coarse grain psychology can be realized by different many
fine grained realizers. 

Ok.  I’m having a
tough time following through with this metaphor.  I get the idea that one might distinguish
vision, color vision, trichromatic vision, normal trichromatic vision as
psychological properties of decreasing grain. 
But, how is the grain story supposed to go for the potential biological
realizers?  It can’t be brain cells
(neurons + glial cells), neurons, pyramidal cells, etc.  Suggestions?


  1. Brandon N. Towl

    If I recall correctly, the metaphor of “grain” has to do with how narrowly mental kinds or types are individuated, and likewise how narrowly brain state types are individuated. So, for example, most philosophers are willing to say that pain, as mental state, is had in common by humans, chimps, mollusks, etc. Surely there are differences between my pain and mollusk pain (for example, mollusk pain typically causes an octopus to squirt ink; not so with me or a chimp). But, despite behavioral differences and different causes, they each count as pain. But when philosophers consider brain states (again, according to B&M) they individuate much more narrowly, such that different species (or even different individuals within a species, or different time slices of a single individual) have very different brain states. And so, say B&M, if mental state tokens have just a very loose resemblance, of course it will appear that very different systems can share one and the same brain state.

    It is hard to see how one would get a more “coarse” grain with brain states, but it is not impossible. For example, a “coarse” brain state might be a brain state described by an abstract network property, while a more “fine” individuation would make reference to actual configurations and interconnections of neurons in given networks. For some inspiration, you might want to look at the literature on population coding– there is quite a lot on pop. coding in motor cortices (particularly work by Georgopoulos and colleagues) and in visual cortex as well. One of the upshots of this research is that there can be *a lot* of individual variation when it comes to the actual location and connections of neurons in a network, and yet the network has some very stable statistical properties (which happen to correlate very nicely with certain mental states, such as movement in an intended direction).

    Does this help at all? I can send more info your way if it does.

  2. kenneth aizawa

    Recall that the grain problem is supposed to be a diagnosis of what MR looked plausible (when it is false or unjustified).  The idea is that Putnam, Fodor, et al., where looking at coarse grain psychological properties and fine grain brain state properties.  But, surely Putnam, Fodor, et al., were not thinking of, or foisting upon us, these high falutin’ sciency brain state properties.  So, I’m looking for a non-metaphorical account of what these lower level “grains” are supposed to have been that has lead so much of the profession astray.

  3. Richard Brown

    Hi Ken,

    I think the idea is supposed to be that Putnam had in mind physico-chemical properties of the brain, where what neuroscientists actually look at (according to B&M) is “activity in the same part or conglomarate of parts”

  4. kenneth aizawa

    This too does not seem to me to capture the notion of grain.  One can say that Putnam was looking at A, when he should have been looking at B.  That’s one thing.  It’s more to say that A is a finer grain (version of?) of B.  I can take it that trichromatic vision is a finer grain category that chromatic vision, because the former is a species of the later.  But, I don’t see how to apply this analysis to your case nor for that matter to any other case.  It could be that I’m just philosophically myopic (in addition to my other myopia).  Or is there some other account?  What is the account of what “grains” are according to which A is a finer grain of B? 

    Just to look ahead, possibly.  There are other constraints on articulating this theory of grain.  B&M apparently think that there are lots of grains and that one can get univocal realization just by picking the right grain. 

  5. Brandon N Towl

    On the contrary, I think Fodor and Putnam /were/ hoisting such “high falutin’ ” properties on us. What Fodor and Putnam had in mind when disucssing brain states were highly detailed, determinate states of neural systems. They may not have known what exactly those look like, but then again the details do not matter to their argument. But Fodor and Putnam would, I think, consider a detailed specification of a neural network as a “brain state”, though maybe not an abstract property like vector encoding (for example).

    So, to use your example,
    vector code:specific neural network::
    vision:tri-chromatic vision

    (BTW, I think Polger discusses grain in his book, and I vaguely remember Shapiro saying something about it in his J. Phil paper– perhaps these guys know more than I do. Well, they do know more than I do!)

  6. kenneth aizawa

    I think that the title of my post is potentially misleading here.  What I am after is how to cash out the metaphor of grain for neuroscientific properties.  At the psychological level, this seems straightforward.  The psychological properties stand in a genus-species relation or a determinable-determinate relation.  That looks like at least a vaguley plausible account of pain, human pain, the human pain associated with migraine.  But, how does it go with neuroscience? 

    Your analogy has the same problem for me as does Richard’s example.  You can say that Fodor, et al., focused on this thing A, when they should have focused on that B.  I get that, but how does this apparatus of “grain” work?  It doesn’t seem to me that vector code: specific neural network stand in a genus-species relation or determinable-determinate relation (though I would be open to hearing how they in fact do). 

    Incidentally, I reviewed the passages in Polger’s index that refer to grain.  This was helpful, since it in fact reinforces my sense that there is an expository problem here for B&M.  On pp. 21-29, for example, Tom does not mention how the story goes for neuroscience, only for psychology.  Moreover, he sticks with this metaphor of grain.  That’s precisely what I want to move beyond.  (He does, however, eventually seem to equate or alternate between grains and levels.  Levels, however, seem to be another matter.)

    This additional material on grains seems to be important to Bechtel and Mundale over and above a diagnosis, since they seem to want to use it to somehow guarantee that for any psychological property, one can find a unique realization.  (Check out the text in the second half of the last paragraph on p. 202).

  7. kenneth aizawa

    Polger apparently thinks this is ok, since he does it in his book. 

    But, I don’t think so.  Are “pain” and “octopus pain” at different levels?  (This is the B&M example on p. 202 of their paper.)  Not in the same sense in which, say, amino acids and cells are at different levels.  The B&M notion of grain has something to do with “abstracting away” and “taxonomic refinement”, which suggests this idea of genus-species or determinable-determinate.  This works toerably for “pain” and “octopus pain”, but how about for neuroscience?  I was going for this kind of thing in my “crazy” example of brain cell, neuron, pyramidal cell.  These at least stand in some sort of relation of “abstraction” and “taxonomic refinement.”

  8. Hi Ken,

    The first thing that comes to my mind in thinking about this stuff is that a coarse grained way of carving up psychological stuff would be “memory” and comparatively fine grained way of carving the brain would be into subsystems or parts like pre-frontal cortex, hippocampus, etc.

    The gist of the B&M point, I take it, is that a more apporpriately fine grained partitioning of the psychological stuff might be working memory, long-term memory, etc.

    Another example:
    coarse grained: seeing
    fine grained: seeing what, seeing where

    at the course grained level, seeing is mutltiply realized by dorsal stream and ventral stream cortical processing. But at the finer grain, psychological stuff has unique neural realizers.

    That, at least, is the picture I think B&M had in mind.

    Becthel advised the dissertations of both me and Mundale, so I hope I’m not screwing this up!

  9. kenneth aizawa

    Your idea seems to be that the neuroscience hierarchy should be something like:
    brain, occipetal lobe, visual cortex, V1.

    I don’t think this can be exactly right either.  Note this quite important passage:

    One can adopt either a coarse or a fine grain, but as long as one uses a comparable grain on both the brain side and mind side, the mapping between them will be correspondingly systematic.  For example, one can adopt a relatively coarse grain, equating psychological states over different individuals or across species.  If one employs the same grain, though, one will equate activity in brain areas across species, and one-to-one mapping is preserved (though perhaps further taxonomic refinement and/or delineation may be required.)  Conversely, one can adopt a very fine grain, and differentiate psychological states between individuals, or even in the same individual over time (B&M, 1999, p. 202).

    So, on the psychological side, you apparently have something like the following set of grains:
    Pain, vertebrate pain, mammalian pain, primate pain, human
    pain, Caucasian pain, Pete Mandik pain, Pete Mandik on June 6, 2007 at 12:00 EST pain.

    Now, B&M’s (surprising) suggestion in the passage just quoted is that you can pick whichever psychological grain you want and still be able to find a correspondingly coarse or fine realization of this grain.  (This is a much different kind of claim that one about why Fodor, et al., went wrong.)  It sounds like we have some sort of guarantee that every one of these psychological properties is going to be univocally realized.  That’s big news.

    So, there seems to me to be some very pressing need to say just what these different brain grains are.

  10. Hi Ken,

    I will admit to being sloppy in my previous comment, but I don’t endorse the analyses of grain you initially attribute to me (one in which the neural grain heirearchy is mereological). I think the way to ultimately spell this out is perhaps something you agree with: that both grain heirchies need to be conceived as heirechies of kinds, sub-kinds, sub-sub-kinds, etc.

    If were are focusing on states or events, then we need parallel heirarchies of kinds of states or events. If we are focusing on processes, then we need parallel heirachies of kinds of processes. And so on.

    To flesh out one of my earlier examples, if we are focusing on seeing, then the psychological heirearchy will be something like seeing, visual object identification, visual object identification by color. The corresponding neural grain heirarchy will be something like post-retinal neural processing, post-retinal neural processing in the ventral stream, post-retinal neural processing in V4.

  11. kenneth aizawa

    I did take you to be defending a mereological theory of grain, but I’m perfectly happy to go with your hierarchy of kinds.  That makes some sense to me. 

    Nevertheless, it seems that for B&M, in the passage cited and before that in their text, their notion of grains has to admit of coarseness sufficient unto bringing humans and octopi under one rubric, e.g. animal hunger, and fineness sufficient unto Pete’s hunger at time t.  Moreover, this notion of grain must work at both the psychological and the neuroscientific level.  So, among other things, you need a “trans-specific’ account of these grains, which it seems to me that you do not have.  I think this problem arises whether you have a mereological conception, or a hiearchy of kinds conception.

Comments are closed.