Specificity and Multiple Realization

who wish to challenge the multiple realization of psychological properties sometimes
suggest that it is more likely that a more general property, such as being in
pain, is multiply realized than it is that a more specific property (of that
general kind), such as human pain, is multiply realized.  There are more ways to make a powerful engine
than there are to make a powerful V8 engine. 
There are more ways to make a powerful V8 than there are to make a 350
hp V8.  The finer the grain of
description, the more constraints there will be on the modes of construction, hence
the fewer realizations there will be.

a view, however, does little, by itself, to undermine the view that
psychological properties are multiply realized. 
For one thing, this leaves open at least the possibility that there are
scientifically legitimate general properties, such as being in pain, and that
they are multiply realized.  For another,
it leaves open the possibility that even the fine grained, specific properties
are multiply realized, even though this is less likely than it is for general
properties.  After all, we know that there
are many ways to design a 350 hp V8.  Different
cylinder displacements, different compression ratios, different fuels can be
used to make such an engine.  Returning
to the example of pain, even if pain is more likely to be multiply realized
than is human pain, that leaves open the possibility that human pain itself is
multiply realized.  Something stronger is
needed to press this kind of challenge to the multiple realization of
psychological properties.



  1. Frankie Egan

    It is a purely logical point that there are more ways to multiply realize a determinable (e.g. colored, pain) than a determinate property (red, pain in humans). But you’re right that there is nothing here to challenge the multiple realization of psychological properties.

  2. kenneth aizawa


    I’m glad we agree here.  (Bleg: as a purely information point, has anyone noted this in print?)

    But, you know, Bechtel and Mundale, 1999, late in their paper, seem to think they can muster an argument for unique realization out of this kind of thing.  It’s not completely clear, to me at least, how this goes.

    So, what this suggests is that some critic of MR might want to try to find a plausible stronger claim to use to challenge MR. Maybe one might poke around in the neighborhood of the idea that absolute determinates are uniquely realized.

  3. gualtiero piccinini

    Not in general.  If the absolute determinates are of a fundamental property, it’s plausible.  If they are of a higher level property (like a certain shade of blue), it’s implausible.

  4. Christopher Richard Wade Dettling

    It is a view?
    It is a view, or it is not a view, or it is nonsense.
    If you say it is a view, then it must be a view, since you must know the difference.

  5. Ken,

    I think you are right. But here’s an argument that absolute determinates need not be uniquely realized. Suppose that among the absolute determinates are the fundamental properties of, say, physics. And suppose that the fundamental properties of physics are all dispositional, and suppose that all dispositional properties have to have non-dispositional qualitative grounds. Then the determinates could be realized in ways that were qualitatively different. (I don’t agree with any of this, just wonder if it might fit the bill.)


  6. Brandon N. Towl


    What you say seems right to me: even though “more determinate” properties face more constraints, and are thus *likely* to have fewer realizations, there is nothing in the argument to bar multiple realization.

    Two thoughts:

    First, some folks argue that the only properties there are turn out to be maximally determinate (Gillett and Rives make this argument, as does John Heil). For them, the less determinate (more determinable?) “properties” are just less-specific properties that refer to more determinate properties (given a context).

    Second– although the “grain” issue is orthogonal to issues of multiple realizability (which, I take it, was one of the points you were getting at– correct me if I misunderstood), it is usually offered as an explanation as to why people (e.g. Fodor) thought we had MR properties. This doesn’t establish that MR properties do not exist (other arguments are supposed to do this). But the argument does go some way to explaining why folks thought there were MR properties: because they confused coarse-grain descriptions of singly-realized properties with some sort of MR property.

    I hope these comments were helpful!

  7. kenneth aizawa

    Good question.

    I’m thinking of realization as I think it is found in psychology and neuroscience, which I think is captured by Gillett’s Dimensioned view of realization and as articulated in some of his and my joint papers.

    So, a realized property would be at one level and the realizers at another.  So, maybe a psychological absolute determinate would be making the normal color match of a specific frequency of light.  This would be MR if there were, say, multiple ways of doing this with neurons.  So, what is absolute is at one level, what is MR is at another.

    Now, one does sometimes get hear something like the idea that, given diversity at the neuronal level, you could/should let that tell you that there should be more discriminations at the psychological level.  But, that, let me say, leads to a mess.  For details, you’ll have to wait until I finish my paper. =)

  8. kenneth aizawa


    I think that if G&R are correct, then the issue of the MR of absolute determinates becomes more pressing.  If, the default is that scientifically legitimate properties are absolute determinates, then one should really care more about whether this leads to unique realization.

    I have not done a real lit review on specificity and MR, but this “grain” issue goes much farther back than Bechtel and Mundale. (I am sceptical of the B&M analysis of Fodor’s putative confusion.)  Sp, for example, in Neurophilosophy, Churchland mentions domain specificity as holding out hope for reduction.  (Cf. pp. 356ff).  So, I’m thinking that the issue of specificity and MR has been used, not to argue for UR, but to give hope to UR. 

    So, maybe the fan of UR, will want to find a plausible way to strengthen this hope into something more, where the fan of MR, like me, will want to crush this hope.

  9. kenneth aizawa


    I’ve not ignored your comment, but I just don’t know what to say about dispositional properties.  I don’t know anything about them.  I should, I guess, but I don’t.

  10. kenneth aizawa

    Ok.  So, suppose you have a high level absolute determinate property, e.g. the ability to match a specific color of light.  How is that even possible?  Bear in mind that you can change lower level properties that don’t matter, e.g, say, the density of the myelin sheath of neurons, but those causally irrelevant properties aren’t realizing properties.  (This is, the dime version of one of Larry Shapiro’s ideas.)  So, how can you change genuinely realizing properties without changing the realized property, especially when the realized property is very specific, as in an absolute realizer.

    (I have an answer myself, but it would be nice if this were not a totally obvious answer.)

  11. Brandon N. Towl

    Hi Ken,

    I think you are spot-on with your take on the “grain problem”. And so, good luck with the crushing! I don’t find the argument convincing myself– then again, I think that the fan of UR/reduction has stronger legs to stand on. But perhaps I’m wrong in this.

    BTW, doesn’t Polger make use of the grain argument a good deal in his book? I don’t have the book ready to hand, but I’m sure that one of his later chapters deals with this argument.

    One last comment– do your arguments really require a dimensional view of realization? Not only is this a minority view about realization (or so some people tell me…), but it would seem to me to be easier to defend MR on a *non*-dimensional view. Discussing this fully would be a bit off topic, but I’d be open to an email exchange.

  12. Carl Gillett

    Brandon, a quick comment, since one of your points raises some important background issues. And because it draws out the approach Ken and I are using in our series of joint papers on this stuff.

    Basically, I am a little concerned by your suggestion about using the account of realization that most easily gets you MR. I think that is utterly wrong-headed. Here is the appraoch I favor.

    One starts by trying to find the account of realization that best fits the wider object phenomena — presumably in this case the features of scientific explanations and the concepts they deploy, the nature of scientific methodology and practice, and so. Having decided what the best account of realization in the sciences is, THEN one goes about using that account in one philosophizing on relevant issues, whether the nature and existenc of multiple realization, the character of scientific reduction, the existence of emergence etc.

    I think a lot of people are missing the wider context of issues against which one decides what the best account of scientific realization actually is. (And that is even before we mention the fact that still more people are failing to see that there are different projects for which an account of realization can be given). Ken and I are only interested in a project in the philosophy of science.

    It is also worth noting that one of the larger points Ken and I basically argue for in our series of papers on mr is that the dimensioned view of realization provides a better fit with scientific findings and methodology than competing accounts, such as Larry Shapiro’s. The discussion is still ongoing, but I find our points persuasive. (For example I think we can resconstruct scientific practice, whereas other accounts cannot). I also think there are some pretty powerful underlying reasons to think that what you term the *majority* views, however widely held, are clearly inadequate in the scientific cases. But that is a longer story.

    So, you have to do your philosophy of science first before you even get to talk about mr. Best, Carl

  13. Carl Gillett

    Hi Peter, the time may have past for a reply that gets read, but I was reading through the thread and I liked your question. And I thought about this a while ago. So here is an answer I like which amplifies some of Ken’s suggestions.

    First, the compositional hierarchy in the sciences is distinct from the determinate-determinable hierarchy.
    I take the realization relation in the sciences to be species of composition relation, it is the compositional relation posited between property instances. And usually the properties which are realizers are not related by determinate-determinable relations to the properties they realize — unlike the situation in the contrived cases often considered by philosophers. So to see the point, perhaps looking at a real scientific case (just for a change…) will help to show what I mean and sort things out.

    Take the property of having a Knoop hardness index of exactly 380 kg/mm2. I think that is an absolute determinate property and a commonly posited prperty in the sciences. But it is also, I would suggest, a multiply realized property. For human teeth have instances of the property, and certain metal alloys used in fillings also have instances of this property — for obvious reasons! But the sciences tell us that the lower level properties and relations that realize the property of having a Knoop hardness index of exactly 380 kg/mm2 in teeth are different from those in fillings. In the fillings the realizers are certain properties/relations of metal atoms; in enamel they are different properties/relations of the molecules found in a ceramic.

    So, here we have a multiply realized absolute determinate property in the sciences.

    Now note that having a Knoop hardness index is not related by a determinable-determinate relationship to bonding and spatial relations in metals or ceramics. And I think this is the rule. In such real, concrete scientific cases, the determinate-determinable relation does not marry-up (at least most of the time) with the realizer-realized relation. Just as Ken suggested.

    Hope that helps, best, Carl

  14. kenneth aizawa

    Yes, I don’t think the kind of “absolute determinates are UR” argument is likely to be that strong.  On this one, I’m kind of beating the bushes looking for an argument.  But, I think that there are times when folks write things that suggest they believe something like this.

    Incidentally, what do you think is the best argument for UR of psychological properties?

    Polger mentions “grain” a couple of times in his book, but I’m not really happy with that term, since it seems to me it goes back to Bechtel and Mundale, where they have that “diagnosis” and the claim that all properties are uniquely realized, if you only look around.

    I don’t know if my arguments require the Dimensioned view, but I think the view is the right one for the cases I’m working with.  And I think that the arguments are pretty easy.  So, in truth, this is part of the positive project of pushing the Dimension view and its companion theory of MR.  It would be great if the Dimensioned view were the minority view now.  Carl and I hope to change that over the next five years or so. 

  15. Hi Ken,

    Great post. An interesting question is how to phrase the contrast you describe. As you say, there do seem to be more ways to realise a specific property (being a 350hp V8 engine) than a less specific property (being a V8 engine). However, there also seem to be an unlimited number of ways to realise the properties in either case (unlimited numbers of different engine materials, different displacements, etc.). So perhaps the contrast should not be drawn in terms of raw numbers. Perhaps instead it should be drawn as follows: once the facts are fixed concerning the determinable property (V8 engine), there are many further ways in which it could have or fail to have the determinate property (350hp V8). In contrast, once the facts are fixed concerning the determinate property they are automatically fixed concerning the determinable property. This seems to capture that thought that there are ‘more’ ways to realise the determinable than the determinate, but without committing one to the comparison of two unlimited numbers.

    Something that might be helpful to arguing for unique realisation of increasingly specific properties is a contrast between efficient and non-efficient ways of realising a higher-order property. If one allows any kind of realisation of the higher-order property, efficient or non-efficient, then there is little prospect of unique realisation, since one could imagine all kinds of Rube Goldberg arrangements that produce the higher-level property. Perhaps one might argue that, in our environment, with our resources, and facing the challenges we do, there are only a small number of ways of efficiently realising a higher-order property like colour discrimination, with neurons. Similarly for other high-level psychological properties: perhaps there is only one way, or a handful of ways, in which they could be realised with neurons given our contingent resources, limitations, and environment.

    The disadvantage is that such a restriction might be perceived as changing the topic. Thoughts about multiple realisation often allow us to consider environments where other materials are plentiful, where considerations of efficiency or practicality of the realisation don’t apply, and where even laws of nature may differ. Ruling these considerations out by fait may seem a cheap way of gaining unique realisation… Anyway, I don’t hold any of these arguments myself but thought that they may be a way of fleshing out the thought that the absolute determinates (even of high-level properties) are unique realised (or realised in only a handful of ways).

    All the best,

  16. kenneth aizawa

    Precisely.  I got interested in the Dimensioned view, because it made perfect formal sense to me of what I took to be implicitly at work in the putative unique versus multiple realization of memory consolidation in biochemistry.  That’s what my 2007 Synthese paper is about.  So, Bickle and I were having this debate about MR in this area.  The Dimensioned view makes sense of it for me.  That makes the Dimensioned view attractive.  And since then it has gone from strength to strength for me.

  17. kenneth aizawa

    So, what I am thinking about is what more might be said to someone who is sceptical about what Carl says in the third paragraph.  Can two materials really have precisely the same hardness?  How is that possible?

  18. kenneth aizawa

    So, for example, in one paper by Gillett and Aizawa, we claimed that two green cone opsins had the same absorption spectrum.  Bickle took except to the idea that there really exactly the same.  I want to try to spell out how this can happen, although in the case of the cone opsins I’m betting there is no science journal article spelling it out.  Maybe the best that can be done with John is that I make him a little less indignant about this claim.  … Nah.

  19. kenneth aizawa


    The issue of infinitely many realizations of both specific and general properties did cross my mind, but to no initial effect.  Now I think that it might be easiest just to relate the two in set theoretic terms: the specific is a property subset of the general.

    I do think that adding a consideration of “efficiency” would risk changing the subject.  In recent discussions, there has a been shift away for multiple realizability to multiple realization, from a modal to a non-modal notion.  I think that critics of MR, e.g. Bickle and Bechtel and Mundale, have done this mostly because they think they can simply report on actual scientific practice to establish unique realization.  I think the move to M realization is a risky gambit for the critic of MR.  The most they can hope to do is undercut actual evidence for M realizability.  By contrast, what Gillett and I do (in arguing for M realization) is to show (I think, at any rate) that there is actual M realization, so that we get M realizability for free. 

Comments are closed.

Back to Top