The Multiple Realization Book: Précis

Thomas Polger and Lawrence Shapiro, The Multiple Realization Book (OUP, 2016)

First, we’d like to thank John Schwenkler for giving us the opportunity to talk about The Multiple Realization Book (OUP 2016) on Brains. For about twenty years we’ve each been trying to understand the phenomenon of multiple realization and its importance in philosophy of mind and philosophy of science. In the mid-1990s we each, independently, became dissatisfied with the state of debates over multiple realization in the philosophy of mind and the cognitive sciences, e.g., those that focused on philosophical reconstructions of scientific categories or that centered around the question of what counts as reductive explanation. We each concluded that the very idea of multiple realization had been insufficiently articulated, and we each became convinced that a better understanding of the phenomenon was necessary, and we each had the hunch that the result would be favorable to brain-based or mind-brain identity theories. The initial parts of the inquiry went into Shapiro’s “Multiple Realizations” paper and his book The Mind Incarnate (2004), and into Polger’s paper “Putnam’s Intuition” and his book Natural Minds (2004). When writing our first books we became aware of each others’ work, and by 2005 or so we had begun to collaborate. The Multiple Realization Book is a consolidation and extension of the arguments we’ve been working on during this time.

We are philosophical naturalists. Of course ‘naturalism’ is a term that covers many sins. For us the main upshot is doubly methodological: First, philosophical questions are questions about the world, broadly construed, and should be approached by the same methods and standards of inquiry that are used for any other questions about the world. In particular, we think that all claims about the world can be evaluated in terms of their evidential support. Second, inquiry into the nature of minds (or anything else) does not begin de novo, but rather against the background of what we (i.e., we human beings) already know about the world. And what we know, we think, is that all the phenomena in the world that we know about are broadly physical. It’s famously hard to say just how physicalism as a general doctrine should be defined. But in the special case of the nature of minds, it’s generally clear what physicalist theories look like. (If we follow our approach and end up with a borderline case, then we’ll cross that bridge when we come to it—and we anticipate some of that bridge crossing in the second to last chapter of The Multiple Realization Book.)

Given our methodological commitments concerning where to start our inquiry and how to proceed, we agree with Jaegwon Kim when he says: “the mind-body problem—our mind-body problem—has been that of finding a place for the mind in a world that is fundamentally physical” (1998: 2). And since the mid-twentieth century, the main choices have been brain-based approaches (e.g., the mind-brain identity theory), varieties of behaviorism, and varieties of functionalism or other realization-based theories.

As will be familiar to many readers, the functionalist or realization theories have been ascendant since the late 1960s. Ask any undergraduate who has taken an introductory course in philosophy of mind why these theories became popular, and you will get a story about how behaviorism was refuted by Putnam and Chomsky, and how multiple realization refutes the brain-based theories. Functionalist or realization theories are “the only game in town.” But it is often forgotten that the reason for favoring realization-based theories is that they purport to do the best job of balancing the demands to account for the reality and explanatory saliency of psychological processes, on the one hand, and the diversity of psychological beings, on the other hand. Realization theories are preferable because the world is full of psychological unity despite physical disunity. This phenomenon is multiple realization—the hypothesis that a variety of physically distinctive creatures (systems) can—despite their physical dissimilarities—nevertheless have or be in the very same psychological states, e.g., pain, belief, desire, hunger, and so on. That multiple realization is actual or ubiquitous, or even possible (i.e., that psychological states are multiply realizable even if not multiply realized), is widely thought to be incompatible with brain-based theories of the mind, and therefore to strongly favor functionalist and other realization-based theories. (Behaviorism is generally thought to falter by providing an inadequate basis for the reality and explanatory saliency of psychological processes.)

It’s at this point that the other component of our methodological naturalism kicks in. Historically philosophers have sometimes thought that the “mind-body problem” is different in kind from other explanatory projects. But from our current vantage point, “our mind-body problem” (as Kim says) is just like any other explanatory challenge, and we have well established methods for approaching it. Once we recognize that the choice among theories of the nature of minds is a challenging but otherwise ordinary example of theory choice in the sciences, then we can bring our familiar methods for making such choices to bear. Chief among those is to carefully evaluate the evidence supporting various options.

The project of The Multiple Realization Book, then, is to closely examine whether the evidence for multiple realization or multiple realizability is all that it’s cracked up to be. In the end we conclude that the evidence for widespread multiple realization or multiple realizability is much weaker than is commonly supposed. Our argument proceeds in three parts.

First, we have to formulate an account of multiple realization that is precise enough to enable evaluation. Above we glossed the idea of multiple realization as, “the hypothesis that a variety of physically distinctive creatures (systems) can—despite their physical dissimilarities—nevertheless have or be in the very same psychological states.” But what makes two psychological states “the very same”? And what makes their bearers “physically distinctive”? In our view, much past debate over multiple realization has depended on common sense or ad hoc conceptions of sameness and difference, with the result that advocates and critics of multiple realization have often ended up talking past one another or dogmatically asserting their pretheoretical views. In order to make the idea of “same but different” precise, we insist on taking seriously the idea that if multiple realization were as well supported as it is supposed to be, then it would—quoting ourselves—“be incompatible with brain-based theories of the mind and therefore to strongly favor functionalist and other realization-based theories.” This requirement provides something like a job description for multiple realization: whatever kinds of sameness and difference make for multiple realization, they have to be such that they constitute at least a prima facie obstacle to brain-based theories of the mind and other “reductive” theories, viz., the mind-brain identity theory.

This job description for multiple realization turns out to be harder to satisfy than has been generally appreciated. Considering it leads us to the following criteria for multiple realization. Things of types A and B are multiple realizers of some kind if:

  • As and Bs are of the same kind in model or taxonomic system S1;
  • As and Bs are of different kinds in model or taxonomic system S2;
  • the factors that lead the As and Bs to be differently classified by S2 are among those that lead them to be commonly classified by S1;
  • the relevant S2-variation between As and Bs is distinct from the S1 intra-kind variation between As and Bs.

As a gloss: multiple realization requires that realizers be the same but different (i & ii), and differently the same (iii & iv).

We illustrate our criteria with examples from natural sciences as well as with examples of artifacts, such as corkscrews. Two waiters corkscrews with different colored handles are not multiple realizers of waiters corkscrew (or corkscrew) because the color of a corkscrew is irrelevant to its being a corkscrew—so they fail to meet criteria (iii). Two waiters corkscrews with handles of slightly different lengths are not multiple realizers of waiters corkscrew (or corkscrew) because although handle length is relevant to being a corkscrew (or waiters corkscrew)—so (iii) is satisfied—small differences in handle length are individual or intra-kinds differences, not differences in kind, so (iv) is not satisfied. Of course one could challenge our criteria, both in terms of the job description we attribute to multiple realization and in terms of the adequacy of our specific criteria. We’re happy to admit that our criteria might not be the final word in every detail. But we think we’re in the right ballpark and much closer than our competitors. We defend these criteria by arguing that varieties of variation that are weaker than this do not do the job of multiple realization—they are not even prima facie obstacles to the mind-brain identity theory or other “reductive” theories.

Second, having set out our criteria, we are in a position to evaluate the evidence for multiple realization. There are two phases: What is the evidence for actual multiple realization? Then, What is the evidence for multiple realizability in the absence of evidence for actual multiple realization? To answer the former, we examine what we take to be the most favorable cases that are supposed to demonstrate actual multiple realization; and we find that many of the most commonly cited examples fail to satisfy our criteria. To answer the latter, we consider what sort of evidence could be used to support theoretical claims that imply that psychological states are multiply realizable even if there are fewer examples of actual multiple realization than advocates expected. The three theoretical commitments that we consider involve (a) neural plasticity and reuse, (b) convergent evolution, and (c) artificial intelligence. In each case we argue that the evidence does not support those theoretical commitments.

Finally, third, we consider the fallout from our arguments against multiple realizability. We have to make sure that our arguments against multiple realization don’t undermine other theoretical desiderata. Is our theory eliminativist or antirealist rather than realist? (No.) Can we account for causal explanation of and by the mental? (Yes.) Does our account fit the actual practices and methodologies of the cognitive and brain sciences? (Yes.) And can we justify the explanatory value of psychology and the cognitive sciences? (Yes.)

In subsequent blog posts we explain a bit more about our approach, our arguments, and their wider implications. But of course the devil is in the details—and the details are in The Multiple Realization Book.

Note: All citation details can be found in the references of The Multiple Realization Book.

8 Comments

  1. Kenneth Aizawa

    So, here is a typo, I think. You write,
    Things of types A and B are multiple realizers of some kind if:

    As and Bs are of the same kind in model or taxonomic system S1;
    As and Bs are of different kinds in model or taxonomic system S2;
    the factors that lead the As and Bs to be differently classified by S2 are among those that lead them to be commonly classified by S1;
    the relevant S2-variation between As and Bs is distinct from the S1 intra-kind variation between As and Bs.

    This makes the conditions sufficiency conditions, but I think you mean for them to be necessary conditions. I’ve noted this before, since the account is of sufficiency conditions in your “Identity, variability, and multiple realization in the special sciences”.

  2. Kenneth Aizawa

    In “Identity, variability, and multiple realization” the four conditions are conditions for
    “Multiple realization of one kind by another occurs when:”

    In this post, we the four conditions are conditions for
    “Things of types A and B are multiple realizers of some kind if:”

    So, are the four conditions conditions for i) when two kinds A and B realize another kind or ii) when two things of types (members of kinds?) A and B realize another kind?

    (In The Multiple Realization book we just get the four conditions.)

  3. Larry Shapiro

    Hi Ken, the four conditions are meant to describe conditions sufficient for A and B to be multiple realizations of some kind, call it K. So, A and B multiply realize K when, judged from one taxonomy, they are instances of K, but, judged from some other taxonomy, they are distinct. So, e.g., perhaps simple and compound eyes are, according to one kind of typing, the same kind, but, according to another kind of typing, distinct. I’m not sure why you think there’s a typo in our account. We intend the four conditions to be jointly sufficient. I’m also prepared to say that each is individually necessary, but Tom might not be so bold.

  4. Kenneth Aizawa

    Thanks, Larry. So, let me being with this: “I’m not sure why you think there’s a typo in our account. We intend the four conditions to be jointly sufficient. I’m also prepared to say that each is individually necessary, but Tom might not be so bold.”

    The main reason that I think you mean “only if” (i.e. that the conditions are supposed to be individually (and perhaps jointly) necessary) is that this is how you seem to deploy them on many occasions. You make argue that such-and-such case is not an instance of MR, because the case fails to satisfy one of i)-iv). So, for example, the slow watch and the fast watch (TMRB, p. 71). You argue that they are not multiple realizations of watch because they fail iv). Or take a case near and dear to my heart. You claim that the cone opsin case does not satisfy condition iv): “But consider the last criterion that the variation in the opsins must meet if it is to qualify as multiple realization: (iv) The differences among the opsins are distinct from individual differences among trichromats.: (TMRB, p. 109) Also, Polger, Thomas W.; Shapiro, Lawrence A. (2016-07-07). The Multiple Realization Book (Kindle Locations 2330-2333). OUP Oxford. Kindle Edition. I think there are others as well.

    But, there is also this:
    In the previous chapter we defended our Official Recipe for multiple realization. There are at least four jointly necessary conditions for multiple realization to occur:
    (i) As and Bs are of the same kind in model or taxonomic system S1.
    (ii) As and Bs are of different kinds in model or taxonomic system S2.
    (iii) The factors that lead the As and Bs to be differently classified by S2 must be among those that lead them to be commonly classified by S1.
    (iv) The relevant S2-variation between As and Bs must be distinct from the S1 intra-kind variation between As and Bs.
    (TMRB, p. 89) Or, Polger, Thomas W.; Shapiro, Lawrence A. (2016-07-07). The Multiple Realization Book (Kindle Locations 1914-1921). OUP Oxford. Kindle Edition.

  5. Kenneth Aizawa

    So, “the four conditions are meant to describe conditions sufficient for A and B to be multiple realizations of some kind, call it K. So, A and B multiply realize K when, judged from one taxonomy, they are instances of K, but, judged from some other taxonomy, they are distinct. ”

    This sounds like you are dropping the talk of members of types/kinds, which appeared in the post. Right?

    But, more importantly, it seems that you are taking A and B to be kinds. What you want to provide, I think, is an account of when kind A and kind B constitute multiple realizations of kind K. Right?

    • Tom Polger and Larry Shapiro

      Kinds and taxonomies are important. We think they go along with explanations or models. And they are how we find out in a non ad hoc way which differences between individuals are relevant and which are not, which are type (kind) differences an which are individual differences within a type. So, yes, to a first approximation, when the individuals that are co-classified by one taxonomy (i.e., of the same kind) are cross-classified by a different taxonomy (i.e., not of the same kind), then—provided the sameness and difference meet our criteria—we have an example of multiple realization.

  6. Tom Polger and Larry Shapiro

    Ken, we hedge sometimes because there’s always the chance that we’ll be convinced that there are additional criteria we should add. But, sure, satisfying these criteria is necessary and sufficient for being evidence of MR as we currently understand it. Sometimes we make the weaker claim, but that is consistent with the stronger claim. You’re right that there are small differences between the formulations between the IVMR paper and TMRB. We’re happy to continue to refine the formulation; the bigger question is whether we can convince people that we have the right general approach. If MR has nothing to do with reduction, explanatory generality or relevance, and the choice among different theories… then there’s a bigger problem for us.

  7. Kenneth Aizawa

    Ok. So, how does the Official Recipe being *today*
    Kind A and Kind B are multiple realizers of some kind K iff …

    or

    Things of kind A and kind B are multiple realizers of some kind K iff …

    The initial post seemed to be the second disjunct, but then Larry seems to think that it is the first disjunct.

Comments are closed.

Back to Top