There’s no overestimating the significance of the multiple realization thesis in the past fifty years of theorizing about the mind’s relationship to the brain. The idea behind the thesis is simple enough, and most easily explained in terms of a comparison. Suppose you thought that the relationship between the mind and the brain is like that between water and H2O. This latter relationship involves an identity. To say that water is H2O is to claim that the kind water just is the kind H2O. Where there’s one, there’s the other; the words ‘water’ and ‘H2O’ are just different ways of referring to the same thing. Similarly, you might suppose, where there are minds there are brains, and the words ‘mind’ and ‘brain’ are just two different ways of talking about the same thing.
It’s the idea that minds are brains that the multiple realization thesis is meant to challenge. However, unlike the dualist, who denies mind-brain identity on the grounds that minds are a sui generis kind of spooky substance—entirely distinct from physical substance—advocates of multiple realizability deny the identity in a way that’s friendly to anyone with a broadly physicalist outlook. Minds—that is, mental states and processes—are functional kinds, defined not in terms of what they are made of but in terms of what they do. Mental capacities like memory or perception are to be specified in terms of operations or activities, and, as such, can be implemented—a.k.a., realized—in various ways. Being a mind is like being a watch. Given that watch is a kind defined by the function of keeping time, an object can be a watch whether realized by a digital or analogue mechanism. So too, the realization theorists expect, mental capacities can be realized by different sorts of devices— those in the heads of human beings, but also those in Vulcans or artificial intelligences.
In fact, some realization theorists have gone so far as to claim that minds can be realized in virtually an infinite number of ways. Consequently, psychology, qua science of the mind, is assured of its distinctness and independence from neuroscience which (while no doubt useful for understanding how brains work, they might concede) cannot possibly provide us with information about how minds work.
In short, the relationship between mind and brain is not like that between water and H2O, because while only one kind of thing can be water—namely, H2O—many different kinds of things can realize minds—can do the job of minds. And so, the realization theorist argues, minds are not identical to brains.
We take a close look at the thesis that minds are actually multiply realized or, perhaps, even possibly multiply realizable. Surprisingly, given that claim of multiple realizability has enjoyed nearly unanimous assent among philosophers of mind and psychology, it has never been clearly articulated. As far as we know, until we began our work on multiple realization no one thought to specify conditions for what it is required for a kind to be multiply realized, such that the realizing kinds could do the work required by the realization theorist intent on denying mind-brain identity. Having thought long and hard about these conditions, we now believe that the multiple realization thesis is false, or, at any rate, hardly the obvious truth that it has for so long seemed to so many.
The route to this conclusion begins with an insight about the burden that realization theorists must meet. They must provide an account of what it means for different kinds of realizers to be the same kind of thing. Just as a realization theorist about, e.g., watches, must explain why digital watches and analogue watches are different in one sense but the same in another, so too the realization theorist about minds must be able to justify the claim that mental kinds can be in one sense different but the same. We contend that on a reasonable formulation of what it means to be different but the same, realization theorists fail to make their case.
Our guiding idea is that the “sameness” in the formula must be sufficiently robust, and, similarly, that the “differences” must be sufficiently relevant. The reason that an hour glass does not count as a way of realizing a watch is that it is not sufficiently “watch-like.” Likewise, the reason that two analogue watches, alike in all respects except that the minute hand on one is slightly longer than the minute hand on the other, do not count as distinct kinds of realizations of a watch is that their differences are not sufficiently relevant.
Of course, making more precise what counts as sufficiently similar and relevantly different is not easy. However, we hope that it should strike you as obvious that such a task should be pursued prior to dismissing the possibility of mind-brain identity on the grounds that minds are multiply realizable. Even if we are wrong, in the end, about how to understand similarity and difference, we are quite sure we are right that any realization theorist who rejects mind-brain identity on the basis of the multiple realization thesis owes such an account.
So how do we propose to analyze the similarities and differences that constitute multiple realization? As follows. 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.
The appeal to taxonomic systems or models in (i) and (ii) provides a broad characterization of the multiple realization formula. Digital and analogue watches are the same kind of thing relative to the way watches are distinguished from toilets or toasters; yet they are different kinds of things relative the way a jeweler might think about how to repair them. Conditions (iii) and (iv) are designed to capture the right kinds of similarities and differences between distinct realizers. The digital and analogue watches both perform the same watch-qualifying function by means of different kinds of mechanisms (condition (iii)), and the differences between them are more than those one would expect to see between variants of the same kind, such as that between two analogue watches with minute hands of different lengths (thus satisfying condition (iv)). Thus, on our analysis, digital and analogue watches count as different realizations of the kind watch; but digital watches and hour glasses do not;nor do two analogue watches that differ only in minute-hand length.
We next apply this analysis of multiple realization to an assortment of cases that philosophers have used to support the multiple realization thesis: cases involving neural plasticity, artificial intelligence, and instances in which evolution appears to converge on different neural architectures to produce the same psychological functions. Stacked up against our analysis, we argue that these lines of evidence often fall well short of exhibiting definitive cases of multiple realization. Of course, we’re in no position to assert that no evidence for the multiple realization of psychological states will ever be forthcoming; and we’re even open to the possibility that good evidence for some cases of multiple realization already exists. Our concern is the more modest one of defending a limited sort of identity theory, where we’ll find that neuroscience motivates identity claims between some psychological capacities and neurological processes, while permitting the possibility that other psychological capacities may well be multiply realized.
In the end, we propose a picture of the relationship between psychology and neuroscience that is rather more complicated than those that either identity theorists or realization theorists have traditionally painted. Some (but not all) psychological kinds will be identical to neural kinds; some (but not all) psychological kinds will be multiply realized in neural or other kinds; some (but not all) psychological kinds will be identical to neural kinds which in turn will be multiply realized in more basic molecular biological kinds; some (but not all) psychological kinds will be multiply realized in neural or other kinds, some (but not all) of which might be identical to more basic molecular biological kinds. It’s a messy and perhaps inelegant conception of how psychological kinds stand in relation to other scientific kinds. But we ought to be long past thinking that book of nature will be an easy read.
With a more sophisticated rendering of the connections between psychology and other sciences, questions about how to understand the autonomy of psychology might seem to become more pressing. However, we think that the issue of autonomy can be addressed straightforwardly. The reason to retain psychological models and explanations, despite our belief that parts of psychology might be reducible to more basic sciences, emerges from the fact that psychological processes are in fact difference makers. That is, psychologists and cognitive scientists regularly demonstrate that interventions on psychological processes make differences to an organism’s behavior. This remains true even if interventions on psychological processes are also ipso facto interventions on the neural processes with which they are sometimes identical. This suffices, we argue, to furnish psychology with everything one can ask for from an autonomous science.
We end up a view of psychological kinds as autonomous, but sometimes reducible. As real, but not always realized.
Featured image: Ocean whirlpool. CC0 via Pexels.