DRAFT ONLY; COMMENTS WELCOME!

Realization and the Formulation of Physicalism

Andrew Melnyk

University of Missouri

In an important paper now 25 years old, Richard Boyd proposed that physicalists about the mental don’t have to hold that mental phenomena, whether types or tokens, are identical with physical phenomena; it’s enough for physicalism about the mind if, as Boyd put it, “in the actual world all mental phenomena are physically realized” (my italics; Boyd 1980, 87). In view of Boyd’s proposal, philosophers interested in how best to formulate a doctrine of physicalism tout court (not merely about the mental) will want to consider what we may call realization-based formulations of physicalism. Such formulations are those that elaborate the following schema:

(RBFP) Everything that exists is either identical with the physical (in a certain narrow sense of “physical”) or realized by the physical.

Elaborating this schema obviously requires many things, but one of them is a clear account of the all-important relation of realization.

In a recent book, I have developed a realization-based formulation of physicalism, realization physicalism, that appeals to a particular account of the realization relation (Melnyk 2003, Chs. 1 and 2). But other accounts of the realization relation may be found in the literature. So what happens if you try to develop a realization-based formulation of physicalism that appeals to one of these other accounts of realization? That’s the question this paper will address.

The plan of the paper is as follows. In the first section, for the purposes of subsequent comparison and contrast, I’ll provide a partial sketch of realization physicalism. In the second, I’ll investigate the prospects for a realization-based formulation of physicalism that uses an account of realization suggested by Ernest Lepore and Barry Loewer (LePore and Loewer 1989). In the final section, I’ll investigate the prospects for a realization-based formulation of physicalism that uses an account of realization recently suggested by Sidney Shoemaker (Shoemaker 2001). My conclusion will be that while the first of these alternative accounts of realization doesn’t yield a promising rival to realization physicalism, the second one, if suitably refined, perhaps does.

I

Turning the schema (RBFP) into a substantive formulation of physicalism requires giving an account not only of realization but also of two other crucial matters: (i) the exact scope of the phrase “everything that exists,” and (ii) the precise character of “the physical.” But since the detailed accounts of these matters that realization physicalism provides are inessential for present purposes, I will neglect them in the sketch of the view that follows.

Exposition of the aspects of realization physicalism that are essential for present purposes must begin with its account of realization. According to the account, realization is a relation that holds not between types (e.g., between properties or event-types) but between tokens of types (e.g., property-instances or event-tokens). Moreover, a realized token can be realized only if it’s a token of a functional type, i.e., a type whose tokening just is the tokening of some or other type that meets a specific associated condition, C (e.g., by playing a certain role). Correlatively, a realizing token realizes a token of a given functional type by being a token of some or other type that meets the special associated condition for that functional type (e.g., by playing the requisite role). Here, then, is realization physicalism’s account of realization:

(RPR) Token x realizes token y (or: token y is realized by token x) iff

(i) y is a token of some functional type F (i.e., some type whose tokening just is the tokening of some or other type that meets a certain condition, C);

(ii) x is a token of some type that in fact meets C; and

(iii) the token of F whose existence is logically guaranteed by the holding of clause (ii) is numerically identical with y.

I should perhaps further explain the definite description in clause (iii), “the token of F whose existence is logically guaranteed by the holding of clause (ii).” If, as clause (ii) asserts, x is a token of some type that in fact meets condition C, and if, as clause (i) implies, it suffices metaphysically for the tokening of functional type F that some or other type that meets C be tokened, then the existence of x logically guarantees the existence of a certain token of F; and it is that token of F to which the definite description in clause (iii) refers.

Here is how the account applies to a concrete case. Suppose that my present headache is realized by some simultaneous physical occurrence in my brain. In that case, according to the account, a headache must turn out to be a functional event-type of some specific kind, the physical occurrence in my brain must be of some type that in fact meets the special condition that characterizes the functional nature of a headache, and my present headache must be the very headache whose existence is necessitated by the physical occurrence in my brain. However, whether my present headache is identical with the physical occurrence in my brain is a question left open by the fact that the former is realized by the latter. Moreover, nothing in my account of realization, nor anything in realization physicalism more generally, gives any reason to expect an a priori way of determining the functional nature of any folk psychological state. If folk psychological phenomena—or for that matter meteorological phenomena—turn out to be functional phenomena, then that will have to be discovered a posteriori.

With realization physicalism’s account of realization in hand, let us turn next to its account of what it takes for a functional token to be realized by the physical, as schema (RBFP) has it. Realization physicalism understands the property of being realized by the physical as that of being physically realized, where this latter property is defined as follows:

(RPPR) A token y of a functional type, F, is physically realized iff

(i) y is realized by a token of some physical type, T; and

(ii) T meets the special associated condition for F solely as a logical consequence of (a) the distribution in the world of physical tokens and (b) the holding of physical laws.[1]

We can safely ignore the reasons for favoring this account of being physically realized; also the precise understanding of what laws should count as physical laws for the purposes of its second clause (see Melnyk 2003, Ch.1).

Realization physicalism, however, can at last be stated, adequately for present purposes, as the following thesis:

(RP) Every token of any type is either itself a physical token or else a functional token that is physically realized.

II

Let’s now explore the prospects for a formulation of physicalism based on the understanding of realization offered by LePore and Loewer. That understanding is expressed in a single paragraph, quotations from which I will discuss in detail (LePore and Loewer 1989, 179-80; all quotations to follow come from this passage). LePore and Loewer begin by asking, “Exactly what is it for one of an event’s properties to realize another?”, which perhaps makes it sound as if they conceive of realization as a relation that holds between types such as properties or event-types. But not so, for they begin their answer to the question by saying that “The usual conception [sc. of realization] is that e’s being P realizes e’s being F iff…”, which makes it clear that they conceive of realization as, or primarily as, a relation between tokens, presumably events. Iff what? “…iff e is P and e is F and there is a strong connection of some sort between P and F.” And they immediately add: “We propose to understand this connection as a necessary connection which is explanatory [their italics]”.

It’s hard to say exactly what this explanatory necessary connection amounts to for LePore and Loewer. The part about a necessary connection between P and F is clear enough: an explanatory necessary connection between P and F requires that it be physically necessary that PàF. However, the further requirement that this necessary connection be explanatory isn’t clear, though it is clear that the envisaged explanandum is an event (or state), not a regularity, because they speak of e’s being P explaining e’s being F. According to Lepore and Loewer, then, e’s being P can explain e’s simultaneously being F. But how precisely do they envisage this sort of synchronic explanation of an event by an event?

They say two things relevant to this question. The first is that “For e’s being P to explain its being F it may be necessary for there to be a system of connections between realized and realizing properties of property kinds to which P and F belong [their italics]”. This remark suggests the following more or less deductive-nomological view of the explanation in question: e’s being P can be explain e’s simultaneously being F, given that (a) e’s being P and e’s being F fall under the physically necessary law that PàF and (b) that this physically necessary connection between P and F isn’t a lone, isolated connection, but rather one element in a system of such connections between P-type properties and F-type properties.

Unfortunately, even if e’s being P can in this way explain e’s being F, the realization relation as understood by Lepore and Loewer—at least thus far—doesn’t seem strong enough to serve physicalist purposes. To see this, suppose that P is a physical property, that F is a mental property, and that e’s being P realizes e’s being F in LePore and Loewer’s sense. Then two conditions are met: (i) it’s physically necessary that PàF, and (ii) this physically necessary connection between P and F is but one element in a system of such connections. But the meeting of these two conditions doesn’t give physicalists what they want, because it doesn’t seem to ensure that e’s being F is in any significant way physical. Why? Because no sense appears to have been given in which e’s being F is constituted by e’s being P; and no sense appears to have been given in which the claim that e is F is made true by the physical way things are.[2] That neither of these conditions is met would be plain, I think, if the necessary connections invoked between P-type properties and F-type properties were nomologically (not physically) necessary. For in that case realization in the style of Lepore and Loewer would be consistent with F’s being an entirely non-physical property that is instantiated in accordance with certain brute laws of emergence whenever a suitable physical base property is simultaneously instantiated. But what difference could it make if the necessary connections between P-type properties and F-type properties were physically necessary? Apparently none, though I will shortly return to this question.

LePore and Loewer say a second thing that bears upon the question how precisely they envisage the sort of synchronic explanation by which e’s being P can explain e’s being F. “[I]t may require,” they say, “that the central laws and principles governing the realized properties be explained by [i] the connections between basic and non-basic properties and [ii] laws governing the basic properties.” Can this additional requirement save LePore and Loewer’s understanding of realization from the charge that it doesn’t give physicalists what they want? It would appear not. The original difficulty was that even a system of physically necessary connections between P-type properties and F-type properties doesn’t appear to yield a sense in which the F-type properties are physical (even though the P-type ones are). But requiring that the F-type laws be explained by appeal to those very same physically necessary connections plus the P-type laws does nothing to reinforce or supplement those connections. So the additional requirement still leaves LePore and Loewer’s understanding of realization unable to yield an adequate realization-based formulation of physicalism.

However, their understanding of realization still has merit. It’s true that realization is a physically necessary connection that’s explanatory. But this truth, I suggest, is a consequence of the truth of the account of realization assumed by realization physicalism. So LePore and Loewer’s understanding of realization can be incorporated into a realization-based formulation of physicalism. But since that formulation is realization physicalism, it is obviously no rival to realization physicalism. Let me now justify these remarks.

Suppose, as before, that P is a physical property and that F is a mental property. But now suppose that e’s being F is physically realized, in the realization physicalist’s sense, and realized, in particular, by e’s being P. Then e’s being F just is e’s having a certain functional property—the property of having some or other property that meets associated condition C. And e’s being P realizes e’s being F because being P meets that associated condition. But, and here’s the crux, since being P meets that condition because of the laws of physics, it’s physically necessary that if e is P, then e has some or other property that meets C; but it’s metaphysically necessary that if e has some or other property that meets C, then e is F; hence, it’s physically necessary that if e is P, then e is F. So the realization physicalist’s understanding of realization can explain why it’s physically necessary that PàF, given that e’s being P realizes e’s being F. And that physically necessary connection between e’s being P and e’s being F was something that sorely needed explaining. For LePore and Loewer assume throughout that P and F are distinct properties, physical and mental; and obviously no physical law connects a physical property to a mental one. So it was prima facie mysterious all along how it could be physically necessary that PàF. But this can be explained if what LePore and Loewer say about realization is viewed as a consequence of realization as understood by realization physicalism.

Furthermore, if realization is as realization physicalism understands it to be, then LePore and Loewer also turn out to be right to hold that if e’s being P realizes e’s being F, then e’s being P explains e’s being F. For if e’s being P realizes e’s being F, then we can offer the following explanation of e’s being F: e is F because it’s P, the physical laws ensure that P meets condition C, and a thing’s being F is its having some or other property that meets C. An explanation of this kind certainly appeals to a metaphysically necessary truth (the identification of F with a certain functional property), but the only contingent facts it appeals to are physical. Hence, I am prepared to say, an explanation of this kind counts as a physically reductive explanation of e’s being F.[3]

II

Let’s turn now to the prospects for a realization-based formulation of physicalism that uses an account of realization recently suggested by Sidney Shoemaker (Shoemaker 2001). Although Shoemaker considers himself to be a kind of functionalist, his account of realization differs from that of the realization physicalist in three important ways. First, whereas the realization physicalist understands realization as a relation between tokens, Shoemaker views it as a relation between properties; for example, between a physical property and a mental property (see, e.g., 2001, 86). Secondly, whereas the realization physicalist understands a functional property as a higher-order property (i.e., as the property of having some or other property that meets condition C), Shoemaker understands a functional property as a first-order property that essentially confers certain causal powers on the objects that possess it (2001, 77). For example, he suggests that “mental properties should be characterized in terms of their causal relations to one another and to inputs and outputs” (original italics; 2001, 84). Finally, and again in contrast to realization physicalism, Shoemaker defines the realization relation between two properties by appeal to a relation of inclusion between the causal powers conferred by the realizer property and those conferred by the realized property.

Shoemaker sketches his ensuing account of realization as follows, where a ‘conditional power’ is a power of a thing to cause some effect if that thing possesses some further property or properties (2001, 77):

“…property X realizes property Y just in case the conditional powers bestowed by Y are a subset of the conditional powers bestowed by X (and X is not a conjunctive property having Y as a conjunct). (2001, 78)

Two remarks. (1) The parenthetical condition is designed to prevent conjunctive properties from counting as realizers of their conjuncts, which, it seems intuitively, they shouldn’t; but since this condition is irrelevant to present purposes, I’ll henceforth ignore it. (2) Because this account of realization says “subset”, and not “proper subset”, it applies to the special case where the conditional powers conferred by X and by Y are identical, so that, given Shoemaker’s principle that “no two properties confer exactly the same conditional powers” (2001, 78), X and Y are one and the same property. Accordingly, Shoemaker adds that, if Y is multiply realized, then the conditional powers that it confers must be a proper subset of the conditional powers conferred by any property that realizes it (2001, 79). Since we are interested in precisely those cases where realization is invoked because type-type identity cannot be, we can rephrase Shoemaker’s account of realization, suitably for present purposes, as follows:

(SR) Property X realizes property Y without being identical with Y iff

(i) every conditional causal power conferred by Y, the realized property, is identical with some conditional causal power conferred by X, the realizing property; and

(ii) some conditional causal power conferred by X, the realizing property, is distinct from every conditional causal power conferred by Y, the realized property.

Now let’s try to formulate a version of physicalism based on Shoemaker’s understanding of realization. Recalling the schema for realization-based formulations of physicalism,

(RBFP) Everything that exists is either identical with the physical (in a certain narrow sense of “physical”) or realized by the physical,

we might at first propose the following, in which “realized” is obviously to be understood in line with (SR):

Every (actual) property is either identical with a physical property or realized by a physical property.

But even as a formulation of physicalism whose scope is restricted to properties, this proposal is inadequate. The problem is that it doesn’t say how often every non-physical property is realized by a physical property; whereas, for physicalism to be true, every non-physical property must always be realized by some or other physical property. To avoid this problem, we might therefore amend as follows:

Every (actual) property is either identical with a physical property or, on every occasion on which it’s instantiated at all, realized by a physical property.

This suggestion is better, but we need to be told what it is for a property to be realized by another on a particular occasion, i.e., what it is for a property-instance to be realized. We need an account of realization as a relation between tokens—in this case, property-instances—that remains faithful to the spirit of Shoemaker’s account of realization as a relation between types.

The following Shoemakerian account naturally suggests itself:

(SRT) a’s being X realizes a’s being Y (where X≠Y) iff

(i) a is X and Y;

(ii) every conditional causal power conferred on a by Y is identical with some conditional causal power conferred on a by X; and

(iii) some conditional causal power conferred on a by X is distinct from every conditional causal power conferred on a by Y.

This account of the realization of property-instances can now be used to generate a final version of physicalism based on Shoemaker’s understanding of realization:

(SP) Every (actual) property is either identical with a physical property or such that every (actual) instance of it is realized by an instance of some or other physical property.[4]

In what follows, I’ll confront this proposed formulation of physicalism with a dilemma. The formulation will be inadequate unless the dilemma can be avoided. But I’ll argue that when the formulation is suitably refined, perhaps it can be avoided.

The dilemma arises because clauses (ii) and (iii) of the Shoemaker-inspired account of realization as a relation between tokens—(SRT)—can be read in two different ways; and this results, of course, in two different ways of reading the proposed Shoemaker-inspired formulation of physicalism, (SP). When the first way of reading (SRT) is used to formulate (SP), then (SP) fails to state a condition that is, intuitively, sufficient for physicalism. But when the second way of reading (SRT) is used to formulate (SP), then a puzzle arises, which needs somehow or other to be solved.

So what are the two ways of reading clauses (ii) and (iii) of (SRT)? Let me focus only on clause (ii), since the application to clause (iii) will be obvious and in any case won’t much matter. Clause (ii), it’ll be recalled, says that

every conditional causal power conferred on a by Y is identical with some conditional causal power conferred on a by X.

Now, presumably talk of causal powers admits of the type/token distinction. Suppose that both a and b have the power to cause an explosion of exactly the same kind. Do a and b have the same causal power? Yes and no. Yes, because they each have a causal power of the same type; but also no, because a’s power to cause an explosion isn’t the same power-token, or power-instance, as b’s power to cause an explosion. So the first thing clause (ii) could mean is that

(A) every type of conditional causal power conferred on a by Y is identical with some type of conditional causal power conferred on a by X.

Given reading (A), the Shoemaker-inspired account of realization as a relation between tokens—(SRT)—becomes this:

(SRT-A) a’s being X realizes a’s being Y (where X≠Y) iff

(i) a is X and Y;

(ii) every type of conditional causal power conferred on a by Y is identical with a type of conditional causal power conferred on a by X; and

(iii) some type of conditional causal power conferred on a by X is distinct from every type of conditional causal power conferred on a by Y.

But this conception of token-token realization—(SRT-A)—leads to an inadequate formulation of physicalism when presupposed by (SP), inadequate because insufficient for physicalism as conceived intuitively. For suppose X is a physical property and Y is a mental property distinct from any physical property. Then even if a’s being Y is realized by a’s being X in the sense of (SRT-A), all that follows is that a possesses two properties both of which confer certain types of conditional powers on any objects that possess them, where it so happens that the power-types conferred by the first property on any object that possesses it form a proper subset of the power-types conferred by the second property on any object that possesses it; in this circumstance, then, a presumably gets certain causal powers conferred twice over, once because it possesses X and then again because it possesses Y. What doesn’t follow from the fact that a’s being Y is realized by a’s being X in the sense of (SRT-A) is that a’s being Y is intimately enough connected to a’s being X for a’s being Y to count as a physical affair, even in the broad sense we’re trying to explicate. It doesn’t follow, for example, that a’s being Y is constituted by a’s being X; nor does it follow that an ascription of Y to a is made true by a’s being X.

Upon reflection, this result is unsurprising. Shoemaker’s original account of realization—realization in the absence of type-type identity—in effect treats the realization relation between properties as a degenerate case of the identity relation between properties: whereas identity between property P and property Q requires identity to hold between between the set of power-types conferred by P and the set conferred by Q, realization of P by Q requires merely the proper-subset relation to hold between the set of power-types conferred by P and the set conferred by Q. An account of realization thus grounded in an account of property identity perhaps shouldn’t have been expected in the first place to yield an account of realization as a relation between property-instances.

So much for the first thing that clause (ii) of (SRT) could mean; let’s turn now to the second:

(B) every token of a conditional causal power conferred on a by Y is identical with some token of a conditional causal power conferred on a by X.

I suspect that (B) expresses how we construe clause (ii) when (SRT) strikes us as an account of realization that a physicalist could use. When (B) is plugged into the Shoemaker-inspired account of realization as a relation between tokens, the result is this:

(SRT-B) a’s being X realizes a’s being Y (where X≠Y) iff

(i) a is X and Y;

(ii) every token of a conditional causal power conferred on a by Y is identical with some token of a conditional causal power conferred on a by X; and

(iii) some token of a conditional causal power conferred on a by X is distinct from every token of a conditional causal power conferred on a by Y.

And (SRT-B) could obviously be used to generate a version of (SP), the Shoemaker-inspired formulation of physicalism.

But the version of (SP) thus generated gives rise to a puzzle. For if a’s being X realizes a’s being Y in the manner described by (SRT-B), then we have a state of affairs in which, for at least one power-token, X and Y both confer it on a; they both confer on a the very same power-token. How come? After all, if X and Y had belonged to distinct objects, then they would have conferred distinct power-tokens; likewise if X and Y had both belonged to a, but at different times. So how come, when X and Y belong to the same object at the same time, they don’t confer two causal power-tokens?

Let me consider three ways of answering this puzzling question. One way is to suppose that, whenever a’s being X realizes a’s being Y in the manner described by (SRT-B), then a’s being X just is a’s being Y; for if this were so, then on this occasion X and Y couldn’t fail to confer the same power-tokens on a. But this way of answering the question faces a couple of serious problems. First, because it makes a mental/physical token-identity claim, it violates the spirit of realization-based formulations of physicalism, which aim to formulate physicalism, as Boyd said explicitly, without holding that apparently non-physical phenomena, whether types or tokens, are identical with physical phenomena.[5] Secondly, and decisively, this way of answering the question has the consequence that if clause (ii) of (SRT-B) is satisfied, then clause (iii) can’t be satisfied. Clause (iii) requires that some token of a conditional causal power conferred on a by X be distinct from every token of a conditional causal power conferred on a by Y. But if a’s being X and a’s being Y are identical, then X and Y must both confer on a the very same set of power-tokens; so there can’t be one conferred by X but not by Y.[6]

Here’s a second possible explanation of why, when a’s being X realizes a’s being Y in the manner described by (SRT-B), X and Y don’t confer two causal power-tokens on a. It appeals to the idea that power-tokens are individuated by exactly three things: the power-types of which they’re tokens, the objects that possess them, and the time at which they’re possessed. Suppose that X confers a power-token on a by belonging to a at t, while Y also confers a power-token (of the same type) on a by belonging to a at t. Then, because there is only one object, a, only one time, t, and only one power-type, the power-token that X confers on a by belonging to a at t must be the power-token that Y confers on a by belonging to a at t.

This second explanation, however, leads to a putative formulation of physicalism that’s insufficient for physicalism as conceived intuitively. To see how, note that, according to the explanation, the individuation of power-tokens doesn’t depend at all on the individuation of the property-instances that confer them; after all, it’s precisely because the identities of power-tokens are alleged to be independent of the identities of the conferring property-instances that two distinct property-instances can manage to confer a single power-token. But if power-tokens, on the one hand, and the property-instances that confer them, on the other, are really as independent of one another as this explanation suggests, then the fact that two distinct property-instances confer on some object the very same causal power seems to imply the holding of no physicalistically-interesting metaphysical relationship between the two property-instances. Suppose that X and Y are distinct properties, physical and mental, respectively, and that a’s being X realizes a’s being Y in the manner described by (SRT-B). Then every power-token conferred on a by Y is identical with a power-token conferred on a by X. But according to the present explanation of how this can be, this identity just reflects the facts that (1) a is both X and Y at the same time and (2) the set of power-types conferred by Y is a proper subset of the set of power-types conferred by X. No further conclusion of interest seems to follow concerning how a’s being X is related to a’s being Y. It doesn’t follow that a’s being Y is somehow constituted by a’s being X; and it doesn’t follow that an ascription of Y to a is made true by a’s being X. As far as I can see, therefore, even if a’s being X does realize a’s being Y in the manner described by (SRT-B), it doesn’t follow on the present explanation that a’s being Y is physical, even in the broad sense we’re trying to explicate.

Finally, let’s consider a third explanation of why, when a’s being X realizes a’s being Y in the manner described by (SRT-B), X and Y don’t confer two causal power-tokens on a. Whereas the second explanation opened up a gap between power-tokens and the property-instances that bestow them, the key idea of the third explanation is to close that gap—indeed, to reduce it to nothing—by identifying property-instances with something like bundles of power-tokens of particular types.[7] Thus, suppose that a exists and has X; necessarily, then, and solely in virtue of possessing X, a possesses a bundle of power-tokens of certain types; but if a subset of those power-tokens forms a bundle of power-tokens the possession of which just is possession of the distinct property Y, then a necessarily has Y too. Moreover, given that a has X—and therefore Y—for the reason just explained, it follows that every power-token conferred on a by Y is identical with a power-token conferred on a by X (though the converse doesn’t follow, and won’t in fact hold if the subset relation is a proper-subset relation). And the reason why a’s being X and a’s being Y don’t confer two power-tokens is that a’s being Y (i.e., a’s possessing a certain bundle of power-tokens) partially constitutes a’s being X (i.e., a’s possessing a certain bundle of power-tokens that includes the first bundle as a proper subset).

The Shoemaker-inspired account of realization as a relation between tokens, (SRT-B), when conjoined with the metaphysical idea that property-instances are something like bundles of power-tokens of particular types, can usefully be regarded as constituting a third Shoemaker-inspired account of realization as a relation between tokens:

(SRT-C) a’s being X realizes a’s being Y (where X≠Y) iff

(i) a is X and Y;

(ii) any instance of X = some bundle of power-tokens of (collective) type T1, and any instance of Y = some bundle of power-tokens of (collective) type T2; and

(iii) the power-tokens in the bundle that = a’s being Y form a proper subset of the power-tokens in the bundle that = a’s being X.

And this account of realization can be plugged into the schema for a Shoemaker-inspired formulation of physicalism, (SP), to yield an actual formulation of physicalism. Such a formulation of physicalism provides an intriguing alternative to the physicalist ways of handling instances of non-physical properties already familiar from the literature. It doesn’t identify each instance of a non-physical property with an instance of a functional, i.e., higher-order, property; nor does it identify each instance of a non-physical property with an instance of a physical property. Instead, it identifies each instance of a non-physical property with a part of an instance of a physical property. Identification of something with a part of a property-instance obviously assumes that property-instances can have parts; but they do have parts if, as this view holds, property-instances are bundles of power-tokens; their parts can be sub-bundles of such bundles. Moreover, this formulation of physicalism would certainly be a rival to realization physicalism, since there are incompatibilities between the two formulations. For example, while realization physicalism treats mental properties as higher-order properties, this formulation treats them as first-order properties; it treats an object’s possession of a mental property as simply its possession of a particular bundle of first-order powers of the right types. One consequence of this difference is that realization physicalism can’t allow an object to possess just mental properties, even in a non-physicalist world; even there, an object with mental properties must also possess realizer properties of some kind. The present formulation of physicalism, however, can allow an object to possess just mental properties, even in a non-physicalist world; the object need only possess the right bundle of causal powers. So how promising a rival to realization physicalism is this formulation? The verdict I’ll indicate is that it’s fairly promising, though there are problems it must address.

This formulation of physicalism is attractive because it seems to entail three things that any formulation of physicalism should entail. First, if all instances of non-physical properties are realized by instances of physical properties in the way described by (SRT-C), then good sense is given to the idea that instances of physical properties (or at least parts of them) constitute instances of non-physical properties. Secondly, this formulation of physicalism also seems to entail that true ascriptions to objects of non-physical properties are made true by those objects’ instantiation of physical properties. Thirdly, this formulation promises to entail an explanation of how the physical properties of an object determine its mental properties, i.e., the supervenience of the mental on the physical. (Neglect ‘wide’ mental states for now.) Thus, (1) an object possesses a given physical property; (2) the instantiation by the object of that physical property just is the object’s possession of a certain set of power-tokens; (3) the possession by any object of a certain subset of those power-tokens just is the possession by that object of a certain mental property; so (4) the object must—metaphysically must—possess that mental property also. Now the details of this explanation need refinement, since surely not just any old subset of a given set of causal power-tokens forms a property-instance (see Shoemaker 2001, 85-86); some further condition must be met. But the proposed explanation of the supervenience of the mental on the physical will only work if this further condition is a purely physical condition (e.g., the holding of physical laws).

On the other side of the ledger, however, are a number of questions that the present formulation of physicalism must plausibly address. Let me conclude by mentioning three such questions.

The first question asks whether the formulation is capable of giving an account of the (broadly) physical character of certain kinds of non-physical properties. The question arises because, on the account of realization, (SRT-C), being assumed, a property-instance is a candidate to be physically realized only if the property-instance can be identified with a bundle of causal power-tokens. But for some properties, it’s implausible to identify their instances with bundles of causal power-tokens, since such properties don’t seem to be purely causal. For example, some properties seem to be such that possession of them requires having not just causal powers but an actual causal history of some particular kind; perhaps being a member of a species and being a mother are such properties. Other properties seem to be such that possession of them requires standing in some actual relation (which needn’t be causal) to some particular external thing or things; perhaps being a genuine US quarter and residing in Bowling Green, Ohio are such properties.[8] Now this first question is surely answerable in principle, if only because a supporter of the present formulation could always retreat to a hybrid formulation of physicalism which incorporates some other notion of realization (e.g., that of realization physicalism) to account for instances of the non-causal (or not completely causal) properties that it can’t handle. But an answer that avoids introducing such a massive disunity into the overall view would obviously be better.

The second question asks what account is to be given of the (broadly) physical character of entities other than property-instances, e.g., of the objects that possess properties. The question must be tackled eventually, or else the proposed formulation will be incomplete. As with the first question, of course, there are options. One might be to supplement the present formulation with the thesis that every object-token is identical with an object-token that is physical in the narrow sense. Another option might be to supplement the present formulation with the thesis that every object-token is such that every causal power-token it possesses is one conferred on it by its possession of a physical property; this second option wouldn’t require a claim of token-token mental-physical identity.

The third question enquires about the plausibility of the metaphysics underlying the present formulation of physicalism. Let me sketch two worries, both familiar, but serious nonetheless. The first worry concerns what causal power-tokens are supposed to be. Obviously they can’t be analyzed into property-instances, since according to the metaphysics underlying the present formulation of physicalism they’re meant to be the parts of property-instances. So, for example, no construal of power-tokens on Humean lines will do, since such a construal would have to identify an object’s power to cause so-and-so with its instantiation of a certain regularity, but objects instantiate regularities in virtue of possessing properties; so either the construal would be circular or it could never be completed. An alternative view of the nature of causal power-tokens would treat them as the sui generis categorical bases for the holding of certain counterfactuals about the objects that have them. But such a view obviously raises pressing questions. Are these sui generis categorical bases supposed to be items to which we’re somehow already committed? If so, how are we committed to them? And if they’re not, then what warrant we do have—or could we have—for positing their existence?

The second metaphysical worry is whether it’s coherent to hold, as the metaphysics underlying the present formulation of physicalism does hold, that all property-instances are bundles of causal power-tokens. After all, causal powers are powers to cause specific things, and powers differ one from another precisely because these specific things differ. But what sorts of things are they? You might plausibly answer: properties. But on the present metaphysics, these are just more causal powers. And that would be fine if there were some way to individuate these new causal powers. But on the present metaphysics, if we appeal again to properties, then exactly the same problem arises again; and we have to appeal again to properties because there’s nothing else to appeal to. So unless we allow the existence of some properties that are not powers, there is nothing to constitute the difference between one power-type and another.

REFERENCES

Boyd, Richard. 1980. “Materialism Without Reductionism: What Physicalism Does Not Entail.” In Readings in the Philosophy of Psychology, vol. 1, ed. Ned Block, 268-305. London: Methuen.

LePore, Ernest, and Barry Loewer. 1989. “More on Making Mind Matter.” Philosophical Topics 17: 175-91.

Melnyk, Andrew. 2003. A Physicalist Manifesto: Thoroughly Modern Materialism. New York, NY: Cambridge University Press.

Shoemaker, Sydney. 2001. “Realization and Mental Causation.” In Physicalism and Its Discontents, eds. Carl Gillett and Barry Loewer, 74-98. New York, NY: Cambridge University Press.

---

[1] And possibly also, to account for perforated tokens, (c) the existence in certain spacetime regions of nothing.

[2] For more on these conditions of adequacy for formulations of physicalism, see Melnyk 2003, Chs. 1 and 2).

[3] For elaboration and defense of this understanding of reduction, see Melnyk 2003, Ch. 3.

[4] Its scope is still restricted to properties, of course.

[5] Shoemaker himself opposes identifying mental property-instances with the physical property-instances that realize them (Shoemaker 2001, 80).

[6] This second problem, by the way, looks like a particular case of the oft-noted general problem of explaining how an instance of property P1 can manage to be identical with an instance of property P2 while P1 and P2 are not themselves identical.

[7] This suggestion is one way of fleshing out Shoemaker’s rather concise discussion (2001, 80-81) of realized property-instances as parts of realizing property-instances.

[8] Shoemaker explicitly allows that non-functional properties can be realized in his sense; but his only example of such properties is that of determinables realized by their determinates, where, to judge by the frequently-mentioned red/scarlet case, he seems happy to treat such properties as causal (see Shoemaker 2001, 74 and 78).