|Comments on Melnyk’s A Physicalist Manifesto
Joseph Levine, UMass Amherst
Andrew Melnyk’s A Physicalist Manifesto is not only a “thoroughly modern materialism”, as the subtitle advertizes, but is also a thoroughly admirable book. It reminds me a lot of Jeff Poland’s Physicalism, a book to which Melnyk acknowledges a debt, in its clarity, systematicity, and comprehensiveness. One has the feeling after reading it that yes, here is the case for physicalism, and we now know what’s on the table and why we need to take it seriously.
Melnyk defends what he calls “realization physicalism”. This is the doctrine that all objects, properties, and events are either physical or physically realized. What counts as “physical” is of course controversial, and Melnyk has an extremely interesting and provocative line on that, to which I’ll return. His account of realization briefly is this. X realizes Y just in case Y is a token of a functional type that is characterized by some associated condition C (such as playing a causal role of a certain sort, though that’s not the only kind of condition allowed), and X is a token of a type that meets that condition. (There’s more to it, but that’s enough for my purposes.) So a token electronic state of a computer realizes a computational state because the latter is definable by reference to a functional role and the electronic state plays that role.
The basic idea of realization physicalism, then, is this. There is a fundamental level of reality, the physical. Above that are phenomena that are essentially characterizable in terms of roles of various sorts. There is in fact a hierarchy of levels of roles, some realizing, or implementing, others. But in the end, the ultimate role fillers are the fundamental objects, properties, and events of physics.
A couple of points to keep in mind for later discussion. First, though realization, according to Melnyk, isn’t identical to supervenience, it does entail it. That is, if A-phenomena are realized by B-phenomena, then A-phenomena supervene - metaphysically supervene - on B-phenomena. Second, Melnyk does allow for the existence of emergent phenomena, though they only obtain (if they do at all), according to him, within the domain of the physical, the definition of which we will now address. According to Melnyk, “physical” refers to whatever objects, properties, and events are postulated by our best current theories in physics. Melnyk is boldly grabbing one horn of what’s known as “Hempel’s Dilemma”, which is: either by “physical” we mean what’s postulated by current physics or by some ideal future completed physics. If the former, then physicalism fairs no better than current physics, and since the latter is very likely false or incomplete, the former is likely false as well. On the other hand, if we take the other horn and define “physical” by future physics, then we don’t really know what physicalism is saying, since we don’t know what entities might be postulated in the future. Furthermore, it might turn out that what we take to be clearly non-physical objects or properties now are postulated as fundamental in a future physics. If this can happen, physicalism has no substantive content at all.
As I said, Melnyk boldly grabs the first horn of this dilemma. He doesn’t deny that defining “physical” as it occurs in the characterization of realization physicalism in terms of current physics renders the doctrine likely to be false. What he does deny is that its likely falsehood undermines our reasons for adopting it. His argument is that there is no reason to think the doctrine of physicalism should be any more epistemically secure than physics itself, which, after all, is the theory most working physicists adopt despite knowing it’s likely to be shown false or incomplete in the future. Why burden physicalism with a greater threshold for adoption than physics, or any other empirical theory?
As I said above, I find this an interesting, and certainly provocative, move. Other philosophers have sought to distill the essence of physicalism, and insulate it from the inevitable changes in physical theory, by appealing to extensions of current physics that are “modest” in the requisite sense. The idea is that some changes in physics don’t really seem to undermine the spirit of physicalism, whereas others would. So let’s pin our doctrine on those extensions of physics that maintain that spirit. But this is just the problem, as Melnyk notes. How do we characterize the requisite sense of “modest extension”, or the spirit of the doctrine? That just comes to the same thing as saying what we mean by “physical”. Melnyk’s line at least squarely faces up to the problem.
The problem I find with Melnyk’s position is that I think there really is a determinate content to the spirit of physicalism, one that survives most foreseeable modifications in current physical theory. I think it is this general doctrine to which most physicalists owe allegiance, and is one they rightly believe is likely to be true - not just less likely to be false than relevant alternative theories. So if you can’t define “physical” by reference to either current or future physics, for the reasons given above, how does one characterize the doctrine?
My view is that we don’t really need to define “physical” to characterize the doctrine. The idea is this. Physicalism is not best seen as a positive doctrine about a fairly well-defined set of entities to the effect that they constitute the realization base for everything else. After all, as Hempel’s Dilemma makes clear, how could philosophers insist on a doctrine that is so clearly subject to empirical determination as the identity of the fundamental set of objects and properties? It’s not that I think that philosophers can’t have views with empirical import; far from it. Even on my characterization of physicalism, there is clear empirical content to the doctrine. But some empirical issues just don’t seem apt for philosophical dispute, and precisely which entities will ultimately populate the physicists’ pantheon of the fundamental seems to me to be one of them.
What I think is apt for philosophical dispute, however, is the claim that the phenomena in some particularly significant domain, such as the mind, or life, are physically realizable. Of course this doesn’t help if we still need to define “physical”. But the point is we don’t. When it comes to philosophically significant phenomena like the mental and the living, our concern is whether they constitute special, emergent phenomena in their own right, or whether they are “naturalizable”, characterizable as ultimately the same as phenomena not within those domains. So physicalism with respect to the mental and the living is the doctrine that they are realizable by the non-mental and the non-living, whatever they turn out to be. This latter question, what the non-mental and the non-living turn out to be, is not really the issue.
On this interpretation, then, physicalism is not a single doctrine, a claim about how all of reality is organized. Rather, physicalism about the mind is the claim that mental phenomena are not “strongly emergent” (to use Melnyk’s phrase), or, to put it another way, not basic. Physicalism about life is the claim that life is not strongly emergent, or basic. What set of entities is basic, this is for physics to tell us and is not of concern to the philosopher of mind, say.
Now, if it turns out that physicists, for their own reasons, posit minds as fundamental entities, then physicalism about the mind will indeed have been refuted. But two points regarding this must be kept in mind. First, that physics will not in fact take this turn is arguably much more likely to be true than that physics will not add to the current stock of fundamental entities. So it’s not unreasonable to think that physicalism about the mind, or about life, is very likely true. Second, what philosophers of mind - dualists and physicalists - are disputing is whether what we now know about the mind gives us reason to think it isn’t realized in the non-mental. If from another quarter, research in physics, we come to learn that the mental is basic after all, then it seems to me the right attitude is that all bets are off. The original dispute has now been shown to be beside the point. On this view both dualists and physicalists base their positions on the assumption that there do exist clearly non-mental (or non-living) phenomena, and the only question is whether the mental (or the living) is realizable in that.
Now one might claim that my alternative characterization of physicalism does leave something out. Though perhaps my formulation suffices to capture the disputes between physicalists and non-physicalists in particular domains, it fails to capture a significant commitment on the part of physicalists; namely, the unity of nature. That is, one might argue that in addition to the claim that the mind, or the living, is not basic in nature, there is also an important claim that nature itself consists of a hierarchy of levels with the physical at the bottom, and all the others realizable by it. To make this claim, one might argue, one needs a substantive characterization of the physical, one that goes well beyond merely being non-mental and non-living.
But what precisely is meant here by the unity of nature? At one point Melnyk characterizes the relevant sense of unity as follows:
“there is some science, S, distinct from the totality of all the sciences, such that every token...is either a token of a type mentioned as such in the laws and theories of S or an S-ly realized token of some or other functional type” (page 235)
Physicalism is then the conjunction of this principle with the claim that that science is physics. But notice that the relevant sense of unity here is carried by the first part, just quoted - that whatever science is fundamental is less than the totality of all the sciences. If this is what matters, then my version of physicalism captures that too, since it excludes psychology and biology from the science that serves as the realization base for all others. So long as we have some hierarchy, which we do if we claim that the mental and the living are realized by the non-mental and non-living, we seem to have met the unity requirement.
Perhaps though there is a more robust sense of unity that Melnyk and others are after. Here’s one possibility. Suppose there is a fundamental level of objects, such that all other objects are either realized by or constituted by (if there’s a difference here) these fundamental objects. Then let’s say that nature is strongly unified if it turns out that every property of every object above this fundamental level is realized by properties instantiated by objects at the fundamental level. Articulation and defense of this claim would seem to entail a precise characterization of the objects and properties at this fundamental level, and thus, assuming the level in question is the physical, a substantive characterization of what we mean by “physical”.
If one were committed to such a thesis, then indeed one’s physicalism would be much stronger than the doctrine I characterized above. But this can’t be Melnyk’s position, since the position I just described, this strong form of unity in nature, amounts to the denial of any strongly emergent phenomena. As I pointed out above, Melnyk allows the existence of strongly emergent phenomena. That is, he allows that there are non-basic objects that instantiate basic, non-realized properties. He doesn’t see this as inconsistent with physicalism, so long as the objects and properties in question fit his definition of the physical.
But now I lose a grip on what is at stake here. Is it a matter of what the phenomena are called? That is, I see a principled philosophical issue in the question whether that some sciences are realization-reducible to others or they are all on a par, and I see a principled philosophical issue in the question whether phenomena in some particularly significant domain are realized or basic. I also see a principled philosophical issue in the question whether there exist strongly emergent phenomena, in any domain. But beyond that, I don’t see what there is, as philosophers, to dispute. That is, whether the strongly emergent phenomena we allow happen to be called physical, chemical, or even ectoplasmic, why care? If that’s so, then we lose any need to define the physical.
I suppose, as these last remarks suggest, part of my uneasiness with Melnyk’s position is that I see a significant difference between a philosophical thesis and a purely scientific one. It’s crucial to Melnyk’s proposal that we shouldn’t expect a different attitude toward physicalism than we have toward current physics. This really is to take the naturalistic attitude in philosophy quite far. I am pulled in that direction myself. For one thing, philosophy can’t be done in a vacuum; one must constantly be attentive to scientific results. For another, if one buys the Quinean attack on the a priori, it’s hard to see a principled basis for distinguishing philosophical from scientific theses.
Nevertheless, it seems to me one can go too far in the naturalistic direction. There is a difference - though I realize without appealing to a substantive a priori it is hard to characterize it precisely - between a philosophical question and, maybe not a scientific question, but a “merely scientific” question (no condescension intended). For many philosophical purposes, just how the list of fundamental objects and properties in physics turns out just isn’t that significant. But whether physicalism is true, or dualism, or some other metaphysical position, this seems a genuine philosophical question. Thus, I’d like a way of characterizing the question so that it is insulated from the “merely scientific”. I think my way of doing it accomplishes that.
The difference between our attitudes toward physicalism that I’ve been trying to articulate also has consequences for the modal status of the doctrine. On the one hand, Melnyk emphasizes its contingency; he even describes it as “strongly contingent”. Given his understanding of physicalism as very much a scientific hypothesis on a par with physics itself, it makes sense that it shares physical theory’s modal status. On the other hand, Melnyk also emphasizes that the realization relation, as mentioned above, involves metaphysical necessity. The tokening of some physical type that realizes a certain functional type is metaphysically sufficient for the latter’s tokening. So you can’t have two physically identical worlds that differ with respect to any of their objects, properties, or events that are physically realized.
Now, one might wonder how physicalism could be contingent if the physical base metaphysically necessitates all that is realized by it. In fact, this is an issue on my way of construing physicalism. But for Melnyk it’s straightforward. One way physicalism can fail to hold in a possible world is if it contains tokens that are neither functional nor physical. Another way is if some of its functional tokens are realized by non-physical tokens, as would be the case if they were realized by the famous “ectoplasm” (Melnyk’s example). That physicalism is a posteriori on Melnyk’s view of course goes without saying.
However, on my way of understanding physicalism, it’s harder to see how to render it contingent. (What’s more, and more controversial I think, I am not so clear even about it’s a posteriori status, but I’ll come back to that.) I share with Melnyk the idea that realization is a metaphysically necessary relation. So let’s consider the two ways a world could turn out to be non-physicalist just presented.
First, suppose some world contains tokens that are neither functional nor physical. Well, what are these tokens? Can they be minds or living things? If physicalism is true of our world, both minds and living things are functional. If they are functional in this world, then I don’t see how they can be non-functional in any other world. After all, this is an identity thesis (a functional identity thesis), and, following Kripke, identities are necessary. (Melnyk clearly agrees with at least this much.) But if these non-functional non-physical tokens aren’t minds or living things, then they are just more stuff: stuff that isn’t captured by our current (or maybe even future) physics, but not the sort of stuff to refute physicalism as I’ve interpreted it. Hence this doesn’t count as a world in which physicalism is false.
Okay, so we need a world in which mental or living tokens are realized non-physically. This was the second way a world could be non-physicalist mentioned above. So what about having a mind that’s realized by ectoplasm? Wouldn’t that do it? Well, unfortunately, not for me. The problem is that, unlike Melnyk, I place no weight on a positive conception of the physical. So aside from various associations with the name, what is supposed to be so non-physical about ectoplasm anyway? If it’s merely that it isn’t part of current physical theory, well we know that’s not going to do it. I think what people have in mind by ectoplasm is a substance that instantiates either mental or living properties in a basic, non-realized way. If there were that kind of stuff in a world, then indeed it would violate physicalism. But the problem is that if mental and living properties are identical to functional properties (similarly for objects), then there is no possible world in which ectoplasm of that sort exists. Therefore, on my way of construing physicalism, as opposed to Melnyk’s, it turns out not to be contingent after all. That is, like other a posteriori necessities, it’s necessary if it’s true at all.
This brings us to another topic, the a posteriori character of physicalism. While I’m certainly sympathetic to Melnyk’s claim that physicalism is an a posteriori doctrine, I think the issue is complicated, and gets entangled with other issues that bear on the ultimate acceptability of the doctrine. Perhaps the best way into the issues I want to bring up is to start with Melnyk’s discussion of the question whether physicalist realizationism is “retentive” or not.
Physicalism is retentive, for Melnyk, if it successfully identifies special science or folk science types with functional types that are physically realized. The idea is that since physicalism claims that only what is physical or physically realized exists (except for abstract objects and the like), unless one can show that minds, living things, ordinary macro objects, and their respective states and properties are identical to physically realized functional tokens, one has to give up belief in their existence. This would be to go non-retentive, or eliminativist. Melnyk adopts a two-pronged strategy here. On the one hand he argues that it is at least quite plausible that physicalism is retentive, at least in the cases we care about. On the other hand, he argues that even if it isn’t, that isn’t such a bad consequence.
Melnyk executes the first strategy by looking at three cases: Fodor’s identification of representation with the relation of nomic covariation subject to the asymmetric dependence condition, the identification of life with a collection of certain capacities, and the identification of genes with certain functional units that determine phenotypes given certain environments. In all of these cases he makes the very plausible case that the relevant entities with which the special science/folk science types are identified can be construed as physically realizable functional types. What’s more, he claims, we now know of no principled reason these identifications can’t be right.
On the other hand, if for some reason life, say, can’t be identified with a collection of functionalizable capacities, so we have to go non-retentive about life, this isn’t really so bad. After all, we aren’t thereby saying either that living things simply don’t exist, or that there aren’t any interesting differences between paradigmatic examples of living things, like us, dogs, and trees, and paradigmatic examples of non-living things, like rocks. It’s just that these differences consist in the presence or absence of physically realizable functional properties, and not some further property that is allegedly something over and above these properties. It’s this further type that is claimed not to exist, not the objects that were thought to be tokens of it.
There are two interconnected issues that arise from this discussion: the alleged a posteriori status of physicalism and its vulnerability to objections based on conceivability considerations. Though answering conceivability arguments against physicalism is not a major focus of Melnyk’s book, he does address the issue, and has written specifically on that topic elsewhere. The basic line, one he shares with many adherents of physicalism, is that the conceivability of a non-physical and non-physically realized special or folk scientific entity is not a threat to physicalism so long as it’s not really, metaphysically possible, and there is no inferential route that takes one from conceivability to possibility.
Defense of the claim that there is an “iron wall” between the conceivable and the possible rests on the premise that physicalism is an a posteriori thesis. In particular, the identities of special and folk science types with functional types, on which the metaphysical impossibility of their being non-physically-realized rests, are a posteriori. Thus the fact that one can imagine life, say, being non-physically-realized, is merely a reflection of the fact that its identity with some functional type is not knowable a priori.
So let’s look again at these identities. As mentioned above, Melnyk provides three examples to make it plausible that important special and folk scientific types are indeed identical to functional types, and thus physically realizable. In the context of this discussion he doesn’t say much about whether the identities in question are a posteriori, relying on earlier arguments to the effect that they are. But if you look at the actual examples, it’s not obvious that they are a posteriori. Take Fodor’s account of the representation relation. What sort of empirical evidence is it based on? One might say this, and this does seem to be the general form of the answer Melnyk gives to this question: we find lots of cases where the representation relation and also the asymmetrically dependent nomic covariation (ADNC) relation are coinstantiated, and so therefore we infer, empirically, that they are the same thing. But the cases are not observed instances of coinstantiation; rather, they are thought experiments. We describe situations in which we judge that there is a representation relation and we find that it fits the ADNC relation. We also lay down conceptually derived constraints, like the one that where there is representation there must be the possibility of misrepresentation, and determine whether ADNC satisfies it. I find it hard to see how a constraint like that is the result of an empirical discovery.
I think the same goes for the identification of life with a certain syndrome of functionally characterizable capacities. Again, one might say that this is purely a matter of identifying a kind through pointing to various paradigmatic examples and then discovering empirically what the underlying nature is, as with the case of water and H2O. Of course it’s true that when we point to paradigmatic living things, such as dogs, cats, us, and trees, we do find that they manifest these functionally characterizable capacities. But this isn’t really enough to show that their doing so is what their being alive consists in. If one resists the identification of life with these capacities despite the biological discovery that all these creatures manifest them, one presses the point by asking what more one wants for something to be called “alive”. That is, we convince the recalcitrant party that this really does answer to what she had in mind all along. If she continues to resist, what empirical discovery is going to convince her?
Lest I be mistaken for an Australian here, let me qualify my remarks. I am very sympathetic to the Quinean attack on the a priori, so I don’t want to claim that these identities are clear cases of a priori knowledge (if they are true at all, of course). But I do think that they are philosophical theories, and, as with the doctrine of physicalism itself, I don’t think we can just assimilate philosophical doctrines to what I above (and again, non-pejoratively) dubbed “merely scientific” theories. I don’t know how to properly categorize this difference, but I don’t therefore want to ignore it.
One way perhaps to capture the epistemic status of these identities - one I’m not really happy with in the end - will serve to bring us back to the conceivability objection. We might say this about the identification of life with the relevant functionally characterizable capacities. We start with a folk conception of what it is to be alive, which includes a list of paradigmatic exemplars. We discover what capacities these creatures have, how they are realized, and then this exhausts what is empirically discoverable about the creatures we consider to be alive. A further question now arises, essentially a semantic question: does our term, or concept, “being alive”, refer to that syndrome of capacities or not? Now, consulting what we would say in various circumstances, the method of thought experiments, helps to determine an answer to this question. Notice, one needn’t claim that this involves access to a priori knowledge, strictly speaking, since we can take the semantic thesis itself to be an empirical question - not about biology but about our psychology - and our intuitions about cases to be the relevant empirical data. The model here would be the way linguists use acceptability intuitions as data for syntactic theories.
Suppose we accept this dual model for how we come to accept such identities (I imagine this is really applicable only to folk scientific types, but these of course are the cases that interest us most). Then Melnyk’s two-pronged strategy for defending physicalism against conceivability objections (though he doesn’t quite put it this way) fits well. He can say to the recalcitrant anti-functionalist about life that they face a dilemma. Either they can see that the set of functionally characterizable capacities answers to their concept of what it is to be alive, or they can insist that there is some further property, something over and above, as we say, that something must have in order to count as alive. If they insist on the latter, though maybe they are right about the concept of life they possess (though again, I think people can be wrong about that too), it’s now an empirical question whether anything in the world, including the alleged paradigmatic exemplars of life, possess this extra feature.
In other words, the non-retentive, eliminativist prong of the strategy comes into play at this point. Melnyk can plausibly claim that the very things that are claimed to be alive according to the recalcitrant anti-functionalist have been empirically discovered not to possess any feature that is not either physical or physically realized. Hence, if one wants to insist that being alive, as one conceives it, is not identifiable with a functional type, one must be prepared to find out that, on this conception of what it is to be alive, nothing in our world actually is alive. So in the end physicalism does get vindicated empirically.
It might seem, then, that whether or not one considers the epistemic status of the relevant functional identity claims to be purely a posteriori, there is a purely empirical route to justifying physicalism after all. Rather than insisting there is an iron wall between conceivability and possibility, one can admit that conceivability considerations might show that there is a distinction in properties to be made here, but then one treats this other property, the one found to be non-functional, non-retentively. That is, one claims either that there just is no such property - that one’s conception fails to refer at all - or that the relevant property isn’t instantiated in our world. I’m not sure whether Melnyk would be happy with this way of looking at it (I’d bet good money he wouldn’t), but it’s certainly not unfriendly to his project in the end.
Well, not unfriendly for most cases, like life. I think one can legitimately say to the recalcitrant anti-functionalist about life that either she really doesn’t have a conception of anything over and above the relevant functional capacities, or, if she does, what she has in mind doesn’t really exist. But what about the case of the mind? Here I think things get trickier. Suppose an anti-materialist about the mind argues as follows. Fodor’s theory of representation can’t work, and the reason is that representation just isn’t a functionalizable relation. For some state or token R to represent X is just for it to be about X, nothing more, nothing less. This is a basic, non-analyzable notion, and it certainly isn’t captured by any nonsense about asymmetric dependence.
How does the Melnyk two-pronged strategy work in this case? First, the materialist would attempt to convince the recalcitrant anti-functionalist that her notion of representation does in fact allow for reconstruction in functional terms, perhaps via a nomic covariation account like Fodor’s. The materialist could point out how paradigmatic examples of representation seem to fit the account, how crucial features like the possibility of misrepresentation are preserved, and the like. Perhaps that would work.
But suppose it didn’t. Suppose she stuck to her guns. Now we try the alternative, non-retentive, or eliminativist ploy. We say, this representation relation you insist on, the one that isn’t functionally characterizable, this may well be what you have in mind, but it isn’t instantiated in our world. Don’t worry, representations, or the paradigmatic things you called “representations” - like natural language strings and thoughts - they still exist. It’s just that they don’t have this feature you thought they had, namely, being about something, where this is understood in a non-functional way. In other words, we treat her just like the recalcitrant anti-functionalist about life.
In the life case, where the recalcitrant anti-functionalist insisted that by “life” she meant some “vital force” type thing over and above any list of functionalizable capacities, it seems quite right to say that no such thing exists. But of course one doesn’t want to say, at least not nearly so easily as in the life case, that no such thing as representation exists. Our thoughts really are about something. While I might be prepared, under some circumstances, to give that up, to go eliminativist, I don’t think saving materialism is one of those circumstances. The situation in the representation case then just isn’t the same as the life case. Strategy 2 isn’t nearly as plausible.
Of course one might say this on behalf of the materialist, and against the recalcitrant anti-functionalist about representation. First, they should have succumbed to strategy 1. That is, we might feel warranted in claiming that our functionally characterized relation really does capture everything we want out of the representation relation. Then, if the opponent claims nevertheless that there is something to representation over and above what is functionalizable, it might not be too bad to insist that that relation, the one she insists is non-functional, doesn’t exist (i.e. isn’t instantiated) in our world. What was supposed to make this implausible was the denial that our thoughts were really about anything. But the materialist can say, no they are about something, at least in my sense, but I don’t find this inconsistent with the relation of being about something being itself a functional relation. If you insist it isn’t, then whatever it is you’re talking about isn’t found in our world.
Again, in the end nothing terribly unfriendly to Melnyk’s project, the defense of physicalism, has emerged. But notice how the “conversation”, as it were, has shifted. A lot of it now is about how we conceive of the target property or relation, what it is we have in mind, what needs capturing by a functional characterization. The purely a posteriori character of the endeavor, the defense of physicalism, has certainly been sullied with some a prioristic impurities. My point is that I don’t see how we can really avoid this.
But now I reach a point that really is unfriendly to Melnyk’s defense of physicalism, a point I’m sure he suspected I’d reach eventually. Suppose the mental feature in question is not representation but consciousness: there being something it’s like for me to occupy various mental states. Again, I don’t see how a strictly a posteriori method is going to totally convince me that this feature of my mental life is identical to some functional property or relation. Certainly mere coinstantiation, even universal coinstantiation, isn’t going to do it. There are of course theoretical reasons to go along with that, largely surrounding the issue of the causal relevance of consciousness. But if the recalcitrant anti-functionalist about consciousness insists that despite all these reasons, to be conscious just isn’t the same thing as being in a functional state, how does the two-pronged strategy play out here?
Of course we can try to convince her she’s wrong, and that very feature she has in mind is really a functional property after all. But now how are we going to do this? If we go the route of what Chalmers calls “type A” physicalism, then we indulge in some a prioristic reasoning and analysis. This of course is not what Melnyk has in mind. Rather, the idea is to say that we find out empirically that these states we pick out as our paradigmatic conscious experiences are, it turns out, identical to functional states. Again, I think some recalcitrant anti-functionalist about consciousness - might even be me this time - can say that we have something more substantive in mind than just whatever it is those states turn out to be, ostending the paradigmatic conscious experiences. There is a property they have, and it’s just unclear how to understand the claim that it really is identical to a functional property.
Here’s where push comes to shove. Either we acknowledge that this recalcitrant anti-functionalist really does have something in mind that isn’t functionalizable, or say she (or if it’s me, he) is confused. If not the latter, then we move on to the second, non-retentive, eliminativist prong of the strategy. But can we buy it this time? I could buy it about life, easy. I can even see buying it about representation, given how we spun the eliminativism above. But I really can’t buy it about consciousness. My conscious experience just does seem present to me, subjectively, in a way that doesn’t seem amenable to elimination in the way these other properties/phenomena have been.
In the end, then, I think the success of Melnyk’s enterprise, a thoroughgoing defense of physicalism, depends to a large extent - at least in the case that matters most to me - on the extent to which subjectively accessible phenomena can be incorporated into the physicalist framework. Unlike Melnyk, I don’t see this enterprise as sufficiently a posteriori in character so as to be really comparable to other, “merely scientific” enterprises, though I admit I have no general account of philosophy’s epistemic status. This difference, concerning the epistemic status of the doctrine of physicalism, yields another difference, concerning how it should be characterized, as discussed above. Finally, partly also owing to the difference concerning epistemic status, I am less optimistic about physicalism than Melnyk is.
I’ll end on one ironic note. Given Melnyk’s formulation of physicalism, it turns out he thinks the doctrine he’s defending is less likely to be true than I think the doctrine I’m attacking is. Go figure.