Draft – please do not quote
According to the zombie conceivability argument, phenomenal zombies are conceivable, and hence possible, and hence physicalism is false. Critics of the conceivability argument have responded by denying either that zombies are conceivable or that they are possible. Much of the controversy hinges on how to establish and understand what is conceivable, what is possible, and the link between the two—matters that are at least as obscure and controversial as whether consciousness is physical. Because of this, the debate over physicalism is unlikely to be resolved by thinking about zombies—or at least, zombies as discussed by philosophers to date.
In this paper, I explore an alternative strategy against the zombie conceivability argument. I accept the possibility of zombies and ask whether that possibility is accessible (in the sense of ‘accessible’ used in possible world semantics) to our world. It turns out that the question of whether zombie worlds are accessible to our world is equivalent to the question of whether physicalism is true. By assuming that zombie worlds are accessible to our world, supporters of the zombie conceivability argument beg the question against physicalists. I will then consider what happens if a supporter of the zombie conceivability argument should insist that zombie worlds are accessible to our world. I will argue that the same ingredients used in the zombie conceivability argument—whatever they might be—can be used to construct an argument to the opposite conclusion. If that is correct, we reach a stalemate between physicalism and property dualism: while the possibility of some zombies entails property dualism, the possibility of other creatures entails physicalism. Since these two possibilities are inconsistent, one of them is not genuine. To resolve this stalemate, we need more than thought experiments.
In this paper, I look at some conceivability arguments from a new vantage point. My objection differs from other, more standard objections to conceivability arguments because unlike other objections, I don’t attempt to deny that the allegedly conceivable things are conceivable or possible. Instead, I argue that even if the premises of conceivability arguments are accepted, their conclusion doesn’t follow, because it remains to be shown that the envisaged possibilities are accessible (in the sense of ‘accessible’ used in possible world semantics) to our world. By assuming without argument that the possibilities in question are accessible to our world, this type of accessibility argument begs the question.
Although I will focus on the zombie conceivability argument, the most important observations and morals of this paper generalize to conceivability arguments that are analogous to the zombie conceivability argument in relevant respects.
Physicalism is the view that everything is physical. We will be mostly concerned with phenomenal physicalism, according to which phenomenal consciousness is physical. The following is a popular and useful explication of physicalism in terms of supervenience:
Physicalism: every property supervenes on physical properties
Phenomenal physicalism: phenomenal properties supervene on physical properties
In other words, physical truths entail (fix, determine) all truths, including phenomenal truths. Two entities cannot differ without differing physically. This formulation is still imprecise, but it will remain our working formulation of physicalism. Later, I will discuss one attempt at making physicalism more precise and its shortcomings.
What is relevant here is primarily the aspect of physicalism that pertains to phenomenal properties. Roughly speaking, these are properties instantiated by entities in virtue of their possessing phenomenal states (or possessing phenomenal consciousness, or being conscious). According to this aspect of physicalism—phenomenal physicalism—phenomenal properties supervene on physical properties.
In the recent philosophical literature, zombies are usually defined as creatures physically identical to us that lack phenomenal consciousness. The zombie conceivability argument goes as follows (cf. Chalmers 2002, Chalmers 1996, p. 122):
(1) Zombies are conceivable
(2) Therefore, zombies are metaphysically possible
(3) Therefore, the physical facts do not entail the phenomenal facts
According to the zombie conceivability argument, zombies are conceivable, and hence possible, and hence physicalism is false. For zombies are physically identical to us and yet lack consciousness. Hence, if they are possible in an appropriate sense, what they lack—consciousness—is not, and cannot be, determined by the physical facts.
The zombie conceivability argument entails a form of property dualism about consciousness, according to which phenomenal facts do not supervene on physical facts but are either further primitive facts or facts that supervene on further non-physical facts, such as laws that connect the physical and phenomenal facts.
Critics of the conceivability argument have responded in two ways. Some deny that zombies are conceivable. Others concede that zombies are conceivable but deny that they are possible. But the zombie conceivability argument has been surprisingly resilient in the face of criticism. Much of the controversy hinges on how to establish and understand what is conceivable, what is possible, and the link between the two—matters that are at least as obscure and controversial as whether consciousness is physical. Because of this, the debate over physicalism is unlikely to be resolved by thinking about zombies—or at least, zombies as discussed by philosophers to date.
In this paper, I explore an alternative strategy against the zombie conceivability argument. I will accept the possibility of phenomenal zombies and ask whether that possibility is accessible to our world. As it turns out, the question of whether zombie worlds are accessible to our world is equivalent to the question of whether physicalism is true. By assuming that zombie worlds are accessible to our world, supporters of the zombie conceivability argument beg the question of physicalism. But suppose that zombie-philes should insist, with or without argument, that zombie worlds are accessible to our world. I will argue that, when relevant matters are properly clarified, the same ingredients used in the zombie conceivability argument—whatever, exactly, they might be—may be used to construct an argument to the opposite conclusion. If that is correct, we reach a stalemate between physicalism and property dualism: while the possibility of some zombies entails property dualism, the possibility of other creatures entails physicalism. Since these two possibilities are inconsistent, one of them is not genuine. To resolve this stalemate, we need more than thought experiments.
Accessibility between Worlds
The step from (2) to (3) is generally taken to be uncontroversial, even by physicalists. But we can gain insights into both physicalism and conceivability arguments from questioning it. So although I am sympathetic to many of the concerns that lead people to resist the first premise and the first step, I will accept them for the sake of the argument and offer reasons against the second step.
To evaluate the step from (2) to (3), it is convenient to consider interpretations of (2) and (3) in terms of possible world semantics. This is consistent with the way the debate is framed in the literature.
To avoid begging questions, it is important to be clear on how possible world semantics works. To evaluate a modal statement at a world, we look at which statements are true at the set of possible worlds that are accessible to that world. For instance, “P is possible” is true at world w iff there is a possible world w’, accessible to w, where “P” is true. “P is necessary” is true at world w iff “P” is true at all possible worlds accessible to w.
How do you determine which worlds are accessible to which? Accessibility between worlds is a device invented by philosophers to evaluate modal claims (see Kripke 1963a, b; Hintikka 1963). Hence, to determine whether one world w is accessible to another world w’, we need to determine which facts, including which modal facts, obtain at w’. To suppose otherwise is to put the cart of semantic machinery before the horse of semantic truth.
To illustrate, consider the case of physical necessity. To evaluate physical modalities, we cannot consider all possible worlds. For there are possible worlds, e.g., worlds according to Newtonian physics, in which things occur that are impossible at our world. As a consequence, we need to restrict the range of possible worlds that are accessible, i.e., relevant to evaluating physical modalities at a world. The two most popular ways of doing this are to restrict accessibility either to worlds where the same physical laws obtain or to worlds where the same physical truths obtain (Hermes 2004).
By the same token, just as the relevant physically possible worlds are those where the same relevant physical truths obtain, the possible worlds relevant to the truth of physicalism are those where the same relevant truths obtain that obtain at our world.
With these preliminaries in place, we can go back to physicalism and conceivability arguments. We are asking whether (3) is true at our world. In terms of possible world semantics, physicalism (i.e., the denial of (3)) may be interpreted as follows: in any possible world that is accessible to the actual world for the purpose of evaluating the relevant modality, the physical facts materially imply the phenomenal facts. For that to be true, none of the relevant worlds can be such as to be phenomenally different from our world while being physically the same.
But (2) may be interpreted to say that there are possible worlds that are physically the same as our world while being phenomenally different. Therefore, if any of those worlds are accessible to our world for the purpose of evaluating the relevant modality, then (3) follows from (2), and physicalism is false. The question is, are any of the zombie worlds described in (2) accessible to the actual world for the purpose of evaluating the relevant modality? As far as I can tell, this question has received very little explicit attention.
The obvious problem here is that, having granted (1) and (2) for the sake of the argument, the question of whether zombie worlds are accessible to the actual world is equivalent to the question of whether physicalism is true. In other words, to simply assume that zombie worlds are accessible to the actual world begs the question of physicalism.
A way around the accessibility problem would be to assume that the correct logical system for metaphysical modality is S5, where all worlds are accessible to one another, and that physicalism is a claim of the relevant metaphysical type (to be assessed within S5). If these two assumptions are granted, then the step from (2) to (3) becomes valid. Many philosophers agree that S5 is the correct logic for metaphysical modality, but some have argued against it (e.g., Salmon 1989).
Now a zombie-phile may reply that she takes the correct logic for metaphysical necessity to be S5, so that all possible worlds are accessible to all others. This, of course, requires argument, which as far as I know no one has given. (Notice that there has been almost no discussion of Salmon’s argument in the literature.) If zombie-philes and other supporters of conceivability arguments want their argument to go through, they may want to consider replying to Salmon’s arguments.
But this is only the easy step in addressing the problem. For even if S5 is the correct logic for metaphysical modality, it remains to be shown that physicalism is a claim of the relevant (unrestricted) metaphysical type. This, I submit, is going to be a tall order.
Why shouldn’t physicalism be a claim of the relevant (unrestricted) metaphysical type? After all, the claim is that all properties supervene on physical properties as a matter of entailment. Why would any possible world be irrelevant? If all possible worlds were relevant, this would answer both the question of the relevant modality and that of the relevant logical system in a way that lets the zombie conceivability argument go through.
But the problem with this option is well known. If all worlds were relevant, then physicalism would turn into an unrestricted necessity. If true, physicalism would hold in all possible worlds. But physicalists are not that ambitious. They accept that physicalism might have been false. For instance, Cartesian dualism might have been true instead. So physicalism needs to be formulated as a contingent thesis. That is, there need to be possible worlds in which physicalism fails.
standard way that physicalism has been formulated as a contingent thesis can be
traced to a paper by Terence Horgan (1982). In honor of its originator, I will call this
formulation Horganic physicalism, but I will adopt
Horganic physicalism: any possible world that is a minimal physical duplicate of our world is a duplicate simpliciter of our world.
Horganic phenomenal physicalism: any possible world that is a minimal physical duplicate of our world is a phenomenal duplicate of our world.
A minimal physical duplicate of a world is a world in which all the same fundamental physical truths obtain but no extra, non-physical substances or principles obtain—same microphysics, no new spooky stuff (Jackson 1998, p. 13). A duplicate simpliciter of a world is a world in which all facts, including all higher-level and higher-order facts entailed by the fundamental facts obtain.
An important motivation for Horganic physicalism is to allow for the possibility of worlds where physicalism fails. For instance, Horganic physicalism allows for the possibility of Cartesian worlds, where Cartesian dualism obtains, because those worlds are not minimal physical duplicates of our world and hence, according to Horganic physicalism, their possibility does not refute physicalism. So, physicalism turns out to be contingent as desired.
Although Horganic physicalism restricts accessibility to some worlds where physicalism is false, it does not prohibit access to zombie worlds. Since some zombie worlds are by definition minimal physical duplicates of our world, Horganic physicalism counts them as relevant to establishing the truth of physicalism. Thus, if Horganic physicalism correctly captures the content of physicalism, the step from (2) to (3) is valid.
For (2) says that zombies are metaphysically possible. Hence, given that zombies are physical duplicates of us and given the standard possible world semantics for modal claims, it follows that there are possible worlds that are physical duplicates of our world but contain zombies, and hence are not phenomenal duplicates of our world. Ergo, Horganic phenomenal physicalism fails, and Horganic physicalism tout court fails with it.
So here we have an explanation of the fact that the accessibility of zombie worlds has received so little attention: it is buried under the carpet of Horganic physicalism, and Horganic physicalism is widely accepted as an adequate formulation of physicalism. But the wide acceptance of Horganic physicalism as a formulation is a historical accident, which should not prevent us from carrying our investigation further.
At this point, while we are still granting (1) and (2), we face two options: either we accept Horganic physicalism as a good formulation of physicalism and hence reject physicalism, or we keep the question of physicalism open by giving up Horganic physicalism as a formulation of physicalism. So far as I know, the second option has not been explored explicitly. This is surprising, for there is an obvious reason against Horganic physicalism as a correct formulation of physicalism: Horganic physicalism is refuted by the zombie conceivability argument.
Notice that Horganic physicalism, when suitably generalized to apply to arbitrary worlds (this is how Lewis 1983 formulates it), counts zombie worlds as worlds in which physicalism is true. For zombie worlds certainly have no minimal physical duplicates that are not also phenomenal duplicates. But this is clearly against the spirit of physicalism. Physicalism is intended to include phenomenal physicalism, and phenomenal physicalism is false at zombie worlds, at least in its intended interpretation, according to which the phenomenal facts of our world are entailed by the physical facts of our world. Hence, we should revise Horganic physicalism.
The Contingency of Phenomenal Physicalism
There is an independent reason to amend Horganic physicalism: it doesn’t fully capture the contingency of physicalism. To capture the contingency of physicalism, as we have seen, a good formulation of physicalism must allow for possible worlds in which physicalism fails. Horganic physicalism is explicitly designed to allow for some possible worlds in which physicalism fails. For instance, it allows for worlds that contain disembodied minds in addition to everything that is contained in our world. But it rules out other worlds where physicalism fails, and for no principled reason.
The problem might become clear if we take a step back. As a first pass at a formulation of physicalism in terms of supervenience, Horgan, Lewis, and Jackson considered the following:
Generic supervenience physicalism. Any two worlds that are physical duplicates are duplicates simpliciter.
Generic phenomenal supervenience physicalism. Any two worlds that are physical duplicates are phenomenal duplicates.
The natural objection to this formulation is that it turns physicalism into a necessary thesis. As Lewis puts it, however, “two worlds could indeed differ without differing physically, if at least one of them is a world where Materialism [i.e., physicalism] is false” (Lewis 1983, p. 35). But the same objection applies to Horganic physicalism: a world w with phenomenal properties and its zombie duplicate w’ could indeed differ without differing physically, if at least one of them is a world where physicalism is false. But which of the two? The moral usually drawn is that either w’ is impossible or physicalism fails at w (the world with phenomenal properties). But it is equally legitimate to conclude that physicalism fails at w’ (the zombie world). If physicalism fails at w’, then there is no reason to deem it impossible (as Horganic physicalism does). All a physicalist needs to do is let w’ be possible while deeming it just as inaccessible to our world as a Cartesian world.
To put the point another way, if phenomenal physicalism is contingent, allowing for the possibility of, say, Cartesian worlds, it should be contingent all around. It should also allow for the possibility of property dualist worlds. But by ruling out zombie worlds, Horganic phenomenal physicalism (suitably generalized to apply to any worlds, as Lewis 1983 formulates it) rules out the possibility of property dualist worlds. For a property dualist world is one that has zombie duplicates. By ruling out zombie duplicates, Horganic physicalism rules out property dualist worlds. But there is no principled reason why physicalism should be consistent with the possibility of Cartesian worlds but not with the possibility of property dualist worlds, and hence of zombie worlds. All of those refute a generic formulation of physicalism if they are accessible, so a consistent physicalist should deny access to all of them.
Perhaps not everyone will be immediately convinced that Horganic physicalism turns property dualism into a necessary falsehood. But given Horganic physicalism, whichever argument the physicalist employs to rule out zombie worlds as impossible can be generalized to apply to any world where consciousness is associated with physical states, including any property dualist world.
Consider a possible world w in which consciousness is associated with physical states (e.g., a putatively parallelist or epiphenomenal world). Since the phenomenal facts at w are not entailed by the physical facts, there should be a physical duplicate of w where the physical-phenomenal laws are different and because of that, where there are no phenomenal states. So relative to w, there are zombie worlds. But we may imagine w to be very much like our world. In fact, since we don’t actually know the physical facts that (according to the physicalist) entail the phenomenal facts or how the entailment goes, we can imagine w to be arbitrarily close to our world. We have to suppose that w is different from our world in those physical respects that entail consciousness; otherwise, presumably, w’s physical duplicates will have to have phenomenal properties too. But since we don’t know what those physical respects are, we can vary the physical facts in w arbitrarily, covering all physical facts that we know. In the end, we seem to be forced to admit that we can imagine a zombie duplicate of our world. But that would refute physicalism, so a physicalist must deny that there are any zombie worlds that are (minimal) physical duplicates of w, and hence, conclude that w is actually impossible.
Let’s try to work in the opposite direction. According to the physicalist, there is no zombie duplicate of our world, for any physical duplicate will share the phenomenal facts of our world. So any world w where there are phenomenal facts and whose physical facts are sufficiently like our physical facts to entail the phenomenal facts will also lack zombie duplicates. What about worlds where the physical facts are sufficiently different from ours that no consciousness can be physically produced? Perhaps they can have consciousness by virtue of extra physical-phenomenal laws, and be property dualist worlds, and hence have zombie duplicates? Or perhaps their physics, though different, is enough to entail the phenomenal facts in that world by virtue of analogues of whatever physical principles entail the phenomenal facts in our world. If the physical facts entail the phenomenal facts in our world, surely it is equally possible that they entail the phenomenal facts in those other worlds. But then those worlds must be physicalistic too, and they cannot have zombie duplicates. For otherwise, if those worlds have zombie duplicates, there is no reason why ours shouldn’t. After all, the physical facts are different, but we cannot point at any relevant difference, let alone explain why it’s relevant. Hence, under Horganic physicalism, the physicalist is committed to denying the existence of any zombie duplicates of worlds where consciousness is associated with physical states. That is, the physicalist is committed to the necessary falsehood of property dualism, i.e., the necessary falsehood of views such as psycho-physical parallelism and epiphenomenalism. But that, of course, is more than physicalists can hope to defend, and it’s not what physicalists should ever want to claim. So Horganic physicalism needs to be amended.
The flip side of this observation is that given Horganic physicalism, the zombie conceivability argument proves too much. For the zombie conceivability argument, if sound, would prove that phenomenal physicalism is a necessary falsehood. The zombie conceivability argument is a perfectly general argument; it holds at all worlds (except for zombie worlds, which according to Horganic physicalism conjoined with the possibility of zombies, are the only physicalistic worlds; what irony). Consider an arbitrary world w. Suppose (hypothesis for the reductio) that physicalism holds at w (hence there are no non-physical things and properties there to distract us). Now consider a physical duplicate of that world where there are no phenomenal facts. By the logic of the zombie conceivability argument, the second world exists and shows that physicalism is false at w. Hence, if the zombie conceivability argument is sound and Horganic physicalism is a correct formulation of physicalism, phenomenal physicalism is a necessary falsehood. But phenomenal physicalism strikes me as no more of a necessity than physicalism in general. I think any sane physicalism would understand it in the same way.
The moral is simple. If we want phenomenal physicalism to be contingent, as we should, we need to reject Horganic physicalism as a formulation of physicalism. Horganic physicalism needs to be supplemented with a further restriction on which worlds are accessible, so as to rule out zombie worlds as irrelevant.
Hence, not all possible worlds, not even all possible worlds that are minimal physical duplicates of our world, are relevant to determining whether physicalism is true. The relevant ones are those that are accessible from the actual world. And the worlds that are accessible from world w are those that are possible given the facts at world w. If the facts at world w include Cartesian or property dualism, then zombie worlds are accessible to w. But if the facts at w include physicalism, then zombie worlds are inaccessible to w, regardless of how great a degree of possibility they otherwise enjoy.
Conceiving against Dualism
At this point, a committed property dualist might reply that phenomenal physicalism is different from physicalism tout court. Whereas physicalism tout court is contingent, phenomenal physicalism is necessary. At most, phenomenal physicalism is true in zombie worlds, though of course, it’s only vacuously true for lack of phenomenal properties. And in light of the zombie conceivability argument, physicalism is necessarily false (or vacuously true only at zombie worlds). As a consequence, according to this line, the denial of physicalism—property dualism—is necessarily true (or at least true at all worlds where phenomenal properties are instantiated).
Of course it would be desirable to have an independent argument for the necessity of phenomenal physicalism and property dualism—an argument not relying on the soundness of conceivability arguments. But while we wait for such an argument, we can still show that no such argument is going to save the zombie conceivability argument.
If property dualism is necessary, it is true iff it is true at all worlds where phenomenal properties are instantiated. It cannot have counterexamples. It must be understood along the following lines. According to property dualism, the phenomenal facts are not fixed by the physical facts. Perhaps they are primitive, or perhaps they are fixed by some other facts, such as laws correlating the physical to the phenomenal facts. Now call whatever non-physical facts fix the phenomenal facts according to property dualism basic facts. Here is how property dualism needs to be formulated (analogously to Horganic physicalism):
Property dualism: any possible world that is a phenomenal duplicate of our world contains some (non-physical) basic facts.
Given this formulation, it is easy to show that, by the logic of the zombie conceivability argument, property dualism is no less vulnerable to refutation by conceivability arguments than physicalism is. This is instructive because it allows us to neutralize the zombie conceivability argument by its own logic, without appealing to any special considerations involving concepts, modality, or reference. Instead of attempting to determine who is right about those complicated matters, we can construct a conceivability argument that, while following the same logic of the zombie conceivability argument, reaches the opposite conclusion:
(1*) Z’s are conceivable
(2*) Therefore, Z’s are metaphysically possible
(3*) Therefore, the physical facts entail the phenomenal facts
The conclusion of the argument is the opposite to that of the zombie conceivability argument, but the logic is the same. This is possible because Z’s are different from ordinary zombies in relevant ways. By definition, Z’s are such that the phenomenal facts about them supervene on the physical facts. Thus, their possibility is incompatible with property dualism as defined above. The step from conceivability to metaphysical possibility is also the same as in the zombie conceivability argument. What remains to be shown is that Z’s are conceivable, or at least no less conceivable than ordinary philosophical zombies.
A good example of Z’s is human beings as the physicalist conceives of them. According to the physicalist, the phenomenal facts about us are entailed by the physical facts. In fact, according to most physicalists, the physical facts explain the phenomenal facts—a condition stronger than entailment, for the physical facts might entail the phenomenal facts without there being any way to explain the latter in terms of the former. But to make the Z conceivability argument go through, we need not assume that physicalists are right. All we need is physicalism to be conceivable and hence possible in the same way that the dualist claims zombies to be conceivable and hence possible. In other words, all we need to assume is that there is no incoherence hidden in the physicalist position.
The result of running the two arguments in parallel is paradoxical. For now we have two conceivability arguments that start with premises that are intuitively plausible, follow the same logic, but reach mutually inconsistent conclusions.
To avoid this paradox, a defender of property dualism cannot consistently reject the logic of the Z conceivability argument, on pain of undermining the zombie conceivability argument, which is the justification for her view. The only way out is for the property dualist to reject (1*). But on what grounds? (1*) is supported by the same kind of a priori intuitions that support (1). The supporter of the zombie conceivability argument maintains that the only legitimate ground for rejecting (1) would be that it contains a hidden contradiction, and the burden to find that contradiction lies on the physicalist. So now, in the face of the Z conceivability argument, the consistent dualist must maintain that (1*) contains a hidden contradiction. But the burden is now on the dualist to find the contradiction hidden in (1*).
Notice that it’s not open to the dualist to claim that (1*) is contradictory because it leads to (3*), which is inconsistent with (3). The grounds for (3) is precisely (1), namely, a premise with no more a priori ground than (1*) itself. Hence, the dualist cannot use the intuitive plausibility of (1) to rule out (1*) any more than the physicalist can use the intuitive plausibility of (1*) to rule out (1).
The dialectical situation may be described as follows. Supporters of the two conceivability arguments are in epistemic divergence, which means their disagreement cannot be resolved by the methods they employed in reaching their conclusions. Epistemic divergence can be defined more precisely as follows. A method of investigation is public iff it can be applied to answer the same questions by all investigators, and when it is so applied, different investigators obtain the same answer. For instance, ordinary scientific observational and experimental methods are public in this sense, because different investigators can replicate each other’s results by following the same methods. By contrast, receiving divine revelations, introspecting one’s mental states, and reading tea leaves are not public methods. The reason is either that different investigators cannot apply the same methods (e.g., divine revelations occur only sporadically and unpredictably, to say the least), or different investigators cannot apply the same methods to the same questions (e.g., we can only introspect our own mental states), or different investigators that apply the same methods to the same questions obtain incompatible answers (e.g., different people looking at the same tea leaves make different predictions). Epistemic divergence occurs when people answer questions by means of methods that, for one reason or another, are not public. And since their methods are not public, they cannot use those same methods to resolve their disagreement.
In our case, the
question is whether physicalism or property dualism is true, and the method
being applied is the method of possible cases.
This method consists of consulting one’s a priori intuitions about what
is conceivable and hence possible (
The advantage of the present argumentative strategy is that it need not discuss in any detail notions of conceivability, possibility, and whether the former is a guide to the latter. As a consequence, it need not locate the flaw of conceivability arguments in any particular place. This is an advantage because the relevant notions of conceivability, possibility, and inference from the former to the latter are at least as contentious as the notion of zombie and the possibility of physicalism.
The present strategy is to show that conceivability arguments lead to either inconsistency or epistemic divergence, neither one of which is desirable. Unless we wish to get involved in an endless clash of intuitions about which possibilities should be taken more seriously, we should conclude that this kind of conceivability argument is inconclusive. If we are going to understand the nature of consciousness, we need something over and above a priori intuitions.
Balog, K. (1999). "Conceivability, Possibility, and the Mind-Body Problem." The Philosophical Review 108(4).
Bealer, G. 1994. Mental properties. Journal of Philosophy 91:185-208.
Bigelow, J. and R. Pargetter (1991). Science and Necessity.
Campbell, K. K. (1970). Body and Mind. Doubleday.
Chalmers, D. J. (1995). "Facing up to the Problem of Consciousness." Journal of Consciousness Studies 2.
Chalmers, D. J. (1996). The Conscious Mind: In Search of a Fundamental
Chalmers, D. J. (2002). Does Conceivability Entail Possibility? Conceivability
and Possibility. T. Szabó Gendler and J. Hawthorne.
Chalmers, D. J. (2003). Consciousness and Its Place in Nature. The Blackwell Guide to the Philosophy of Mind. S. P. Stich and T. A. Warfield. Madden, MA, Blackwell: 102-142.
Chalmers, D. J. (2004). "Imagination, Indexicality, and Intensions." Philosophy and Phenomenological Research 58(1): 182-190.
Garson, J. (2003). "Modal Logic." The Stanford Encyclopedia of Philosophy (Summer 2005 Edition). E. N. Zalta, Ed. URL = <http://plato.stanford.edu/archives/sum2005/entries/logic-modal/>.
Hermes, C. M. (2004). "Two Concepts of Nomic Accessibility." Southwest Philosophy Review 20(2): 87-94.
Hintikka, J. (1963). "The Modes of Modality." Acta Philosophica Fennica 16: 65-79.
Horgan, T. (1982). "Supervenience and Microphysics." Pacific Philosophical Quarterly 63: 29-43.
Jackson, F. (1998). From Metaphysics to Ethics: A Defence of
Kim, J. (2005). Physicalism,
Or Something Near Enough. Princeton,
Kirk, R. 1974. Zombies vs materialists. Proceedings of the Aristotelian Society (Supplementary Volume) 48:135-52.
Kripke, S. A. (1963a).
"Semantical Analysis of Modal Logic I:
Kripke, S. A. (1963b). "Semantical Considerations on Modal Logic." Acta Philosophica Fennica 16: 83-94.
Kripke, S. A. 1980. Naming and Necessity.
Levine, J. (2001). Purple Haze: The Puzzle of Consciousness.
Lewis, D. K. (1983). "New Work for a Theory of Universals." Australasian Journal of Philosophy 61: 343-377.
Linsky, L., Ed. (1971). Reference
Loux, M. J., Ed. (1979). The
Possible and the Actual.
Lynch, M. P. (2004). "Zombies and the Case of the Phenomenal Pickpocket." Synthese.
Marcus, E. (2004). "Why Zombies are Inconceivable." Australasian Journal of Philosophy 82(3): 477-490.
McDermott, M. (1999). "Counterfactuals and Access Points." Mind 108(430): 291-334.
McGinn, C. (2004). Consciousness and Its Objects.
Melnyk, A. (2001). Physicalism Unfalsified: Chalmers's Inconclusive
Conceivability Argument. Physicalism and Its Discontents. C. Gillett and
Mitchell, S. D. (2003). Biological Complexity and Integrative
Papineau, D. (2002). Thinking about Consciousness.
Polger, T. W. (2000). Zombies Explained. Dennett's Philosophy: A
Comprehensive Assessment. D. Ross, A. Brook and D. Thompson.
Rescher, N. and R. B.
Brandom (1979). The Logic of Inconsistency: A Study in Non-Standard
Possible-World Semantics and Ontology.
Rowlands, M. (2001). The Nature of Consciousness.
Salmon, N. (1989). "The Logic of What Might Have Been." The Philosophical Review 98(1): 3-34.
Searle, J. R. (1992). The
Rediscovery of the Mind.
Stalnaker, R. (2002). What is it Like to be a Zombie? Conceivability
and Possibility. T. Szabó Gendler and J. Hawthorne.
Stoljar, D. (2001). "The Conceivability Argument and Two Conceptions of the Physical." Philosophical Perspectives 13: 393-413.
Stoljar, D. (2005). "Physicalism and Phenomenal Concepts." Mind and Language 20(5): 469-494.
Stoljar, D. (2006).
"Zombies and Actors." Content and Modality: Themes from the
Philosophy of Robert Stalnaker. A. Byrne and J. Jarvis Thomson, Eds.
Tennant, N. and F. Jackson (1994). "Logic and Physicalism." Philosophy and the Cognitive Sciences: Proceedings of the 16th International Wittgenstein Symposium. R. Casati, B. Smith and G. White, Eds. Vienna, Hölder-Pichler-Tempsky: 113-126.
Worley, S. (2003). "Conceivability, Possibility, and Physicalism." Analysis 63(1): 15-23.
Yablo, S. (2002). Coulda, Woulda, Shoulda. Conceivability and
Possibility. T. Szabó Gendler and J. Hawthorne.
 So long as “physical” can be explicated in a way that does not trivialize physicalism, my paper does not depend on the details of how “physical” is explicated. I will assume that there is a suitable explication of “physical.”
 I took this formulation of physicalism from Chalmers 1996 and Jackson 1998, who offer extended defenses of its adequacy. Chalmers and Jackson go on to offer a more precise formulation along the lines of Horgan 1982 and Lewis 1983, but I will postpone discussion of that formulation until later. Keeping with the more generic formulation is consistent with what other writers do (e.g., Stoljar 2001, 2005, 2006). Thus formulated, physicalism is weak enough to encompass both reductive and non-reductive forms of physicalism. For instance, according to functionalism about consciousness, facts about consciousness supervene on physical facts even though the nature of consciousness is not physical but functional. Zombies are also discussed in debates specific to functionalism, where they are sometimes defined as creatures behaviorally or functionally identical to us who lack phenomenal consciousness (Searle 1992, Polger 2000). For present purposes, since functional properties are generally assumed to be realized by physical properties, nothing will hinge on the contentious distinction between physical and functional properties. This notion of physicalism, either reductive or non-reductive, is intended to be in opposition to property dualism, not pluralism (of the kind defended, for instance, by Mitchell 2003).
 People who run conceivability arguments (with or without zombies) include Bealer 1994, Campbell 1970, Chalmers 1995, 1996, 2002, 2003, 2004, Kirk 1974, and Kripke 1980.
 E.g., Melnyk 2001, Worley 2003, Marcus 2004, Kim 2005.
 E.g., Balog 1999, Yablo 2002, Stalnaker 2002, Lynch 2004, Levine 2001, Rowlands 2001, McGinn 2004, Papineau 2002.
 I am working with the notion of accessibility that is endorsed by all commentators on the issue of accessibility that I have come across. (1) Kripke 1963a, b does not even use the expression “H2 is accessible to H1” but rather “H2 is ‘possible relative to’ H1,” which he explains by saying that “every proposition true in H2 is possible in H1” (1963a, b, p. 64 in Linsky 1971). (2) Hintikka uses the term “alternative world” and explains it thus: “whatever is possible must be true in some alternative world and ... whatever is necessary must be true in all the alternative worlds ... only these genuine alternatives really count” (Hintikka 1963, p. 67 in Loux 1979). (3) Hughes and Cresswell say the accessibility of a world is a matter of what is conceivable. This may seem to help the zombie-phile, until we notice that what is conceivable is not unrestricted: it must be restricted by the way the actual world is. So even under this view, the question of whether zombie worlds are accessible to the actual world reduces to the question of whether physicalism is true. (4) Salmon (1989, cf. also Garson 2003) argues that accessibility is a matter of what is possible at a world. Again, the question of accessibility of zombie worlds reduces to the question of physicalism. (5) Bigelow and Pargetter (1990) say it would be desirable if what is accessible to a world would depend on the intrinsic features of that world (cf. McDermott 1999); once again, the question of accessibility reduces to the question of whether physicalism is true at our world.
 For instance, Nathan Salmon has argued that S4 (and hence S5) contains an axiom schema some of whose instances are false (Salmon 1989). The axiom schema in question says that necessary propositions are necessarily necessary, or “ðfÉððf”. Salmon’s counterexample involves the composition of material artifacts. According to a widely shared intuition, any particular artifact could have originated from matter that is slightly different from its actual original matter; for instance, a particular statue s could have originated from a lump of bronze m* slightly different from the lump m, which it actually originated from. But the same artifact could not have originated from matter that is too different from its actual original matter; a particular statue could not have originated from a lump of bronze m’ that is sufficiently different from m while remaining the same statue. If this is accepted, it follows that for some matter m’ sufficiently different from m but sufficiently overlapping with m*, s could have originated from m* but could not have originated from m’; but at the same time, if s had originated from m*, then it would have been possible for s to originate from m’. Hence, although it is necessary that s did not originate from m’, it is not necessarily necessary that s did not originate from m’. This example violates the axiom schema mentioned above. Hence, according to Salmon, S4 (and S5 with it) is fallacious. Salmon believes that the correct logic for metaphysical necessity is probably T (ibid., p. 4).
If critics of S5 are right, we cannot assume that zombie worlds are accessible to the actual world even if we hold the question of physicalism to be a matter of straightforward metaphysical necessity. Rather, we need to discuss which possible worlds are accessible to the actual world. What is the reason to think that the zombie worlds are accessible? The answer is important, for as we have seen, even if we assume that physicalism is a metaphysical claim of the relevant type, determining whether zombie worlds are accessible is equivalent to determining whether physicalism is true.
 Following Horgan, a similar formulation was endorsed by David Lewis (1983), and then by Frank Jackson (1993, 1994, 1998) and David Chalmers (1996, p. 42).
At this point,
Another way to put the point
is by saying that whether the entailment holds at all worlds depends on what
kind of entailment we are appealing to.
Different notions of entailment range over more or less comprehensive
sets of possible worlds. If
 It may seem that rejecting Horganic physicalism as a good formulation of physicalism on these grounds begs the question against the anti-physicalist. But from the present perspective, things are really the other way around. Horganic physicalism begs the question of which zombie worlds are accessible to which worlds in favor of the anti-physicalist. That is, by accepting Horganic physicalism, the physicalist took up a huge burden of proof, which makes the task of the dualist easier by comparison. If we want to redress the balance, it is better to accept the possibility of zombie worlds and ask whether they are accessible to our world. Hence, Horganic physicalism should be amended.
At any rate, Horganic physicalism was a tentative formulation, which may be superseded by better formulations. If we find out that it was flawed—and being refuted is quite a flaw—this is reason enough to look for a better formulation.
 For instance, Horgan and Philip Pettit told me they think phenomenal physicalism is a contingent truth (personal communication)
 The tension between Horganic physicalism and the zombie conceivability argument was already noted by Neil Tennant (1994). Tennant correctly pointed out that another option open to the Horganic physicalist is to conclude that we are zombies. Tennant took this to be a reductio of the view that Horganic physicalism is a correct formulation of physicalism, and so do I.
 This qualification is especially germane in light of the dispute, related to the debate about conceivability arguments, about kinds of possibility and necessity. For if, as some allege (Jackson 1998, Chalmers 1996), there is no important distinction to be drawn between logical and metaphysical necessity, and hence if the metaphysical possibility of a world may be established on purely logical grounds, then it becomes very important to distinguish between those possible worlds that are metaphysically accessible to one world from those that aren’t. For if physicalism and other metaphysical doctrines are to remain contingent, as they should, the set of possible worlds that are relevant to establishing their truth at a given world must be appropriately restricted.
 In personal correspondence, David Chalmers told me he professes such a view.
 Those considerations are the bread and butter of the current debate about conceivability arguments. The problem is, it’s difficult to reach consensus on any of the relevant notions.
 For a more explicit definition and detailed discussion of method publicity and epistemic divergence, see The Author 2003.