20th WCP: Conceivability Arguments or the Revenge of the Zombies

The feeling of an unbridgeable gulf between consciousness and brain-process: how does it come about that this does not come into the considerations of our ordinary life? This idea of a difference in kind is accompanied by slight giddiness - which occurs when we are performing a piece of logical sleight-of-hand.

(Wittgenstein, Philosophical Investigations, §412)

Today I want to take on the question of what a class of arguments, usually called the conceivability arguments, have to say about the mind-body problem. These arguments have been used to argue for two claims. One is that phenomenal consciousness is not identical, realized by, or supervenient on, physical properties. The other is that there is an explanatory gap between phenomenal and physical levels of description, that doesn't exists with respect to other higher level descriptions, and that may or may not have metaphysical ramifications. My claim is that these arguments do not succeed in establishing these conclusions. It is because (and I take this to be the primary lesson of the conceivability arguments) what they reveal does not have to do with phenomenal consciousness itself, it rather has to do with the nature of phenomenal concepts.

In what follows, I will focus on the most elaborate and sophisticated version of the conceivability argument for dualism. First I provide a general exposition of the structure of conceivability arguments, then I proceed to describe in greater detail Frank Jackson's and David Chalmers' New Conceivability Argument. Finally I construct a reductio that at the same time reveals where the arguments went wrong.

1. INTRODUCTION

Phenomenal consciousness - the what its like feature of experience - can appear to a scientifically inclined philosopher to be deeply mysterious. It is difficult to conceive of how the swirl of atoms in the void, the oscillation of field values, or anything physical can add up to the smells, tastes, feelings, and so forth that constitute our phenomenal experience.

The most important argument for the claim that there is no place for phenomenal consciousness in a completely physical reality relies on considerations of conceivability. The argument, which goes back at least to Descartes, begins with the premiss that we can conceive of any physical or functional facts obtaining without there being any phenomenal experience at all. This is sometimes expressed by saying that zombies (i.e., beings that are our physical and functional duplicates, but possess no phenomenal experiences) are conceivable. From this assertion of conceivability it is inferred that zombies are genuinely possible. And this conclusion is incompatible with physicalism as that doctrine is usually understood.

The claim that zombies are conceivable does not have to do with our powers of imagination, or our psychological constitution in general, but rather with the nature of physical and phenomenal concepts. The relevant notion of conceivability is this:

(Con) A statement S is conceivable, if it is consistent with the totality of conceptual truths, i.e., if -S is not a conceptual truth.

Conceptual truths (or analytic truths) are truths in virtue of meaning. It is usually assumed that if S is conceivable then it is knowable a priori that S is conceivable. That is, that someone who can entertain the thought that S can come to know, at least in principle, on the assumption that he\she is an ideal logician, whether or not S is conceivable without empirical investigation.

To support the premiss that zombies are conceivable, it is claimed that there is no contradiction, detectable a priori, in describing a possible world as being physically exactly like our world, yet containing no experiences. Some philosophers have denied this, offering behaviorist and functionalist analyses of phenomenal concepts; but these analyses are rather implausible. When I think (same I submit for you) I am in pain I am not thinking that I am behaving or disposed to behave in some way; or that I am occupying some particular neurophysiological state or functional state. Of course, this is not to say that the property of being in pain is not a physical or functional property; but rather that the concept pain is not a functional or physical concept. Whatever the ultimate nature of phenomenal experience, when I judge that I am having an experience of a particular sort on the basis of actually having that experience, the concept I invoke is not a behavioral, physical, or functional concept. Rather, it seems to be a concept that I apply directly and spontaneously to the experience. In any case, in the following I want to grant to the proponent of the Conceivability Arguments as much as possible; so I will grant that there is nothing in our concept of consciousness that would allow us to rule out a priori the existence of zombies; zombies are conceivable.

From this conceptual claim, however, it is further argued that the existence of zombies is a genuine metaphysical possibility. This is a powerful result. If it is correct, and if, as I will assume throughout this paper, there are phenomenal facts, then physicalism is false. For it would mean that the totality of physical facts obtaining in our world, including nomological and causal facts, does not necessitate the phenomenal facts that obtain in our world.

But there is an obvious objection. On the face of it, the mere fact that a state of affairs is conceptually possible does not entail that it is metaphysically possible. For example, it is conceptually possible (at least it was before the 18th century) that water is not H2O, but it is not really metaphysically possible for water not to be H2O, since water is H2O, and we know (from Kripke 1972) that identities (where the terms of identity are rigid designators) are necessary. But during the last three decades the relationship between conceptual possibility and metaphysical possibility had been greatly clarified (again, building on Kripke's insights) so as to take these objections into account.

This has lead to a revival of interest in Conceivability Arguments and sophisticated versions of these arguments have been developed by Kripke (1972), Nagel (1974), Robinson (1993), White (1986), Jackson (1993), Chalmers (1996), and others. Like their predecessors, these arguments rely on there being a link between conceivability and metaphysical possibility, but in the formulation of this link they now take into account that conceivability does not always imply possibility. The proponents of these new Conceivability Arguments claim that while the conceivability of water not being H2O fails to imply that it is metaphysically possible for water not to be H2O, the conceivability of a zombie-world does imply that a zombie-world is a genuine possibility.

Now I will look in greater detail at the new Conceivability Arguments due to Frank Jackson (1993) and David Chalmers (1996). While their arguments are my particular focus, my criticisms extend to the other Conceivability Arguments as well, since I will be attacking the link between conceivability and metaphysical possibility which they all presuppose. These arguments are all refutable by a version of the argument I call "the Zombie Refutation". The reason they fail has to do with the very nature of phenomenal concepts that give rise to the conceivability of zombies. Because of the special nature of these concepts, the link between conceivability and possibility that the conceivability arguments rely on turns out to be self-refuting.

2. THE ARGUMENT

Jackson's and Chalmers' arguments are similar. Their definitions of physicalism are almost identical, as is the semantical framework in which they formulate their arguments. Although they employ slightly different formulations of the crucial premiss linking conceivability and possibility, since it can be shown that Chalmers's premiss entails Jackson's, for the present purposes I will attribute the argument I will be setting out to both of them. I will be mainly following Jackson's exposition.

(a) Preliminaries

One caveat. Whereas Chalmers eagerly embraces the dualist conclusion of the argument, Jackson has a more cautious attitude. He himself presents the argument as a challenge for the physicalist, rather than a straight refutation of physicalism. But, on plausible assumptions, shared by Jackson's earlier self (in the Knowledge Argument), it can be easily turned into a refutation. And this is how I will treat it.

In a nutshell, the argument is the following. Physicalism requires that a phenomenal statement, like 'Frank is experiencing a yellow sensation', must, if true, be necessitated by truths expressed in the language of physics. Jackson and Chalmers argue that this necessitation must itself be a priori and that such a priori truths must be grounded in the nature of phenomenal and physical concepts. However, phenomenal concepts do not support such a prioricities. It follows, assuming that there are phenomenal truths, that Physicalism is false. Let us now look at the argument a little more closely.

Physicalism

Jackson observes that physicalism, at a minimum, requires a commitment that

P. Any world which is a minimal physical duplicate of our world is a duplicate simpliciter of our world.

Two worlds are physical duplicates if and only if they agree on all the true statements expressed in the language of physics. A minimal physical duplicate of our world is what we would get if we used the physical nature of our world (including, of course, the laws) as the sole ingredient in making a world; so, a minimal physical duplicate of our world is, by definition, a physicalistic world.

Jackson intends this to capture the idea that there is nothing over and above the physical stuff in our world. He suggests that his formulation of physicalism (P) is equivalent to the claim that every truth about our world, be it physical, chemical, biological, psychological, etc., is necessitated by a statement of physics that gives the full physical description of the world, and is true in all and only the minimal physical duplicates of our world. For the purposes of this paper, I will use the following definition:

(E) For any true statement T,

o(KÉT),

where K is a very long conjunction, expressed in the language of physics, giving the complete physical truth (including truths about the laws of physics) about our world.

Conceptual explanation

According to Jackson and Chalmers, the necessities 'KÉT' (where T is a truth) cannot be brute facts; they need explaining. Jackson observes that if T is, e.g., a psychological statement then analytical functionalism has a story to tell about why the statement is necessary. As he puts it:

....it is the very business of conceptual analysis to explain how matters framed in terms of one set of terms and concepts can make true matters framed in a different set of terms and concepts. (p. 32)

Jackson's view is that in the absence of a conceptual story of how the purely physical makes the psychological true, the entailment would remain an "impenetrable mystery". Both he and Chalmers thinks that the explanation has to be, in an appropriate sense, conceptual. They argue that if physicalism is true then 'KÉT' is not only metaphysically necessary, but is also an a priori conceptual truth; i.e., they argue that if physicalism is true then all truths are a priori derivable from the full physical description of the world. I will call this the

A Priori Entailment Thesis:

if (E) is true, then, for any true T, statements of the form

KÉT

are conceptual truths.

This is the key premiss in Jackson's and Chalmers's argument against physicalism; it provides the crucial link between conceivability and possibility.

Why think that the A Priori Entailment Thesis is true? I am not going to provide here the full motivation for the claim; I simply want to point out that it is immune to the criticism we made earlier with respect to the naive conceivability-possibility principle. Although it is conceivable simpliciter that water is not H2O, its conjunction with the full physical description of the world is not conceivable. Building on Kripke's argument, Jackson observes that, arguably, in all bona fide cases of identity statements where the denial of the identity statement is conceivable (e.g., 'water is not H2O'), there are contingent truths, which, together with conceptual truths involving the terms in question (here, the term 'water' and 'H2O') entail the identity statement.

For example, on the assumption, roughly, that H2O is the unique thing that plays the water-role, the statement that water is not H2O is not conceivable, since it is a conceptual truth that the unique thing that plays the water-role is water. Jackson generalizes this observation and claims that the denial of all true statements, in conjunction with the full fundamental truth about the universe, is inconceivable. His idea is that the full fundamental description of the universe always provides enough background information to fix the reference of any concept in terms of fundamental concepts, and so it is always possible to derive any true statement from it.

(b) The argument

If the A Priori Entailment Thesis is true, the physicalist faces trouble vis a vis fitting psychological, and especially phenomenal, properties into the physical world. The reason is that there are no suitable conceptual analyses of phenomenal concepts for the relevant supervenience claim

KÉx feels pain (or any other statement expressing a phenomenal proposition),

to be a priori.

The derivation of 'KÉWater is H2O' depended on the conceptual truth 'Water is the clear, odorless, etc... liquid'. The availability of such conceptual truths is essential to the kind of derivation we are considering, since the derivation works by finding a contingent statement linking the description to a term of a lower level theory, and ultimately to a term of microphysics. Now consider the statement

KÉx feels pain.

To derive 'x feels pain' a priori from K, there must be some conceptual truth connecting 'pain' with a non-phenomenal description such that satisfaction of the description is a priori sufficient for 'feels pain'. But, arguably, there are no such conceptual truths. For any such non-phenomenal description we can conceive of its being satisfied without anyone feeling pain. 'Pain' is, as Loar (1990) calls it, a direct recognitional concept; we do not apply the term, at least in our own case, on the basis of any evidence, sensory, behavioral, or physical, distinct from what the term picks out, i.e., distinct from the experience itself. 'Pain' refers to pain directly, or rather, via an essential feature of it, say, painfulness. But it follows from the A Priori Entailment Thesis that if 'x feels pain' cannot be derived a priori from K, then, given that 'x feels pain' is true, physicalism is false. To put it more formally:

1) If physicalism is true, then for any true T, statements of the form

KÉT

are conceptual truths.

2) There are some true statements Q to the effect that phenomenal conscious experience occurs (eliminativism about phenomenal experience is false).

3) If Q is a phenomenal statement , then 'KÉQ' is not a conceptual truth.

4) Physicalism is false.

3. ZOMBIES DECEIVED

I now introduce the Zombie Argument. This argument will show that the Conceivability Argument as formulated by Jackson and Chalmers is self-undermining; that is, that with the addition of some plausible further premisses we can derive a contradiction from it. Suppose that Jackson's argument is sound. Its conclusion, that physical facts do not necessitate phenomenal facts, would then be true. And it would follow that there is a possible world which is exactly like our world physically, but in which no phenomenal, or other, non-physical facts obtain.(In fact, it would be a minimal physical duplicate of our world.) Let me emphasize: I make this assumption only for the sake of a reductio. Of course, if physicalism is true, as I think it is, then such a world is impossible. But my strategy is to show that the very assumption that there is such a world undermines the argument that lead to positing the existence of such a world.

In the world we are imagining there exists a zombie-Jackson, physically just like Jackson, but not the subject of any phenomenal states. Professor zombie-Jackson appears to give a series of lectures in zombie-Oxford (as Jackson did in Oxford) arguing for the A Priori Entailment Thesis. What are we to make of his words?

First of all, plausibly zombie-Jackson will have intentional states. When he talks, his words are not mere meaningless sounds. I will argue that it is plausible to assume that zombie-Jackson has intentional states even if he lacks phenomenal states. Moreover, I will argue that it is plausible to assume that zombie-Jackson's intentional states will be identical with Jackson's intentional states except for intentional states that, in Jackson, involve phenomenal concepts. Those of zombie-Jackson's intentional states that, in Jackson, involve phenomenal concepts, will refer to states of affairs present in zombie-Jackson's world. On this view, zombie-Jackson's argument will be just as meaningful as Jackson's, though not quite identical to it. Although the argument is word by word identical to Jackson's argument, some of the words (those that express phenomenal concepts in Jackson's language) have different meanings in Jackson's and zombie-Jackson's mouth. I marked these words with an '+'. 'Pain+', e.g., stands for a term of zombie-Jackson that corresponds to Jackson's term 'pain'. They will use the same words to express different concepts; whereas Jackson's concept is phenomenal, zombie-Jackson's concept, by assumption, is not. We come back to the exact nature of the difference shortly.

Zombie-Jackson's argument will go like this:

1*) If physicalism is true, then for any true T, statements of the form

KÉT

are conceptual truths.

2*) There are some true statements Q+ to the effect that a phenomenal+ state occurs (eliminativism about phenomenal+ states is false).

3*) If Q+ is a phenomenal+ statement , then 'KÉQ+' is not a conceptual truth.

4*) Physicalism is false.

My plan is the following. Given the assumptions I have made, I will argue that if a premiss of Jackson's argument is true, the corresponding premiss formulated by zombie-Jackson will be true as well. We know, however, that the dualist conclusion of zombie-Jackson's argument is false in the zombie-world. Remember, physicalism, on the notion that is relevant here, is not a necessary doctrine; it is true about some worlds and false about others. The zombie-world, by stipulation, is a minimal physical duplicate of our world, so its minimal physical duplicates will be duplicates simpliciter of it. According to our formulation of physicalism, this is what it takes for a world to be physicalistic. Consequently, we know that zombie-Jackson's argument cannot be sound. Since, given that it is meaningful, it is clearly valid, one of its premisses have to be false. It follows then that one of the premisses of Jackson's argument have to be false as well.

To put it in another way, this is a reductio of Jackson's argument. It turns out it is possible to derive a contradiction from Jackson's original premisses, taken together with a few plausible additional assumptions. I will argue that these assumptions are indeed plausible, and that, given these assumptions, premisses 1*-3* follow from Jackson's original premisses. But then the dualist conclusion of zombie-Jackson's argument follows as well, which contradicts the claim, also a consequence of Jackson's original argument, that the zombie-world exists that it is physicalistic.

The fact that one can derive a contradiction from the original argument together with the added premisses, shows that one of the premisses must be false. Since, I argue, my claims about the meaningfulness of zombie-Jackson's argument are extremely plausible, the fault must lie with one of the premisses of zombie-Jackson's argument (and, consequently, with the corresponding premiss in Jackson's argument). While this does not necessarily mean that the dualist conclusion is false, it does mean that the argument used to establish it is not effective.

Let's now formulate these auxiliary assumptions more precisely. I am going to state them briefly right at the start; they will be discussed and defended in detail later after the argument is given. Let me point out here that these assumptions do not by themselves imply physicalism (indeed, a dualist might very well accept them).

Assumption 1

Jackson and zombie-Jackson share most of their intentional states except those involving phenomenal concepts.

Assumption 2

Those concepts of zombie-Jackson that correspond to Jackson's phenomenal concepts, will refer in the zombie to some (physical) state of the zombie.

Assumption 3

A prioricity for thoughts supervenes on the conceptual roles of its constituent (and related) concepts.

Let's see how these assumptions, together with Jackson's premisses 1-3 will suffice to derive zombie-Jackson's premisses. As we said, zombie-Jackson, being Jackson's physical twin, offers an argument that is identical, word for word, to Jackson's argument. On Assumptions 1, Jackson and zombie-Jackson mean the same by their words, except where phenomenal terms are involved. This means that Premiss 1* in the zombie's language expresses the A Priori Entailment Thesis, which, if true, is necessarily true, so if it was true in the actual world, it will be true in the zombie-world as well. On Assumption 2 we get Premiss 2*, i.e., the claim that eliminativism about phenomenal+ properties is false. Given Assumption 3, Premiss 3* of zombie-Jackson's argument,

(3*) If Q+ is a phenomenal+ statement, then 'KÉQ+' is not a conceptual truth,

has as much claim to be true as Premiss 3 in Jackson's argument. While 'KÉQ+' has a different meaning from 'KÉQ', it can be shown that if Premiss 3 is true then Premiss 3* is true as well. Jackson's phenomenal concepts, and zombie-Jackson's 'phenomenal+ concepts have parallel conceptual roles. On Assumption 3, a prioricity, or conceptual necessity supervenes on the conceptual roles of the relevant concepts. That means that if 'KÉQ' is not derivable from conceptual truths then neither is 'KÉQ+' derivable from conceptual truths. Q+, like Q lacks conceptual links to physical, functional and behavioral concepts sufficient to ground the a prioricity of 'KÉQ+'.

I would like now to consider some objections. First, Assumption 1: one might object to it that zombies do not have intentional states at all. Presumably the reason would be that having phenomenal states is essential for having intentional states. In other words, one might object that because zombie-Jackson does not have phenomenal states, he does not really have bona fide intentional states either, and so cannot put forward any argument. The most prominent exposition of this view is due to Searle (1992, Ch. 7); he attempts to establish that consciousness is necessary for intentionality. His argument is based on considerations about the inscrutability of reference originally formulated by Quine (1960, Ch. 2). Searle puts his thesis in the following form:

The notion of an unconscious mental state implies accessibility to consciousness. We have no notion of the unconscious except as that which is potentially conscious. [emphasis in original] (Searle 1992, Ch. 7, p. 152)

Contrary to Searle, a good case can be made that zombie-Jackson does have intentional states. Zombie-Jackson communicates with his colleagues: he answers questions, his utterances convey information, his actions are made intelligible by the assumption that he has beliefs and desires, etc. His cognitive organization seems to be essentially the same as Jackson's. On a more technical note: on all the extant theories of meaning, zombie-Jackson will count as a thinker. On a Davidsonian interpretationist account, zombie-Jackson will have intentional states: he is just as interpretable as Jackson is. Similarly with other theories, like the informational account (e.g., Dretske 1988), the causal-historical account (e.g., Kripke 1972), the counterfactual account (e.g., Fodor 1990), the teleosemantic account (Millikan 1989, Papineau 1993), etc. Zombie-Jackson's brain states (putting the problem of phenomenal vs. phenomenal+ states aside for the moment) carry the same information as Jackson's brain states; they have the same causal history linking them to entities in the world as Jackson's brain states do; the same counterfactuals hold about them as about Jackson's brain states, etc. The only account on which zombie-Jackson will not count as a genuine thinker, is the account on which phenomenal consciousness is essential for intentionality. Although the idea is not absurd, the argument for it do not seem to be very strong, and the contrary assumption seems far more intuitive. Moreover, the proponent of the Conceivability Arguments has to hold that phenomenal consciousness is both non-physical and it is essential for intentionality; but then we are owed an explanation of how causally inert, non-physical properties can play a role in endowing mental symbols with meaning. In any case, since the thesis poses a serious challenge to my argument, I would like to come back to it after I considered some other objections.

One might also object to Assumption 2, i.e., the claim that the zombie's term 'pain+' refers to a (physical) state of the zombie. There are two ways in which Assumption 2 could be false: first, if 'pain+' referred to non-physical phenomenal pain, a property alien to the zombie-word; second, if 'pain+' referred to nothing. Either of these scenarios would be damaging to my argument: if zombie-Jackson's term 'pain+' referred to phenomenal pain, then Premiss 2* would be false, since all phenomenal+ statements would be false in the zombie-world. If, on the hand, 'pain+' didn't refer to anything, then Premiss 2* would be meaningless. Either way, my reductio would not go through. Let's look at these scenarios one by one.

On the first scenario, the term 'pain' and the term 'pain+' have the same meaning. This is not Chalmers's view: he argues that the term 'pain' and the term 'pain+' have different meaning..He claims (Chalmers 1996, pp. 207-208) that, e.g., in spectrum inverted physical twins, phenomenal terms must have different meanings. Since, the twins being physically identical, the difference must be due to acquaintance with different phenomenal properties, the zombie's term has to be different from both since the zombie is not acquainted with any phenomenal properties. This applies equally in the case of pain or any other phenomenal property.

But these considerations aside, we can see that this scenario is very implausible. By assumption, phenomenal properties are alien to his world. It is quite implausible to assume then that when Jackson says 'that feels good', referring to the phenomenal feels produced by a back-rub, zombie-Jackson also refers to a phenomenal feel, even though there is none in his world. Of course, I am not saying one can never have terms that lack actual reference. The term 'winged horse', e.g., has reference. All I am claiming is that, in the particular case of phenomenal+ terms, like the term 'pain+', the reference could not be non-physical qualia.

'Pain+', like 'pain', is a simple term; its reference is not fixed via a description. What could make it the case that it refers to a non-physical property? None of the possibilities one can think of would give the result the objector has in mind. It is unlikely that on an interpretationist account zombies could come out referring with their term 'pain+' to non-physical properties alien to their world, as it would make all their phenomenal+ statements false. There are no suitable causal, counterfactual, or lawful relations between non-physical phenomenal pain and the term 'pain+' either. Since there are no pains in the zombie-world, there could not be any causal relations between pain and the term 'pain+', as such relations would require the existence of laws connecting physical and non-physical entities in the zombie-world; but by stipulation, there are no such laws there. The case is the same with counterfactual relations. Another way for reference to be fixed in some direct, non-descriptional manner is for it to be fixed by a relation of acquaintance. Chalmers (1996, p. 197) claims that, in the case of phenomenal concepts, reference is constituted by acquaintance with the referent, where acquaintance is not to be cashed out in terms of causal, counterfactual, or lawful relations. However, this would not help in the zombie-case: zombie-Jackson is just not acquainted with phenomenal experiences in any sense of the word. As the above options exhaust the existing possibilities, we can conclude that the zombie's simple term 'pain+' couldn't refer to a non-physical property.

The other objection to Assumption 2 was that, even if the zombie has intentional states in general, his term 'pain+' in fact does not refer to anything. In my view, this is wrong. This position has the counter-intuitive consequence that all of the zombie's phenomenal+ talk lacks truth-value. This would be a very uncharitable interpretation of zombies: it would imply that zombies are massively deluded about their mental life. Zombies not only seem to use phenomenal+ terms to give reports about their inner states, they also seem to use them to give explanations of each other's behavior, just like we give explanations of each other's behavior in phenomenal terms. So, for example, zombie-Jackson's friend apparently explains why zombie-Jackson takes an aspirin by referring to his head-ache+. Also, their phenomenal+-utterances and non-phenomenal+ utterances have intelligible connections - they say, e.g., "when I had a tooth-ache+ last time, I went to the dentist", etc. These explanations and reports seem to be accurate, since whenever, e.g., zombie-Jackson says that he is in pain+, he is in a brain state or functional state that is reliably correlated with his term 'pain+' (the same brain state or functional state that is reliably correlated with Jackson's term 'pain'). The natural candidate for the reference of zombie-Jackson's term 'pain+' is this very brain state or functional state. (In fact, Shoemaker (1998) has similarly argued that zombies will refer to a brain or functional state by their phenomenal+ concepts. He uses the point to a different effect, however; he argues for the view that our phenomenal concepts also refer to physical state, since we are physically identical to our zombie-Twins.) This means that whenever Jackson's statement 'I am in pain' is true, zombie-Jackson's statement 'I am in pain+' will be true as well, being about a brain or functional state he is in. The plausibility of this claim might be obscured by the fact that, although zombie-Jackson's statement, e.g., 'I am in pain+' attributes some brain or functional state to himself, of course, he does not conceive of it in this way, i.e., he doesn't think of this state qua brain or functional state. But I think it is clear that the burden is on those who claim that the above facts are not enough to establish that zombie-Jackson's pain+ is a legitimate concept.

Finally, one could object to Assumption 3, that is, the assumption that a prioricity supervenes on conceptual role. Here is my defense of it. If a prioricity did not so supervene then it would be possible that sometimes we cannot tell, even in principle, after a lot of thinking, and doing many thought-experiments, of an a priori truth that it is true. If a prioricity did not supervene on actual and potential inferential relations, then we could not claim any special access to a priori truths; a paradoxical situation. Moreover, this would undermine whatever certainty we have in Premiss 3, i.e., the claim that,

for any true phenomenal statement Q, KÉQ is not a conceptual truth.

Denying Assumption 3 would undermine the Conceivability Arguments by making Premiss 3 highly contentious.

I would like now to return to the objection to Assumption 1, that is, the claim that zombies have intentional states. The objection was that (phenomenal) consciousness is essential to intentionality, and since the zombies, by assumption, do not have phenomenal states, they cannot have thoughts either. Even if it was true that, at least potential, (phenomenal) consciousness was necessary for intentionality, it would not damage my argument. My argument can be run in a way that would make the objection irrelevant.

In fact, zombie-worlds are only introduced for expository convenience. They are not essential to refute Jackson and Chalmers' argument. My argument against them only presupposes that it is conceivable to refer to a brain-state directly. I presume there is nothing incoherent about the idea of referring to a brain-state directly, without the mediation of any physical, functional, or abstract concept, and even without the mediation of a phenomenal feel figuring as mode of presentation or reference fixer. Even on the assumption that phenomenal consciousness is essential to intentionality, this will allow me to construct an analogue of the Zombie Refutation.

One way to do this is to consider a world where there are partial zombies. If Jackson and Chalmers is right that qualia are non-physical, then there is a world that is a physical duplicate of our world, but in which there are creatures that have only some of our phenomenal experiences. These creatures will feel pleasure whenever we do, but will feel no pain at all. Since they do have phenomenal states, and, we might even stipulate, all of their intentional states are accompanied by phenomenal consciousness, there is no reason to deny that they have intentional states. However, on considerations discussed in reply to earlier objections, the most natural thing to say is that they their term 'pain+' refers to a brain state.

There is another way to make the point in a slightly different way. I submit that the following scenario is at least conceivable - and so, on Jackson's and Chalmers' view, possible. Imagine a world where there are creatures in many respects like us. They have the same physical and mental constitution as we have, except that there are some among them that are capable of forming concepts we are not capable of; let us call these people yogis. The yogis are capable of directly detecting certain states of their brains, even though they do not conceive of these states as brain-states. In some ways, these yogi-concepts will work the way our phenomenal concepts work; they are applied to their referents directly, without the mediation of any physical, functional, or abstract concept. What is peculiar to them is that in the case of the yogi-concepts reference is not even mediated by a phenomenal feel.

I think there is nothing inconceivable about this scenario. The yogis will notice that they are capable of detecting some inner state of theirs, even though they do not have any idea how they are doing it. In this they will be somewhat similar to actual blind-sighters; the difference is that while blind-sighters are blindly detecting some feature of their environment, yogis blindly detect some feature of their own brain. For the sake of simplicity, let us suppose that there are two different states of their brains that they can detect, state A and state B, and they use the term 'flurg' and the term 'florg' to refer directly to these states.

Yogis can formulate a variant of the Conceivability Argument. Using the A Priori Entailment Thesis, and the fact that truths involving the yogi-concepts, e.g., 'flurg occurred', are not derivable a priori from the full fundamental (physical, or if dualism is true, physical cum phenomenal) description of their world, they argue that there is a possible world exactly like theirs physically and phenomenally, but where no flurgs occur. But such a world is impossible, since, by stipulation, the term 'flurg' refers to brain state A. The yogi's argument is unsound. But among its premisses the only contentious one is the A Priori Entailment Thesis.

So, even if the objection that phenomenal consciousness is essential for intentionality were sound, it would not succeed in disarming my refutation of the Conceivability Argument. The Zombie Refutation, and its analogues, the partial-Zombie Refutation, and the Yogi Refutation, show that there is something wrong with the Conceivability Argument. It is plausible even on the Zombie Refutation that the premiss that has to be given up is the A Priori Entailment Thesis; but on the Yogi Refutation this conclusion is inevitable. My arguments then not only show that the conceivability arguments are unsound. They also prove that Jackson's and Chalmers's principle linking conceivability and possibility is false, and moreover, they help diagnose where it goes wrong. The Yogi argument has the advantage over the Zombie Refutation of making it even clearer that the conceivability of zombies arise not out of any feature specific to phenomenal consciousness, but rather because of a certain peculiarity of our phenomenal concepts. This peculiarity, that is, referring to a state directly, can plausibly be shared by concepts undisputably referring to physical states, and so with regard to these concepts the A Priori Entailment Thesis is inapplicable.