Are Zombies Logically Possible?

And Why It Matters

A philosophical zombie is a being physically indistinguishable from an actual or possible human being, inhabiting a possible world where the physical laws are identical to the laws of the actual world, but which completely lacks consciousness. For zombies, all is dark within, and hence they are, at the most fundamental level, utterly different from us. But, given their definition, this singular fact has no direct implications about the kind of motion, or other physical processes, the zombie will undergo within its own world. Under quite standard physicalist assumptions, such as certain assumptions about the 'initial conditions' of the zombie's world and that of the causal closure of the physical¹, a zombie's behaviour, as well as its underlying physical state, should be indistinguishable from the behaviour and physical state of a genuine human being.

The first case envisaged above is that of the beloved 'zombie duplicate' -- in particular, my philosophical zombie is one which is physically indistinguishable from me. This is the case most usually invoked in discussion, since it can be granted that I, at least, definitely do possess consciousness. But if we grant that human beings in general are conscious beings then the second case will serve our philosophical thought-experimental purposes just as well, while avoiding the (perhaps ultimately irrelevant) complications that perfect physical duplication might involve. For example, it is arguable that the only possible world that could have a perfect physical duplicate of me would have to be totally physically identical to this world and hence might be the very same possible world. This problem could arise independently of any concerns we might have that consciousness is somehow a relational property, via an assumption of the complete causal inter-connectedness of the physical realm.

It is, of course, far from clear that any two worlds that are physically indistinguishable are thereby the same world, but assuming otherwise at this stage seems to come close to begging one side of the question to be explored here. However, I don't think this worry is really very plausible, even if we accept physicalism, since it appears quite possible for there to be physical things which are completely causally isolated from each other. Many modern cosmological models allow for this, or even demand it. If so, possible worlds could differ physically in ways that have no effect on the physical states of certain parts of them, and hence there could well be a physically different world with a physically indistinguishable duplicate of me in it.

Before considering the mere logical possibility of philosophical zombies, I want to digress briefly on the matter of their actuality or nomological possibility. A real zombie would be an actual being who is either physically identical to some human being, or is physically identical to some genuinely possible human being but who is utterly lacking in consciousness. To assert that zombies are nomologically possible would be to assert that in some world that shares all of its laws with the actual world there is a being identical to some actual or genuinely possible human being who is utterly lacking in consciousness. Of course, the existence of a real zombie would entail that zombies are nomologically as well as logically possible, but the reverse entailments do not hold.

The robust sense of reality so necessarily lacking when discussing the logical possibility of zombies, should instantly reassert itself if we ask, even while, for the moment, granting the possibility of zombies, whether there are, or are likely to be, or ever have been, any real or even just nomologically possible zombies. Clearly, the question of whether there are any philosophical zombies actually lurking among us is a form of the venerable problem of other minds. I take it that this question deserves the same kind of answer as other distinctively philosophically skeptical questions, such as whether the world might have been created five minutes ago, or whether there is any 'external' world at all.

Note an important difference here between skeptical questions like those of other minds and external reality and what I take to be quite non-skeptical, though distinctively philosophical, worries, such as the problem of the freedom of the will. The problem of freedom depends upon a tension between commonsense, intuition and certain interpretations of what we know (or think we know) about the laws of nature. This tension is such that the intuitive appeal of belief in the existence of free will seems to be in prima facie conflict with scientific knowledge, and the familiar arguments against the existence of freedom exploit this tension in various ways. The skeptical hypotheses are not like this. They conflict equally with all of intuition, commonsense and what we know about the laws of nature. Thus trying to defend seriously either the actual existence or the nomological possibility of zombies would require denying those laws of nature which seem to link physical states to states of consciousness. Of course, we don't know very much about these laws, but it is already abundantly clear that there are any number of quite particular regularities between neural systems, states and processes and varieties of conscious experience, and many of these are already being scouted out by our rapidly developing neurosciences (for just one striking example see Tong et.al. 1998).

So the worry that there are real philosophical zombies somewhere hereabouts, or that they are even nomologically possible is a kind of skeptical worry. How, in general, should we respond to specifically skeptical challenges? It would be great if we could show that the challenge was incoherent, but that's a rare treat. More typically, it seems evident that the skeptical worry is a worry at all just because it is prima facie coherent, that is, it seems to be at least logically possible. For example, I don't think there is much doubt that the hypothesis that, for example, the universe was created five minutes ago in precisely the state it was in five minutes ago is coherent, and that is tantamount to saying that it is logically possible that the world was created five minutes ago. Similarly, it seems fairly obvious that it is logically possible that I am in the matrix right now, or being deceived by the evil genius about the very existence of an external physical world.

But I don't believe that any of these hypotheses are true; in fact I regard them as spectacularly unlikely, and in violation of certain physical laws which I take to hold in the actual world. You probably agree with me on this. Furthermore, I think that I know that the world was not created five minutes ago, despite the fact that I concede that a 'radically young' universe is logically possible. Philosophically speaking, I think that any epistemological theory which led to the conclusion that I did not know that the world was very old solely on the basis of the logical possibility of the opposite and utterly independent of whether or not it is true that the world is very old, would be a deeply flawed account of knowledge (that's not to say such flawed account have never been offered).

Roughly speaking -- though philosophy has shown there are many subtle niceties to the subject -- a proper epistemology should endorse a kind of 'conservatism of belief', which can be pugently expressed via the mechanic's cliché 'if it ain't broke, don't fix it'. My current belief that the universe is quite ancient serves me well. It fits in with many other of my beliefs. It does not lead me astray and is amazingly well able to integrate with new evidence that I pick up every day talking with others (say about what happened yesterday), reading the newspaper or browsing the web science pages. This belief has passed all the epistemic tests it needs to count as knowledge² (though it may begin to fail these tests at any time). One might complain that just fitting in with a general system of belief and evidence cannot be enough for knowledge. Quite so. What is needed in addition is simply the truth of the belief. But, since the world is in fact very old, I do know that fact.

We are in exactly the same sort of epistemic position with regard to the question of real or even just nomologically possible zombies. I don't believe there are any. Nobody can give me, nor is there any, evidence that it is zombies that surround me rather than fully conscious human beings. So my belief is well supported, stable and unassailable. I am under no epistemic pressure to change or even examine this belief and can remain secure in my knowledge that there are no philosophical zombies and that in fact the laws of nature which link brain states to states of consciousness rule out zombies as nomologically impossible.

But this knowledge does not show that philosophical zombies are logically impossible. It is much harder to show any such a thing. Nonetheless, some philosophers contend that philosophical zombies are logically impossible, and this paper will focus on one argument recently advanced by Robert Kirk (1999). It is claimed that there simply are none of the possible worlds invoked in the zombie definition. No being physically identical to me, in the appropriate sense, could lack consciousness³. This is a strong claim. It is far stronger, for example, than the claim that zombies are nomologically impossible, for this only asserts that in any possible world that shares all its 'natural laws' with our world any physical duplicate of me will be as conscious as I am. The claim that zombies are nomologically impossible is also distinct from the claim that they are physically impossible. Distinguishing nomological and physical possibility assumes that the realm of natural law might (logically might) outrun that of mere physical law. Notice that if in fact all natural laws are physical laws or logically supervene upon physical laws then there are no worlds that agree on our physical laws but differ in some natural law. Then, since there are evidently laws which link physical states and states of consciousness, the logical possibility and the nomological possibility of zombies would come to the same thing, for then any world that differed in its natural laws from our world would also differ in some physical law. But, as we shall see, there is very little reason to collapse this distinction⁴.

On the other hand, if the laws which link physical with non-physical properties are 'natural' without being, or logically supervening upon, physical laws then zombies could arise by breaking these natural laws without breaking the physical laws. If mental properties are non-physical, then it seems quite reasonable to claim that laws linking the mental and physical are themselves not physical laws (in just the way it would be reasonable to deny that laws of economics are physical laws). The issue would then come down to whether or not such natural, but non-physical laws, logically supervene upon the physical laws. But, obviously, simply to assume that all natural laws logically supervene upon physical laws would beg the question against the logical possibility of philosophical zombies. Or, in other words, it would suffice to show that zombies are logically impossible to show that all natural laws logically supervene upon the actual physical laws⁵. But that looks to be extremely hard to show. If any kind of 'non-physical world' is logically possible and if it is logically possible for such a realm to enjoy its own set of natural laws then the supervenience claim would obviously be in jeopardy. I have no idea how one could even begin to argue that such lawful but non-physical realms are logically impossible, and am equally at a loss see how the laws of such realms, if their possibility is granted, would have to be logically supervenient upon the laws of our physical realm. For one thing, since such realms would agree on all their physical laws (trivially) the supervenience claim entails that they would have to agree on all their laws -- so at most one such realm would be possible. There would still be many different possible non-physical worlds since they could presumably differ in their 'initial conditions'. But they would have to share their laws. One might claim that such realms could differ in their 'substance' while agreeing in their 'laws' but I really have no idea what such a claim really means. Nor do I see how 'mixed realms' that contained both physical and non-physical components would, of logical necessity, have any and all of the laws which govern the non-physical side of things depend upon the physical laws. So this seems to be a very ambitious and hence not a very promising way to attack the idea of philosophical zombies.

And, contrary to the opinions, or at least the hopes of many, the idea of zombies is important. For to claim that zombies are logically possible is to deny a very common form of physicalism. I want to emphasize this point: it is the mere logical possibility of zombies that refutes physicalism. If, say, my philosophical zombie is logically possible then there is a possible being which shares all my physical properties but does not share all my mental properties. Thus mental properties are non-physical properties and the physicalist assertion that everything is ultimately physical is false. An ontologically liberal functionalist cannot escape either, since the idea of a 'functional zombie' is a simple extension of the idea of a 'physical zombie'. In fact, if, as seems reasonable, we assume that functional properties logically supervene upon physical properties, then the possibility of a philosophical zombie refutes functionalist physicalism no less than it refutes 'bare physicalism' (see Chalmers 1996, especially chs. 3 and 4).

This is a real problem. I've sometimes heard it said that the idea of zombies -- like that of certain other bizarre, purely philosophical thought experiments -- is so weird that we just don't know what to say about them, or that we just have no way to assess their logical possibility. And therefore (therefore?) we needn't spend any time worrying about them. Not good. This is tantamount to saying that we don't know what to say about the truth of physicalism and have no way to assess its truth or falsehood. Good friends of physicalism ought not to take this line. They know perfectly well what to say about zombies: such monstrosities are not logically possible. The question is whether there is any way to convince someone who is neutral about physicalism of this without begging the question.

I've also heard it said that physicalism is not meant to be such a 'strongly metaphysical' claim that it would have such exotic implications about the nethermost regions of logical space; rather, physicalism is supposed to be a kind of quasi-scientific, empirical claim that the creatures of this world, and in particular human beings, are purely physical creatures. And, just as physics doesn't care that it is (or might be, so to speak) logically possible for quarks to have radically different properties in some merely logically possible worlds, philosophers shouldn't care about hypothetical possibilities of Cartesian minds, or whatever other wild psychical or ectoplasmic metaphysics one might dream up. This sort of reply misses the point, and the strength of the zombie challenge.

Consider that the mere denial of physicalism does not entail that zombies are possible. The former is entailed by the existence of Cartesian possible worlds -- worlds in which there are non-physical entities and properties: mental substances possessing mental properties as postulated by Descartes. But there could be Cartesian worlds even if philosophical zombies are logically impossible⁶. So it is not at all the case that the zombie hypothesis is just a fancy way of dressing up anti-physicalism.

At the same time, I would think that philosophers are right to be wary of asserting that physicalism is logically necessary: that there are absolutely no possible worlds that contain non-physical entities and/or non-physical properties. The metaphysical visions of Leibniz or Spinoza, for example, do not appear to be flat out impossible.

Such modesty does not weaken the zombie argument, in fact it is irrelevant to it. For the possibility of zombies just shows that consciousness is not itself a physical property, nor even a property which logically supervenes upon such.

Of course, we don't have a very clear idea of exactly what makes a property a physical versus a non-physical property. But we can use the rationale of the zombie argument itself to help us get clearer about such properties. Here is a first pass at a sufficient condition for being what I will call a radically non-physical property:

(R) if some thing, x, has a property, P, which is such that it is logically possible for something physically indistinguishable from x to lack that property then P is a radically non-physical property⁷.

My philosophical zombie, if it is logically possible, will reveal that there are mental properties of consciousness which are radically non-physical properties by this criterion. And that seems intuitively correct. However, some properties which intuitively present no difficulties for a physicalist outlook are hereby declared radically non-physical. Any relational property will come out as non-physical by this criterion. For example, a Canadian one dollar coin has a possible physical duplicate which is not money at all. However, it is easy to extend our criterion of non-physicality to cover relational properties thus:

(R*) if some thing, x, has a property, P, which is such that there is a possible world which is physically identical to the actual world but in which x's physical duplicate lacks P, then P is a radically non-physical property.

Note that the specification of the possible world that tests for non-physicality might have to include the total history of the world. If the actual world was in fact created five minutes ago then it may be that there is no real money in it (only lots of perfect counterfeits), if one defines money in terms of being printed by a legitimately authorized mint (as opposed to miraculously appearing out of nowhere with the rest of the world and being treated as money by the similarly newly created denizens of that world). Thus test worlds may have to be totally physically identical to the actual world. This identity might even have to extend through 'future history' as well. For example, it is conceivable that the property of inertia is a relational (and physical of course) property depending upon the total distribution of matter throughout all of space and time. So if -- as seems unlikely to me -- properties of consciousness are in fact relational properties then it will be the logical possibility of whole 'zombie worlds' that will reveal that such properties are radically non-physical.⁸

We might now define a rather weak form of physicalism simply as:

(P) There are no radically non-physical properties.

Now it is clear that the mere logical possibility of zombies is enough to refute physicalism insofar as it reveals the radically non-physical nature of at least some mental properties and insofar as any reasonable version of physicalism (such as (P) for example) rejects the existence of such radically non-physical properties.

Showing that philosophical zombies are logically impossible would thus make the world safer for physicalism. But on the other hand, a proof of the impossibility of zombies which rested upon a premise asserting or implying the truth of physicalism would be worthless. That would just beg the question.

Robert Kirk has recently (1999) presented an argument attempting to demonstrate that philosophical zombies are logically impossible. I think the argument does not succeed, but its failure is instructive, although in the end not surprising. Kirk proceeds by reductio, assuming that a zombie is possible and from that deriving a contradiction. His method involves supposing that a zombie suddenly acquires qualia. To prevent misunderstandings based on this notoriously troublesome term, let me reassure the reader that no pernicious assumptions about the nature of qualia are at play here: the term 'qualia', both in Kirk's paper and throughout this paper, just stands for the properties of subjects that make them, in the constitutive rather than causal sense, qualitatively conscious. Qualia are simply the properties we have and zombies lack just in virtue of our being conscious. The acquisition of qualia seems quite straightforwardly to transform the zombie from a non-conscious to a conscious being, and in fact into a being just like us. Kirk's argument then proceeds:

4. But my zombie twin is by definition unaffected by anything non-physical.

5. So he is unaffected by acquiring non-physical qualia.

6. In order to become conscious as a result of his acquisition of non-physical qualia he would have to be in some way affected by them

7. Since he is not so affected, he, at any rate, does not become conscious. (1999, p. 6).

Of course, in this part of the argument premise 5 is highly problematic, if not incoherent. It is true that the zombie is not physically affected by his sudden acquisition of qualitative mental states, but it does not follow from this that he is unaffected tout court. To suppose that all affection is physical affection is just to assert physicalism⁹. That is, to a first approximation we can define 'being affected (by x)' as the acquiring or the losing of a property (because of x). The zombie, by hypothesis, gains qualia, that is, certain properties of qualitative consciousness. Thus he is affected. But, also by the hypothesis of Kirk's reductio, these properties are radically non-physical. So to deny the non-physical affection of the zombie is to deny that there are any such properties to gain, i.e. to assert a physicalism at least as strong as (P). We already know that if physicalism is true there can be no zombies so we have here no argument against the existence of zombies independent of the assertion that physicalism is true. No friend of zombies will deny that physicalism implies that zombies are impossible, but equally no friend of zombies will be impressed by an argument from physicalism to the impossibility of zombies.

But Kirk is not offering such a simple-minded and hopeless argument against zombies however. Points 4 through 7 are just the warmup. He wishes to deduce from the fact that the zombie will not be physically affected by the acquisition of qualia the conclusion that the zombie will not notice or think about his transformation or the qualia that produce that transformation. From this he attempts to deduce further that if zombies are possible then we too, being in effect zombies with qualia, as in the thought experiment, cannot notice or think about our qualia, or even tell the difference between different qualitative experiences.

Let us first consider the claim that the zombie cannot notice the qualia that it acquires. Here the assumption, dormant until now, that the physical laws governing the zombie (or the zombie world) are the same as the actual physical laws, comes into play. It has to be further assumed that these laws guarantee the causal closure of the physical world. As noted above, this is the assumption that every physical event is determined entirely by physical conditions and events to the extent that it is determined at all. We don't know for a fact that the causal closure of the actual physical world is true, but very fundamental conservation laws in physics seem to suggest that it must be if our physics is at all on the right track¹⁰. I'm happy to grant physical causal closure for the sake of the argument. Given this assumption, it indeed must be the case that the acquisition of non-physical qualia cannot have any indirect or downstream physical effect on our zombie any more than it can have an immediate physical effect.

Kirk's argument then proceeds as follows:

8. Telling the difference between the subjective character of two perceptual experiences requires detecting them; and that requires ... being sensitive to or affected by them.

9. Since zombies are by definition unaffected by anything non-physical, my supposed zombie companion cannot tell the difference between non-physical qualia ...

11. Since the friends of zombies maintain that I am nothing but a compound consisting of a 'zombie companion' and non-physical qualia ... their position entails that nothing can detect differences between non-physical qualia, hence nothing can tell the difference between the subjective character of smelling tea, and that of smelling coffee.

12. But many of us can tell the difference between the smell of tea and the smell of coffee.

13. Therefore we are not compounds of the sort the friends of zombies maintain we are, and zombies in their sense are not genuinely [logically] possible. (p. 9)

This argument compresses several intermediate inferences. First there is the claim that if zombies are logically possible then we (ordinary conscious human beings) are compounds of a purely physical zombie plus non-physical qualia. Although rather a perverse way to put the point it must be correct. I say 'perverse' since on the assumption of the possibility of zombies the only possible difference between us and them is the non-physical qualia that we possess and so of course we must be 'physical zombies' plus qualia. Or, to put it more sensibly, if we think of ourselves minus qualia we end up with zombies and if zombies are logically possible then it is coherent to think in this way.

Now, I (and Kirk) take it that being able to tell the difference between, for example, the taste of tea and that of coffee involves more than mere behavioral discriminative abilities, for the zombies possess these but are nonetheless said to be unable to 'tell the difference' between the taste of tea and coffee. What they lack, then, is no physically manifestable ability but rather the kind of knowledge of what tea, or coffee, taste like which conscious experience affords (readers will recognise that we are very close to Jackson's famous Mary argument here, see Jackson 19??). Normally, we think that it is this knowledge that grounds the ability to discriminate the taste of tea from coffee but that is true, at best, only in normal cases of human perception. We could, in principle, devise a machine that could chemically discriminate coffee from tea with as much accuracy, and similar categorizing dispositions, as human beings, but we don't think that such machines thereby necessarily possess any qualitative consciousness of tastes. Non-conscious discriminators are certainly possible (in fact there seems to be a great many of them, from thermostats to sunflowers). Zombies are an extreme case of non-conscious discriminators, which possess exactly the same discriminatory powers that we do, but without the consciousness that normally attends such powers.

Thus, of course, to all appearances, zombies are able to tell the difference between the taste of tea and coffee. My zombie, for example, will utter sounds which can be interpreted as English statements about how different the taste of tea is from that of coffee, and in what ways these tastes differ. Do these utterances have 'real meaning'? I don't think the answer to this question is at all obvious, but let us suppose that zombies can speak real English (or a homonymous and synonymous variant of English) so that when they claim, for example, that they are conscious they are making a genuine claim, a claim which is, unfortunately for them, simply false.

If they can make such claims, then presumably zombies can have real beliefs. Again, it is not really obvious to me that this is the correct interpretation of zombie behavior and neural activity¹¹. But let's suppose that this is OK, so that zombies do have genuine beliefs about various things; most significantly they have beliefs about qualitative states of consciousness. One might wonder, I suppose, how zombies could acquire concepts of states of qualitative consciousness in the complete absence of consciousness, even if it is granted that they can possess concepts in general. But blind people can have concepts of visual consciousness despite lacking it, so it is difficult to claim there is a difficulty of principle here. Many of the zombies' beliefs about qualia are true beliefs, such as the belief that the smell of roses is a pleasant smell, or the belief that there is a certain thrilling feeling that attends riding a roller-coaster. But all their substantive beliefs about their own conscious experiences are uniformly false (althought they have, as do we, various logically trivial beliefs 'about' self-referred experiences which are true by default, such as the belief that right now I am either having the experience of tasting cinammon or I am not). Of course, zombie beliefs are one and all non-conscious beliefs, whereas many of our beliefs are conscious. But there doesn't seem to be anything incoherent in the idea of non-conscious belief.¹²

These zombies will even have, or appear to have, ostensive beliefs of the form: so strawberries taste like that (just after 'tasting' their first strawberry). The ostensive act here will fail simply because of a lack of a referent for the demonstrative. However, there is nothing particularly mysterious in themselves about such beliefs, if they are beliefs, nor anything about them which especially concerns the zombie problem. We have all at one time or another had this sort of false belief, or 'cognitive failure'. It is of course very weird that zombies are so systematically mistaken even about such intimate judgements about their own (putative) experience. But no one claims that zombies aren't weird; lots of logically possible things are weird (like flying pigs, the worlds of Harry Potter or Frodo Baggins, etc.). The question is whether zombies are as weird as square circles.

Notice that if it was logically impossible for something to have beliefs about things that could not physically affect it, then zombies could have no beliefs about these presumed non-physical properties of qualitative consciousness. Then it might seem there would be a quick route towards the conclusion that zombies are logically impossible. The argument would go like this. Zombies could not have any beliefs about qualia, either before or after their transformation from qualia-less to qualia-enjoying creatures. But then they could not tell the difference between distinct qualia just because 'telling the difference' involves, at least, having some beliefs about qualia. But we are, or are equivalent to, zombies that have acquired qualia. So therefore we can't tell the difference between distinct qualia. But since we can, we get a reductio of the claim that zombies are logically possible.

This is not a very convincing argument however, because it rests on the extremely dubious premise that it is impossible to have beliefs about what cannot physically affect one. Beliefs about mathematical objects or abstract entities in general present obvious difficulties here, as do beliefs about possible physical things that we cannot be in physical contact with (e.g. events outside our 'light cones'). In general, the difficulty with this line of argument is to present a case for this restriction on belief, and concept, formation which is independent of the assumption of physicalism, an assumption which would in this context once again beg the question. In any event, I don't want to explore this avenue further here, since it plays no part in Kirk's argument.

Having granted that zombies can have genuine beliefs, let us see how the rest of Kirk's argument fares. I think it rests on a basic mistake in epistemology. It is a mistake to think that there has to be any physical change in a believer to transform beliefs into knowledge. I think that however knowledge is to analysed, it will be possible for all the conditions of the analysis to be met save for that demanding the truth of what is believed. This is obviously true for common theories of knowledge. On justified true belief accounts, it is easy to imagine situations where the only thing preventing a belief from attaining the status of knowledge is the fact that the belief is false. There are lots of justified false beliefs which, if they had been true, would have amounted to knowledge. Suppose you tell me you were in Simcoe Hall yesterday afternoon. You are generally trustworthy and I understand your utterance, and thus form the belief that you were in Simcoe Hall yesterday. Because of a slip of the tongue however, you mispoke yourself; you were actually in Simcoe Hall the day before yesterday. So obviously I don't know that you were in Simcoe Hall yesterday, but I would have known if you had been (everything else in the siutation being kept the same). A similar point clearly holds for reliability theories of knowledge and, I would venture to assert, any other reasonable account.

I think zombies are in an epistemic situation akin to my example, though one as all encompassing and strange as befits their peculiar nature. They have every reason, so to speak, to believe that they have qualia (always recalling that we have granted that they can have and, in virtue of their understanding of English and physical indistinguishability from normal human beings, do have beliefs about qualia). They are, to use another model, epistemically virtuous, at least they are as virtuous as we are, and if we can manage to have knowledge about qualia then so could they -- if they only had any qualia to have knowledge about.

So what happens when, as Kirk imagines, a zombie suddenly acquires consciousness? First, the zombie is affected, but non-physically, just by gaining certain properties that it lacked before. Of course, nothing physical changes, by hypothesis. And yet the zombie's false beliefs about its own mental states suddenly become true, and they become knowledge too. Friends of zombies will say that the zombie now knows what things taste like in virtue of actually having the conscious experiences which carry this information, whereas the earlier, non-conscious beliefs about tastes, referred to nothing, hence were false, hence could not constitute knowledge about how things taste. Zombies are, so to speak, very close to being able to tell the difference between tea and coffee, lacking only one, alas essential, component; they are like the man who says 'I'd be rich if I only had money'.

Consider this from the side of the zombie. Suppose a zombie is asked to think about or 'internally compare' the tastes of coffee and tea. The zombie thinks for a while, and carefully sips some tea and coffee, then soliloquizes about this 'difference' for awhile, doubtless saying many things that are true of the tastes of tea and coffee. But the zombie's remarks are grounded in utterly false beliefs simply because the requisite experiences are just missing. Of course, this kind of zombie would never seriously entertain the idea that it is a totally non-conscious being. It's epistemic situation is the same as mine and if I have no good reason to wonder whether I am a zombie then neither does it. But that doesn't eliminate the possibility that it is a zombie, any more than the fact that I have no good reason to believe that I am a brain in a vat shows that it is logically impossible that I am a brain in a vat.

One might wonder whether a zombie could know it was a zombie. I don't see anything absolutely preventing a zombie coming to believe correctly that it was a zombie, but I doubt that would be knowledge, since it would be an irrational belief¹³. What evidence could there be in favour of the zombie hypothesis? Since the zombie hypothesis requires the idea that the zombie inhabits a very peculiar sort of possible world, I don't think there could be a route towards a rational belief that it was a zombie. What could ever lead it to rationally conclude that it was in such a bizarre, merely logically possible, world?

But doesn't this whole analysis imply that it is logically possible that I am a zombie, right now, even as I, apparently, consciously think about my own consciousness? Not at all -- at most it shows that from your epistemic point of view it is logically possible that I am a zombie (and if zombies are logically possible at all this is hardly surprising -- the problem of other minds is a skeptical problem which is not incoherent). For I have conscious beliefs about my experience, and they are what form the basis of my knowledge that I am conscious. If you are wondering whether you are a zombie, I ask you just to consider whether you are conscious. You have conscious experiences; you are conscious that you are not a zombie, and that entails that you are indeed not a zombie. A conscious belief that one is conscious is self-validating; it cannot be a false belief¹⁴.

It is instructive here to compare my knowledge that I am not a zombie with my knowledge that the world is ancient, and was not created five minutes ago in the exact state it actually was in five minutes ago. Although I do know that the world is very old, it is logically possible, and perhaps more, that I am wrong. The fact that knowledge that P logically implies that P is true does not threaten the claim that what is known might be false. It is an analytic fact about knowledge that:

NEC(Kp ==> p),

but of course it does not follow that