1 De-modalizing anti-physicalism: the priority of the knowledge argument (DRAFT OF 14 SEPTEMBER, 2004) Howard M. Robinson robinson@ceu.hu István A. Aranyosi fphari01@phd.ceu.hu Philosophy, Central European University 1. The modal understanding By a modal understanding of anti-physicalism we mean an argumentative startegy of the sort that appeals to epistemic and metaphysical modal notions, like conceivability and possibility. The general recipe for formulating a modal argument against physicalism is along the following lines. You start with epistemic modals, and assert that it is conceivable, that is, not ruled out a priori that there be a world where everything from our world is duplicated as regards matters physical, but there is somehting different from our world as regards the mental phenomena. Then you appeal to a principle that would link epistemic modality and metaphysical modality, and deduce that it is possible that there be such a situation. Finally, you deduce a truth about the actual world, namely that mentality is contingently related to the physical phenomena. The fact that this sort of argument has become the most popular and discussed nowadays is related to the fact that physicalism has recently received a standard and widely accepted definition, which is formulated in such modal terms. The definition is part of the approach to physicalism put forth by Frank Jackson (1992, 1994, 1998) and David Chalmers (1996). Jackson defines physicalism by way of minimal physical duplication of worlds. Physicalism is true at a world iff a minimal physical duplicate of that world is a duplicate simpliciter of it. By minimal physical duplicate we should mean a world which duplicates everything physical and nothing more. For instance, if the original world which is to be tested for the truth of physicalism 2 contains some strange stuff, like ghosts, then the existence of such things will not affect the truth of physicalism, given the Jacksonian definition. Of course, we are interested in the actual truth of physicalism, and therefore we should apply the definition to our world. Chalmers appeals to the notion of logical supervenience. Physicalism is then the doctrine according to which any truth at the actual world logically supervenes on the totality of actual physical truths. To logically supervene means to be a priori entailed, so physicalism is explicitly formulated as the inconceivability of a physical duplicate of our world that is not a duplicate simpliciter. Jackson subscribes to the a priori entailment commitment of physicalism as well, and we agree with them on the issue. However, our agreement with them motivates our dissatisfaction with the focus of the recent deveopments of the debate over physicalism. These developments have been characterized by the exclusive focus on the issue of apriority and necessity. Most physicalists' main preoccupation is to deny the move from epistemic premises to metaphysical conclusions, while most dualists try to defend this link. We would like in this paper to show, among other things, that these modal issues are far from being so clear, that dualists should not appeal to such arguments not because of the feebleness of the conceivability-possibility link, but because of the feebleness of the conceivability intuitions themselves, and that there is a classic and clearer argument, Jackson's knowledge argument, that can do the anti-physicalist work with more effectiveness. 2. Elements of the problem. Once upon a time there was a fairly straightforward (but not, of course, uncontroversial) argument against physicalism, called the `knowledge argument'. It was the most focused of the qualia objections to physicalism. Discussion of this argument, it seems, however, has become enmeshed with - perhaps even swallowed up by - discussion of other closely related arguments that import extra complexities that perhaps damage the clarity of the original argument. These 3 other arguments concern, for example, the conceivability of zombies, the explanatory gap that there is between physical theory and mental phenomena, and the question of whether mental concepts can be conceptually analysed in physicalist terms. Our objective is to investigate what difference, if any, these sophistications make to the knowledge argument, and we shall do this by considering three important papers that have contributed to this diversification. (In fact the locus classicus for bringing these things together is chapter 3 of David Chalmers' The Conscious Mind, but we think it is in (some) discussions of Chalmers, rather than the original, in which they get run together as if there was no significant difference.) To clarify the situation, we will set out a brief taxomony of the various arguments involved. There are now, in the arena together, the following beasts threatening to devour the poor physicalist. The Explanatory Gap (EG): That no amount of information about the working of the brain etc can explain how there come to be conscious states associated with it. Conceptual Analysis Thesis (CAT): That (i) the gap between brain and consciousness could only be closed if there were available a (correct) conceptual analysis of mental concepts in something like functional terms, and (ii) there is no such correct analysis. The Knowledge Argument (KA): This argument purports to prove that total physical knowledge does not constitute total knowledge of the empirical realm, so there is something in experience which is non-physical. The Conceivability of Zombies Argument (CZ) : Because there is no conceptual analysis of conscious or phenomenal states in physicalist or topic neutral (eg. functional) terms, we can 4 conceive of our having physical twins which lack conscious or phenomenal states. The possibility of such `zombies' shows the falsehood of physicalism. In an obvious way, all these arguments are making the same point, namely that there are fundamental features of consciousness that physicalism cannot accommodate: they are all arguments for property dualism. How they are interconnected is intuitively clear. But is it not the KA that provides the argumentative force behind all the others? Let us remind ourselves of the argument. Harpo is a scientist who has been stone deaf from birth, but who has become the supreme authority on all physical aspects of the hearing process, from sub-atomic physics, through neurology, to computational and behavioural analysis. But he still does not know what it is like to hear, or what the phenomenal nature of sound - eg. C-sharp - is. And when, by the wonders of a new surgery - perhaps developed by himself - he gets hearing for the first time, he learns something new that no amount of physical knowledge could have given him. The subject matter of this new knowledge, therefore, must fall beyond the range of the physical. It is what the KA, says is left out by or from physical knowledge is also that which cannot be conceptually analysed in the appropriate way, and the fact that it cannot be captured in this way is what produces the explanatory gap. It is also because it cannot be captured by physical knowledge or by analysis that we can imagine it absent and so conceive of zombies. What are the other arguments adding? In order to understand this, it will help to formulate a more cautious version of CAT. Weak Conceptual Analysis Thesis (Weak CAT): That (i) the gap between brain and consciousness can be closed only if it is possible to infer a priori from the physical facts to the presence of psychological ones, given that one possesses the psychological concepts: (ii) in fact, this cannot be done. The difference between the original and the weak versions is rather like that between a reductive behaviourism that thinks that a strict analysis of individual concepts in behavioural 5 terms can be given, and a Wittgensteinian behaviourism that holds that the behaviour is logically sufficient for the psychological, but piecemeal reduction is not possible. In fact, the weak version allows for a general behaviouristic gloss of concepts, and if one is prepared to call that an analysis, there is no significant difference between the theses. Armstrong, for example, insists that his causal theory of the mind does not give meaning, in any strict sense, of psychological vocabulary, but ageneral rationale or gloss. From now on I shall use CAT to refer to the weaker thesis. CZ can be treated simply as a separate argument towards the same conclusion. Most of David Chalmers work on possibility and conceivability is perhaps best seen in this way. But there is a tendency in at least some of the literature to treat them as standing or falling together, which would not be the case if they were different arguments to the same conclusion. The tendency is to treat CZ as 'where it is at' in discussing KA. How does this connection work? I think it is as follows. One response to KA is to say that Harpo acquires when he gets his hearing is a form of knowledge how, not a form of knowledge that. This is the behaviourist/functionalist response, and currently the consensus is that this will not do1. So the context of developments is an acceptance that the KA is sound in the sense that it establishes that Harpo gains some new real knowledge, which is factual in at least the negative sense that it is not just a capacity, then to try to reconcile this with physicalism. We can now explain how the arguments relate in the following way. (1) KA presents a very strong threat to physicalism because it seems to require property dualism. But (2) it presents such a threat only if physicalism entails that there is a conceptual analysis, in the weak sense. 1 Though one of us believes that this consensus is unjustified. See Aranyosi (2004a, chapter 3) for a defense of the commonsense functionalist response to KA. 6 This is so because (3) in the absence of an analysis - entailment from physical facts to mental - there is bound to be some kind of explanatory gap and, hence, scope for something new for Harpo to learn when he starts to hear. But, therefore, (4) if physicalism does not require analysis, it will not be afraid of the consequences of the lack of an analysis, that is the existence of an explanatory gap, and the conclusion of KA. The suggestion is that a physicalist need not be threatened by a weak explanatory gap. It is at this point that the anti-physicalist is tempted to fall back on the conceivability of Zombies, for that involves a real property dualism. So (5) if one can show that zombies are a real possibility, the conclusion of KA must be strong enough to establish property dualism. There has, therefore, been a scuffle between those defending zombies and those intent on showing that physicalism does not require an a priori component. But proving that there could be zombies is not the only way of showing that the KA's conclusion is strong enough. It would also do to show that physicalism is really committed to weak analysis (Weak CAT), for everyone agrees that KA is incompatible with that. We want to consider three papers that have as their common thread the notion that the physicalist is not obliged to provide some kind of analysis of mentalistic language in order to accomodate it to his ontology, however austere the latter may be. The strategy of the first paper, by Block and Stalnaker (1999), is to deny that reduction in general involves conceptual analysis and hence to claim that its absence in the case of mind signifies nothing. The next two, by Balog and Loar, respectively, have a more precise common thread. Their authors hold that the physicalist can treat phenomenal concepts in a purely demonstrative fashion, as does the dualist. They do not claim, as do Block and Stalnaker, that there is not a conceptual analysis 7 requirement for normal scientific reductions: it is the special demonstrative nature of the experiential case that does the work. The argument for this really comes in Loar's paper, though they both rely on it. Balog's paper argues that CZ is too promiscuous, as it could be used to prove things that are ex hypothesi false. Loar's paper attempts to close the explanatory gap by refuting the knowledge argument. It is the only one of the papers that deals with the KA as an independent argument. (3) Block and Stalnaker on why explanatory gaps are harmless. It is a principle adhered to by Smart, Armstrong and functionalists that, if one is to claim that mental states are physical, one must give an analysis - or at least a gloss - of mentalistic terms which makes it intuitively clear how something physical could qualify as mental. If this is required in the case of the reduction of the mental to (or its identification with) the physical, it must be a general requirement for similar identifications or reductions. Such a requirement would be equivalent to the provision of a topic neutral account of the concept to be reduced. In the case of mental states that would most probably be something along the lines of Mental state M is that state - whatever it is - that is typically brought about by stimulus S or ...typically causes response R. In the case of other states - for example being water, it might be something of the form Water is that kind of stuff, whatever it is, that underlies the stereotype that, on Earth, we call 'water'. Block and Stalnaker deny CAT(i), claiming instead that: (i) The absence of conceptual analysis does not, in general, entail the presence of an explanatory gap. (ii) There is no conceptual analysis involved in purely physical cases - eg water = H2O. 8 We would like to first point out that (ii) could be interpreted in two ways, the first of which makes the qualia/water analogy be irrelevant from the point of view of defending physicalism, and the second self-refuting. The first interpretation could be that we have only a weak analysis or just a gloss on how water-truths are determined by the totality of physical truths This interpretation is of no use for defending physicalism, because the qualia objection is precisely that we lack even this weak analysis or gloss over how physical or neural states can metaphysically determine phenomenal truths. The second interpretation could be that even though the water-truths are purely physical we lack an analysis of them in physical terms. This interpreatation is even more problematic. You cannot really say both that: (a) The water = H2O case is purely physical and (b) We do not know with certainty, given by an a priori deduction, but only by fiat that the identity holds. If the argument is meant to show that there are the same explanatory gap problems with standard scientific identities as in the mental-physical case, than that is an argument at most to the non-physicality of scientific identifications. We agree here with David Lewis, who pointed out, against this kind of argumentation, that it implies that we are not sure about the physical nature of, say, boiling, which is absurd. As Lewis puts it ([1994] 1999, p. 296): “It may seem that when supervenience guarantees that there are physical conditions sufficient for the presence or absence of a given mental item, the sufficiency is of the wrong sort. The implication is necessary but not a priori. 9 You might want to say, for instance, that black-and-white Mary really did gain new knowledge when she first saw colour; although what she learned followed necessarily from all the physics she knew beforehand, she had remained ignorant because it didn't follow a priori. A short reply to this objection from necessity a posteriori is that if it did show that boiling was irreducible to fundamental physics, it would likewise show that boiling was irreducible to fundamental physics – which is absurd. For the identity between boiling and a certain process described in fundamental physical terms is necessary a posteriori if anything is.” As a general observation, we should also point out a difference between Block and Stalnaker's approach to the issue and that of the classic identity theory of J. J. C. Smart (1959). While the latter appealed to Occammist considerations for establishing identities, but these were taken, precisely because of these considerations as contingent identities, and the approach was coherent, the former seem to hold a hard to understand and even incoherent view, according to which we should assert that the water- H2O case is a necessity, yet we should also assert that we don't know for sure that it holds. We believe one cannot have it both ways: once you commit yourself to asserting that there is a necessity, you cannot appeal to Occamist considerations and assert that you don't know it for sure. As we said, Smart's theory was at least coherent on this point. The scientific identities are postulated on grounds of general method - simplicity. This can be done, according to Block and Stalnaker just as well for mental state-brain state identities. The implication is that, in so far as KA establishes emergent properties, they can be integrated into the body of science by nomological fiat, as happens in other cases. Their position is, in this respect, similar to Searle's. They argue this, partly with Levine in mind, but mainly with Jackson and Chalmers as 10 their target. To set up what they are attacking they quote and paraphrase Levine as follows. Levine holds the following, in common with Jackson and Chalmers. "While it is conceivable that something other than H2O shuld manifest the superficial macro properties of water...it is not conceivable...that H2O should fail to manifest these properties (assuming, of course, we keep the rest of chemistry constant)." In other words, there cannot be such a thing as 'zombie water' - something like water in all its micro physical properties (with the same laws and conditions) and lacking the same manifest properties. By contrast, about pain, he holds (in Block and Stalnaker's words): "However, according to Levine, one cannot say the analogus thing about conscious states and their neural correlates. Even if pain turns out to be perfectly correlated with pyramidal cell activity, and even if we decide that pain is (necessarily) pyramidal cell activity, it will remain possible (Levine contends) to conceive of pyramidal cell activity without pain." Part of this - namely the contrast with water - is common with Jackson and Chalmers, but the possibility that there is nevertheless identity and hence necessity is particular to him. For present purposes, we shall ignore it. It seems to us that Block and Stalnaker's response to the core argument rests on a fairly simple mistake, namely that they are confused about what is involved in there being a conceptual connection between micro and macro. Their claim is that there is no analysis of water such that it follows a priori that H2O, given its properties, is or constitutes water. Let us look first at the arguments and conclusion found on 8-9, during their discussion of Levine. Their first argument is as follows. “Let C be a complete description, in microphysical terms, of a situation in which water (H2O) is boiling, and let T be a complete theory of physics. Can 11 one deduce from T, supplemented with analytic definitions, that H2O would boil in circumstances C? To see that one cannot, suppose that the deduction is taking place on Twin Earth. The stuff they call "water" is XYZ, and the process they call "boiling" is a process that superficially resembles boiling, but that involves a different physical process. Just as they would say (truly) "water is XYZ, and not H2O (and if there were H2O, it wouldn't be water)," so they would say, truly, "If there were H2O, and it were behaving like that, it wouldn't be boiling." They could hardly deduce ‘H2O would boil in circumstances C’ if on their meaning of ‘boil’, H2O can’t boil at all. (We assume that boiling is a natural kind concept. If you don’t agree, substitute some other process term that does express a natural kind concept.)” The argument appears to be that 'one' - i.e. we - cannot deduce that H2O boils in C because a Twin Earthian could not deduce that it twin-boils in C; and the slippage from boiling to twinboiling is legitimated by the fact that 'boil' is a natural kind concept. This seems wrong in at least two ways. (i)If boil is a natural kind concept, twater does not boil, so the slippage is not legitimate. That is, the fact that H2O does not twin-boil does not mean that it does not boil and it does not mean that a Twin-Earthian acquainted with the concept of boiling could not work out that H2O boils. (ii)Natural kinds are irrelevant. A description of what both water and twater do is enough. There must be such a description, otherwise there would be no sense in calling the other earth our twin. What can be deduced, even by a properly informed twin-earther, is that H2O will do that under C. Their second argument is as follows. “We don't really need a Twin Earth story to make our point. Consider a person on actual Earth, who does not know the story about how water boils -- perhaps 12 she doesn't even know that water is made up of molecules. One presents her with the theory T, and a description (in microphysical terms) of a water boiling situation. Can she then deduce that if T is true and a situation met conditions C, then the H2O would be boiling? No, since for all she knows the actual situation is like the one on Twin Earth. Perhaps, if she were told, or could figure out, that the theory was actually true of the relevant stuff in her environment, she could then conclude (using her knowledge of the observable behavior of the things in her environment) that H2O is water, and that the relevant microphysical description is a description of boiling, but the additional information is of course not a priori, and the inference from her experience would be inductive. (i) The appeal to her ignorance about whether she is on Twin Earth is ineffective because the Twin Earth version of the argument has been shown to be hopeless, and because B. and S. provide their own reply to it.” (ii) There is a mistake about what has to be shown a priori. It is not that H2O under C is water boiling, but that it manifests all the properties we believe water to possess when it boils. This is an important point to which we will recurr. Immediately so, because of what B and S say in their conclusion. All we reject is the a priori, purely conceptual status attributed to the bridge principles connecting the ordinary description of the phenomena to be explained with its description in the language of science. What is actually deduced in such an explanation is a description wholly within the language of science of the phenomenon to be explained. For this to answer the original explanatory question, posed in so--called folk vocabulary, all we need to add is the claim that the phenomenon described in scientific language is the same ordinary phenomenon described in a different way. But if the closing of an explanatory gap does not require an a priori deduction of the folk description of 13 the phenomena, then it has not been shown that unavailability of a conceptual analysis of consciousness need be an obstacle to the closing of the explanatory gap between consciousness and the physical. B. and S. are right that it cannot be deduced a priori from science and the properties of H2O that H2O is water. But it can be deduced a priori that H2O is an adequate candidate to be water: ie, if H2O were water it would explain all the relevant observable facts about water. This kind of adequacy is precisely what neuroscience lacks in respect of consciousness: we cannot deduce from the nature of pyramidal cell activity, as characterised within neuroscience, that, were there to be something of that sort, it would it would give rise to or constitute a pain-like sensation. B and S confuse the non-deducible brute fact, asserted on the grounds of general scientific simplicity, that it is H2O (not, eg, XYZ) that is the waterish stuff in our vicinity, with a supposed brute fact, supposedly asserted on general scientific grounds, that H2O manifests itself in a waterish way under certain conditions. The latter is not brute, nor asserted on general scientific grounds, but is deducible from the properties of H2O, the laws of nature and the conditions. These considerations enable us to answer B and S's objections to Jackson and Chalmers, we think. Perhaps Chalmers invites B and S to make their mistake by the way he states the relationship between macro properties and micro base. He supports the full CAT, when all that is needed is Weak CAT. Chalmers initially defines logical supervenience as follows: B-properties supervene logically on A-properties if no two logically possible situations are identical with respect to their A-properties but distinct with respect to their B-properties. (The CM, 35) The problem with this is that there may be 'ghostly' B-properties, in some world, over and above those that supervene on the A-properties. The brunt of the rather obscure arguments in B and S 16-20 is that there can be no a priori argument from microphysical facts to the absence of such 14 ghostly properties, so there is no general a priori move from how things are at the micro level to how they are at the macro. But the situation is that such micro facts are a priori sufficient to guarantee the presence of the macro facts. Whether they are the actual explanation of all the macro facts of the relevant kind is an empirical matter. I would reformulate Ch's definition: B-properties supervene logically on A-properties (holding the A-level laws of nature constant) iff the presence of A-properties is a priori sufficient for the presence of Bproperties [the presence of B-properties can be deduced from the presence of Aproperties] and the A-properties provide an a priori explanation of the presence of those B-properties for which they are sufficient / the existence of which can be deduced from them. There is no longer any need to deduce the whole macro set up, including negative conditions from the micro facts. So Block and Stalnaker's arguments may show that there is no adequate conceptual analysis of 'water', such that it follows that there is no more to water than H2O. It does not show that we cannot work out a priori that the properties of the base - H2O - are sufficient to explain all the observable properties of water, or ones indistinguishable from them. Given this, we are methodologically free to assert identity. This a priori sufficiency is absent in the mind-brain case, so the assertion of identity is not warranted on general methodological grounds - certainly not with as solid a rationale as in the case of water. (4) Balog's reductio of the zombie argument. Balog's strategy is to show that an argument exactly parallel to Jackson's would allow Zombie's to prove that they are not purely physical. As it is ex hypothesi that they are, there must be something wrong with this general form of argument. 15 Jackson's Argument (1) If physicalism is true, then for any true T, statements of the form K -> T are conceptual truths. (K is the total of true descriptions in physics.) (2) There are some true statements Q to the effect that phenomenal conscious experience occurs... (3) If Q is a phenomenal statement then `K -> Q' is not a conceptual truth. So (4) Physicalism is false. Zombie-Jackson's Argument (1*) If physicalism is true, then for any true T, statements of the form K -> T are conceptual truths. (2*) There are some true statements Q+ to the effect that a phenomenal+ state occurs... (3*) If Q+ is a phenomenal+ statement, then `K -> Q+' is not a conceptual truth. So (4*) Physicalism is false. As it is ex hypothesi that physicalism is not false for zombies, the original argument must be somehow flawed. (2) and (3) are not in dispute, so (1) can be rejected. The natural objection to this argument is that (3*) is false, because the only sense that could be given to a zombie's `phenomenal+' states would be a functional one and functional statements are entailed by K. Balog denies this, because what is and what is not a conceptual truth depends on the conceptual role of their constituents, and the conceptual role of the elements in zombie thoughts are just the same as those in ours. This raises two questions. First, whether conceptual role semantics is correct and second whether there really is an identity of role between the 16 zombie and human concepts. Balog does not supply much argument here. On the former of these two points she says If a prioricity did not so supervene...[end para] Of the latter ...Jackson's phenomenal concepts and zombie-Jackson's phenomenal+ concepts have parallel conceptual roles. Q+, like Q, lacks conceptual links to physical,functional, and behavioural concepts sufficient to ground the a prioricity of `K -> Q+'. It is this latter assertion that seems controversial. It is natural to think that the meaning of the human Q depends essentially on its connection with the real phenomenal qualities that zombies lack, and that, therefore, any parallel concepts that we might be prepared (if only for purposes of argument) to admit that zombies possessed, would have to have another source of meaning, and this could only be functional role2. The crux can be brought out by comparison with what Balog calls `yogic' concepts which are the same as blindsight ones. A yogic concept is one by which we can refer to something without knowing anything about the nature of that to which we refer and is, in a sense, the kind of concept that a zombie would possess. It is a form of topic neutrality without anything that guides the reference contextually. Yogic concepts (as opposed to yogic access to states we otherwise understand - as with blindsight) are peculiarly empty. In fact without some kind of associated content, they are no better than dispositions to meaningless utterances. We want to make three points about yogic concepts. Balog’s argument is yet another point of both agreement and disagreement between the authors of this paper. Both of us agree on this interpretation of Balog’s thought experiment as the correct one, namely, that the parallel zombie concepts that we would be prepared to accept should be thought of as functional ones. The disagreement is that one of us (HR) thinks this points to property dualism, while the other (IA) that this shows that deep down zombies are not conceivable since we cannot avoid functional role concepts in analyzing the situation, so that commonsense functionalism can safely be concluded from the thought experiment. 2 17 (1) Balog says that they can be treated in the same way as zombie concepts, giving a restricted version of the 'zombie argument'. But these concepts are topic neutral - that is, the yogis have the resources to realise that they know nothing about the reference of these terms and so nothing rules out that reference being in fact physical. And the physical facts do entail that they make the reports. So the physical is sufficient and the conceptual analysis thesis is not avoided. (2) Blindsight concepts, in so far as they actually occur, have their sense from the context in which the subject is not blindsightedly but naturally perceiving. It is not obvious to me that there could be purely direct referential blindsight concepts - which is what yogic concepts are. Without either contextualising descriptive gloss or a qualitative nature to the reference, they are peculiarly empty: so empty, I think, as to be meaningless. (3) How does Balog think that phenomenal concepts, on her account, differ from yogic ones? They are directly referential, are not explicitly of something physical and do not invoke nonphysical qualities. How does this differ from yogic concepts? To understand this, let us consider Loar's classic paper. (5) Loar's account of concepts that are both phenomenal and demonstrative. Loar's position can be stated as follows. (i) The new knowledge Harpo gets deploys special phenomenal concepts but does not relate to new properties or qualities. (ii) These concepts are ostensive, not given descriptively, and this explains why there is no functionalist analysis of them. (iii) It thereby also explains why no account in terms of physical theory - into which these ostensive concepts are not integrated - can explain the role of these phenomenal concepts. 18 So this answers the KA and shows why no conceptual analysis is required, and the EG either harmless or nonexistent, depending on how you construe it. If correct, it would also undermine the zombie argument. That argument, as currently expressed, depends on Chalmer's distinction between the primary and secondary intension of terms. The primary intention of 'water', for example, is what 'water' shares with 'twater'; the waterish stereotype that can be the face of either H2O or XYZ, or an infinity of other things in other worlds. The secondary intention is H2O or XYZ, or whatever, depending on the world. The role of this distinction is to show that Kripke's a posteriori necessities cannot really restrict the range of metaphysical possibilities. There is no possible world in which water is not H2O, but this is only a semantic restriction, not one which restricts the range of possible combinations of properties: there are waterish things that are not H2O, but you cannot call them 'water'. Physicalists argue that the relation between eg. pain and some brain state is necessary a posteriori, like that of water and H2O. Developing Kripke's arguments, and deploying the notions of primary and secondary intentions, Chalmers shows that it cannot be. But if the primary intention is topic neutral, without a contextualising description, then it does not present a face which could be the face of a variety of possible worlds. Its hidden real essence - its secondary intention - is the only positive content that it possesses. Everything turns, therefore, on the legitimacy of this bare notion of topic neutrality. 6. Three Problems for the Loar Solution. (i) Does his theory, even if correct, solve the problem? Loar's solution is, in some ways at least, a variant on a solution already proposed by others, namely that what Mary originally lacked and came to possess was a new mode of access to something about which she already knew in another way. The replies to this are well-known. In Frank Jackson's words; ' What is immediately to the point is not the kind, manner, or type of knowledge [DS] has, but what [he] knows.' Richard Warner illustrates the point: we may either see or hear that a certain person is 19 in the room and which we do is indifferent to the content of our discovery. It is not difficult to appreciate the prima facie appeal of the idea that a difference of mode in knowledge could explain DS's case. The idea behind the 'mode' theory is, presumably, that if the phenomenal is merely another mode of access to respectable physical facts, then it does not itself constitute an extra fact: it is not an addition to ontology, but an extra mode of access to part of the previously accepted ontology. A moment's reflexion will, however, show this to be sophistical. Call the set of physical facts P and the scientific modes of access which DS has to these facts S, and the experiential mode of access he lacks H. If we regard S and H as external to P, then the addition of a new mode of access will not alter P. But the physicalist hypothesis is that everything relevant is included in P: modes of access to physical facts are themselves simply physical processes and are included in physical facts. Therefore, if DS knows all the relevant members of P, he should know all the facts about H: he should know, that is, all that is real about applying phenomenal concepts. Can Loar avoid this objection by appealing to the particular nature of phenomenal concepts, as he explains them? This would involve arguing that one can apply such concepts only if one has the experience oneself, so only those who have the experience can have the knowledge which is cast in that way. There is an ambiguity or equivocation here. The early Mary's defecit was not simply that she lacked certain pieces of information, but that there was a certain type of proposition which she was not in the position to properly understand: she lacked the concepts, not merely the knowledge. Uncontroversially, one cannot exercise these concepts to gain knowledge through them unless one has the experiences. But knowledge about the process of applying phenomenal concepts - how they work and why - is, if physicalism is correct, something that Mary knew about all along. Only if there is something about how those concepts work which is only available from the first personal perspective, should she have lacked any understanding. But this is the sort of thing the physicalist cannot allow. Only if the first-personal perspective reveals something essential about their reference not otherwise available would understanding be so 20 dependent. (longer footnote on Nagelian approaches, see my chapter II) (ii) Is it possible to locate Loar's position? There are three accounts, which I can understand, of how reference to phenomenal states might work, and which are brought into the discussion. (a) One identifies them by some quale they present, or which constitutes them. (b) One identifies them by a kind of bare reference that is analogous with blindsight. (c) One identifies them by some description which locates them in virtue of a contextual (eg stimulus - response) property. It is agreed that the first of these is the dualist option that the physicalist is trying to avoid. The second, it is agreed, leaves out the experiential element altogether. The third, it is agreed, is what Kripke refutes. Loar's own theory seems to be a mysterious combination of (a) and (b): it omits the quale, without omitting consciousness of the experience. I cannot detect the tertium quid between these theories. By contrast, Loar finds it obvious. The main point is by now more than obvious. Whatever the antiphysicalist has said about these [ie phenomenal qualitites, feelings of them and blindsight] cases the physicalist may say as well. The way that one picks out the phenomenal quality cramp feeling by way of a particular feeling of cramp (or image, etc) is hardly incompatible with holding that that phenomenal quality is a physical property. The contrast between phenomenal concepts and self-directed blindsight concepts and cramp concepts finds physicalist and antiphysicalist equally able to say something sensible.(604-5) The only sense that I can make of this is that the difference between normal awareness and the blindsight kind of case, is that the grasp in the former is via something that involves a feeling or image, the latter does not. But do not feelings and images fall within the range of those qualia-type entities that we are supposedly explaining? It is, after-all, the difference 21 between these two cases that constitutes the difference between 'lights on' and 'lights off', as far as consciousness is concerned. If I understand him, Loar is making the very mistake Kripke criticises. (The pain is not what lies behind the feeling, it is the feeling.) He is treating the phenomenal quality as something that lies behind the experiential states. The quality, as Loar conceives it, is common to `blindsight' and full consciousness, and what constitutes the latter is treated as a form of imagistic conceptualisation, which he also describes as `a particular feeling [of cramp]'. But, (a) there is no adequate account of imagery possible in this context. (`Adequate' meaning that it is both physicalistically pure and will do the job of constituting what the `blindsighter' lacks). (b) This means that the imagistic conceptualisation is the phenomenon. Recent discussion of one of us (HR) with Kati Balogh has shown that the following is her position, and perhaps it is Loar's. The difference between the phenomenal case and the yogic/blindsight case is that in the former, but not in the latter there is a phenomenal quality present, but that quale simply is identical with a brain state. There is, I think, no reduction of the quality involved here, and no topic neutrality in our knowledge of the phenomenal aspect of the quale. I think this takes us back to B and S, with the assertion of brute identities between qualities on grounds of simplicity alone. This will seem plausible only if this is how other reductive identities operate, and we have seen how it is not. (iii) The direct anti-physicalist approach that we favour, and which we will expound in the next section, consist of our having a clear and distinct, to use some good old Cartesian phrase, idea of a conceptual disparity between physical and phenomenal notions, namely the former being connected to properties of the dispositional and relational logical type, while the latter to those of the categorical and monadic logical type. If Loar wants to accommodate the directness, and by that transparency of phenomenal reference, and to also be a realist about phenomenal properties, then he has to assert that one and the same property is of both logical types that the two kinds of concepts conceptualize it, which is contradictory. Therefore, we do not see how to be a 22 phenomenal realist and assert physical/phenomenal property identities3. 7. Why modalization is problematic In this section we would like to be more explicit about why modalization of the physicalism/anti-physicalism debate is not desirable, and why the what we call ‘direct or nonmodal arguments’, KA being one of them, should better be focused on, in order to avoid lack of clarity. In one of his recent papers Chalmers (2002) merges the KA and the ZA under the same heading of “conceivability arguments”. As he formulates it, the KA is an argument requiring a weaker premise, the negative conceivability of zombies, while the ZA a stronger one, the positive conceivability of zombies. Positive conceivability is a strong kind of conceivability, similar to Descartes clear and distinct perception. It requires an act of clear imagination of a scenario, as opposed to negative conceivability, which only requires the lack of a defeater of a scenario. We do not agree with this merging. Before arguing why the KA is not a negative conceivability of zombies argument, we touch upon the original ZA. The main issue that has been discussed in connection with the soundness of the argument is whether conceivability entails possibility. Positive conceivability is the kind of conceivability that presumably entails 3 Correspondence of one of us (IA) with David Chalmers revealed that the principle that direct reference to qualia imply transparent reference to them, with the consequence that the type of conceptualization should be isomorphic to the type of property referred to, would in fact be easily denied by one who follows Loar’s approach. We don’t think a good case can be made for this. The only counterexample to this principle seems to be the kind of bare demonstrative reference, when a cognizer has no idea whatsoever what she is referring to by a term, that is, does not associate any property to the property she is referring to. This is not the case with qualia: here we implicitly associate properties that make them fit for being subsumed unde more general logical categories, e.g. monadicity and categoricalness. For another problem of conflict between qualia realism and physical/phenomenal identities see a recent article by Jose Louis Bermudez (2004), where he points out vagueness – which is characteristic to the phenomenal and is absent in the case of physical properties-- as a factor that makes impossible such identification. 23 possibility. At the same time, it puts a lot more burden on anti-physicalists, since it is a demanding notion of conceivability, namely imaginability4. This is one reason we would advise dualists not to appeal to the ZA. But the positive conceivability based ZA should also be less appealing to defenders of phyhsicalism. There are two main approaches within the physicalist camp, a posteriori and a priori physicalism, and it seems that none of them can gain too much from properly attacking the ZA. The proper attack on ZA is, we believe, the denial of positive conceivability, since once one accepts that zombies are positively conceivable, one should also accept entailment of possibility5. But attacking positive conceivability is not something either of the two kinds of materialists would find comfortable doing. The a posteriori materialist is committed to the denial of the conceivability-possibility entailment in the case of zombies, so she has to attack that premise rather than the conceivability premise. The a priori materialist will, of course, attack the positive conceivability premise, but as a result of his much stronger commitment to denying even negative conceivability. According to a priori materialism, phenomenal properties and physical/functional properties are analytically related, so she will not bother with only attacking one of the terms of such a fine-tuned distinction like the positive/negative conceivability distinction. She will deny conceivability in its weakest form. Therefore, the modal approach represented by ZA is not very advantageaous from the point of view of either physicalism or dualism, since it makes things more confused. Let us then turn to the KA. As we have mentioned, we do not agree with David Chalmers’ putting the KA under a negative conceivability of zombies argument. The KA’s two important parts are: (1) that one could know alll the physical facts about colour experience, and (2) that one would not thereby know everything about colour experience. From these it follows that there is something left out by physical knowledge. The fact that something is left out by the complete physical knowledge is pretty clearly equivalent with the fact that there are non- 4 On how demanding this kind of conceivability can be see Stephen Yablo (1993) and Peter Van Inwagen (1999). 5 This is because positive conceivability is precisely a notiion elaborated for the purpose of possibility entailment. 24 physical facts, which is just the negation of physicalism. First, the KA does not even implicitly require any act of conceiving of a non-actual world. Second, it does not have to appeal to principle of modal cognitive reliability. Third, the fact that one could augment the argument or alter it to include the negative conceivability of zombies, does not mean that the argument needs such an augmentation or alteration in order to be effective. We argue, in fact, that the KA is independent from the ZA, and it can be put under the heading of direct or non-modal arguments against physicalism. An alternative direct argument against physicalism that could be formulated is a reformulation of point (iii) that we used to criticise Loar’s position in the previous section. The reformulation would mainly consist in drawing a clearly property dualist conclusion from the disparity of physical/phenomenal concepts/properties, according to which no physical property can be identical to any phenomenal property. This sort of argument is, we believe, more effective than the conceivability arguments in attacking physicalism, and also clearer than the conceivability arguments, so that they can be more properly discussed by physicalists6. 8. Conclusion. Starting, as is our intuition, from the KA, we shall try to piece together the area. Let us begin by asking (1) What does the knowledge argument show? Two possible responses to this are (2) That some concepts are not integratable into science. (3) That some properties are not integratable into science. 6 For more advantages of theis kind of property dualist arguments over the classical ones, see Aranyosi (2004b). 25 Physicalists need to deny the latter. But they can hardly rest content with affirming the former because of the following is hard to deny (4) The phenomena - qualia as they are experienced - are not just concepts or ways of conceiving, but things or states that get conceptualised, ie. the objects of concepts. If one accepts (4), as one must, how might one avoid (3)? The traditional answer has been (5) The objects/contents of phenomenal concepts are grasped topic neutrally via some contextualising analysis, gloss or paraphrase. In the present context, this is something everyone agrees to reject, for this is what is proposed by those who think there is a conceptual analysis available. So one might try (6) The contents are grasped topic neutrally without any analysis etc. This was the solution that lead us to wonder (a) how they had any content at all and (b) how they differed from `yogic' concepts. The reply appears to be (7) The contents are not grasped neutrally, but as having a qualitative content, but this content is (token?) identical with a brain state. (8) But how can there be brute identities between non-neutrally and non-contextually conceived properties? (9) At least we can say that if the Conceptual Analysis Thesis was false, in the sense that there was no a priori sufficiency of the base in normal, non-controversial cases of reduction (as Block and Stalnaker affirm) then it would be difficult to see anything special about mind-brain relations. So KA does require that version of CAT. 9. General reflections. There seems at first sight to be a wide range of contemporary physicalisms, stretching from 26 eliminativism at one extreme, through Dennett, functionalism, and Searle, to those who regard it as true but essentially mysterious - McGinn, Strawson, Nagel. In fact there seems to me to be a lot more in common here than one might think at first sight: at least in the forms these theories have come to adopt as the discussion has developed. A surprising range of philosophers accept the following three propositions. (1) Of course, there are the mental states - conscious and intellectual - that we think there are, and they are broadly as they seem to us. We are not denying the obvious. (Even Patricia Churchland says this about consciousness.) (2) What these states in fact are, are complex micro structures that realise the behaviourinfluencing role of the mental states in question. There is serious disagreement about whether complex micro structures with these functional roles have to be in certain kinds of matter or can, in principle, be in anything, but there is agreement that functional organisation in something however restricted or not it must be - is all there is to mental states. (3) There is an endemic failure of imagination that makes it impossible - or more or less so - for us to see how what is described in (2) can amount to what is described in (1). It at least seems to us that we can conceive things as in (2) without their amounting to (1). There is a real emergence relative to our understanding; a kind of epistemological emergence. This is why Mary learns something new. But this is due to a lack in our understanding - a conceptual gap between two ways we have of thinking about things - not to there being two forms of reality. There are various accounts of what underlies this endemic failure of ours, some making it less insuperable than others. To support our claim that belief in these principles is quite general... (1) Even Patricia Churchland claims to believe this with respect to consciousness. The common affirmation is that, as far as our experience of life is concerned (if not all our `theories' for explaining our behaviour), it is as it seems. Whether this really makes sense for an eliminativist is another matter. 27 (2) Not only functionalists, (which includes computationalists and eliminativists) believe this. Searle does not deny closure under physics, and is an internalist, so he believes that neural micro-structres drive behaviour, and that the presence of structures of the right kind, in the right kind of stuff, is all there is to the existence of mind. He simply holds the odd view that emergence of conscious and intentional states allows an interactionism that does not threaten that closure, because such emergences are not peculiar or threatening. (3) is the new-ish one, and stems from the now almost universal belief that no reductive analysis - however broad, loose or schematic - will seem intuitively to capture the nature of consciousness from the first person perspective. Trying various ways of living with this fact whilst sticking to physicalism is now the name of the game. Searle, and Block and Stalnaker believe that the `emergence' is a harmless feature of the general relations between `levels', which explain qualitiative emergence only in a brute way that is, via a theory which simply says it does. Balog and Loar explain it as a feature of the bare demonstrative nature of phenomenal reference., though perhaps also relying on the same point as S, B and S. McGinn says it is because the eye cannot see itself. And Galen Strawson and Nagel think we need a radically new understanding of physics. They alone think that physical science itself must change radically for the problem to be solved. The relation of ZA to EG and CAT is also supposed to be clear: ZA says that, if there is no conceptual analysis of the mental in physical-compatible terms, there follows not only an explanatory gap from physical to mental, but also that the existence of the one without the other is conceivable. We can express these two points as follows (1) KA -> CAT & CAT -> EG 28 and (2) CAT -> EG & ZA From this it follows (3) KA -> ZA In other words, the Zombie (or Conceivability) Argument follows from the Knowledge Argument, and, therefore, if the Zombie/Conceivability Argument can be refuted so can the Knowledge Argument. Similarly, if the Conceptual Analysis Thesis or the Explanatory Gap can be refuted, so is the Knowledge Argument. This explains how the discussion has developed. But are these entailments correct? In particular, do the others really entail the zombie argument? ZA takes us from the philosophy of mind to the philosophy of modality, and one might well want to stay as clear of that as possible: or, to speak more carefully, be discriminative about those aspects of the issue that are strictly relevant to the mind-body problem, and those that more properly belong with issues on modality. For example, the proponent of the KA might want to resist the ZA because he thinks he can make his basic point without getting into the deep and muddy waters that surround issues of conceivability and the relation between metaphysical and logical possibility. Or he might hold something like Strawson's view of the person, which rules out zombies but would acknowledge the force of KA against physicalism. In neither case would CAT imply ZA. The general structure of the reservations about the connection between the knowledge argument and the conceivability of zombies could be put as follows. Let us assume that the KA works, in that it shows that there are qualia and their nature qua qualia - what it is like to have them - cannot be worked out from knowledge of physical states. Someone who thinks that the KA entails ZA thinks that the very fact that there is this qualitiative nature with no kind of a priori connection with the physical shows that there is a possible world in which the physical can occur without the quale. But someone might accept the KA, as establishing property dualism and still maintain either: 29 (a) Qualia have other essential properties than those they seem to have and these other properties could be physical, and the relation between the qualia and those physical properties is such that neither could, metaphysically, occur without the other. (b) Having a quale has other properties essentially than those given in introspection, and these could be physical and neither the state of having the quale nor those physical properties could, metaphysically, occur without the other. Whether there could be such an association between a quale, understood as something the nature, qua quale, of which Mary only came to know after she left the laboratory, and physical properties, is further question from that of whether there are qualia that instantiate nonphysical properties. Some philosophers hold, for example, that metaphysical necessity is constrained by natural necessity, so that natural necessities are held to be metaphysical necessities. A nomological connection between qualia or and brain states would then establish (a), but the KA would still succeed in showing property dualism. One might want to treat the correct account of metaphysical necessity as another matter. (b) is different from (a), in that it ties the connection to the act of having the state, not the content: but the same considerations arise. I mention it because it might seem more plausible than (a). A physicalist who accepts Mary but who wants to deny property dualism would have to affirm (c) Qualia have other properties essentially associated with them than those they seem to have, these could be physical, and the qualia could be nothing over and above these other properties: that is, the qualia could be identical to the physical properties. The attempt to combine (c) with accepting Mary's revelation, is Balog and Loar's position. (a) or (b) without (c) is a form of dual aspect theory, with metaphysically necessary supervenience: (c) alone is physicalism proper. This does, however, expose another reason for transfering the weight of argument from KA to ZA. One might argue that only the success of ZA would show that (c) cannot be true. Once, that is, there is any dispute about what KA really proves, (granted its success in showing that Mary learns something) only ZA can show that the conclusion is a 30 genuinely anti-physicalist one. To put it another way, only ZA can show that the content of Mary's new knowledge involves a new property and not just the exercise of a new concept. ZA is certainly, in context, a sufficient condition for making the strong interpretation of KA's conclusion, ruling out all of (a), (b) and (c). But it is not a logically necessary one and it is certainly not necessary for ruling out (c) and one might hope to defend that on other grounds, having no faith in showing anything conclusive in the philosophy of modality.