identity criteria, telling us where one thing of the given sort is to be counted as leaving off and perhaps another counted as beginning.
To anyone familiar with the names, Catholicism denotes an institution, but Saint Peter’s Basilica an edifice. The identity criteria for institutions and for edifices, telling us how much one can be reformed or remodeled and still itself remain there, and not a replacement, may not be entirely clear. But surely it is at least clear enough that reduction to a pile of rubble is enough to make for the destruction of an edifice, while survival of all the people involved with beliefs and practices unchanged is enough to make for the continued existence of an institution.
But now comes the important point. Arguably, once we have these judgments about classification and criteria in place, the modal claim, and with it the bomb scenario, may become superfluous. Isn’t the fact that Catholicism and Saint Peter’s Basilica are things with very different identity criteria, and therefore very different sorts of things, already enough by itself to establish that they are very different things? If so, any assertion to the effect that they are quite literally one and the very same thing must be a piece of nonsense.
Now the neuralist holds that being in pain, as a type, is the same as having C fibers that are firing, and that my being in pain now, as a token, is the same as my having C fibers that are firing now. Let me concentrate on the tokens, leaving the reader to think through the case of the types. To the dualist, it seems clear that my being in pain now is my having a feeling now, and that one doesn’t understand what the word pain means unless one understands this, and also that my having C fibers that are firing now is my undergoing an electrochemical process now, and that one doesn’t understand what the phrase C fiber firing means unless one understands this. These things are as clear as or clearer than that Catholicism is an institution and that Saint Peter’s is an edifice.
To the dualist, it seems clear that the identity criteria for tokens of being in pain involve who is having them, and when, and what it is like to have them — are they dull or sharp? steady or intermittent? to what body-part, real or phantom, are they referred? — but nothing about physical particles. Likewise it seems clear that the identity criteria for tokens of undergoing C fiber firings involve the motions of physical particles down to individual electrons, and nothing about what anything feels like. Indeed, to the dualist these claims seem clearer than any claims one might make about the identity criteria for institutions or for edifices.
To the kind of dualist I am imagining, the conclusion seems inevitable that since the identity criteria are radically different, the things are of radically different sorts, hence are radically different things, so that the neuralist’s assertion that they are one and the same thing is absurd, like a childish confusion of church in the sense of religious denomination with church in the sense of house of worship.
IV Conceivability versus Possibility
The indirect attack on the supervenience question by a direct attack on the mind-body problem fails, or anyhow bogs down, obliging us to take a more direct approach to the issue of the possibility of a zombie world.
Sorts and Identity Criteria
The principle that different identity criteria make for different things will, to dualists who accept it, make short work of either physicalistic view in philosophy of mind. For what has just been said in the preceding chapter against neuralism would apply a fortiori against the functionalist claim that to be in pain is to be in a second-order state, since though it may be obscure what are the identity criteria for being in a second-order state, it seems clear enough that they are if anything even farther away than are the identity criteria for undergoing electrochemical processes from the identity criteria for feeling sensations.
In the background here is a view including such components as these:
(1) Things come in sorts.
(2) Sorts come with identity criteria.
(3) For things of the same sort, the identity criteria for things of that sort determine whether they are — or count as — the same.
(4) Things of different sorts never are — or count as — the same.
This view is not unquestioned in philosophy, but it can be found invoked in areas far from philosophy of mind, so at any rate it is not an ad hocinvention of a certain kind of dualist.
As I have already said earlier, this book is not the place for any extended excursion into philosophy of mathematics, but it may be mentioned that a notable example of appeal to a view about identity criteria similar to that just expressed in (1)-(4) is found in Crispin Wright’s Frege’s Conception of Numbers as Objects, put forward as a solution to the so-called Julius Caesar problem, the answer to Gottlob Frege’s question how we know that the number two is not Julius Caesar. It is because the identity criteria for numbers, as set forth by Frege in his Grundlagen der Arithmetik and endorsed by Wright, are so very different from those for human beings, Caesar included, that a human being can never count as a number.
Needless to say, antidualists of either kind, adherents of neuralism and functionalism alike, can be expected to reject the whole foregoing line of thought. But is there any debatable or discussable issue here? The situation is very largely one of the dualist in effect saying that the antidualists have misunderstood the very meaning of the word pain and for that matter of the word identity as well, and of the antidualists maintaining that, no, they haven’t. And we seem to see here little more than a mere slinging back and forth of intuitions.
Back to Supervenience
This brings me back to my official topic, the supervenience thesis, which has been left pretty much in the background for some time now. One reason for interest in the supervenience thesis is precisely that it does not seem to provoke in the dualist the reaction that I have been describing in the case of identity claims: the reaction of immediate dismissal as an absurdity. To claim identity between items with different identity criteria is one thing, merely to claim necessary correlation is another. Supervenience therefore seems to promise us a more debatable and discussable issue to replace mere intuition-slinging. The modal turn seems to offer a way around an impasse. How far it really does so remains to be examined.
Let us then set aside the issue of identity criteria and turn from the attempt to settle the issue of supervenience indirectly, by a direct attack on the mind-body problem, to the converse project of a direct attack on the supervenience question, which might indirectly settle the mind-body problem, at least in part. For a refutation of supervenience would amount to a refutation of physicalism and a proof of dualism, while a proof of supervenience would amount to a refutation of dualism and a proof of physicalism, though the mere statement of the thesis of supervenience in itself does not favor one form of physicalism over another.
I said earlier that I am ultimately going to question the interest and importance of the supervenience issue. I will ultimately express three different grounds for suspicion on this point, and I can now say that the first suspicion I will express is that the modal turn will not, after all and in the end, give us a way around an impasse, but rather will prove to be a pointless detour, because in order to settle the supervenience issue, we will need to come back to the very issues about identity criteria that I have just temporarily set aside. But such issues will reemerge only slowly, and there will be a good deal else to discuss before we can usefully come back to them.
A Trio of Notions of Necessity
The supervenience thesis involves three concepts: the mental, the physical, and the modal. Something has been said about the scope and limits of the mental in my earliest chapter. Something will be said about the scope and limits of the physical in a later chapter. I will devote the present section of the present chapter to some discussion of various flavors of modality, in order to distinguish the one involved in the issue before us.
It will be well to begin with what has been said about the matter of different varieties of modality by linguists without ulterior philosophical motives. I rely largely on the Cambridge Textbooks in Linguistic volume on Mood and Modality, by Frank R. Palmer. The category of mood and modality refers to the distinction between what is affirmed as actual or real and what is not, insofar as that distinction is grammaticalized, as one says: represented in language by either modifications of the main verb of a clause, or the use of special auxiliary verbs. The term grammatical mood in a narrow sense refers to the first of these means of expression, and in a broad sense to both.
As for distinctions in verb form, that between what are called the indicative and subjunctive moods, which still flourishes in related languages such as German, survives only vestigially in English. As for auxiliary verbs, in English those used to express mood or modality overlap with those used to express tense or temporality. They consist of the pairs will and would, shall and should, may and might, can and could, plus a few others. Notoriously, there are differences as regards the first two pairs between usage in the United States and usage in South Britain, with further variation in other Anglophone regions; but fortunately these differences seem to relate more to temporal than to modal uses.
The various flavors of modality recognized by linguists are expressed by very different means in some world languages, so there is no question but that a conceptual distinction can be drawn; but in English the same auxiliaries, and especially the pair consisting of must for necessity and may for possibility, do duty for all of them. Palmer recognizes a trio of flavors of modality that he calls deontic and epistemic and dynamic.
Deontic modality is represented by the contrast between she may go and he must stay meaning she is permitted to go and he is obliged to stay. Deontic necessity and possibility thus amount to obligation and permission. This flavor is not what will be at issue for us.
Epistemic modality is represented by the contrast between she must have seen it and he may have seenit,meaning given what is known, she must have seen it and for all that is known, he may have seen it. As knowledge varies from person to person and time to time, so does epistemic modality. The epistemic possibility of an afterlife — our not knowing for sure that there isn’t one — is the starting point for Blaise Pascal in his famous wager. It is also as far as Descartes gets in his Second Meditation. But it is far short of Descartes’ ultimate goal: this flavor is again not what will be at issue for us.
Dynamic modality, the remaining flavor, pertains to what I or you or he or she or they can do or could have done, meaning am/are able to do or was/were able to do. For our purposes what will be of most interest is an impersonal variant: what it could have happened that, or what it could have been the case that. Kripke, uses possibility without distinguishing adjective — simpliciter, tout court, sans phrase — for this kind of possibility, what could have been, or more precisely, what is or isn’t but could have been,and necessity without distinguishing adjective for the correlative notion what is and couldn’t have failed to be.
This omission of any distinguishing adjective is inconvenient when so many others are using possibility and necessity in so many other senses, and inspired by an off-hand remark of Kripke’s, in the literature the distinguishing adjective metaphysical is commonly applied to these modalities. Here I will let metaphysical be tacitly understood, as the default case, when no other distinguishing adjective appears, but will sometimes add it explicitly for emphasis or to avoid ambiguity.
A Trio of Notions of Firm Truth
So much for the deontic and epistemic and metaphysical flavors or modality. Turning fully from linguistics to philosophy now, there is a traditional trio of notions to be considered there as well, of which [metaphysical] necessity is again one. For as W. V. O. Quine, a notorious archenemy of modal thinking, writes in his best-known book:
[P]hilosophical tradition hints of three nested categories of firm truth: the analytic, the apriori, and the necessary. Whether the first exhausts the second, and the second the third, are traditional matters of disagreement [Word and Object, 59].
It is a commonplace that the three notions are of different character. Necessary as opposed to contingent truth, in the sense intended, has as noted been called metaphysical, pertaining to being. Aprioriness or apriori truth as opposed to aposteriori truthmay be called epistemological, pertaining to knowledge. Analyticity or analytic truth as opposed to synthetic truth may be called semantical, pertaining to meaning. While the epistemically possible was characterized earlier as what may be for all that is known, the apriori possible or epistemologically possible may be characterized as what may be for all that is knowable independently of sense-experience, and the analytically possible or semantically possible may be characterized as what may be for all that is knowable just from the meanings of words.
As to matters of traditional disagreement, as one moves from Leibniz to Kant to Frege to Rudolf Carnap and P. F. Strawson and other contemporaries of Quine, there is a tendency for the [metaphysically] necessary to be conflated with the apriori, and then the apriori with the analytic, explained in terms of so-called semantic rules or linguistic conventions. Quine himself famously attempted to persuade his colleagues do away even with the analytic-synthetic distinction, leaving none of the traditional categories standing.
Quine did not persuade many, and on the contrary, one distinction formerly collapsed has since Quine’s heyday come to be restored to respectability, largely by examples of so-called aposteriori necessities in Kripke’s 1970 Princeton lectures entitled Naming and Necessity. By contrast, there has been less tendency to revive the notion of the synthetic apriori.
This leaves us with two degrees of firm truth, the [metaphysically] necessary on the one hand, and the analytic or apriori — not distinguishing the two — on the other hand. A notion of firm truth brings with it a correlative soft notion of not being firmly untrue, and from some points of view it is more illuminating to start with the soft rather than the firm notions.
A Quartet of Soft Notions
In fact, we will have to take at least brief note of no fewer than four soft categories. Let me first list them, then explain and exemplify them. I will give each a one-word label, among which possible will be reserved for the metaphysical notion, along with an adverb that may be added for emphasis when contrasting one of the four with another; but I will also give each a two-word label in which the second word will be possible.
superficially supposable = suppositionally possible
coherently conceivable = conceptually possible
genuinely possible = metaphysically possible
outright noncontranomic = nomologically possible
The superficially supposable or suppositionally possible is that which, unlike a round square or married bachelor, we can at least begin to consider without encountering anything obviously analytically or apriori false. The coherently conceivable or conceptually possible is that whose falsehood is not analytic or apriori, not even unobviously so. The genuinely possible or metaphysically possible is what could have been, that whose falsehood is not a [metaphysical] necessity, not even one discoverable only aposteriori and synthetically. The outright noncontranomic or nomologically possible is what could have been without violation of the laws of nature.
An example of something supposable but not conceivable would be anything that can be ruled out by analytic or apriori considerations, but only unobvious ones, such as a definition arrived at only after a long and difficult dialectic of analysis, or a theorem arrived at only after a long and difficult deductive proof. A case would be a counterexample to the result conjectured by Descartes’ contemporary Pierre de Fermat but only proved by our contemporary Andrew Wiles. The existence of a counterexample is not coherently conceivable because of Wiles’s proof, but is superficially supposable because the proof is long and difficult.
We can begin, but only begin, to picture at least the external aspects of what it would be like for such a counterexample to exist. The mathematical gadfly Doron Zeilberger once announced on his blog that just such a counterexample had been found — supposedly involving numbers so large they took several supercomputers to store — and complained about journal editors who were refusing to publish an announcement about the discovery. Reportedly a number of readers failed to note that the date of the posting was April 1, and were taken in for a time. One can picture a play or movie inspired by the blog post, on the order of David Auburn’s Proof. The movie would show the excitement and consternation in the mathematical community, and perhaps bits and pieces of the calculation. What could not be shown would be a complete calculation, since to show that would amount to showing a counterexample, and there is none.
Examples of things conceivable but not possible are provided by Kripke’s examples of aposteriori necessity, which we will be examining soon enough. More precisely, if p is necessary but aposteriori, not p is impossible but conceivable, while if q is impossible but conceivable, not q is necessary but not knowable apriori, and so either knowable but only aposteriori or else unknowable.
As I have already said twice earlier, this book is not the place for any extended excursion into philosophy of mathematics, but mathematics does provide a putative case. It is often held, on the one hand, that mathematical statements are necessary if and only if they are true, and on the other hand, that they are apriori if and only if they are provable. Work of Kurt Gödel, Alfred Tarski, and others on the relationship of truth to proof in mathematics has suggested to many that there may be truths that are not provable — not merely unprovable from or relative to this or that accepted system of axioms, but in some sense absolutely unprovable. Such an unprovable mathematical truth would be a necessary truth not knowable apriori. It might simply be unknowable, or it might instead be well supported by computational verification of special cases and various kinds of heuristic argumentation, and so arguably knowable aposteriori.
It is even sometimes suggested that Goldbach’s conjecture — to the effect that any even number greater than four is a sum of two primes — may be an example of an aposteriori necessity of this kind. For illustrative purposes, let me pretend it is. This is not one Kripke’s primary examples, but he briefly mentions it in passing in Naming and Necessity as a case of a kind about which there had been some amount of discussion even prior to his own intervention.
Examples of things possible but not noncontranomic are arguably provided by many science-fiction fantasies, such as gravity shields or faster-than-light spaceships. A zombie world would be another case according to dualists, who claim it is possible, but accept that it would be contrary to the laws of nature.
The four possibility-like notions come with four correlative necessity-like notions, a quartet of nested categories of firm truth to replace Quine’s trio, as well as with four associated implication notions, all defined as follows:
A little thought shows that conceptual necessity works out to be analyticity or aprioriness, also called — as if there weren’t enough terminological clutter here already — conceptual truth. The phrases nomologically impossible and nomologically necessary may be abbreviated to contranomic and nomic.
This has been a lot of jargon, but far from all the jargon to be met with in reading around on our topics. An unlimited number of other firm necessity-like and soft possibility-like notions can be found mentioned in the literature, which fortunately there will be no need to consider here. (In particular, the labels “logical necessity” and “logical possibility” have been very widely and very loosely and very diversely used in the literature, so much so as to make them at present quite serious impediments to understanding. I will avoid them.)
The Status of the Zombie World
Four questions now arise about the existence of zombies, or rather, about the zombie world hypothesis. Let me list them and what I take to be the answers offered by dualists and neuralists and functionalists.
(1) Is it superficially supposable, like Fermat counterexamples?
Dualist: Yes Neuralist: Yes Functionalist: Yes
(2) Is it coherently conceivable, like Goldbach counterexamples?
Dualist: Yes Neuralist: Yes Functionalist: No
(3) Is it genuinely possible:
could it have existed even if only by violation of the laws of nature?
Dualist: Yes Neuralist: No Functionalist: No
(4) Is it outright noncontranomic:
could it have existed without any violation of the laws of nature?
Dualist: No Neuralist: No Functionalist: No
I will not say anything more about (1) and (4), on which there appears to be agreement; they are significant only as foils, for contrast with the disputed questions. I will devote the remainder of this section and the next section to question (2), whether the zombie world is conceivable. I will thereafter assume that it is, as maintained by dualists and neuralists, and begin in the section after next an extended discussion of question (3), whether the zombie world is possible.
Arguments in the literature against the conceivability of zombies include two kinds, one of which I will quickly dismiss. By definition, a zombie world would contain an exact duplicate of each physical object, including each human body, to be found in the world as it actually is. Therefore there would be for each of us something in the zombie world moving around and emitting sounds just as we do, and therefore moving around and emitting sounds just as if it were a conscious subject with beliefs of various kinds. What I am calling a hardcore dualist will insist that though there undeniably would be beings with beliefs in this as if or metaphorical sense, there would be no beings with beliefs in a literal sense. Some arguments against the conceivability of zombies begin with the contrary assumption that zombies would indeed have beliefs and knowledge, the same as we do in the world as it actually is. Such arguments a hardcore dualist will dismiss out of hand and at once.
Their very existence is a warning. It seems that whenever we get drawn too far away from sensation and affect, from red looking like or anger feeling like THIS, and allow belief (or more generally “intentionality”) to be brought into the argument, our attention is being slyly misdirected, so that some materialist sleight of hand can make what is truly distinctive about the mental, conscious experience, disappear. We have seen an instance of this kind of legerdemain already in talk about itches as representations.
But I have to admit that though the arguments I am dismissing are not cogent ad rem, as the traditional phrase has it, not cogent as directed to the issue, they may be cogent ad hominem, as the companion phrase has it, cogent as directed to certain opponents, namely, those I am calling softcore dualists. These, while rejecting any physicalistic account of sensation and affect, incautiously concede some kind of physicalistic account — perhaps neuralist but more likely functionalist — of belief. More serious consideration, however, is due to those functionalist arguments that do not rely on the softcore dualists’ incautious concessions.
An Inconceivability Argument Reconstructed
Let me try to reconstruct what seems to me the main such argument. Recall now the definitions of , , , associated respectively with pain, its neural correlate, its functional role, and the second-order state of being in a first-order state that realizes that functional role, so that functionalism maintains that is , neuralism maintains that is , and dualism maintains that is neither.