It Necessarily Ain't So

By Mark Johnston
Princeton University


1. You finally meet the love of your life. After an exciting series of dates, he decides to propose. There you are together at Lespinasse enjoying a simply ravishing meal. Just before the desert is served he produces a small box, the sort of box which could very well contain an engagement ring. You open it with anticipation, only to find inside a nugget of gold and some chimney soot. Your heart sinks. You knew he was a philosopher, but you realize with growing alarm exactly what this means as he offers the following speech:

You consider whether its worth pointing out that what you expected was not just Diamond, as if Carbonado, a sort of Diamond slag, might have done. You expected a cut diamond, a precious stone, and some Soot is definitely not a precious stone. But you glumly decide that you are being snowed, and that nothing could penetrate this mumbo-jumbo. You suspect that if you insisted on a cut diamond he might well produce a piece of chiseled coal. After all, since he thinks that Diamond = Soot, he probably also thinks that Diamond = Coal, and if Diamond = Coal then a diamond or a piece of Diamond is going to be hard for him to distinguish from a piece of Coal. Despite your steely grimace, your would-be husband continues:

You have had enough. As you leave him in the restaurant, you point to the check, saying "Please take care of that aggregate of tree molecules."


2. Your would-be husband is a bit of a nerd, but I believe his little sophism is quite instructive. The conclusion he aimed for, namely that Water Vapor is Snow and Diamond is Soot is of course absurd. However, the absurdity can be validly derived from (2), (3) and (4) below along with (1), the most commonly cited philosophical example of an identity linking a manifest kind, i.e. a kind whose instances we identify and re-identify merely on the basis of its manifest properties, with a chemical kind, i.e. a kind individuated by its perhaps recherche chemical properties.

(1) Water = H2O.

(3) If Water = H2O then Diamond = C (Carbon).

(4) If Diamond = C then Soot = C.

Together yield:

(5) Water Vapor = Snow and Diamond = Soot.

Why is (5) absurd? Well, Water Vapor is a kind of vapor whose instances - - fogs, mists, jets of steam, etc. -- are quantities of vapor, whereas Snow is a kind of powdery material whose instances are quantities of powder. Diamond is a kind of stuff whose instances are stones and lumps (of Diamond), whereas Soot is a kind of powdery material whose instances are quantities of (sooty) powder. If kinds are to turn out to be one and the same then so must their instances. But a quantity of powder cannot be identical with a fog or a mist or a jet, nor can a stone or lump be identical with a quantity of powder.

So why is the little sophism at all instructive? Surely we can account for the falsity of (2) and (3) in a way which does not reflect badly on (1). No; here I aim to show that this cannot be done. The right account of why the argument is a sophism requires that we understand the relation between manifest kinds like Water and chemical kinds like H2O as constitution rather than identity. It is (1) which is false.2


3. A first thought on behalf of the friends of (1) is this: Instances of chemical kinds come in different states or conditions; solids, liquids, gases, powders, plasmas, etc. On the other hand, some ordinary kind names like "Water Vapor", "Snow", "Diamond" and "Soot" are somehow more demanding as to the condition or state of the things in question. So we might say that Water Vapor is H2O in a vaporous condition, Snow is H2O in a powdery condition, Diamond is Carbon in a (largish-scale, clear, well-formed) crystalline condition and Soot is Carbon in a powdery condition. Water however can be understood as an inclusive kind; something can be an instance of that kind by being Water Vapor or by being Snow or by being Ice or by being Liquid Water.

Therefore (1) should be replaced by


(7) Water Vapor = H2O in a vaporous condition and Snow = H2O in a powdery condition.


(8) Diamond = C in a crystalline condition and Soot = C in a powdery condition.

From all of which, we hope, nothing untoward follows. This account suggests why (2) and (3) are bogus steps. They omit any reference to the respective states crucially associated with Water Vapor, Snow and Diamond.

But how are we to understand a phrase like "H2O in a vaporous condition?" In particular, is this true

(9) H2O = H2O in a vaporous condition.

If so, we have our reductio ad absurdum back again, for if (9) is true then so is

(10) H2O = H2O in a powdery condition.

Thus it follows that

Clearly we only advance if (9) is false. (I am inclined to think that (9) is true, at least on its most natural reading. After all holy water is water and Occupied Paris was Paris. But let us try to advance anyway.) What then is H2O in a vaporous condition, and what is its relation to H2O sans phrase?


4. Let us tailor the reading of "H2O in a vaporous condition" to suit. We want

(9) H2O = H2O in a vaporous condition.

to be false and

to be true. There is another constraint. Just as a diamond is essentially crystalline, a vapor is essentially vaporous. If a mist should liquefy or solidify the mist is no more, just as when a diamond is completely crushed so that its crystalline form is destroyed the diamond is no more. (See the remarks below about the yellow fog.)

Our tailored understanding of "H2O in a vaporous condition" thus needs to imply that if something is an instance of H2O in a vaporous condition then it is essentially in a vaporous condition. On this understanding of "H2O in a vaporous condition" (7) is plausibly taken to be true, so long as we treat "H2O in a powdery condition" in the same way. (9) is certainly false and indeed it now follows that Water Vapor is not identical with Snow. In the original sophism (2) and (3) were the bogus steps.

We now have what the friends of (1) want. The cost is supposing that quite generally there is a way of forming complex names out of simpler names and prepositional phrases where the condition described in the phrase enters into the essence of the thing picked out by the complex name.

In the sequel we shall see that this is a severe cost which in fact purchases nothing in the way of a principled resistance to kind constitution.


5. Now we may say something useful about the relation between H2O and H2O in a vaporous condition, so understood. Let us begin with this: Is an instance of the second kind, i.e., a particular vapor made up of H2O, accidentally or contingently an instance of that kind? Could that very vapor that particular fog or mist have been or come to be made up of say H2SO4? No, it seems that a particular vapor's chemical constitution is part of what makes it that very vapor. (This of course is entirely compatible with supposing that other vapors are made up of H2SO4 or, indeed, of the philosophers' compound XYZ.)

Next, and more important for what follows: Could that very vapor have been a solid or a liquid? I don't mean to ask whether the stuff that makes up the vapor comes in solid or liquid form. I know that it does. I mean to ask whether, say, T.S. Elliot's yellow fog which rubbed its back against the window pane could have been or could come to be identical with a block of ice, or a puddle of water. I don't mean to ask whether the fog could solidify into a block of ice or liquefy into a puddle of water. I know that it could. I mean to ask whether the fog could be either of these things. And it seems to me that the fog couldn't be either of these things. For when the fog solidifies or liquefies that is the end of the fog. A vapor like the fog is essentially a vapor. And this fact is not at all at odds with the fact that some liquid might be the result of a vapor liquefying and that some ice might be the result of a vapor solidifying. Iron oxide is the result of iron rusting, but it is not iron. A corpse is the result of killing a living human being, but it is not that living human being.

To summarize: the instances of H2O in a vaporous condition, i.e., the water vapors, are essentially instances of that kind. But whenever we have a water vapor we have a quantity of H2O, i.e., some instance of the kind H2O. But obviously the very same quantity of H2O say a particular mole of H2O --- can be first in the form of a vapor, then in the form of a liquid and finally in the form of a solid. So whereas a water vapor is essentially vaporous, the quantity of H2O which makes it up is not essentially vaporous. Therefore no water vapor is identical with the quantity of H2O which makes it up. The relation in question is material constitution, not numerical identity. When the instances of one kind, e.g H2O, in this way materially constitute the instances of another kind, e.g. H2O in a vaporous condition, I will mark the intimate relation between the kinds by saying that the first kind constitutes the second.


6. Where are we? The friend of

(1) Water = H2O

can only defend it by maintaining that there are kinds of stuff denoted by such designators as "H2O in a vaporous condition", kinds distinct from the kinds picked out by the simpler names they contain. Of such kinds they then must say things like

(12) H2O in a vaporous condition is not identical with but is constituted by H2O.

meaning that every instance of the first kind is constituted by but not identical with an instance of the second kind.

It is now entirely unclear why at the very beginning we shouldn't have captured the relevant fact of chemistry not by (1) but by

(13) Water is constituted by H2O.

Or better, given the facts about heavy water and tritium3:

Taking this rather than (1) to be the content of the relevant chemical discovery also avoids having to suppose that there are such distinct kinds as H2O at 72 F and H2O at 72.00003.. recurring F and so on and so forth. This endless multiplication of kinds is the consequence of the special construal of "H2O in a vaporous condition." forced on the friends of (1). For in order to defend that special construal they need to claim that quite generally there is a way of forming complex names out of simpler names and prepositional phrases where the condition described in the phrase enters into the essence of the thing picked out by the complex name.

Better to drop that special construal and treat the relevant designators as functioning like "Occupied Paris". Then we do not have a distinction between kinds for every distinction we can capture by any prepositional phrase. Accordingly, "H2O at 72 F" and "H2O at 72.00003..recurring F" both pick out the kind H2O. That is the other advantage of recognizing the constitution of kinds where others see only identity. We avoid an endless and implausible multiplication of kinds.4


7. At the end of section 5. I used a form of argument which is not respected in some quarters. I argued against identifying a particular vapor with the quantity of H2O which makes it up on the grounds that the vapor is essentially vaporous, e.g. ceases to exist when solidified into ice, while the same quantity of H2O could be first in a vaporous form and then in a solid form. Some philosophers attempt to resist this kind of argument by finding a disguised relativity in modal predication. A characteristic form of resistance would here amount to this: the vapor qua vapor is essentially vaporous, but so also is the H2O qua vapor. Elsewhere I have tried to discredit this whole line of thought.5 Here I simply wish to point out that the locus classicus for the claim that Water = H2O implicitly repudiates this line of thought. For in Naming and Necessity, Saul Kripke argues against the identity of a statue and the matter that makes it up on the grounds that the statue but not the matter is essentially statuesque.6

My point is simply that the very same consideration applies in the case of kinds of stuff. Kinds are identical only if they have just the same instances. Modal arguments precisely paralleling the argument about the statue and its matter apply to the instances of manifest kinds and the instances of the chemical kinds which make them up. Those arguments exhibit a distinction among instances which implies a distinction among the kinds they instantiate. As against the grain of Naming and Necessity, we should grant the same rights to kinds of material stuff as we do to kinds of material object. If a statue is not identical to the matter which makes it up then an instance of a manifest kind like Water is not to be identified with the instance of H2O which makes it up. Since kinds are identical only if their instances are, ordinary manifest kinds of stuff are not to be identified with underlying chemical kinds.


8. That would be a significant result even if it admitted of a few exceptions. In this spirit, some have suggested to me that Water is a special case. On their view, because we know Water in liquid, solid, vaporous and powdery forms we are when using "Water" picking out the kind which chemistry tells us could be in any one of these forms, i.e. H2O.7 "Water" carries with it no restriction as to the chemical state or condition water is in, so it can be outright identified with H2O. On this view, (1) is true, but not (2) or (3), so that the absurdities do not follow. For, while being an instance of the kind Water is not demanding with respect to form, nevertheless being an instance of the kind Snow requires being a powder, being an instance of the kind Water Vapor requires being a vapor, being an instance of the kind Diamond requires being a lump and being an instance of the kind Soot requires being a powder. Lamentably, according to this view, philosophers have presented the atypical case as central.

Though it concedes almost all I wish to say, this view is still too sanguine. What does the advocate of the view say about the relation among Water, Snow and Water Vapor? He had better not say that Snow = Water in a powdery condition and Water Vapor = Water in a vaporous condition. For, once again, how are we to understand the designators on the right hand sides of these identities? The dilemma, by now familiar is : Either we invoke the special construal in which case we push the bump to another point in the carpet, having to admit that Water merely constitutes Water in a vaporous condition or we face the absurd consequence that Water Vapor = Snow.

Neither will do. Therefore, on the view under discussion "Water" had better be a name for the stuff which merely constitutes Water Vapor and Snow. The view can now be restated as follows: We know Water as the stuff that constitutes or makes up Water Vapor and Snow and Ice and so on, hence there is no reason not to identify it with H2O, which chemistry tells us makes up these kinds.

Yet something embarrassing for this view is still hidden behind the sofa. How does Water so understood stand to the familiar liquid that fills lakes and rivers and in which we bathe. Does it merely constitute that liquid or is it identical with that liquid?

Here is a way of highlighting the embarrassment. It looks to me as though whatever reason there is for saying that a water vapor is essentially vaporous is a reason for saying that a watery liquid is essentially a liquid. After all, when a liquid is completely evaporated the liquid is no more. The liquid does not survive as the resultant vapor. 8 A particular liquid is essentially a liquid, just as a particular fog is essentially a vapor. So the advocate of the view that Water constitutes Water Vapor, Snow and Ice had better not say that Water = the kind Liquid Water (a.k.a. The Liquid, Water.) For then he should have to maintain that instances of Water are all of them liquid and essentially liquid.

Why would that be a bad thing to say? Well, this latter claim is in tension with the claim that Water constitutes Water Vapor. For together these claims entail that whenever we have, say, a jet of steam, the water which constitutes the steam is a liquid, which is clearly false. To avoid that consequence the view under discussion had better reduce to this: Water is a name for the stuff which merely constitutes Water Vapor, Snow, Ice and Liquid Water.

That is bizarre. Who would have thought that Water merely constitutes Liquid Water, so that none of their instances are identical. No, surely the truth is that every instance of Liquid Water is an instance of Water. This is why the view under discussion is too sanguine and is not a plausible way to save (1). It is H2O and not Water which merely constitutes Liquid Water. So H2O is not identical with Water.

The simple truth about the kind Water is that it is neither identical with nor does it constitute any of the kinds Snow, Ice, Water Vapor or Liquid Water. Rather, these are sub-kinds of the kind Water. This means that each instance of these kinds is an instance of the kind Water.

What about saying that each instance of Snow, Ice, Water Vapor or Liquid Water is an instance of H2O? No, that would be false. Instances of H2O are quantities of H2O. One of these quantities, e.g. a mole of H2O, can first be in a powdery, then a solid, then a liquid and finally a vaporous form. But that is not true of any of the instances of Snow, Ice, Water Vapor or Liquid Water. Their form is more essential to their being the particular powders, solids, vapors and liquids that they are. The right thing to say is that these kinds are constituted by H2O, i.e. their instances are constituted by quantities of H2O. They are not sub- kinds of H2O. For they share no instances with H2O.


10. There remains another apparent way to defend some version of the identification of Water with H2O, a way which although initially attractive, in fact leads nowhere other than to the conclusion that Water is not to be identified with H2O, but is only constituted by it.

Perhaps, so this initially attractive line of thought goes, in conceiving of the claim that Water is H2O as an identity statement, involving rigid designators for kinds Naming and Necessity was making a little too much of the analogy between identifications involving names for individuals and identifications involving names for kinds.

Accordingly, theoretical identities like Water is identical with H2O, Gold is identical with the element with Au, and so on might be rewritten in this form

(15) For any x (x is water if and only if x is H2O).

whereas the claim that Water Vapor is H2O can only mean

(16) For any x (if x is water vapor then x is H2O).

Likewise, the claim that Snow is H2O comes to this

(17) For any x (if x is snow then x is H2O).

From which nothing absurd about the relation of Water Vapor and Snow follows.

Obviously, since I think that Water, Water Vapor and Snow are merely constituted by H2O I agree with this as a way out of the problem. For if "x is H2O" simply means that x is constituted by H2O then (15), (16) and (17) are indeed true. (Forgetting about whether XYZ and so forth could be another kind which constitutes water.)

But equally obviously, so interpreted (15), (16) and (17) are much too weak to capture the force of the theoretical identifications of kinds in which Kripke was rightly so interested. Throughout the Third Lecture of Naming and Necessity Kripke is concerned with examples like Gold = Au and Water = H2O partly to bring discredit on the Identity Theory of Mind as exemplified in such claims as Pain = C-fibers firing. No Identity Theorist would have been satisfied with claiming that Pain is merely constituted by C-fibers firing.

Is there some stronger truth in the offing that preserves the import of the Identity Theory? Suppose we say

modeled on

Then clearly the relevant claims about Snow and Water Vapor will take the form of


from which nothing absurd about the relation between Snow and Water Vapor follows. That is why, in the original sophism, (2) and (3) were the bogus steps.

If this is the reader's preferred way out of the sophism with which we began then I apologize for making him sit on his hands until now. (I too hate to wait for what I consider to be the real point.) But I have an excuse. This is not a way to save the identification of Water with H2O, even when this is understood as amounting to (18). Claims (18), (20) and (21) are false and for reasons we have already canvassed.

Elliot's yellow fog was an instance of the kind Water Vapor. It was also made up of a quantity of H2O, an instance of the kind H2O. But contrary to (21) these instances are distinct. Suppose the fog suddenly liquefies. Then we still have the very same quantity of H2O as made up the fog, but now it is in a liquid form. (After all part of the point of talking of a chemical kind like H2O is to allow that particular quantities of it can first be in one state and then another.) Not so with the fog. When the fog liquefies the fog is no more, just as when the gold ring is rolled into a ball, the ring is no more. The fog is not identical with the quantity of H2O which makes it up and that is the relevant instance of the kind H2O here. So Eliot's fog is a counterexample which shows that (21) is false.9 Mutatis mutandis with Snow and (20).10

Now (18) can look for a moment as if it is immune to this criticism. For after all in (18) we are concerned with the inclusive kind Water, the kind which includes as sub-kinds Water Vapor and Snow and Ice and Liquid Water. And when the fog liquefies we still have some water around.

Let us look more closely. Something is an instance of the inclusive kind Water by being an instance either of the kind Water Vapor or of the kind Snow or of the kind Ice or of the kind Liquid Water. The inclusive kind Water is a determinable kind. Things get to be instances of it by being instances of its determinates. Compare Red: things get to be red by being some specific shade of red. In the case of red there couldn't be a red that is non-specific as to shade. But maybe the inclusive kind Water is here different, maybe one wants to allow among the instances of the kind Water, quantities of Water; quantities which can be first in a solid form then in a liquid form then in vaporous form and finally in a powdery form. Fine, but it would be unmotivated to exclude particular fogs as also instances of the inclusive kind Water. And if they are instances of the kind Water then they are counterexamples to (18).

And if they are not? What if we stipulate that the inclusive kind Water has as instances only quantities of water, only things which have no essential tie to any particular manifest state or condition. Grand, now (18) is true, but neither Snow nor Water Vapor nor Ice nor Liquid Water are sub-kinds of the "inclusive" kind Water. For none of their instances are instances of that kind. That means the name "Water" so understood has then been stipulated to be just a name for H2O, whose instances are all likewise only mere quantities. If we take refuge in this stipulation then no thesis is then being offered about the relation between manifest kinds like Water and chemical kinds like H2O.


11. One source of the urge to identify Water with H2O comes from the idea that the designator "Water" functions semantically as if it were explicitly introduced to designate the underlying chemical kind found in demonstrated samples which may include a certain liquid, a certain vapor, a certain powder and a certain solid. For this "as if" story to be apposite it had better not burden ordinary users of "Water" with concepts which none of them nor their "experts" could yet have.

Obviously, as it stands the "as if" story is not apposite. The introduction of the designator "Water" in this way is a performance possible only at a comparatively late stage of human history, since the notion of a chemical kind, a kind individuated by its chemical properties, probably postdates alchemy and may even require some anticipation of the classification of elements found in the Periodic Table. Therefore the semantic account associated with that model of introducing "Water" can hardly be the right account of the meaning of ancient ordinary language names for the manifest stuff which we drink, bathe in and use to irrigate land. In abstracting away from the specific conditions which we find water in and fixating directly on the chemical kind, the semantic account makes "Water" too much like "Californium" or "Einsteinium," names explicitly introduced to denote chemical kinds.


12. There are intermediate cases, cases in which the ancients may be held to have deferred to something which turned out to be modern chemistry. Indeed, "Gold" can look like such a case. A longstanding human anxiety concerns the assaying of the purity and authenticity of alleged samples of Gold. Chemistry gives a definitive account of what distinguishes real Gold from Fool's Gold, and of what Gold's purity amounts to. So here, at least, there is more justice in seeing chemistry as filling out a well recognized gap in our knowledge of how to tell real Gold and thereby showing us what it is to be identical with real Gold.

However, the case for the identity theory of Gold (Gold = Au) is not fully convincing. The ancient interest in the purity and authenticity of alleged samples of Gold was not a brute or non-derivative interest in its underlying chemical structure whatever that might turn out to be, but rather an interest in Gold's enduring manifest properties of malleability, splendor, resistance to corrosion and so on; properties which are not enduring properties of Fool's Gold. Compare Diamond and its manifest properties of hardness, crystalline form and decorative appeal. Even though we are concerned to distinguish real diamonds from fakes and defective stones this does not establish that Diamond = C, for some of the things we wish to set aside as defective or fake are forms of Ballas and Bort, which are also made up of Carbon.

Our interest in Diamond is an interest in a manifest kind. The manifest form of Diamond is crucial to being Diamond. I take something like this to be so in the case of Gold. For if by "golden" we mean something with the manifest form of Gold (which may of course be temporarily hidden by dirt or other encrustation) then, while for the ancient goldsmiths it was probably necessary that

(22) Any quantity of Gold is golden

it is certainly not necessary that

(23) Any quantity of Au is golden.

For single atoms of Au have no manifest form or at very least, if electron microscopes make manifest their form, not a form which is in any way golden. They could not, for example, be splendid or malleable.

Better then to say that

(24) Gold is Au in a sizable lattice structure

and resist the thought that we have here the "is" of identity by asking whether "Au in a sizeable lattice structure" is to be understood so that

(25) Au = Au in a sizeable lattice structure

is true. If (25) is true then treating (24) as an identity statement runs foul of the fact that (22) and (23) have a different status. Alternatively, if (25) is thus false, so that Au in a sizeable lattice structure is only constituted by Au, then we are without any motive for resisting the claim that

(19) Gold is not identical with but is constituted by Au.

As Bishop Butler might have put it: everything, including Gold, is itself and not another thing.


13. One side effect of chemistry telling us more and more about the chemical properties of chemical kinds, is that chemistry shows us which chemical kinds constitute manifest kinds. It does this by showing how the chemical kinds in question account for the causal powers of (instances of) the manifest kinds in question.

There is still room to allow that in special cases a chemical kind could just be a manifest kind. After all, we could imagine that some chemists hypothesize the existence of Kripkenium, a transuranic element with atomic number 290. Thanks to its theorized atomic structure they are, we may suppose, led to speculate that Kripkenium could be stable for an hour or so under very cold conditions. Centuries later middle-sized lumps of Kripkenium are discovered on the surface of Neptune. The name `Kripkenium' was introduced solely to denote a chemical kind, and this very chemical kind turns out to be "observed in nature" much later on. Likewise, it might have been that in the case of Polyurethane a hypothesized molecule with a certain chemical structure first received the name, and then sheets of it were made.

This is for me a bit of a "don't care" but I still find reason to resist. Even if all the above were so, it will still be true that the kind Lump of Kripkenium is constituted by the kind Kripkenium, and that the kind Sheet of Polyurethane is constituted by the kind Polyurethane. For even here what is manifest to us are not mere quantities of this or that chemical kind, as it might be moles or half moles or atoms or molecules, but rather lumps or sheets made of a quantity of the relevant chemical kind. The instances of chemical kinds are mere quantities, quantities capable of being in whatever state the chemical kind can be in, whereas the instances of manifest kinds are more intimately tied to the identity conditions of the lumps or sheets or bodies or powders or vapors that those instances come in.

Here is a speculative consideration in favor of that claim about Kripkenium and Polyurethane, a consideration offered as a stimulus for further debate rather than as a crucial resting point. Consider a single atom of Kripkenium or a single molecule of Polyurethane. These are quantities of, and so instances of, the relevant chemical kinds. But they are not instances of any manifest kind, for they have no manifest features. 11 So, since kinds are identical only if their instances are, then Kripkenium and Polyurethane, as opposed to the kinds Lump of Kripkenium and Sheet of Polyurethane, are not manifest kinds. 12

In any case, however these special cases turn out, I should sum up like this: The scientific image of the world is an image of how manifest kinds and their manifest instances are constituted so as to have the efficient causal powers which they do in fact have. Manifest kinds are thus satisfactorily located in a causal- explanatory view of the world. But in general they are not to be "reduced" by way of identification with chemical kinds. Their manifest form is at least as important to being what they are as is the matter which makes them up.


Appendix For NYU Seminar: On The Explanatory Gap

Saul Kripke was the first to suggest that there is an important asymmetry between theoretical reductions in science, so called theoretical identifications such as Water = H2O, and such Physicalist identifications as Pain = C-Fibers firing. Frank Jackson and David Chalmers have recently made much of this asymmetry. To put their point in its simplest form, when it comes to Water and H2O then an appeal to empirical discoveries and Occam's razor provides premise (1) while the a priori semantic account of "Water" as a natural kind term yields (2)

(1) At an appropriately basic physical level H2O is the unique explainer of the causal powers characteristic of Water.

(2) If at an appropriately basic physical level H2O is the unique explainer of the causal powers charaterisitic of Water then Water = H2O.

(Think of Kripke's natural kind semantics as embodying the following "as if" story: "Water" was introduced to denote the kind which explains, at an appropriately basic physical level, the causal powers of these -- as it turns out Watery -- samples.)

Now if (2) is true then it is true a priori. So (1) a discovery of physical chemistry strengthened by an appeal to simplicity or Occam's razor yields as an a priori matter

(3) Water = H20.

That is to say that it is not conceivable that (1) is true and (3) is false. (This is a more subtle version of what we get from what Colin McGinn calls the Calm structure.) There is really no explanatory gap at all between (1) and (3), for it is a priori that if (1) then (3).

What about the corresponding argument in the case of a conscious state such as Pain? Perhaps neurophysiology along with simplicity considerations yield

The counterpart of (2) is (2')

Together they entail:

(3') Pain = The firing of C-fibers.

However it is not a priori that if (1') then (3'). For (2') is not a priori.
"Pain" is not a natural kind term. There cannot be "fool's pain"; anything that manifestly seems like pain is pain. Here is the alleged explanatory gap: even if it is true that if (1') then (3') we cannot "see" how is can be true in the way we can see, i.e. derive a priori, that if (1) then (3). That is why peculiarly philosophical fears about "Zombies" (who have their C-fibers firing but no pain) live on. They cannot be ruled out a priori. Alex Byrne independently made the same point in his Princeton Ph.D. thesis (1993).

My first observation is this: That cannot be the right account of the explanatory gap, for both (2) and (2') are in fact false. At most H2O constitutes Water and the firing of C-fibers constitutes Pain. (Recall that this means: in the relevant class of cases every instance of the second kind is constituted by an instance of the first kind.)

The truths that replace (2) and (2') are

My second observation is this: both of these truths are plausibly taken as a priori, i.e. they follow from what we should mean by "material constitution". The constituting matter of a thing is what (at an appropriately basic physical level) accounts for its characteristic causal powers. There is no explanatory asymmetry when we have the right explananda in mind, viz. that Water is constituted by H2O and Pain is constituted by the firing of C-fibers.

My third observation is meant to relocate the explanatory gap and indeed to generalize it. It is one thing to exhaustively explain the causal powers of a manifest kind. This can be done and when it is done it establishes the relevant constitution claim: the underlying explaining kind constitutes the manifest kind. It is quite another thing to identify the qualities of the manifest kind with quantitative properties at the explaining level, or to give an a priori derivation of the qualities from the quantities. Except for some very special cases, this can never be done. (I think it cannot be done for manifest shape, nor for slipperyness, nor for redness, nor for wetness.) The manifest qualities supervene on the quantities, but the supervenience is not an a priori matter. What some call the hard part of the mind-body problem is perfectly general and has nothing much to do with special features of mind and body.

Generalizing the mind-body problem in this way may simply increase the ontological anxieties of some thinkers. Speaking for myself, it makes me calmer about the whole issue. I think "That is just the way nature is."

There remains a remnant of the old mind-body problem: The appearing to subjects of manifest objects and their manifest qualities can seem very special. Yet the same ontological structure applies to appearings: they are constituted by but not reducible to patterns of causation among things. The facts about them are fixed by the facts about the patterns. Yet there is no a priori derivation of the appearings from the patterns. One may say with some justice that the existence of appearings in a world of things is a mystery.

It is a mystery, but it is not an insult to any clearheaded explanatory ambition.13 We cannot account reductively for the sheer intelligibility of things. We can only inhabit it.




1. See Saul A. Kripke Naming and Necessity (Harvard University Press, 1980) pp 99, 116,
128-9, 134, 148, 150. A characteristic remark is found on pg 128 "Let's consider how this applies to the types of identity statements expressing scientific discoveries that I talked about before -- say that water is H2O."

2. Some may already think that (1) is a misleading way of expressing a theoretical identification of Water and H2O. If you think that feel free to take a peek at section 10 where another way of taking this identification is found wanting..

3. In 1932 the chemist Harold Clayton Urey discovered that water contains a small amount (1 part in 6000) of so-called heavy water or deuterium oxide (D2O); Deuterium is the isotope of hydrogen with an atomic weight of 2. In 1951 Aristid Grosse discovered that naturally occurring water also contains minute traces of tritium oxide; tritium being the isotope of hydrogen with an atomic weight of 3. Most manifest kinds are mixtures.

4. Compare the argument of my "Constitution is not Identity" Mind, 1992, where it is shown that those who identify a ring with the gold that makes it up will find it hard to resist an implausible result, viz. an endless multiplication of overlapping rings.

5. ibid.

6. See Naming and Necessity footnote 74 and the discussion of persons and their bodies in the surrounding text.

7. David Hilbert urged this.

8. What happens if after evaporating a liquid we then liquify the resulting vapor. I am not sure which of two things to say. One is: we recover the same quantity of liquid but not the very same liquid. The other is that instances of kinds of material stuff can exhibit intermittent existence. Hence we recover the very same liquid. Mutatis mutandis for solidifying a liquid and then melting it. The second seems to me slightly more plausible. Compare this: any high end stereo system is plausibly taken to be capable of intermittent existence, for if it doesn't exist until its assembled from the myriad of components out there in Hi-Fi land then it probably doesn't exist when it is disassembled for transporting or for canibalizing for other systems. If after a round of canibalizing of components I get back my original components and reassemble them, then that is my stereo system once again.

In any case, one thing which is really implausible is to identify some liquid with the solid ice it yields and the vapor it came from. For then we should have no choice but to identify each of these with a quantity of the underlying chemical kind. And a year ago that quantity could have been scattered throughout the Pacific Ocean and the Black Sea. It follows that the ice existed then. Yet this is clearly wrong. \

9. No one should want to say that there are here two instances of the kind H2O which happen to be exactly co-incident at least before the liquefaction, i.e. the fog and the quantity of fog which makes it up. Or if they do they should recognize that they are already in the constitution business: on their view, oddly enough, constitution holds among instances of the very same kind, viz. H2O. Better to say that the fog is not an extra instance of the kind H2O.

10. Why is (19) false? This is not the place to elaborate the point but there can be instances of the state Firing of C-fibers in petri dishes where there are presumably no instances of the state Pain. Only a much more extensive brain state could be a real candidate for theoretical identification with Pain. But even then I think we have constitution rather than identity.

11. Why should every instance of a manifest kind be manifest, i.e. able to be recognized and re-identified on the basis of manifest features? The case to consider is a case where we have two kinds K and K', where K has only manifest instances and K' includes these and still other instances not themselves manifest. I say that K is the manifest kind and is a sub-kind of K', which is not itself a manifest kind. So far this is just stipulative. However in a larger manuscript entitled The Manifest I argue that requiring that all the instances of a manifest kind or quality be manifest gives a satisfactory and indispensable restriction on the dissectiveness of manifest kinds and qualities.

To put it all too briefly, the manifest quality red found on surfaces is often said to present itself as dissective, i.e , such that every part of the red surface is red. At least this is the common and somewhat plausible philosophical interpretation of the manifest homogeneity of redness. However given Scientific Realism it immediately follows that nothing is red since there are small parts of any allegedly red surface which are not red. Intolerable! The way out is to note that since homogeneity is a manifest property of redness it can have only manifest instances, so that the "insensible parts" of a red surface need not themselves be red. In this way the requirement that every instance of a manifest quality be itself manifest does necessary explanatory work. Similar explanatory work is done by the corresponding constraint on kinds, e.g. the kind Red Thing.

12. Suppose that as matter of chemical law Kripkenium had to come in a middle-sized lump. Even then we should have to distinguish the kinds Kripkenium and Lump of Kripkenium. For instances of the former kind would include physically undetachable atomic parts which are not instances of the latter kind.

13. That is to say that this is not one of McGinn's superficial mysteries, a mystery only due to the fact that we as a species lack the intellectual talent required to come up with the reduction.