The Meaning of “Water”: An Unsolved Problem

William G. Lycan

University of North Carolina and EastCarolinaUniversity

WATER. …I. The liquid of which seas, lakes, and rivers are composed, and which falls as rain and issues from springs.When pure, it is transparent, colourless (except as seen in large quantity, when it has a blue tint), tasteless, and inodorous.

--Oxford English Dictionary

…the fact that an English speaker in 1750 might have called XYZ ‘water,’ whereas he or his successors would not have called XYZ water in 1800 or 1850 does not mean that the ‘meaning’ of ‘water’ changed for the average speaker in the interval.In 1750 or in 1850 or in 1950 one might have pointed to, say, the liquid in Lake Michigan as an example of ‘water.’What changed was that in 1750 we would have mistakenly thought that XYZ bore the relation sameL to the liquid in Lake Michigan, whereas in 1800 or 1850 we would have known that it did not.

--Hilary Putnam[1]

1.Over thirty years ago, we all learned from Kripke and Putnam that the extension of a natural-kind term such as “water” is determined by the relevant natural kind—in “water”’s case, the chemical kind, H2O.And, if less obviously, so is that term’s intension determined by that kind: “Water” designates H2O in every possible world in which water exists.(Hence, necessarily, if water exists, water = H2O.)

Putnamian natural-kind externalists refer us to the planet Twin Earth, a molecular duplicate of Earth except for the waterish stuff there being a different chemical substance, XYZ.The externalists hold that the word “water” as used on Twin Earth does not designate what our word “water” does.H2O and XYZ are simply different substances.The sentence “Water is usually liquid” is true in English iff H2O is usually liquid, while the same(-spelled) sentence is true in Twin English iff XYZ is usually liquid, a quite different state of affairs.And our word “water” and the Twin English word “water” generate different functions from possible worlds to extensions.

Moreover, consider a single possible world containing both H2O and XYZ: Our “water” applies at that world just to the H2O and not to the XYZ, while Twin “water” applies to the XYZ but not to the H2O, so the two words can hardly have the same meaning.

2.But what does the English word “water” mean?It is important to see that H2O is not the answer, even though “water” rigidly designates H2O and H2O is the word’s intension (in the sense that “water” specifies a function that spits out H2O at every world).For many English speakers know the word’s meaning perfectly well even though they have never heard of H2O and have no chemical concepts at all.And as Putnam seems to concede in the passage quoted above, “water” meant what it means in English long before Watt and Lavoisier, and we do not suppose that the word’s meaning changed with their discoveries.So we cannot say that the meaning of “water” is its intension.

So, a puzzle:“Water” must mean something different as between Earth and Twin Earth, because the same-spelled words designate different natural kinds in those speech communities.The natural kind H2O is the intension of the (Earth) English word.Yet that intension is not the English word’s meaning.So the fact that the words name different natural kinds does not after all show that “water” means something different as between Earth and Twin Earth—but it was our only reason for believing in such a difference in the first place.

Now, Putnam said in the second epigraph to this paper that “water” did not change its meaning between 1750 and 1850.But we cannot maintain both that meaning determines reference and that linguistic meaning is “in the head,” for the Twin-Earth example shows that what is in the head does not determine reference.We must choose, and for his part Putnam would rather preserve the Fregean idea that meaning determines reference than hold on to the commonsensical notion that meaning is in the head (op. cit., p. 704).He argues as before that English “water” and Twin-English “water” differ in meaning.It follows that they always did differ in meaning, from the day they were both coined.

Our reasons for rejecting the first option—to say that ‘water’ has the same meaning on Earth and on Twin Earth…??…may be illustrated thus:Suppose ‘water’ has the same meaning on Earth and on Twin Earth.Now, let the word ‘water’ become phonemically different on Twin Earth—say, it becomes ‘quaxel’.Presumably, this is not a change in meaning per se, on any view.So ‘water’ and ‘quaxel’ have the same meaning (although they refer to different liquids).But this is highly counterintuitive.Why not say, then, that ‘elm’ in my idiolect has the same meaning as ‘beech’ in your idiolect, although they refer to different trees?(p. 710n)

3.I do not accept the “quaxel” argument.[2]It comes close to begging the question.Of course we grant the referential difference, and the stipulated merely phonemic difference.But everything else on Twin Earth is the same as here, as regards anything anyone has ever thought was pertinent to meaning: everyday verification condition, inferential role, “use” of various sorts.Putnam has in effect just asserted that referential difference entails meaning difference no matter what.[3]

In fact, turning Putnam on his head and simultaneously borrowing his own idea as expressed in the second epigraph, there is an argument to show that “water” on Twin Earth does mean what our “water” means.

1.In 1750 on either planet, there was nothing to make a difference in meaning between English “water” and Twin English “water” but the chemical distinctness of the substances referred to.

2.In 1750 on each planet, the chemical natures and hence the chemical distinctness were entirely unknown, and had no effect on linguistic competence or linguistic practice. 

3.If X has no effect on linguistic competence or linguistic practice, then X cannot make a meaning difference.

\ 4.In 1750 on either planet, there was no difference in meaning between English “water” and Twin English “water.”[1,2,3]

5.The meaning of English “water” did not change between 1750 and the present.

6.The meaning of Twin English “water” did not change between 1750 and the present.

\ 7.There is no present difference in meaning between English “water” and Twin English “water.”[4,5,6, transitivity]

4.These considerations could be taken to show that Putnam’s externalism was wrong from the beginning.That suspicion is supported by the intuitions of many uninitiates:Circa 1973, some professional philosophers were at first quite unconvinced that XYZ is not water; they held that XYZ and H2O are just two different kinds of water, even as (Putnam himself admits) jadeite and nephrite are two different kinds of jade.That intuition soon dwindled to tiny-minority status, though it has not disappeared.[4]Perhaps, as Rob Cummins would say, those who voiced it were just not invited to any more conferences—and the preëminence of the externalist intuition is a sociological artifact of Putnam’s (then) scientism backed by his charisma and professional stature.

In further support of that suggestion, there is the reaction of neophyte undergraduates to being taught Putnam.In my experience at least, they put up considerable resistance.I can get them to admit that there is a sense in which XYZ is not water; but few go any further with Putnam.So it may be that Putnam’s externalism is just mistaken, and the English word (as opposed to the philosophers’ word) does mean roughly what the OED says it does.

But remember Putnam’s previous negative case.[5]Dictionary definitions such as the OED’s do not give meanings, in the sense of the entries’ being correct in virtue of meaning.For the features mentioned in them are not analytically implied by the term defined.[6]Let us look back at the OED’s definition of “water” (first epigraph to this paper):The liquid of which seas, lakes, and rivers are composed…”: It is an entirely contingent fact that there are seas, lakes, and rivers; water could exist perfectly well without them.“…[A]nd which falls as rain and issues from springs”: Likewise.“When pure, it is transparent, colourless (except as seen in large quantity, when it has a blue tint), tasteless, and inodorous…”: These properties depend on the contingent structure of human sensory systems; had we a different sort of taste buds, for example, water might taste in some way to us.Nothing whatever in the OED entry is analytically implied by “water.”

Likewise, remember examples such as “cat”:If anything were analytically true of cats, it would be that they are animals.But as Putnam pointed out,[7] it is not analytic that cats are animals; they could have turned out to be robots originally planted on Earth by Martians. 

And here for your delectation are some of the OED’s definitions for other natural-kind terms highlighted by Putnam:

TIGER. …1. A large carnivorous feline quadruped, Felis tigris, one of the two largest living felines, a catlike maneless animal, in colour tawny yellow with blackish transverse stripes and white belly; widely distributed in Asia, and proverbial for its ferocity and cunning….

GOLD. …I. 1. The most precious metal: characterized by a beautiful yellow colour, non-liability to rust, high specific gravity, and great malleability and ductility.Chemical symbol Au….

elm. …1. The name of well-known trees belonging to the genus Ulmus, esp., in England, the Common or Small-leaved Elm (Ulmus campestris), a tree having rough, doubly serrated leaves, flowers nearly sessile, the fruit oblong, deeply cloven and glabrous; in Scotland, the Witch or Wych elm (Ulmus Montana) or the Cork-barked Elm (Ulmus suberosa); in U.S. the White Elm (Ulmus americana)….

beech. …I. a. A well-known forest tree indigenous to Europe and Western Asia, having fine thin smooth bark, and glossy oval leaves; its boughs and foliage form a dense mass, and it bears triquetrous nuts (called mast) placed in pairs in a rough or prickly involucre.It has several ornamental varieties distinguished by the colour or shape of the leaves, as the purple, Copper, and Fern-leaved Beech….

aluminium. …[I]. A metal, white, sonorous, ductile, and malleable, very light, not oxidized in the air, used for instruments, ornaments, and as an alloy.In Chem. it has the symbol Al., is tetratomic, has alumina as its oxide, and the alums as its chief salts.

molybdenum. … A metallic element (symbol Mo) occurring in combination, as in molybdenite, wulfenite, etc….

Unicorn. …I. 1. A fabulous and legendary animal usually regarded as having the body of a horse with a single horn projecting from its forehead…; the monoceros of the ancients.

lemon. …1. An ovate fruit with a pale yellow rind, and an acid juice.Largely used for making a beverage and for flavouring.The juice yields citric acid; the rind yields oil or essence of lemons, used in cookery and perfumery….

red. …1. Having, or characterized by, the colour which appears at the lower or least refracted end of the visible spectrum, and is familiar in nature as that of blood, fire, various flowers (as the poppy and rose) and ripe fruits (whence the frequent similes red as blood, fire, a rose, cherry, etc.)….[This entry goes on for seven and a half more huge OED pages.Read them and find out for sure what “red” means.]

Thus, Quine was right in his famous quip about dictionaries and encyclopedias.The OED itself may be part dictionary, but it is mostly encyclopedia.Hardly anything—if anything—in an entry for a natural-kind term is analytically implied by that term.

By the same token, David Lewis’ method of platitudes will not work.[8]“Water” cannot mean substance that satisfies most of the commonsense beliefs about water, because there are countless possible worlds in which water satisfies few if any of the platitudes.(Also, some natural-kind terms have no associated platitudes whatever.There are no commonsense beliefs about praseodymium, or about upupa epops, the Eurasian hoopoe.Of course those may be considered theoretical terms, and then Lewis would say that they are implicitly defined by their containing theories, but even if so, ordinary speakers do not know the theoretical definitions.)

What, then, is the pre-chemistry meaning of “water”?This paper’s main contention is that that is a very tough question, and should receive much more attention than it has from philosophers of language.

5. An obvious cue comes from David Kaplan, who (during the same period and not altogether coincidentally) pointed out that we have two types of linguistic meaning, which correspond rather nicely to his technical notions of “character” and “content.”[9]Content is a proposition individuated by truth-condition, a function from possible worlds to truth-values.Character is a function from contexts of utterance to contents.Kaplan invokes character to accommodate deixis, especially indexical pronouns.For example, the character of the sentence “I need sleep now” incorporates the rules that “I” refers to the speaker and that “now” refers to the time of utterance,[10]so that in a particular context the sentence expresses the proposition, that that very person (who spoke) needs sleep at that time.

Kaplan argues that character is a more appropriate bearer of the title of “meaning” than is content.An English speaker who comes into a room and sees a token of the sentence “I have not slept since Thursday” written on a piece of paper obviously knows what the sentence means, even if s/he does not know what singular proposition has been expressed in the context.So perhaps “water”’s meaning is its character.

On this model, the leading candidate would be something indexical, something like “whatever real stuff-kind shares the nature of that there,” where the speaker is pointing at a sample of water (= H2O) to define the term ostensively (p. 702).And Putnam endorses that suggestion.Actually he offers several alternate indexical formulations:“Water” means, “stuff that bears...[the same-nature relation] to the water around here” (p. 710, italics supplied), but that definition is obviously circular.Also, “the same stuff as we call ‘water’”—but that would falsely make the kind term metalinguistic.For purposes of initial discussion I shall stick to the ostensive definition model. 

There are daunting objections to that proposal.First, indexicals are devices of direct reference and commit speakers to the existence of their referents.If I point at the stuff in the glass and say “that,” I thereby refer to that particular sample of water, and make it part of the content of what I am saying, just as when I point at a light switch and say, “That one turns on the porch light.”But it is absurd to think that when anyone uses the word “water,” they are even tacitly referring to any particular sample of water.And since different people’s uses of the word would be grounded in different ostendings, “water” would have a different meaning for each of us.(This is not to deny that Kripkean reference-fixers for “water” contain just such indexical elements; of course they do.But, notoriously, to fix reference is not to fix sense.)

A move of Sellars’[11] will solve that problem.It is to replace the singular reference to a sample by deferred reference via an ostended exemplar.When I point at the stuff in the glass, I refer, not to it, but the kind of which it is a characteristic instance.It is as if in response to your asking what color my new car is, I were to point at a recent copy of Philosophical Issues and say “That color.” In so speaking, I do not refer to the particular sample I am pointing at, but only to the color considered as a type; the sample does not enter into the truth-condition of what I have said. 

But what about empty natural-kind terms, like “unicorn” and “phlogiston”?No one has ever demonstrated a sample of any nonexistent kind, nor have there ever been such samples “around here.”If we assume that empty kind terms have the same type of meaning that referring ones do, the indexicality view fails.However, Putnam might reply in Sellarsian style by appealing to deferred ostension, as when we point to a picture of a unicorn, or an effluvium thought to be phlogiston.The reference to an unexemplified kind might succeed even if the alleged exemplar is unreal and the kind nonexistent.

But further objections loom.

(1) Ordinary uses of “water” are not usually accompanied by demonstrations of any kind, much less of a sample of water.(Though this does not touch the “around here” formulation.)

(2) Even when a use of “water” is accompanied by a demonstration, the “this” or “that there” might, on any given occasion, be demonstrating something that is thought by the speaker to be water but is not, as in Putnam’s own “gin” example.[12]If one’s use of “water” on that occasion happened to have been grounded in such an error, the indexicality view predicts that the word would mean gin rather than water.(That does not touch the “around here” formulation either.)

(3) Tyler Burge (op. cit., pp. 103-05) points out that the proposal suffers from being false.If “water” were indexical, an Earthling who visited Twin Earth, demonstrated some XYZ and said “There is some water” would be speaking truly, for XYZ is the stuff-kind that shares the nature of “that there,” and is also the waterish stuff “around here,” i.e., on Twin Earth.[13]

There is a further indexicality option, though I have never seen it defended as such:Instead of indicating a local sample or otherwise referring to the environment, let the indexical home on the speaker and/or the speaker’s original linguistic community.E.g., perhaps “water” means “the waterish stuff found in my neighborhood of origin,” or (again) “the same stuff as we call ‘water’.”Such proposals avoid all three of our objections, because no sample is demonstrated and those particular indexicals do not shift their references when we get to Twin Earth.

But such a move has a permanent liability.(4) I had earlier complained against the second of the latter proposals that it makes the kind term metalinguistic; indeed, it makes “water” mean something about the word “water” itself, which seems intolerable.The first proposal contains a different sort of substantive characterization, “waterish stuff.”If “waterish” means all or even some of the features cited in the OED entry discussed in the preceding section, then it should not appear as part of the meaning of “water,” because none of those features is analytically implied by the term.

The problem generalizes.If “water” is pegged indexically to me or to my community, it must be so pegged by way of some relational predicate.The predicate will be a substantive one, and inevitably not one that is analytically implied by “water.”(This also afflicts the earlier “around here” version.)

And finally: (5) I know of no linguistic evidence that natural-kind terms contain any indexical element.“Water” bears neither syntactic nor semantic marks of indexicality.(a) As we have seen, “water” bears no discernible semantic relation to any demonstrative or personal pronoun.(b) Nor is there the slightest evidence of syntactic lexical decomposition, such as pronominalization into the term itself (cp. “Bluto vomited, but he cleaned it up promptly”).(c) Unlike “flat” or “tall,” “water” is not a comparative adjective; there is no issue of contextual standards of strictness.(And needless to say it does not take comparatives.)(d) I can think of no context in which an Earthling and a Twin Earthling visitor to Earth could respectively assert and deny the same (otherwise nonindexical and unequivocal) sentence containing “water” without one of them being in error—“Even in the Jornada del Muerto there is some water,” “People drink water every day,” etc., in contrast to “It’s raining here,” “This field is flat.”

To borrow Putnam’s own famous colloquialism:[14] Cut the pie any way you like, the meaning of “water” just ain’t indexical.

Yet substantive predicates, and something like “waterish” in particular, may still figure in the meaning of “water” even though “water” is not indexical and those predicates are not analytically implied.That is the basis of our next attempt to specify the meaning of “water.”

6.There is a now well-known development of Kaplan’s “character” idea that dispenses with indexicals: Frank Jackson’s and David Chalmers’ notion of “A-intension” (Jackson) or “primary intension” (Chalmers).[15]They appeal to the now familiar distinction made by Davies and Humberstone:[16]What we ordinarily think of as the intension of a general term is the usual function from possible worlds to the term’s instances at those worlds; the intension of “water” sucks up a world and spits out all the water, i.e., the H2O, found in that world.But this is only the term’s “C-” (Jackson) or “secondary” (Chalmers) intension, because it is what we get when we ask ourselves “what the term applies to under various counterfactual hypotheses” (Jackson, p. 48).But in addition, it is maintained, the term has an “A-intension,” which is what we get when “we are considering, for each world w, what the term applies to in w, given or under the supposition that w is the actual world, our world” (ibid.).[17]The A-intension is Jackson’s candidate for the meaning of a term, and accordingly is what our “water” and the Twin Earthlings’ same-spelled word have in common despite their disparate referents. 

(Incidentally, I came to this literature through Jackson’s work in philosophy of mind,[18] and because of his retro and evil use of it in that area, followed by Chalmers’ even more pernicious version, I conceived a poisonous dislike of the whole idea.But if we return to the present purely linguistic problem, I find myself much more sympathetic.Indeed I now see why Jackson feels his view must be true:Surely, “water” does have a pre-scientific meaning that presupposes no empirical discoveries.I could almost join in that wielding of the “surely” operator, even before I had thought of my own argument for its complement.) 

The notion of “considering a world as actual” is not at all clear, and has been shown to admit of several importantly different versions.[19]But for present purposes let us say it is this: We stipulate and hold fixed that on Twin Earth and any other relevant twin planet and possible world, “water” is a natural-kind term and works semantically much as it does here.That is, it refers to a chemical substance that is the underlying scientific essence of the familiar liquid in question, but its reference is fixed by superficial descriptions such as are found in the OED (“the liquid of which seas, lakes, and rivers are composed, and which falls as rain and issues from springs”).Everything else about English is held fixed also.

Then we can generate Jackson’s A-intension for “water.”At any world at which the expression “water” is used as we use it except for there being a different underlying substance, it will refer to whatever substance “plays the watery role” there (p. 50), which role is given by standard Kripkean this-worldly reference-fixers for the term (p. 49).So, even for us in the actual world, “water” has a kind of nonrigid intension or meaning along with its normal, rigid referential meaning.Thus, “water”’s nonrigid meaning for us is, stuff that plays the watery or waterish role.And that intension is a world-indifferent kind of meaning, as Dorit Bar-On has put it.

Notice that this is “meaning” in a weaker sense than usual.It does not generate the usual analyticities:It is no longer analytic that water falls as rain and issues from springs, etc.(At best it is analytic that “whatever is actually water falls as rain and issues from springs.”)So we have the idea of a weaker task in the explication of “water”: Not to give a synonym, that preserves the term’s analytic connections, but to point to only a lesser kind of “meaning”—though in a sense of the “m”-word that is arguably more worthy.At present it is roughly, once again, Kaplan’s idea of character; we just specify the function from utterance contexts to intensions.(Though, again, Jackson and Chalmers think this particular kind of character also yields a synonym, once we insert the rigidifier “actual”: “whatever actually ‘plays the watery role’.”)

My problem with all this is that I do not believe that English words have A-intensions.To generate an A-intension, one needs a transworld “role,” as in “plays the watery role.”Such roles are supposed to be constituted by reference-fixing descriptions, that are the same across the relevant worlds.That requires that an English word have a distinctive and stable set of reference-fixers.But the latter pretty clearly does not obtain.Reference-fixers are rarely enshrined in the public language;[20] they are private to individual speakers at particular times.

Of course, dictionary entries contain common stereotypical information about water, and philosophers have had no trouble coming up with sets of stereotypical reference-fixers such as the ones deployed in the OED.There is a loose body of information about water that we 21st-century Americans share with most of our fellow English speakers.That seems to constitute a “watery role.”But there being such a body of information is a highly contingent social and environmental fact.One could introduce the word “water” even if there were neither seas nor lakes nor rivers nor rain nor springs, etc., by reference to some other mode of acquaintance with water.In fact, even the collective body of all the reference-fixers that have ever been mentioned in the Putnam literature is expendable.There would always be other reference-fixers for “water,” so long as there were (real or imaginary) water around affecting us in some way—just by making a trickling sound, or looking purple, or feeling slippery.And “water” would still mean just what it does now, in real-world English, even if its reference were fixed, in some context, by descriptions entirely different from the usual ones.(N.b., In saying that, I am not begging the question by just insisting that “water” means H2O; we already know that “water” does not mean H2O.)

Oddly enough, it is not a linguistic fact that “water” satisfies the descriptions that we find in dictionaries.As we have seen, when it comes to natural-kind terms, dictionaries contain mostly or entirely nonlinguistic, encyclopedic information.The difference between what information turns up in real dictionaries and what information does not is only the difference between information that is, though entirely contingent, widely known among a particular dictionary’s intended readership, and information that is not so widely known.(But even that is not quite right.Dictionaries often actively try to be encyclopedias, in supplying not just obviously contingent information but technical nuggets such as chemical symbols.)

Upshot:If “water” has no distinctive, stable set of reference-fixers, it has no associated stereotypical “watery role.”If it has no such role, it has no A-intension.(More recently, Jackson and Chalmers have themselves backed off the idea that A-intensions are public linguistic meanings or types of meaning analogous to Kaplanian characters.[21]They kick A-intensions upstairs and diffusely so, into individual minds at particular times.I have different sorts of objections to the resulting account, but that is outside the purview of this paper.)

What are our remaining options?We can weaken our demands still further.

7.Putnam’s own mature theory in “The Meaning of ‘Meaning’” expresses the meaning of a natural-kind term in a “normal form”: <Syntactic markers, semantic markers, stereotype, extension> (p. 191).This manifests a still weaker conception of the task, because the normal form does not afford analyticities of any sort at all, even though the stereotype is “part of meaning” in the sense that if you do not know the stereotype, then, to the same extent, you do not know the meaning. 

This is not bad, though given my argument in section 3 above, we would have to toss out “extension.”But as before, I deny that natural-kind terms have fixed stereotypes as parts of their meanings, for the same reason that they do not have A-intensions.

8.As an even further weakening, we might consider giving up the assumption of shared public meanings for natural-kind terms, bizarre as that sounds.We might suppose that such terms refer directly, without senses that go beyond their simple designating function, just as pronouns and (in California) proper names do.They have no meanings but the kinds that are their referents.Their references are fixed on particular occasions by casual and ephemeral ostendings or descriptions in the minds of speakers.That is not to deny that “water” has a public pre-chemistry meaning.Its pre-chemistry meaning is that it refers directly to H2O, even though speakers do not know that.(I am not sure who defends this Proper Names view unequivocally.Scott Soames is the best candidate I know.[22])

Like Putnam’s normal-form position, this theory is largely supported by our OED entries; at least, those entries seem to contain nothing but contingent reference-fixers, just as do dictionary entries for proper names of historical figures.And “water” is grammatically singular. 

But the Proper Names theory faces formidable objections.First, as noted above, it implies that pre-chemistry speakers simply did not know the word’s meaning, which is pretty bad.[23]

Second, the Names view suffers from the same liabilities as do Millian or Direct Reference theories of proper names: chiefly, oversubstitutivity (there being, allegedly, no sense whatever in which coreferring names do not everywhere substitute salva veritate), Frege’s Puzzle, the fact of empty names, and the truth of negative existentials.These are horrible problems, and DR proponents tend to turn a blind eye to their horribleness, though I think the least implausible solutions have been offered by Soames in Beyond Rigidity (op. cit.).

Those problems can be avoided by switching to a Kinder, Gentler, attenuated version of DR, defended by me.[24]That version reintroduces private conceptual/computational roles for names, which are a kind of meaning over and above the referent.But since they are private to individuals at particular times, they inherit the original Names view’s denial that natural-kind terms have shared public meanings. 

As noted above, Jackson and Chalmers have suggested bringing A-intensions back into the mind; “water” will have different A-intensions for different people at different times.But this is less defensible than my computational-role account. For one thing, I question whether, for every English speaker and every time, there is at that time in that speaker a determinate A-intension for “water.”In fact, it seems obvious that there is not.If a person is probed (“What stuff do you mean by ‘water’???Quick, now”), that person will be able to cough up a description,[25] but that does not show that the description existed uniquely as such in her/his mind prior to the probe.

Also, it is repugnant at best to give up publicity.A word’s meaning in a natural language is a meaning shared by at least a number of speakers of that language.

Third:The late Paul Ziff used to maintain that proper names are not lexemes of particular natural languages.I agree.True, “Lycan” is a pre-spelling version of the common Swedish word “lyckan,” and “William” is the specifically English form of a name that occurs in other languages as “Guillaume,” “Guglielmo,” etc.But my name is “William Lycan” anywhere in the world, and it is to be used untranslated by a speaker of any other language.A German cannot translate the English sentence “William Lycan loves Mary Lycan” as “Wilhelm Glück liebt Maria Glück.” 

The same is far from true of natural-kind terms.Of course “water” translates into “das Wasser,” “l’eau,” “vatten” (Swedish), “wai” (Maori), etc.And “Maria hat gern water zu trinken” is simply ungrammatical in German, or any other existing language.[26]

Finally, fourth, the Names view loses the distinction between same-kind terms that are synonymous and those that are not.“Coffee,” “java” and “joe” are synonymous in American English, but “coffee” is not synonymous with the botanical genus term “coffea” (as in “coffea canephora,” “coffea liberica,” etc.); that kind identity had to be discovered empirically. “Gorse” and “furze” are synonymous, but neither means the same as their Linnean binominal “ulex europaeus.”[27]

9.One further suggestion has been made to me in conversation by each of several colleagues: ineffability.The idea is that “water” does have a meaning over and above its reference to H2O—it might be thought of as an abstract Fregean sense—but we have no way of expressing or describing that meaning.No law of logic says that if an English word has a meaning, that meaning can be expressed or even described in other words of English.

The first half of the latter point is importantly right.There is no reason why any one English expression should have another, distinct English expression as a synonym.(This matters, I have argued elsewhere, in certain disputes over free will and over the mind-body problem.[28]But I balk at “or even described.”We have seen the one attempt—Putnam’s normal form—to describe “water”’s meaning without offering any sort of synonym.I have argued that that particular attempt failed.But I see no general reason why, if the word has a meaning, its meaning could not be described at all. 

“Water” does not mean H2O.But “water” does not mean anything else either. Our conclusion seems to be that “water” is meaningless.[29]


[1] “Meaning and Reference,” Journal of Philosophy 70 (1973): 702-03. 

[2] Also, Putnam’s own “indexical” model (to be discussed below) predicts shared meaning??as he himself says, on the model of “I.”He must have been thinking of meanings as intensions.

[3] Tyler Burge joins Putnam in denying that there is any sameness of meaning between our word “water” and the Twin Earthlings’ word (“Other Bodies,” in A. Woodfield (ed.), Thought and Object (Oxford: Clarendon Press, 1982)).He does not argue explicitly for the claim, but he does observe (p. 110) that truth-value is at issue.The truth-conditions of my sentences containing “water” involve H2O, while those of Twin Bill’s sentences involve XYZ.Were I to be transported unbeknownst to Twin Earth and to start flinging “water” sentences, I would say many false things, even though the same sentences uttered by the Twin Earthlings would be true.Does that not establish right there that “water” cannot mean the same thing as between English and Twin English?

No, for much the same reason Putnam’s “quaxel” argument failed.The difference in truth-value shows that the terms differ in intension.But we have already seen that “water”’s meaning (whatever it may be) is not its intension.

Incidentally, even if “water’s” meaning does differ as between Earth and Twin Earth, Burge has given us no pointers to either of those meanings.Indeed, since they are not the differing intensions H2O and XYZ, and since Burge himself has refuted the hypothesis that “water” is indexical (see below), he has made the matter even more mysterious.

[4] J.R. Searle, Intentionality (Cambridge: Cambridge University Press, 1983), Ch. 8; J.F. Rosenberg, Beyond Formalism (Philadelphia: Temple University Press, 1994), Ch. 1; Gabriel Segal, A Slim Book about Narrow Content (CambridgeMA: MIT Press, 2000). 

[5] “It Ain’t Necessarily So,” Journal of Philosophy 59 (1962): 658-71; “The Analytic and the Synthetic,” in H. Feigl and G. Maxwell (eds.), Minnesota Studies in the Philosophy of Science, Vol. III (Minneapolis: University of Minnesota Press, 1966).Kripke endorses such arguments in “Naming and Necessity,” in D. Davidson and G. Harman (eds.), Semantics of Natural Language (Dordrecht: D. Reidel, 1972), Lecture III; for his best developed view of natural-kind terms, see the second of his John Locke Lectures (“Reference and Existence,” Oxford University, 1973). 

[6] That is, not even if one believes in analyticity and is not a Quinean skeptic about it.Let us grant for the sake of argument that the standard examples of analytic sentences are indeed true in virtue of meaning alone.

[7] “It Ain’t Necessarily So,” p. 660.

For the same reasons, I pass over the Wittgensteinian view that natural-kind terms are “family resemblance” or “cluster” expressions, i.e., that to be in the extension of “water” one need only have “enough” of the stereotypical properties on a list.What I take Putnam to have shown is that there could be water that did not have any of the stereotypical properties.(More on this below.)

[8] “Psychophysical and Theoretical Identifications,” Australasian Journal of Philosophy 50 (1972): 249-258.I hasten to say that I do not know whether Lewis would have wanted his method applied to “water.”

[9] “On the Logic of Demonstratives,” in P. French, T.E. Uehling and H. Wettstein (eds.), Contemporary Perspectives in the Philosophy of Language (Minneapolis: University of Minnesota Press, 1977). 

[10]It should at least be mentioned that such character rules have lots of exceptions; “I” does not always refer to the speaker, nor “now” to the time of utterance.The real rules would be very complicated.

[11] Science and Metaphysics (London: Routledge and Kegan Paul, 1967).

[12] “If…I am—unknown to me—pointing to a glass of gin and not a glass of water, then I do not intend my ostensive definition to be accepted” (“Meaning and Reference,” p. 702).Yow!

[13] Burge adds a second argument (p. 105), that “‘water’ interpreted as it is in English does not shift its extension with shifts of speakers….The fact that the Twin-Earthians apply ‘water’ to XYZ is not a reflection of a shift in extension of an indexical expression with a fixed linguistic (English) meaning, but of a shift in meaning between one language, and linguistic community, and another.”I agree that there is no shift that feels at all like an indexical shift; but I cannot make this argument myself, since I hold against Burge and Putnam that there is no shift of meaning either.

[14] “The Meaning of ‘Meaning’,” in K. Gunderson (ed.), Language, Mind, and Knowledge: Minnesota Studies in the Philosophy of Science, Vol. VII (Minneapolis: University of Minnesota Press, 1975), p. 144.

Steve Schiffer has protested to me in conversation that if “water” does mean the same on Twin Earth as on Earth, “water” obviously must be indexical—same meaning, different referent, after all.In a minimal and uninformative sense of the term, he is right.But the question is, what kind of deictic meaning is in play?The “indexicality” is not that of a contained pronominal element, or so I have just argued.(I admit that “water” may pass one test of indexicality. Arguably, direct-to-indirect-quotation fails:On the interplanetary listening device I hear Twin Bluto say, “I need at least two gallons of water”; but I cannot correctly report that by uttering, “Twin Bluto said that he needed at least two gallons of water.”—I am not entirely convinced of that judgment.)

[15]Jackson, From Metaphysics to Ethics: A Defence of Conceptual Analysis (Oxford: Oxford University Press, 1998); Chalmers, The Conscious Mind (Oxford: Oxford University Press, 1996).

[16] M. Davies and L. Humberstone, “Two Notions of Necessity,” Philosophical Studies 38(1980), 1-30, going back both to G. Evans, “Reference and Contingency,” Monist 62 (1979), 161-89, and to two-dimensional modal logic as developed by Kaplan (op. cit.), J.A.W. Kamp, “Formal Properties of ‘Now’,” Theoria 37 (1971): 227-73, and F. Vlach, Now and Then: A Formal Study in the Logic of Tense Anaphora, doctoral dissertation, University of California at Los Angeles, 1973.

[17]Jackson’s “A-intension” is what some two-dimensional modal logicians have called the “diagonal” (see R. Stalnaker, “Assertion,” in P. Cole (ed.), Syntax and Semantics 9: Pragmatics (New York: Academic Press, 1978)).

[18] “Armchair Metaphysics,” in J. O’Leary-Hawthorne and M. Michael (eds.), Philosophy in Mind (Dordrecht: Kluwer Academic Publishing, 1994); Chalmers, op. cit.I have attacked Jackson’s and Chalmers’ antimaterialist argument in“Vs. a New A Priorist Argument for Dualism,” in E. Sosa and E. Villanueva (eds.), Philosophical Issues, Vol. 13 (Oxford: Basil Blackwell, 2003).

[19] R. Stalnaker, “On Considering a Possible World as Actual,” Proceedings of the Aristotelian Society Supplementary Volume 75 (2001), 141-156; also, Scott Soames’ Reference and Description (Princeton: Princeton University Press, 2005), pp. 220ff., which work also offers a comprehensive critique of Jackson’s and Chalmers’ two-dimensionalism.But for a more recent explication, see Jackson’s “Why We Need A-Intensions,” Philosophical Studies 118 (2004): 257-77.

[20]I say “rarely” in deference to examples like “Jack the Ripper,” in which one might think a name is publicly associated with a single reference-fixer.But as a matter of fact, I am not sure there is any such example at all.I recently polled my undergraduate philosophy of language class and asked them to write down the description they associate with “Jack the Ripper”; no two answers were the same.(Not fair, I realize, since it was over a century ago that the Ripper was making headlines.I would like to try the experiment using the name of someone who is currently in the news; unfortunately dummy or attributrive names in the news are nearly always those of heinous criminals who are still at large.)

[21] Jackson and Chalmers, “Conceptual Analysis and Reductive Explanation,” Philosophical Review 110 (2001): 315-60.

There is a further problem about A-intensions, both for public English words and for mental A-intensions.It seems perfectly possible that a term’s reference should get fixed by descriptions that are one and all “near-misses” in Donnellan’s sense (“Reference and Definite Descriptions,” Philosophical Review 75 (1966): 281-304), as in his famous example of indicating a party guest by calling him “the man drinking the martini” when in fact the guest’s martini glass contains plain water with an olive in it.Suppose there is a kind term T that refers to an underlying substance S, but none of T’s reference-fixers is strictly correct.Nothing plays the “T-ish role,” though S does things that approximate that role.But it is supposed to be analytic that S is whatever actually plays the T-ish role.It follows that the term T fails to refer, which is contrary to hypothesis.

[22] Beyond Rigidity (OxfordOxfordUniversity Press, 2002), Ch. 10.Soames calls his view “Extended Millianism,” and he accepts the implication that “simple natural kind predicates that designate the same natural kind…mean the same thing (have the same semantic content)” (p. 278).

The view is prefigured by Ruth Millikan’s Language, Thought, and Other Biological Categories (MIT / Bradford Books, 1984), Ch. 9, if I understand her complex terminology correctly: The “sense” of a kind term is just the substance referred to, while various kinds of “intensions” used by individuals to point to or get at that “sense” are mainly private to those individuals at times (the exceptions being, uninterestingly, “language-bound” intensions such as what “water” stands for or what is called “water”).On the other hand, Millikan allows that a “full-bodied public intension of a public term might…sensibly be calculated by averaging only over those full-bodied intensions that are deemed to be most reliable by laymen, or if laymen would ask experts, deemed reliable by experts or, when possible traditionally handed down by experts” (p. 154); in particular, “names of natural kinds…usually have very definite ‘meanings’ in the sense of having quite definite full-bodied public intensions” (p. 155).On the whole it seems that more than the one letter “k” distinguishes Millikan from the Millian; but her proposal is too complicated for me to evaluate here, especially since I do not know what sort of “averaging” she has in mind.

Much more recently Millikan has defended a purer version of the Names view, in “Where Meaning Is, Since Not in the head,” talk presented at the University of North Carolina, September, 2005.

[23] Millikan (op. cit., pp. 151-56) would opine here that we pick up concepts very easily, exploiting Putnam’s “division of linguistic labor,” just by overhearing and acquiring the superficial use of the corresponding kind terms with the presumption that they refer.On her view, I do know the meaning of “gum-cistus” at least to a degree, without having the faintest idea what a gum-cistus is.(“Gum” suggests that it is a kind of plant, but that is only a guess.)

[24] Modality and Meaning (Kluwer Academic Publishing, Studies in Linguistics and Philosophy series, 1994), Ch. 7.As important a philosophical event as it is unknown to the public.Buy this book now.

[25] This is what is called “the spot-check test” in my textbook, Philosophy of Language: A Contemporary Introduction (London: Routledge, 2000), p. 39.

[26] There are interesting intermediate cases, such as patronymics. Stan Munsat has called my attention to Native American given names that have entirely translatable meanings.But I note that translation is optional.In English one can say either “Tashunka Witko’s circular movement at Little Big Horn was flawless” or “Crazy Horse’s…”; though it would be less usual, a Lacotah could use “Crazy Horse” in the corresponding sentence of Sioux, say if talking with an Anglo bilingual.(And of course neither name implies that its bearer was either crazy or a horse.) 

Note a further nasty consequence of the Names view:No two empty natural-kind terms differ in any semantical way from each other.If a kind term has no meaning over and above its referent, and two kind terms lack referents entirely, then there is no semantic difference whatever between them.“Unicorn,” “testudo aubreii,” “aesophila grahami,” and “jub-jub bird” are semantically identical, which seems to imply that they can be substituted anywhere salva veritate at least.(Of course the exactly parallel complaint can be made in the case of proper names, but for names the consequence is not quite so obviously false.)

[27] It may be thought that this argument begs the question against the Names theory, in that it simply assumes that there is a distinction between pairs of synonymous kind terms and pairs of merely coreferring ones.But (a) there certainly have been kind terms whose co-reference had to be established empirically, as when one and the same botanical or insect species had been christened independently by different naturalists.And (b) there are certainly synonymous kind terms; if the above examples fail to convince, consider cross-linguistic cognates, such as “water” and “das Wasser.” 

[28] “Free Will and the Burden of Proof,” in Anthony O’Hear (ed.), Minds and Persons (Royal Institute of Philosophy Supplement: 53, Cambridge: Cambridge University Press, 2003); “Vs. a New A Priorist Argument for Dualism,” loc. cit.


[29] Thanks for spirited discussion to audiences at the 2004 Philosophy Conference, EastCarolinaUniversity, and at VirginiaCommonwealthUniversity (February, 2005), and especially to Steve Schiffer and Ram Neta.