INCOMPLETE DRAFT 7/2012 – Claudio F. Costa.
The final version of this text was published in the book Lines of Thought: Rethinking Philosophical Assumptions (Cambridge Scholars Publishing, 2014).
The final version of this text was published in the book Lines of Thought: Rethinking Philosophical Assumptions (Cambridge Scholars Publishing, 2014).
On the Meaning of ‘Water’
Unacceptable doctrines that illuminate are like crosses on maps that show where treasures lie hidden.
Limitations of the traditional descriptivist view
Consider first the traditional descriptivist view of the meaning of ‘water’. According to this understanding, this meaning can usually be expressed by descriptions naming phenomenal characteristics such as ‘a transparent, tasteless, odourless liquid that makes up rain, oceans, lakes and rivers’. I assume that we should think this meaning as a Fregean sense, that is, as a mode of presentation, as cognitive content, as an informative semantic content. Moreover, I assume that we should think this meaning (hypothetically) as constituted by semantic-cognitive rules that are implicitly or tacitly instantiated in our own heads when – with an understanding of its sense – we employ the word ‘water’ in speaking and thinking. These are what I find reasonable assumptions.
Certainly, this view contrasts with Hilary Putnam’s influential externalist theory of meaning. According to him, the ‘extensional meaning’ of the word ‘water’ cannot be something instantiated in our heads, since it is external and determined by the existence in the world of great volumes of liquids that all display the same essential subjacent micro-structure of H2O.
A first point that I wish to clarify here is that descriptivism need not be restricted to descriptions of surface phenomenal characteristics, as Putnam and Kripke seem to think. Descriptions can be functional, denoting temporal processes. For example, ‘a compound that reacts with oxygen and iron to cause oxidation’ describes how water behaves as a catalyst under certain conditions. Moreover, it would be mere prejudice to exclude the micro-structure of water from the domain of descriptions, for both ‘a chemical compound made up of two atoms of hydrogen and one atom of oxygen’ and ‘a transparent, tasteless and odourless liquid’ are indefinite descriptions in the same sense of the word. Avrum Stroll, who realized these shortcomings of traditional descriptivism, called attention to the number and variety of descriptions belonging to the informative content of names of natural kinds. He appealed to the following definition of ‘water’, taken from a standard contemporary dictionary:
Water n. The liquid which in a more or less impure state constitutes rain, ocean, lakes, rivers, and which in a pure state is a transparent, odorless, tasteless liquid, a compound of hydrogen and oxygen, H2O, which freezes at 0°C and boils at 100°C. It consists of 11.188 percent hydrogen and 88.812 percent oxygen by weight.
What the entry shows is that the meaning of ‘water’ is given by descriptions like these, suggesting that the correct descriptive theory of the meaning of natural kind words like ‘water’, ‘gold’, ‘tree’… should include in its repertoire not only phenomenal characteristics, but also descriptions of chemical micro-structural characteristics, along with a wide range of dynamic functions.
Assuming all this, I wish to take seriously a commonplace notion that philosophers often seem to forget, according to which dictionaries give us the meanings of words in an abbreviated form. After all, this is what dictionaries are for: to give definitions that concisely state the meanings of words. If we wish to know more about the meaning of the word ‘water’, more about its informational semantic content, we can consult encyclopaedias, which already summarize a great deal of information. And if we need to learn still more about this meaning-content, we can turn to specialized monographs like Philip Ball's The Biography of Water. It is no wonder that all we usually know about the meaning of a word like ‘water’ is a few of its more characteristic properties, the more important and useful. Thus, let us at least once take seriously this very commonsensical view and see where it leads us.
Further developing neo-descriptivism
To elaborate on the neo-descriptivist viewpoint proposed above, it is worthwhile to consider the historical development of the meaning of the word ‘water’, beginning with what stone-age cultures may have understood by the corresponding words in their languages. In this and comparable cases, the descriptive meaning of the word ‘water’ could be something very similar to the meaning that Kripke and Putnam claimed was proposed by descriptivists. I call descriptions expressing these meanings surface descriptions, abbreviated as sD:
sD: a transparent, tasteless, odourless liquid that quenches thirst, extinguishes fire, falls to earth in the form of rain, and fills oceans, lakes and rivers.
Admittedly, the Stone Age was a long time ago! Over the course of many thousand years, this initial meaning has expanded, slowly and gradually acquiring new elements. People have added a number of dispositional properties to those included in the sD, which I will call dispositional surface descriptions, or dD. Here are a few, most of them already known three centuries ago:
dD: a liquid that (under normal atmospheric pressure) freezes at O° C and boils at 100° C. It is a good solvent, but does not mix with oil, and in a pure state is a poor conductor of electricity. When brought into contact with iron and oxygen, oxidation occurs…
Descriptions like these are also functional or dynamic. Thinking of the word water in terms of what it meant centuries ago, we can suggest that it had two sub-cores of informational meaning: sD and dD, so that we can symbolically summarize what was descriptively meant by the word ‘water’ in those times with the formula:
<sD + dD>
The resulting meaning component considered here is the informed commonsensical content of the word ‘water’. I will call this water’s popular core of meaning. This is how people understood the word around the year 1750.
Nevertheless, in the following century a revolution took place in the understanding of the concept of water. In 1768, Lavoisier placed hydrogen and oxygen in a bottle and heated the mixture, producing an explosion that released gas and water. Based on this and other experiments, he subsequently concluded that water is composed of two volumes of hydrogen and one volume of oxygen. In 1781, Cavendish made similar experiments in England using electrical sparks. Then in 1783, Lavoisier experimented with the reverse procedure and decomposed water into oxygen and hydrogen. After that, in 1800, Nicholson and Carlisle obtained the same results, using electricity from a ‘Voltaic pile’ in a process called electrolysis. In 1811, based on his law of gases and on electrolysis, Avogadro concluded that the atomic composition of water was HO1/2, This result was later revised by Berzelius, who in 1821 discovered the correct chemical formula, H2O. Subsequently, more and more was discovered about the micro-structure of water, for example, that it is a dipolar compound that tends to form isomeric chains when in a liquid state.
From our neo-descriptivist viewpoint, these experiments and discoveries have added a new and contrasting core of informational content to the old popular meaning core of ‘water’, which we may call its scientific core of meaning. The scientific core of meaning is chiefly constituted by the sense of the scientific description of the micro-structure of water, its atomic deep description, which can be stated as:
aD: a liquid constituted by molecules made of two atoms of hydrogen and one atom of oxygen (a dipolar compound, etc.).
Nonetheless, I believe that a description of the micro-structural essence of water (which from a descriptivist perspective could still be regarded as a nominal essence) can be seen as expressing only the central sub-core of the scientific core of meaning, but not the whole scientific core, except when used as an abbreviation for it. This is because the description of the essential atomic structure of water is inferentially linked on a ‘horizontal’ dimension with other chemical phenomena, and on the ‘vertical’ and ‘diagonal’ dimensions with surface phenomenal functions. Indeed, all these links can be expressed by dynamic descriptions. Thus, we link H2O with descriptions of chemical reactions like ‘2H2O ↔ 2H2 + O2’, ‘2H2O + O2 ↔ 2H2O2’ and ‘Fe + 2H2O ↔ Fe(OH)2 + 2H’, thereby making horizontal inferences. We also inferentially link H2O with surface descriptions on a vertical dimension. For example, by considering the strong cohesive force of the positive and negative dipoles of water molecules, we can explain the properties of surface tension and the fact that water remains in a liquid state at room temperatures. Moreover, starting with ‘2H2O + O2’ and ‘Fe + 2H2O’, we can also make diagonal inferences leading to surface descriptions of the resulting phenomenal properties of respectively hydrogen peroxide and rust. Finally, descriptions of experiments like those of Lavoisier, Cavendish, Avogadro, Berzelius and others, which provide a variety of dynamic surface descriptions, can be linked by inferential descriptions on diagonal and vertical dimensions with the chemical structure of H2O. It seems impossible to understand much of the informative content of the formula H2O without relying on any of these inferences, even if most are known only by specialists. Detailed information about the scientific meaning of the word ‘water’ is given, if not by dictionaries, then by encyclopaedias and specialized works.
Here it could be objected that we are setting no limits on the possible number of descriptions of inferential relations included in the concept of water. If they are not limited and they form the meaning of the concept-word ‘water’, then there can likewise be no limits on this meaning.
We find an answer to this objection when we compare the given case with the surface descriptions of the popular core of meaning. Indeed, the popular core of meaning is constituted by central descriptions (e.g., a transparent, tasteless and odourless liquid…) surrounded by less and less central descriptions (e.g., a liquid with high surface tension that does not mix with oil, etc.). By analogy, even if the descriptions belonging to the scientific core of information are not clearly limited in number, their relative values differ, explaining why the most central deep descriptions are included in dictionaries, while other, less essential ones can be found chiefly in encyclopaedias or specialized works. Consider, for example, the formula ‘2H2O ↔ 2H2 + O2’. It is central to the scientific meaning of the word ‘water’ – every chemistry student knows this. On the other hand, ‘2H2O + O2 ↔ 2H2O2’ and ‘Fe + 2H2O ↔ Fe(OH)2 + 2H’ are respectively more central to the meanings of hydrogen peroxide (oxygen combined with water) and iron (since it explains rust). As the cognitive content of a term decreases, it merges into the cognitive contents of other terms, gradually absorbing their contents.
I believe that the proposed understanding is in perfect conformity with our intuitions about the nature of meaning as informative semantic content, for meaning is typically not a matter of all or nothing. It has variable magnitudes and usually has blurred and overlapping boundaries. And the lack of clear limits between sets of constitutive inferential descriptions of different concept-words seems to justify these characteristics. In order to summarize the scientific meaning component (informative content) of the word ‘water’, I will introduce the following abbreviations:
aD = (atomic deep description) the most central descriptions concerning the inferential relations between the micro-structure of water, along with horizontally related micro-structural inferential formulas (e.g., ‘2H2O ↔ 2H2 + O2’, ‘2H2O + O2 ↔ 2H2O2’, etc.).
sdD = (surface-deep description) the most central descriptions concerning diagonal and vertical inferential relations between the surface phenomena and the micro-structure of water (as in the experiments of Lavoisier, Cavendish, etc.).
dsD = (deep-surface description) the most central descriptions concerning diagonal and vertical inferential relations between the micro-structure of water and its surface phenomena (e.g., the combination of molecules of iron and oxygen to form rust, and dipolarity, producing high surface tension, respectively).
In this way, we have three semantic sub-cores, which I call aD, alongside of sdD and dsD, where aD is a kind of indispensable convergence-point of the other two (which in my view supports the realist interpretation of its ontological status as result of semantic extension). Together, these central segments of informational content form what we might call the scientific meaning of ‘water’, which can be summarized as:
<sdD + /aD/ + dsD>
As noted above, most people know this scientific meaning of ‘water’ only very schematically. Moreover, this scientific core remained unknown to or was even neglected by descriptivist views of meaning from Locke to Carl Hempel. I see this as the fatal shortcoming of traditional descriptivism. For it is perfectly legitimate to construct the descriptivist account as we do, taking into consideration descriptions of the hidden micro-structure of water as expressing part of this natural kind word’s informative content.
Finally, we can synoptically portray the whole system of descriptions expressing the semantic rules constitutive of the entire informative content of the word ‘water’ as the union of the two semantic cores – popular and scientific – as follows:
<sD + dD> + <sdD + /aD/ + dsD>
I suggest that this is the system of descriptions that forms the relevant informative semantic content or meaning of the word ‘water’ as it is or can be made to be acceptable to our language's speakers. The descriptions forming this system have different values, from the most central to the most peripheral ones, which merge with the stronger values of other related concepts. We can see this uneven semantic content concisely expressed in dictionary entries on water and presented in a more detailed way in encyclopaedias and specialized works.
Characterizing rule for water samples
Although I have offered a concise descriptive elucidation of what can be counted as meaning of the term ‘water’, I believe we also need a more systematic approach in order to explain its referential function, that is, how this natural kind term works to locate its reference. We need what might be called a rule of characterization or of application for the referential application of the concept-word ‘water’. This conceptual rule would establish the how and to what extent descriptions of the whole system of descriptions must be satisfied in order to allow the referential application of the concept-word ‘water’. Concerning the two nuclei of descriptions expressing the content of meaning, there are two obvious possibilities: their conjunction and their disjunction. Without aiming to give a fully satisfactory account, but hopefully to offer a reasonable and approximated one, I would like to construct a disjunctive characterizing rule (DCR) or, simply, a characterizing rule (CR) constituting our most general concept of water. This rule could be stated as follows:
We use the term ‘water’ to refer to a property singularized by the instantiation x of a liquid substance
(i) x satisfies the meaning rules expressed by the popular semantic core <sD + dD> and/or by the scientific semantic core <sdD + /aD/ + dsD> associated with the word
(ii) to an extent that is on the whole sufficient, and
(iii) condition (i) applies to x better than any other competing characterizing rule for liquid substances.
As it will be shown, I do not claim that this is the only concept of water. I am only suggesting that this rule produces the widest, most generic, most inclusive and most liberal sense of the word ‘water’ that we can reasonably conceive. It is what we could call a topic-neutral sense, because it is applicable in nearly all imaginable contexts. (On the other hand, if we replace the ‘and/or’ in condition (i) with an ‘&’, we get a conjunctive characterizing rule – CCR-‘water’ – which provides for the most exclusive or narrow application of the word, but is clearly at variance with common usage.)
The suggested formulation of the characterizing rule for ‘water’ requires some clarification. First, there is a general assumption, namely, that water is a liquid substance (in normal temperatures), limiting its domain to the closest relevant class, that here takes the place of a genus proximus. According to CR-‘water’, if at the ambient temperature and pressure x is not a liquid, but instead a hard black stone that resembles coal, it cannot be water, even if we made the curious discovery that its micro-structure was that of H2O. And this seems intuitive enough. (A still wider concept of water would drop the assumption of liquidity, only demanding that it belongs to the genus of material substances. In this sense, even a black rock similar to coal but with the chemical structure of H2O could be water. Although a degree of arbitrariness is unavoidable here, this option seems more difficult to accept, as well because the whole point of finding the closest relevant class is to narrow the possibilities for identification, in order to make it feasible.)
Something must also be said about the three partial conditions of DC-‘water’. Condition (i) is what we could call the meaning condition, for it concerns the proper informative content of the word ‘water’ that we find summarized in dictionaries. As suggested above, if we use the word ‘water’ in its generic, all-inclusive sense, the two meaning components – popular and scientific – should be connected by an inclusive disjunction. According to this, we can call a liquid ‘water’ if it has the phenomenal properties of water, even if it lacks its subjacent micro-structure, and vice versa. Some may at first see this as an overly liberal way of understanding the concept of water. However, this all-inclusive sense is intuitively acceptable, and it seems clear that we will be correctly understood if we use the term ‘water’ in this way. As we will see below, only meaning condition (i) is able to make the application of the word sufficiently flexible for this.
Condition (ii) is what we call a sufficiency condition, according to which the true descriptions of (i) do not need to be completely satisfied, but – if taken as a whole – merely sufficiently satisfied. Just what ‘sufficient’ means here I will leave unspecified, as part of the vagueness of the concept of water. (It is not inconceivable that under certain conditions even a liquid that is neither transparent nor tasteless nor made of H2O could still be called ‘water’ because it quenches thirst, extinguishes fires, fills oceans, lakes and rivers…).
Finally, there is condition (iii), which we may call the predominance condition. This condition states that in order to be called ‘water’ a sample x of a liquid must satisfy condition (i) better than it satisfies any other conflicting characterizing rule. This is equivalent to saying that the meaning condition for the liquid called ‘water’ must be more completely satisfied than any other conceptual meaning condition for liquid substances competing with (i) (suppose we need to distinguish between water and alcohol: Of the conflicting descriptions of the two liquids, the one that predominates would be decisive).
Furthermore, it is worth noting that conditions (ii) and (iii) cannot be a relevant part of the informational meaning of the term ‘water’: Since they clearly also apply to many other concepts, they cannot differentiate the meaning. Based on this we can even explain the intuitive distinction we sense between concept and meaning: the general concept of water is the entire characterizing conceptual rule – the CR-‘water’ – while the general meaning of the word ‘water’ is the content of meaning condition (i) of this conceptual rule. This is the reason why we can say that a concept has a meaning, but not that a meaning has a concept: the meaning belongs to the concept (the characterizing rule) and cannot be confused with it.
Finally, it is also worth noting that the term ‘water’ in its generic application given by CR-‘water’ also works as a kind of rigid designator in the sense that it applies in all possible worlds where there is some stuff that can be characterized with this natural kind term.
Putting the characterization rule to work
Using a few selected examples, I will show that we can make use of CR-‘water’. Suppose we take samples of water from a swamp and find that this liquid is brown, has an unpleasant odour, a bitter taste, etc., but still looks like water, albeit ‘impure water’. We call it water because it satisfies the three partial conditions that constitute the rule: It satisfies (i) with regard not only to its scientific meaning, but in part also to its popular meaning. It satisfies (ii) because the conditions are sufficiently met (even if it is muddy, it is still a liquid that can extinguish fire, it is mostly composed of H2O, etc.). Finally, condition (iii) is also satisfied, assuming that the CR-‘water’ is more applicable to the case in question than any competing characterization rules for chemical compounds.
By contrast, suppose we have a sample of a semi-transparent blue liquid with high viscosity and a bitter taste. It does not quench thirst, and when exposed to copper the liquid burns, releasing water and oxygen… Its chemical composition is H2O2. Even if it resembles that of water, in this case we would not say that the liquid is water. It satisfies the characterization rule for hydrogen peroxide more completely than it does the characterization rule for the concept of water, and these chemical compounds differ significantly. Consequently, because of condition (iii), we call it hydrogen peroxide and not water. (However, it would be correct to call the hydrogen peroxide commonly sold over the counter in drugstores and pharmacies a kind of ‘water’, since it contains only 3% hydrogen peroxide, as opposed to 97% water. Indeed, it is so named in some languages. For a similar reason we do not call D2O (deuterium oxide, also called ‘heavy water’) or T2O (tritium oxide, also called ‘tritiated water’) water, even if these isotopes are present in minute quantities in tap water.)
Now, suppose that we prepare gelatine in our kitchen. We know that it consists mostly of water, and therefore it satisfies condition (i) of CR-‘water’. However, since gelatine is not a liquid substance, it cannot be called ‘water’ in the strict sense, since it does not satisfy the general assumption of CR-‘water’. Indeed, who would call it ‘water’? However, it is easier to say that it is water when it is still in a liquid state: the warm water that you have just mixed with gelatine powder... A similar case is that of words for states of water like ‘ice’ and ‘steam’. Although ‘ice’ and ‘steam’ are made from water, they are not water in the required sense. The reason is that they are not liquids and thus do not satisfy the general assumption of belonging to the class of liquid substances.
Finally, suppose we make a cup of tea by adding tea leaves to boiling hot water. This new liquid satisfies the condition of consisting of a coloured and flavourful substance in water, but it is nevertheless water. It is both tea and water. And it is both because the characterizing rule for tea (water with the taste and smell of cured tea leaves) does not conflict with the characterizing rule for water, which concerns the general classification as a liquid substance. It only adds something to what CR-‘water’ identifies. Some concepts cannot coexist in their applications, while others can coexist simply because they can both be applied to the same entity.
It seems that the proposed CR-‘water’ captures our chief semantic expectations. It is true that this rule is vague and does not always seem to satisfy them. But some unavoidable arbitrariness of our natural language conventions can be accounted for this.
The cognitive network of language
As stated above, the meaning of ‘water’ has grown in the course of history, as could be expected. Note, however, that it has grown within linguistic communities, and consequently the knowledge of this word’s meaning, like that of many others, does not need to be shared equally by all speakers of the language. Because this knowledge is a property of the linguistic community as a society of speakers, the individual speaker needs only a partial and vague knowledge of the meanings of many concept-words. Although this does not seem to be the case with the word ‘chair’, this is surely the case with the word ‘water’. There are privileged speakers who know a lot about the meaning of ‘water’. Even so, in all its details the meaning as a whole could be known only by an ideal speaker. Ordinary speakers know only the most central descriptions. Many know only the popular nucleus of meaning… Insofar as speakers share enough of these central descriptions, they can be aware that they are speaking about the same entity. Indeed, to use a concept-word correctly, we need to know only a small portion of its meaning, such as the class of things to which it belongs, which is enough to correctly insert it in the network of communicative praxis, in discourse, insofar as we are aware of our lack of knowledge of meaning. And we usually do this assuming that we are inserting the word in a linguistic network composed of speakers and hearers containing some nodes with privileged speakers provided with a sufficient knowledge of the meaning, those able to give the concept its effective reference. We are able to insert concept-words in a dialogical context in a reasonable way, insofar as at least two conditions are satisfied:
A. Condition of grammatical awareness: We are aware of our own ignorance of the full meaning, that is, we know the form of the characterizing rule and we know that we lack knowledge of much of its content.
B. Condition of convergence: We at least associate a convergent description with the concept-word, understanding this as a description that classifies its objects at least under the closest more general relevant category.
If a speaker satisfies these two conditions, he is already able to insert the concept-word ‘water’ into a sufficiently undemanding discursive context, even if the meaning he is able to give to the concept-word is only very incomplete and the way of reference is parasitic.
One example is provided by incorrect but convergent indefinite descriptions like ‘a very big fish in the ocean’ for the concept-word ‘whale’. Although incorrect, this description is convergent, since it at least correctly classifies whales under the more general category of large marine animals. Because of this, the speaker can give at least a partial meaning and a parasitic reference to the concept-word. The meaning is partial because he would not be able to use the concept-word adequately in demanding contexts, like that of a specialized zoological discourse. And the reference is parasitic, insofar as in using the word he trusts that others will understand at least what kind of animal he means with the word and that somewhere in the network of speakers and hearers there will be highly competent speakers, able to give this concept-word a more complete and adequate meaning and reference.
On the other hand, if a speaker does not know the form of the conceptual rule and thus does not satisfy condition (A), or if he associates a concept-word with an incorrect and divergent description, and thus does not satisfy condition (B), he will not be able to give it an adequate sense or use it in a referential way, not even in a parasitic referential way. Suppose, e.g., that a speaker associates the concept-word ‘whale’ with the description ‘the name of a peak in the Rocky Mountains’. This description is not only incorrect, but also divergent, since it classifies ‘whale’ wrongly as a singular term and a kind of geological phenomenon. By using the word ‘whale’ in this way, one fails to satisfy the condition of convergence and thus shows that one is unable to correctly insert the word into discourse.
Here we see a cognitivist counterpart of what Hilary Putnam called a division of linguistic labour. However, I do not understand this division of labour in an externalist way, as Putnam suggested. Rather, I understand it internalistically, because it is a cognitive division of linguistic labour among conscious speakers of a language.
The application dilemma and its contextualist solution
Before we discuss Putnam’s Twin Earth fantasy, it is instructive to apply the proposed view to a well-known dilemma concerning the meaning of ‘water’, which can be solved neither by traditional naïve descriptivism nor by causal-externalist theories. The question that generates the dilemma arises from a mismatch between surface phenomenal properties and subjacent micro-structures, which can be analysed using two scenarios:
(a) Imagine that somewhere in the world people discover a liquid that has all the surface properties of water (it is transparent, odourless and tasteless, quenches thirst, puts out fires, etc.), but nevertheless has a completely different molecular structure, say XYZ. Would we still call the liquid water?
(b) Imagine now that in another part of the world we find an oily red liquid that does not have any of the surface properties of water (it is not transparent, has an unpleasant odour, does not quench thirst, does not extinguish fires, etc.). Nevertheless, to everyone’s surprise, this liquid is found to consist entirely of H2O. Would we say that this liquid is water?
Causal-externalists and descriptivist-internalists have given contrary answers to these questions. Consider, first, the reactions of philosophers like Putnam and Kripke in defending the causal-externalist view of the reference of terms for natural kinds. What really matters for these philosophers is the essential chemical structure of water, H2O, referred to by the scientific core of meaning (SCM). Consequently, their answer will be that the liquid considered in case (a) cannot be water, despite its appearance, since it lacks the right chemical structure, while the liquid considered in case (b) must be water, since it possesses its essential chemical structure.
On the other hand, descriptivist-internalist philosophers like A. J. Ayer and most of Putnam’s critics have reacted with the opposite answers. What counts for them are the surface properties of the watery liquid, such as those referred to by the popular core of meaning (PCM). They answer affirmatively to case (a): If somewhere on Earth we found a liquid with the surface properties of water, but the chemical structure XYZ, we would still classify it as water, albeit water of a different kind. And they answer negatively to case (b): Even with the atomic structure of water, the oily red liquid cannot be water, since it lacks the surface properties we are used to assuming when we speak of water.
Who is right? The causal-externalist or the descriptivist-internalist? Here we find a clash of intuitions. If we think like the descriptivist, the answer seems to be the one; if we think like the externalist, it seems to be the other.
However, the more elaborate version of descriptivism I am proposing opens the way to a more convincing answer. As I see it, the dilemma results from the fact that ‘water’ has two different cores of informative content: PCM or </sD/ + dD> (the popular meaning component), and SCM or <sdD + /aD/ + dsD> (the scientific meaning component). Conceivable situations of this sort satisfy one core of meaning but fail to satisfy the other. In this case, according to the general CR-‘water’ – whose meaning condition demands only a disjunction between the popular and the scientific meaning components – both substances should be water, at least insofar as there is no relevant concurrence with other concepts of liquid substances that could change our intuitions. However, the important point is that this first generic, wide, topic-neutral alternative is not the only one, since an indexical element should be introduced into the semantics of the term. We found that the weight of each meaning component can be affected by contextual variations, changing its meaning according to the context of interests associated with an utterance. I will call a context of interests one that elevates the weight of aspects of a term’s meaning that we need for some purpose to the detriment of other aspects that are of little or no interest, in this way constructing a contextually dependent sub-concept with a more restricted meaning. And I suggest that in the case of the concept-word ‘water’ the contexts of interest could be either popular or scientific, in each case emphasizing a different contextualized meaning.
In order to make this suggestion convincing, first imagine a context of scientific interests involving only speakers versed in chemistry who assemble in a laboratory to carry out experiments using electrolysis. In this case, we have a context of scientific interests, and the scientific core of meaning is privileged. Here the word ‘water’ is treated in a narrow, non-literal way as meaning the same thing as ‘hydrogen oxide’ or ‘hydrogen hydroxide’ or ‘dihydrogen monoxide’ or ‘hydroxic acid’ or ‘oxidane’… which are a few of the many scientific terms for H2O. These terms only refer to the scientific core of meaning – a meaning condition expressed by the description <sdD + /aD-H2O/ + dsD> – and not to the meaning condition that gives the popular core of meaning. In this case, scientifically trained speakers rightly reject the claim that any watery liquid with a different chemical structure, say XYZ, can be water.
Now, imagine that people from a remote fishing village dig a well to obtain water for drinking, washing and bathing. In this case, their context of interests is quite mundane and practical. It is irrelevant to them whether the liquid they are seeking is H2O or XYZ, as long as it serves the intended purposes and do not have any unhealthy effect. In this case, the traditional popular semantic component is what matters, and even if they are informed about the differences in the chemical structure, they will feel no need to replace the word. The context of interest produces a contextually dependent sub-concept of water that has the meaning condition </sD/ + dD> as its sense. However, if they found a liquid with none of the surface properties of water that was of no use for drinking and washing… they would deny that this was a sample of water, even if they were somehow persuaded that its micro-structure was that of H2O.
We conclude that the best solution to the dilemma is one based on the recognition that the semantics of the word ‘water’ are sufficiently flexible to allow us to choose the horn of the dilemma that we prefer to hold to in those contexts where we cannot hold to both. In this view, causal-externalists and descriptivist-internalists are right in each holding to one horn of the dilemma, since they are tacitly considering sub-concepts of water caring with them the respective scientific and popular cores of meaning. But they are wrong in their belief that the other horn cannot be held to.
Putnam’s externalist thesis about meaning
Below, I will contrast our analysis of the meaning of ‘water’ with that offered by Hilary Putnam in his essay ‘The Meaning of ‘Meaning’’. I begin by summarizing Putnam’s argument. Putnam proposed an imaginary situation to show that in its essentials meaning cannot be some sort of psychologically instantiated Fregean Sinn, that is, something in the head. In his fantasy scenario, there is a Twin Earth exactly like ours in all details, except that there the liquid called ‘water’, although having all the surface properties of what we on Earth call water, has a chemical structure entirely different from that of H2O, which can be abbreviated as XYZ. Now, suppose that in 1750, when the chemical structure of water was still unknown, Oscar pointed to a sample of water on our Earth and said ‘This is water’, and on Twin Earth his Doppelgänger, Twin Oscar, also said this. Both had the same psychological state of mind, since both perceived the same transparent, odourless and tasteless liquid. However, we might suppose that they meant two different things with two different extensions. What Oscar meant, we could say, was a sample of a liquid with the chemical structure H2O, while Twin Oscar meant a sample with the chemical structure XYZ. Since the psychological states were the same, the different meanings of ‘water’ could not be in their heads. More generally: The meanings of natural kind terms like ‘water’ (as chiefly conceived) must be outside our heads, extensionally determined by the essential micro-structure of their references.
Putnam concedes that surface descriptions, which he calls stereotypes, express a dimension of meaning that is located in the head; however, he regards this as a secondary dimension. In our proposed characterization rule, the stereotype is expressed by the surface descriptions. And what Putnam calls the semantic marker (natural kind, liquid…) is expressed by the condition that we must be thinking of a liquid (or a chemical) substance.
Some well-known difficulties
There are various well-known difficulties with Putnam’s solution. One is that according to his view, in 1750 the two Oscars could not really have known the meaning of the word ‘water’, since they did not know its essential micro-structure. Nonetheless, it seems clear that they really did know the meaning, since we recognize that they could fully communicate it to other persons, and that they didn’t know the new dimension of meaning, understood as the micro-structure, since it didn’t belong to the meaning at that time. In fact, if we accept Putnam’s view, we need to admit a wide range of counter-intuitive ideas about meaning. If we have a natural kind word referring to a still unknown micro-structure, for example, we must accept that we do not yet know its presently given meaning, even if we can swear that we know it. (It would make sense if we said that we do not yet know its future meaning, namely, the still undiscovered descriptions of its micro-structure, which is consistent with our view.) If we make mistakes in designating the essence and are not aware of them, we do not know the meaning of the corresponding terms, whose meaning we are certain we know, which is incoherent.
Moreover, as Putnam's critics have pointed out, the competing intuition remains acceptable, that Twin Oscar uses the word ‘water’ in the same sense. Avrum Stroll, for example, invites us to imagine that since 1750 people have been able to fly from our Earth to Twin Earth and from Twin Earth to our Earth, transporting water from one planet to the other, using it to drink, wash, cook, and for other purposes. Now, suppose that only in the last year have people made a chemical analysis of the liquids they respectively call ‘water’, discovering that the liquid from the Earth is composed of H2O, while the liquid from Twin Earth is composed of XYZ. Under such circumstances, he notes:
Would the persons moving between the planets stop calling one of the fluids ‘water’? I doubt it. I think that they would say that water is composed of different components, depending on where it is found, or perhaps they would say it comes in two different forms, like jade.
It would be difficult to give this intuition a satisfactory causal-externalist answer. Nevertheless, in terms of how we analyse the semantic content of the word ‘water’, it is easy to explain this intuition by appealing to contextual interests. Here we have a context of use in which the popular sense of ‘water’ is still predominant, and the scientific sense has not yet become sufficiently established to produce any possible confusion between the two.
The right explanation of the Twin Earth fantasy
What I wish to show now is that to explain Twin Earth fantasy intuitions, the proposed internalist-descriptivist view works much better than both Putnam’s theory and traditional descriptivism. As we already saw when considering a parallel dilemma about the application of the word ‘water’, our semantic intuitions oscillate between two alternatives:
(A) With the word ‘water’ both Oscars refer to the same reference and extension, namely, the same transparent, tasteless, odourless liquid (traditional descriptivism, from Ayer to Stroll).
(B) Each Oscar refers to a different substance with a different extension (causal-externalism, from Putnam to John McDowell).
Now, our analysis allows us to save both conflicting intuitions. For what we think the two Oscars meant with the word ‘water’ in 1750 could have a double interpretation, which varies with the perspective we take: that of the two Oscars or our own. Here it goes:
- Perspectival explanation of intuition (A): Suppose, first, that we take the perspective of the two Oscars in 1750. In this case, we explain intuition (A), according to which they were both pointing to the same reference with the same extension. The meaning of ‘water’ at this time was only its popular meaning, given by the surface description </sD/ + dD>, which at that time constituted the whole meaning condition (i) for the word. In this way, we have an explanation for our commonsense intuition that in 1750 the two Oscars used the word water with a shared meaning and also felt certain that they were able to communicate this whole meaning. Considering that they had the same mental state caused by the same kinds of superficial properties, we do not need to suppose that the meaning had to be outside of their heads.
Putnam writes that in 1750 people already knew that water has a subjacent micro-structure, although they still didn’t know what this was. But we can think back in time and imagine two stone-age Oscars who lived at a time when there was no assumption that water could have a uniform subjacent micro-structure, differing from things such as urine, oil, air and dust, which lack this structure. The meaning of the equivalent word for the stone-age Oscars would be almost entirely restricted to <sD> and couldn’t have even a hint of a scientific subjacent micro-structure. This meaning is part of the meaning assigned to ‘water’ by the Oscars in 1750, and is still part of the meaning we assign to this word today, particularly in popular contexts of interest.
- Perspectival explanation of intuition (B): Now, suppose we consider what the Oscars were referring to in 1750 from our own perspective. In this case, we can explain intuition (B). What makes the explanation possible is the fact that references and extensions belong to the extra-linguistic domain, which in the case of external things like water samples exists outside the head, being able to be accessed and understood in different ways. For, if the reference and extension are outside our heads and minds, they can be conceived independently of what persons such as the Oscars have in mind when referring to them. This allows us, naturally and unconsciously, to change the epistemic subject and consequently the perspective of evaluation. In our case we can consider what the Oscars were pointing to in 1750 not from the perspective of what they had in their minds, but from the perspective of what we presently mean with the word ‘water’, namely the chemical substance that satisfies the meaning condition ‘</sD/ + dD> and/or <sdD + /aD/ + dsD>’. From this perspective, we can already say that Oscar was pointing to water consisting of H2O, while Twin Oscar was pointing to water consisting of XYZ. And not only is what we mean in each case obviously different, but also what we have in our heads.
What Putnam implicitly does is something more. Rejecting the popular meaning, he employs the word ‘water’ intending it only in the sense it acquires in a context of scientific interests, a sense tacitly also known to us. If we do this, we are able to distinguish two different concepts of water: Earth water, with the scientific meaning of <sdD + /aD-H2O/ + dsD> and Twin Earth water, with the scientific meaning of <sdD + /aD-XYZ/ + dsD>. Based on this, we are able to conclude that the reference and extension of the word ‘water’ meant (that is, pointed to, referred to) by the Earth Oscar was that of H2O, while the reference and extension of the word ‘water’ meant (that is, pointed to, referred to) by Twin Earth Oscar must have been that of XYZ, for the first reference satisfies the first meaning condition, while the second reference satisfies the second.
Now, I propose that what we inadvertently do when we think of the fantasy in Putnam’s way is the following: We use the Oscars (in 1750, in the Stone Age, etc.) as indexical devices for the projection of our own scientific senses of the word ‘water’, that is, of our own cognitive ways of presenting its reference and extension, and because of this we obtain different references and extensions. Indeed, the Oscars mean different things with the word ‘water’ only in the trivial sense that they point to things with different micro-structures. For we should not forget that in this case we are the ones who are giving the senses, and not the Oscars: When we consider only the scientific core of meaning, what we have in our heads must be a certain mental state when we consider what Oscar refers to with the word ‘water’, and a different mental state when we consider what Twin Oscar refers to with it. And since the difference in what we have in our heads corresponds to a difference in meaning, we are able to conclude that there is no reason to think that anything belonging to the meaning must exist outside our heads. Moreover, we find here different meanings, different informative contents, different modes of presentation determining the difference in the reference and, ultimately, the difference in the extension (all very Fregean).
Indeed, even the Oscars would agree with our alternative explanation. Suppose we put them in a time machine, bring them to our own time and teach them some chemistry. Once we do this, they will be able to agree with us that in 1750 they were pointing to (‘meaning’) different things with different extensions. However, this would only be because they had learned the scientific meaning core of the word ‘water’. They would be able to entertain different meanings, different informative contents, and different modes of presentation instantiated by different mental states, which are, of course, in their heads. These Oscars would now project the different concepts of water they have in their heads into their own indexical utterances in 1750, in order to characterize the two different kinds of stuffs they were pointing to at that time and the different implied extensions. In this way, they would be using themselves in the past as indexical devices for the projection of their own new scientific meanings of the word ‘water’.
A similar reasoning applies to Putnam’s examples of molybdenum versus aluminium, and elms versus beeches, applied to switched things on the Earth and Twin Earth. We project the specialized understanding of metallurgist and of the biologist to find the switched scientific dimension and extension of the descriptive meaning of these words in the samples pointed out by Oscar and Twin Oscar, in the case of the corresponding metals and trees respectively.
According to our explanations, it does not matter how we cut the cake, meanings are always internal. They are where everyone always knew they were, namely, in our heads. Although Putnam’s theory has at least some advantage in terms of simplicity, our theory should be clearly preferred for the reason that it is the only able to accommodate all contrasting semantic intuitions.
Some supplementary comments complete the picture. First, in his argument Putnam benefits from a certain ambiguity in the meaning of ‘meaning’. In an extended sense and in the third person, ‘to mean’ can easily mean ‘to indicate’ or ‘to refer’ (e.g., ‘I mean this chair and not that table’). In a non-literal sense, ‘to mean’ can even refer to the reference in itself or the extension. But in these secondary senses, ‘to mean’ has no relevance for the proper understanding of ‘meaning’ as the informative content of a linguistic expression, that is, as its Fregean sense. Curiously enough, the word ‘sense’ lacks this ambiguity and cannot be used only to express the informative content (I cannot say: ‘I sense this chair and not that table’).
Second, it could be objected that by accepting this conclusion we are denying that in the case of the 1750 Oscars there was a causal input from the outside world – a causal input that would be provided by the essential chemical micro-structure of water. However, it would be an arbitrary prejudice to restrict the attribution of causal input to the micro-structure (why not a micro-micro-structure, such as a system of one-dimensional strings as proposed by string theory in physics?). In 1750, the causal input could be attributed to the liquid substance of water and nothing more.
Third, as neo-Fregean internalists, we would not have any reason to deny that very often, and in ways that are ultimately unavoidable, some external causal input in different ways genetically determines our meanings (senses), which once established determine the nature of the reference and its extension. Such dual movement turns out to be nearly simultaneous in the case of indexical utterances. However, to take the external causal origin of meaning for the external instantiation of meaning is to commit (as Putnam does) a genetic fallacy: the error of confusing the origin with its product.
We conclude that by postulating some kind of extensionally determined meaning outside our heads, Putnam tries to produce a kind of inversion of semantic values, taking a derivative indicative sense of ‘to mean’ as its proper sense. For him, what really matters is a supposedly external sense of ‘meaning’. However, here he is playing a word game. For when he says that reference determines meaning, he either intends something fallacious, or he should mean that reference very often and in some ways unavoidably originates our linguistic conventions and that in the case of indexical utterances the reference directly originates our own modes of presentation, while only external references and extensions remain outside the head. However, these are not discoveries at all, but plain trivialities.
The statement ‘Water is H2O’
Finally, we should say something about the sentence ‘Water is H2O’. According to Kripke, it states a necessary a posteriori identity: it is a posteriori, because it was discovered by chemists using scientific methods, and it is necessary, because ‘water’ and ‘H2O’ are rigid designators, applicable to the same substance in all possible worlds. Our analysis will lead us to the rejection of this view.
According to our analysis, the word ‘water’ can acquire three relevant meanings. Initially we have the generic, wide, topic-neutral meaning (A), conceptually defined by the (disjunctive) characterizing rule CR-‘water’ and abbreviated by ‘</sD/ + dD> + <sdD + /aD/ + dsD>’. But there is also case (B): those sub-concepts with narrower meanings acquired by the word in popular or scientific contexts of interest. When the context of interest is a popular one, ‘water’ means (B1) ‘</sD/ + dD>’ (a transparent, tasteless liquid…), and when the context of interest is a scientific one, ‘water’ acquires the meaning (B2) ‘<sdD + /aD/ + dsD>’, which is the same as the meaning of ‘H2O’ or ‘hydrogen oxide’ or ‘hydrogen hydroxide’ or ‘dihydrogen monoxide’ or ‘hydric acid’… According to this analysis, we can understand the sentence ‘Water is H2O’ in three ways, depending for the speaker (mainly) on what he has in mind and, for the hearer, on the context in which it is applied.
Consider, first, the wide sense (A), in which ‘water’ can mean equally the surface content expressed by ‘</sD/ + dD>’ and/or the deep content expressed by ‘<sdD + /aD/ + dsD>’. In this case, ‘Water is H2O’ cannot express an identity, for ‘</sD/ + dD> and/or <sdD + /aD/ + dsD>’ is not the same as ‘<sdD + /aD/ + dsD>’. Here the ‘is’ of ‘Water is H2O’ is not an ‘is’ of identity, but rather an ‘is’ of composition, as Stroll has rightly noted. In this case, the sentence ‘Water is H2O’ expresses a contingent a posteriori truth, stating that the chemical composition of water is always H2O (‘For all x, if x is water, x is made of H2O’).
Consider now sense (B1) of water, in which it means the same as ‘</sD/ + dD>’. In this case, ‘Water is H2O’ is also an informative sentence, stating that what looks like water consists of H2O, in other words, that the watery liquid is always composed of ‘<sdD + /aD/ + dsD>’ (‘For all x, if x is a watery liquid, x is made of H2O’). In this sense the statement, ‘Water is H2O’ is not analytic. It is the result of an empirical observation, and it is also a contingent and a posteriori truth.
Finally, consider the scientific sense (B2) of water. In this sense, the word ‘water’ means ‘<sdD + /aD/ + dsD>’, or (approximately) hydrogen oxide (to use chemical terminology). In this sense, ‘Water is H2O’ can be interpreted as a true identity statement. However, in this case we do not have a necessary a posteriori identity, for here we use the word ‘water’ in the same sense as ‘hydrogen oxide’, and what we are really saying is:
Hydrogen oxide is (the same as) H2O.
However, since to say ‘hydrogen oxide’ is only another way to say ‘H2O’, what we are saying is that ‘H2O = H2O’, which is a necessary a priori truth, an analytic tautology and not an a posteriori discovered identity. We conclude that in no allowable sense does the sentence ‘Water is H2O’ ever express a necessary a posteriori truth.
 Even in the case of philosophers like Frege, who held Platonist views of meaning, it is hard to see how one could grasp a Platonic entity without some relation of supervenience to an internal psychological event instantiating the application of a semantic rule.
 Hilary Putnam, ‘The Meaning of ‘Meaning’’, in his Mind, Language and Reality – Philosophical Papers (Cambridge: Cambridge University Press, 1975), Vol. 2. See also his book Representation and Reality (Cambridge, MA: Bradford Books, 1988), chap. 1, where he states his view more carefully. For a discussion, see A. Pessin & S. Goldberg (eds.): The Twin Earth Chronicles (New York: M. E. Sharpe, 1996), with Putnam’s introduction.
 Avrum Stroll, Sketches and Landscapes: Philosophy by Examples (Cambridge, MA: MIT Press, 1996), p. 71. This is no exception: good dictionaries in many languages offer similar descriptions.
 If you have any reservations concerning inferences about the views of stone-age cultures, you can easily use as examples cultures that existed as recently as a few thousand years ago or even indigenous cultures before their first contacts with Western civilization.
 This is only a brief summary of a more complicated and controversial history. See, for example, Philip Ball, A Biography of Water (Berkeley: University of California Press, 2001), chap. 5.
 In this way, I think that the kind of descriptivism I am proposing draws on the inferentialist perspective regarding meaning. See, for example, Robert Brandon, Articulating Reasons: An Introduction to Inferentialism (Cambridge, MA: Harvard University Press, 2000), chap. 1.
 The expression ‘rule of application’ or ‘rule of usage’ (Verwendungsregel) was consistently used by Ernst Tugendhat, who (among others) theoretically anticipated the need for a rule to classify the objects subsumed under a general term. See his Vorlesungen zur Einführung in die sprachanalytische Philosophie (Frankfurt: Suhrkamp, 1976), p. 188.
 In contrast, the identifying rule of a proper name (see in this book, Chapter 2) requires that the identified object should satisfy the identifying rule better than any other object. If the same object satisfies two or more identification rules (as in the already exemplified case of Marcial Maciel and Jose Rivas), the rules must be reinterpreted as part of one and the same rule. This is because with regard to any one object, any identification rule should enable us to pick out the particular object that satisfies the rule. Concerning characterizing conceptual rules, however, there is no individuating object. As a consequence, the characterizing rule that is best satisfied by the sample in question proves to be the rule that allows us to pick out its relevant property, insofar as we are choosing among competing rules, and the properties described by one rule exclude those described by others.
 I am not asserting that the conventions of natural language witt always confirm the conceptual rule.
 I don’t think that the philosophical view I have sketched belongs to an old approach that conflicts with the empirical findings of psychologists like Eleanor Rosch, showing that prototypical cases are easier to categorize… To the contrary, it follows from my view that when central criterial configurations derived from a characterizing rule are satisfied, categorization under this rule can be probably more easily and quickly realized. See E. Rosch, ‘Principles of Categorization’, in E. Margolis and S. Laurence (eds.), Concepts: Core Readings (Cambridge, MA: MIT Press, 1999).
 What Putnam calls ‘the division of linguistic labour’ is a phenomenon that has been recognized by other philosophers at least since John Locke, who, with his theory of meanings as ideas, already had an internalist and cognitivist perspective. See A. D. Smith, ‘Natural Kind Terms: A Neo-Lockean Theory’, European Journal of Philosophy 13 (2005), 70-88, here pp. 70-73.
 For a vivid description of this and other meaning-perplexities, see W. L. Lycan, ‘The Meaning of Water: An Unsolved Problem’, in E. Sosa and E. Villanueva (eds.), Philosophical Issues: Philosophy of Language 16 (2006) (a supplement to Nous), 184-199.
 See Hilary Putnam, ‘The Meaning of ‘Meaning’, op. cit. Saul Kripke, Naming and Necessity (Cambridge, MA: Harvard University Press, 1980), pp. 128-9.
 A. J. Ayer, Philosophy in the Twentieth Century (New York: Vintage, 1983), p. 270. See also D. H. Mellor, ‘Natural Kinds’, British Journal for the Philosophy of Science 28 (1977), 299-312; Eddy Zemach, ‘Putnam’s Theory on the Reference of the Substance Terms’, Journal of Philosophy 73 (1986), 116-27; G. M. A. Segal, A Thin Book About Narrow Content (Cambridge, MA: MIT Press, 2000), p. 19.
 As Joseph LaPorte noted, disagreeing with Putnam, something similar occurred in China with the word ‘jade’. The old jade (nephrite) was mostly replaced by a phenomenally identical stone with a very different molecular structure (jadeite). Nevertheless, the same name was applied to the new stone. In my view, this is to be expected, since here the involved contexts of interest mainly privilege the popular core of meaning. See his Natural Names and Conceptual Change (Cambridge: Cambridge University Press, 2004), p. 96.
 In Putnam’s own words: ‘Oscar1 and Oscar2 understood the term ‘water’ differently in 1750, although they were in the same psychological state, and although, given the state of science at the time, it would have taken their scientific communities about fifty years to discover that they understood the term ‘water’ differently. Thus the extension of the term ‘water’ (and, in fact, its ‘meaning’ in the intuitive pre-analytical usage of that term) is not a function of the psychological state of the speaker by itself’. H. Putnam: ‘The Meaning of ‘Meaning’’, p. 224.
 Putnam, ‘The Meaning of ‘Meaning’’, p. 269.
 Avrum Stroll, Twentieth Century Analytic Philosophy (New York: Columbia University Press, 2000), pp. 233-234.
 Putnam: ‘The meaning of ‘Meaning’’, p. 225-227.
 It is noteworthy that this ambiguity was explored in Frege’s merely terminological use of the German word Bedeutung (meaning) to designate the reference of a word, in contrast with Sinn (sense), which lacks this ambiguity.
 In the later and more differentiated formulation of the argument, in the chapter 1 of Representation and Reality, Putnam laboriously holds to a tentative path between the opposite dangers of fallacy and trivialization. Finally, according to John Searle, he gave up his externalist explanation of the Twin Earth paradox.
 Saul Kripke, Naming and Necessity, pp. 128, 134, 150.
 Stroll notes that if we concede that ‘Water = H2O’, then we should also admit that ‘Ice = H2O’ and ‘Steam = H2O’, which forces us to conclude (due to the transitivity of identity) that ‘Water = ice’ and ‘Ice = steam’, which is absurd. See his Sketches of Landscapes: Philosophy by Examples, p. 46 f. See also A. P. Martinich & A. Stroll: Much Ado About Non-Existence: Fiction and Reference (Lanham, MD: Roman & Littlefield, 2007), pp. 122-125.
 Indeed, this would be the case even if ‘Water is H2O’ were here understood as an identity sentence. On the other hand, if the concept of water were understood in a restrictive way as CCR-‘water’, having as its meaning-condition ‘</sD/ + dD> & <sdD + /aD/ + dsD>’, then the sentence ‘Water is H2O’ would be analytic, since H2O would belong to the concept of water.
 It is curious to note that the here developed conceptual analysis of water leads us to a result vindicated by the basic idea of the two-dimensional semantics, according to which so-called necessary a posteriori statements have two intentions: a primary intention (contingent) and a secondary intention (necessary). The first corresponds to our sense (B1) of ‘Water is H2O’, while the second corresponds to our sense (B2) of this sentence. However, it seems to me that the here suggested semantic-pragmatic analysis is methodologically more adequate and potentially more perspicuous in its results, since it allows a selective emphasis on intentions according to different utterances’ contexts of interest. For an introduction to two-dimensional semantics, see David Chalmers, ‘The Foundations of Two-Dimensional Semantics’, in M. Garcia-Carpintero and J. Maciá (eds.), Two-Dimensional Semantics: Foundations and Applications (Oxford: Oxford University Press, 2006).