Philosophy of language :: Lecture notes Philosophy of Language\ {\normalsize{Lecture notes}}

Philosophy of Language
Lecture notes

Sandy Berkovski

Bilkent University
Spring 2007

Abstract

The course falls roughly into three parts. In the first part (Chapters 1-3) we shall deal with central issues in the theory of reference emanating from Kripke's and Kaplan's work on singular terms. In the second part (Chapters 4-5) we shall discuss issues of meaning and truth, where deflationist approaches will play a central role. In the third part (Chapters 6-8) we touch upon modal semantics and its employment in the analysis of proposition and belief. We then move on to discuss theories of metaphor, one application of which lies in dispelling doubts over ontology of possible worlds. Finally, if time permits, in Chapter 9 we shall prepare transition to the philosophy of logic by discussing problems of logical form.

Disclaimer. These brief notes do not constitute an attempt of writing a book. Many of the issues discussed here may have been amended, expanded, or further clarified in the class. Some passages are adapted, perhaps even verbatim, from a variety of published and unpublished sources.

1 Proper names
    1.1 Kripke on descriptivism
        1.1.1 Frege-Russell
        1.1.2 The cluster theory
        1.1.3 Kripke's modal argument
        1.1.4 Kripke's epistemic argument
        1.1.5 Kripke's semantic argument
    1.2 The causal theory of reference
        1.2.1 Baptism and transmission
        1.2.2 Direct reference
        1.2.3 Two types of causation
    1.3 A neo-Russellian response
    1.4 Nominal descriptivism
    1.5 Evans on singular thought
        1.5.1 Frege's input
        1.5.2 Russell's input
        1.5.3 Proper-name-using practices
        1.5.4 The fixation of reference
2 Descriptions
    2.1 Russell's view
    2.2 Strawson's critique
    2.3 Donnellan's account
    2.4 Incomplete definite descriptions
3 Demonstratives
    3.1 Kaplan's account
    3.2 Doubts about Kaplan's view
    3.3 Dthat
    3.4 Self-locating knowledge
4 Truth
    4.1 Shades of the past
    4.2 Tarski's theory of truth
    4.3 Doubts about Tarski
        4.3.1 Antecedent understanding of the object-language
        4.3.2 Projectibility
    4.4 Redundancy
    4.5 Minimalism
    4.6 Thick disquotationalism
5 Meaning
    5.1 Indeterminacy of translation
        5.1.1 The argument from radical translation
        5.1.2 The argument from theoretical indeterminacy
    5.2 Davidson on meaning
6 Modality
7 Pragmatics
    7.1 Comparison to demonstratives
    7.2 Presuppositions
    7.3 Two-dimensional semantics
8 Metaphor
9 Logical form
10 Assignments

Chapter 1
Proper names

When we speak and think about the world, we express thoughts about particular items in it. These may include persons, places, and ordinary things. They may be the ones we encounter presently in perception, or the ones we recollect, or-the vast majority-those we are told of by others. In those cases we have instances of singular reference. Without it all our thoughts would have been at most qualitative. Linguistically, singular reference is established through the employment of singular terms. A question arises about the link between singular terms and the objects to which they refer. We shall now discuss one class of singular terms-ordinary proper names. Our strategy, then, will be thoroughly a-historical, as we start with the more recent work by Kripke in Kripke [1980] and his criticism of earlier theories, and only several lectures later we shall get to the in-depth discussion of those earlier views.

1.1 Kripke on descriptivism

Our immediate goal will consist in getting a grip on the first two lectures of Kripke [1980]. Kripke starts by criticising the orthodoxy of his day, a view he associates with Frege and Russell. Many scholars had subsequently cast doubt on the adequacy of Kripke's reading of Frege and Russell. Their doubts have great importance; however, we shall put them aside for now. We shall rather present the Frege-Russell view-to be labelled `the descriptivist theory'-as it emerges from Kripke [1980].

1.1.1 Frege-Russell

The descriptivist theory is motivated by the phenomenon of informative identities. Consider the statements:

Hesperus is Phosphorus.
Saint-Petersburg is Leningrad.
George Gordon is Lord Byron.
Bacon is Shakespeare.

All of them can be represented in the form of a = b, where `a' and `b' are two distinct names. We are, therefore, dealing with identity statements. Now the question is what the meaning of a proper name is. Suppose the meaning of the name is exhausted by the individual it denotes. That is, its sole linguistic function, we suppose, lies in designating that individual. However, then, for example:

Hesperus is Phosphorus

is a substantive astronomical discovery, whereas

Hesperus is Hesperus

is not. But, on the other hand, those two statements have, by supposition, the same meaning. Similarly for other cases, where we shall have historical or geographical discoveries. Hence a paradox.

The way out, according to Frege-Russell, is to abandon the assumption that the meaning of a proper name is given by the individual it designates. Instead, with each name we associate a description. We may say that the meaning of the name `Hesperus' is `the brightest celestial object regularly seen near the western horizon after sunset' and the meaning of the name `Phosphorus' is `the brightest celestial object regularly seen near the eastern horizon before sunrise'. In this way we explain how the equivalence above may be informative.

According to descriptivism, the meaning of a proper name `NN' is exhausted by a definite description attributing a property to NN. On one version of this view mentioned by Kripke (27, 33), the name of NN can be replaced salva veritate by such a description, so that any proper name is in effect an abbreviation of that description. Thus, in the sentence:

Caesar was childless

we can replace `Caesar' with a description obtaining:

The first Roman emperor was childless.

1.1.2 The cluster theory

One immediate objection that can be raised against the Frege-Russell view is the objection from error. Consider the name `Caesar'. The description most easily associated with it is `the first Roman emperor'. Suppose, though, we are mistaken, and the title of emperor was not introduced in Rome until after Caesar's death in 44BC. We are naturally inclined to say that the statement:

Caesar was the first Roman emperor

is false. And yet, if `Caesar' is a mere abbreviation of `the first Roman emperor', then we would be claiming, absurdly, that:

The first Roman emperor was the first Roman emperor

is false. Let us suppose we try a different description and say that `Caesar' is an abbreviation of the description `the founder of the Roman empire murdered in 44BC'. Again, we may discover that, unbeknownst to all our sources, Caesar survived the assassination attempt in 44BC and continued to live in obscurity somewhere in the Black Sea. Therefore, we would say that:

Caesar was murdered in 44BC

is false. But again, taking `Caesar' to be an abbreviation of `the founder of the Roman empire murdered in 44BC', we would be claiming that:

The founder of the Roman empire murdered in 44BC was murdered in 44BC

is false. The problem here is that whatever description we select, the individual named might not satisfy the individual concept expressed by it. Thus, instead of associating with a name one description, or equally a fixed number of descriptions, we associate with it a cluster of descriptions. The cluster of descriptions associated with `NN' would express exactly the properties constituting the identity of NN. But what are those properties? One candidate would be the property of being a man. Whether or not it is an essential one, or whether it is immune to error, is a complicated issue. Yet, in any event, it is not sufficient to differentiate Caesar from other human individuals, or alternatively, to differentiate the name `Caesar' from, say, the name `Pompey'. Therefore, we have to expand the cluster of descriptions. And then, to ensure the ability to differentiate between Caesar and Pompey, we have deal with descriptions such as `the greatest Roman commander', `the conqueror of the Gauls', and so on. Thus, the cluster of descriptions will include the most important descriptions associated with Caesar.

Kripke levels several objections against the view just outlined.

1.1.3 Kripke's modal argument

Note first one consequence of descriptivism in its cluster version-construed as a thesis about the meaning of proper names. If the meaning of `NN' is exhausted by a set {d₁, ¼, d_n} of descriptions, then we let D be a complex description `the thing of which the claims "It is d₁", ..., "It is d_n" are true'. Then we have that, since:

If D exists, then D is D

is a necessary truth, so must be the statement:

If NN exists, then NN is D.

The implication here works by replacing `D' with `NN' in the relevant occurrences.

And this consequence is unacceptable. Let D_c be `the greatest Roman commander and the founder of the Roman empire'. The statement:

If Caesar existed, then Caesar was D_c

(1.1)

is not necessarily true. Caesar might have spent all his life in Athens studying Greek grammar. In such a scenario, the antecedent of (1.1) is true, but the consequent is false. Hence the whole statement is false in this scenario. Hence the statement is not necessarily true. Therefore, this family of descriptions does not give the meaning of `Caesar'.

More generally, there is no complex description D such that:

The referent of `Caesar' is the unique individual who satisfies most, or a sufficient number, of the simple descriptions in D;
Ordinary speakers associate D with Caesar, believing its referent to be the unique individual who satisfies most of the descriptions in D;
A statement like (1.1) expresses a necessary truth.

If this is right, then the cluster view is false, and names are not synonymous with descriptions associated with them by speakers.

The key notion of the modal argument is the notion of a rigid designator.

Rigid designation.: A singular term t is a rigid designator of an object x iff t designates x in all worlds in which x exists, and t never designates anything else.
Test for rigidity.: A singular term t is a rigid designator iff the individual who is t could not have existed without being t, and no one who is not the individual who is t could have been t is true; otherwise t is non-rigid.

Kripke claims that proper names function semantically as rigid designators, but that no description is such a rigid designator.

The modal argument can schematically be stated as follows:

Proper names are rigid designators.
If a description D gives the meaning of a term t, then: D is rigid iff t is.
No description is a rigid designator.
So, the meanings of proper names are not given by descriptions.

1.1.4 Kripke's epistemic argument

The problem of error already evident in the Frege-Russell view goes further than the cluster-description theorist believes. If the meaning of `NN' is exhausted by D, then, according to the cluster view, anyone who believes the proposition `¼ NN ¼' would also believe the proposition `¼ D ¼'. Furthermore, the proposition `If D exists, then D is D' is knowable a priori. Hence, the proposition `If NN exists, then NN is D' should also be knowable a priori. Kripke argues against these consequences of the cluster view.

Suppose that the only thing I know about Caesar is that he was the great Roman commander murdered in 44BC (Kripke's own example is the Gödel/Schmidt scenario in pp. 83-4). Again, suppose my belief is false. Historians discover that Caesar escaped the assassination in 44BC. So, whereas my belief that Caesar was murdered in 44BC is falsified empirically, my belief that the Roman commander murdered in 44BC was murdered in 44BC is not, and can not, be falsified empirically. Therefore, the propositions `Caesar was murdered in 44BC' and `The Roman commander murdered in 44BC was murdered in 44BC' are not the same.

In many instances most or all of our important beliefs about historical figures can be challenged. A sceptic may say: `All we believe about Moses is inaccurate'. But the cluster-description theorist would have to argue that such a statement is false, whatever evidence is presented. It is by definition-or more cautiously, by linguistic rules-that the `most important' properties are correctly attributed to Moses. And such a claim is absurd.

The epistemic argument can now schematically be stated as follows:

When D is a description concerning important characteristics of the referent of a name `NN', the following does not hold:
1. One knows or believes the proposition `NN is F' iff one knows or believes the proposition `D is F';
2. The proposition `If NN exists, then NN is D' is knowable a priori.
Therefore, descriptions related to important characteristics of the referents of proper names do not give their meanings.
Since these are the descriptions commonly associated with proper names by speakers, the cluster view is false, and the meanings of proper names are not given by descriptions speakers associate with them.

1.1.5 Kripke's semantic argument

This sort of argument is sometimes treated as a variation of the epistemic argument, since the example Kripke uses is the same Gödel/Schmidt scenario in pp. 83-4. For the purposes of illustration it is easier to operate with names for obscure characters. Suppose all we know about Thales is that he was the Greek philosopher who believed that all is water. So the description D for the name `Thales' is `the Greek philosopher who believed that all is water'. But suppose now that Thales did not in fact believe that all was water. He believed that all was air. There was another Greek philosopher, Thales' brother born in Egypt, who believed that all was water. According to the cluster view, in uttering, for example, the statement:

Thales was Greek

the speaker must refer to that Egyptian-born brother. However, more plausibly, we should say that in using the name `Thales' we continue referring to that old Miletian philosopher. We can exclaim: `Ah, now we know that Thales did not believe that everything was water; it was his brother who did'.

An important thing to notice here is that the cluster-description theorist may relax his view slightly and hold that the meaning of a proper name is not spelled out by the description D (or: a proper name is not an abbreviation of D). But the description fixes the reference of the name. Thus two claims emerge:

There is a complex description D such that the name `NN' refers to the object x just in case x satisfies D;
The proposition `If NN exists, then NN is D' is knowable by virtue of understanding the name `NN'.

In other words, the reference of a name is fixed by a description even though we remain neutral on the issue of the meaning of proper names.

The Thales scenario (analogously the Gödel/Schmidt scenario) serves to undermine this view, because the name `Thales' keeps its `old' reference on our lips even after the description D designates the Egyptian-born sibling. We could say that the Greek philosopher who believed that everything was water was Egyptian-born, but we would not conclude that Thales was Egyptian-born. Likewise, it is a matter of historical, not semantic, enquiry to determine that Thales was the Greek philosopher who believed that all was water. These observations, according to Kripke, should generalise to other cases as well.

Readings
Kripke [1980],Lectures I,II
Frege [1948] (until the discussion of subordinate clauses)
Searle [1958]

1.2 The causal theory of reference

After criticising descriptivism Kripke proposes an alternative way of thinking about reference of proper names. He states his theory briefly and with hesitation, complaining about the possibility of it being false. Despite those caveats it is worth discussing it in detail, especially in the light of the ensuing debate. We call his view `the causal theory of reference'. Several people vigorously opposed the mention of causality in this label. We shall, however, stick to it, whilst clarifying the precise role of causality.

1.2.1 Baptism and transmission

The account splits into: the account of the initial baptism and the account of the transmission of the name down the causal chain of name-users. Baptism involves descriptions. The name is attached to the object by a description interpreted in a wide sense: it may include verbal descriptions and demonstratives accompanied by ostension. We may have baptisms like:

Let this child [in front of me] be called `Isaac¢

or:

The man previously called `Jacob¢ will now be called `Israel¢.

Baptism itself need not be of causal nature. We may baptise causally inert mathematical objects. Another important thing to notice is that the description need not be associated permanently with the name. It neither gives the meaning of the name, nor does it serve to determine the name's reference. It may subsequently be forgotten or misstated. So the baptising description does not give the meaning of the name. Perhaps what we should add is that there must be an intention of using the name for the same object as before.

The same restriction applies to the use of the name by later users, those people who received the name from the baptiser(s). But the intention to use the name in the same mode must be understood in a very weak sense. It should not even be conscious. Perhaps a better way would be to say that there must be no intention of using the name to refer to a different object.

Later users must also recognise the expression `NN' as belonging syntactically to the category of names. Again, although the introduction of the name into their idiolect may be accompanied by a description, this description need not be true or accurate. I may hear the name `Stalin' from someone who told me:

Stalin was the best friend of Soviet children.

This is not a true statement. Yet I may subsequently use the name `Stalin' to refer to Stalin, even though my information about him is inadequate.

1.2.2 Direct reference

Kripke is sometimes believed to have adopted the so-called `direct reference view'. On this view, sentences that differ only by the substitution of one name by another name for the same object express the same proposition. If we add a plausible assumption that belief and knowledge are relations between persons and propositions, we may get implausible consequences.

Suppose the speaker knows that Stalin died in 1953. Since Stalin is Dzhugashvili, then the proposition:

Stalin died in 1953

is the same as:

Dzhugashvili died in 1953.

Thus, on the direct reference account, our speaker should know the second proposition as well. And this is obviously not true of many speakers.

On Kripke's causal view, there is no mystery. The two names, `Stalin' and `Dzhugashvili', are connected to the speaker through two different chains. If the speaker has been presented with the expression `Dzhugashvili' and has recognised it as a proper name, then he is able to refer to Dzhugashvili. But the chain connecting him to Dzhugashvili (or to the first use of the name `Dzhugashvili') is not the same as the chain connecting him to Stalin (or to the first use of the name `Stalin').

Therefore, there is no direct reference, according to Kripke. Referring is still mediated by causal chains. Notice, however, that, although such a view provides a reference-fixing mechanism, it remains neutral on the issue of the meaning of proper names. Yet it seems as though we have to have some theory of meaning here: we have to explain how the meaning of `Stalin' contributes to the meaning of the sentences (propositions) where it occurs.

1.2.3 Two types of causation

In an early reaction to Naming and Necessity Gareth Evans outlined some of the more popular strategies of criticising the causal theory from a Fregean and Russellian perspective broadly construed. We may agree that descriptivism is crippled by a highly problematic, `old' philosophy of mind. To hold that the speaker is able to refer to an object x just on the basis of his beliefs about x without saying how he acquired those beliefs should look suspicious. There may be more than object that perfectly fit my beliefs. Then it is impossible to see how I refer to one of them, rather than the other.

For example, there may be one of the two twins with which Jake is in love. How do we know, and how does Jake know, that he is in love with Lea, rather than Rachel? We can rephrase our question, so that it would be explicitly about the reference of a proper name. We can ask: when Jake says:

I am in love with Rachel,

does the name `Rachel' on his lips refer to Rachel, or to Lea? If he says that he is in love with x who is F, G, and H, it may well be that both Lea and Rachel are F, G, and H. There is no way of distinguishing them as far Jake is concerned. Or even worse, all those properties may fit Lea, and not Rachel. Then it is plausible to say that Jake's love for Rachel is constituted by the fact that it was Rachel whom he met at a party last night. That is, it was Rachel with whom he stands in a certain causal relation.

One might object that we still help ourselves with a description `the one I met at a party'. But unless we assign a special role to such causally charged descriptions, they may easily be outweighed by descriptions such as `the one who cooks well', `the one who plays violin', which Jake erroneously attaches to Rachel, but which in fact fit Lea. However many descriptions fit Lea, Jake's love for one of the twins is determined by the causal relation between him and her; hence, Rachel remains the only candidate. (A stronger claim would be that under no circumstances Jake could have fallen in love with someone whom he never met, or listened to, or wrote to, and so forth. The sort of a scenario reported in medieval and romantic fiction when a character falls in love with someone on the basis of a testimony alone would be declared an illusion.)

Descriptivism follows the paradigm of the old philosophy of mind, because it pays attention only to the beliefs of the speakers. It is a crucial advantage of Kripke's view that it restores the primary role of causation in shaping our thought about particulars. Yet, while accepting the deficiencies of descriptivism, we should attend also to the difficulties of the causal theory.

First of all, strictly speaking, it is not always necessary to be causally connected to the individual in order to refer to him. Suppose there is an area where people receive their names according to a fixed rule, perhaps related to the names of their ancestors. Once the speaker learns this rule, he can successfully name the residents of that area without getting in causal contact with them.

Secondly, the causal theory presents the ability to refer to x as a sort of magic. Once the speaker overhears the name for x, he cannot err in using it to refer to x. But the ambiguity of context must be removed first. If I overhear a piece of conversation where the name `Clinton' is uttered, I must be able to subsequently refer to Clinton. This makes sense in the immediate aftermath of the conversation when the contextual ambiguity is not great enough. Later on, though, error may creep in, and I may start believing that Clinton was a womaniser. But if the name `Clinton' in the original conversation denoted Hillary Clinton, then it would seem that I have lost my grip on the name, although the causal link was never interrupted.

Thirdly, there are cases of the change of denotation where the causal account returns implausible results. For example, `Madagascar' used to to refer to a portion of the African mainland. As a result of a linguistic misunderstanding it came to denote the island off the African coast. There is, then, an unbroken chain between the current users of `Madagascar' and its original users. But the current users obviously refer to the island. Similarly the name `Anir' originally denoted King Arthur's son, but is used now to denote his burial place. Even though there is an unbroken chain, the switch of denotation is clear.

Therefore, Evans draws two conclusions:

The intentions of the speaker supply neither a sufficient, nor a necessary condition for the correct denotation of the name;
The causal relation must be between the states of the object and the speaker's body of information (rather than between the dubbing of the item and the speaker's use of the name).

Readings
Kripke [1980],Lecture II
Evans [1973],§§1-4

1.3 A neo-Russellian response

Kripke [1980] presents Russell as advocating the view that each proper name possesses a class of descriptions which determines its reference. But it is never certain what sort of theory Kripke has in mind when he talks about the Frege-Russell view. Thus, in one place he pictures Frege and Russell as adherents of a many-descriptions theory, according to which each name is a disguised form of many or all descriptions that the speaker associates with the referent. The textual evidence may be found in Kripke [1980],134-5 where Kripke contrasts the Frege-Russell with the cluster view. The peculiarity of the latter view lies in that `only enough' number of descriptions needs to apply (see Kripke [1980],135). However, in Kripke [1980],31 the cluster view is interpreted as claiming (and improving on Frege-Russell) that

we can't substitute a particular description for the name... What we really associate with the name is a family of descriptions (his italics).

It follows that the Frege-Russell is not entitled to the claim that the speaker should associate the whole range of descriptions, but only that he should be able to replace the name by one such description. This is what Kripke says in Kripke [1980],30.

Perhaps, these exegetical ambiguities are of little or none value, since Kripke consciously does not offer a properly historical interpretation of either Frege or Russell. It has been stated many times since that he misinterpreted all four major figures whom he attacked, and of course his book is not a source or reference in the philosophy of Russell. But there is a need for clarification of the name-description relation. This gap is filled in by Sainsbury [2002].

Sainsbury classifies description theories according to their relation to names, descriptions and occasions. The strongest theory holds that there is a single (unique) description for all speakers and occasions that relates name to the description for the given speaker on the given occasion. The weakest theory holds that for all speakers and all occasions there is a description that satisfies such relation. Sainsbury claims that Kripke never addresses the weakest kind of description theories. Further, the name-description relation may be of three types: fixing meaning, fixing reference and fixing speaker's thought. Evidently, the third type equally escaped Kripke's attention. Sainsbury's exegetical claim, then, is that Russell adhered to the weakest thought-fixation theory.

This seems to be confirmed by Russell's own words:

Common words, even proper names, are usually really descriptions. That is to say, the thought in the mind of a person using a proper name correctly can generally only be expressed explicitly if we replace the proper name by a description. Moreover, the description required to express the thought will vary for different people, or for the same person at different times. The only thing constant (so long as the name is rightly used) is the object to which the name applies. But so long as this remains constant, the particular description involved usually makes no difference to the truth or falsehood of the proposition in which the name appears. Russell, [1912,54]
What enables us to communicate in spite of the varying descriptions we employ is that we know there is a true proposition concerning the actual Bismarck, and that however we may vary the description (so long as the description is correct) the proposition described is still the same. This proposition, which is described and is known to be true, is what interests us; but we are not acquainted with the proposition itself, and do not know it, though we know it is true. Russell, [1912,57]

Russell's claim, then, relates to the possibility of thinking about an object. We cannot think about an object without attributing a property to him. That is the role of descriptions. Secondly-and here the peculiar Russell's metaphysics is displayed in full glory-our success at communication owes to our awareness of the shared proposition that we describe. Although we will not discuss the issue further, it should be clear that the semantic role of descriptions cannot be disentangled from their metaphysical role.

Readings
Russell [1912],ch. V
Sainsbury [2002]

1.4 Nominal descriptivism

If we think of the semantic role of proper names as consisting in denoting individuals, we may naturally try to regard this role as giving us the meaning of proper names. Such is the view sometimes called `the nominal description theory'. A name `NN' would then be semantically equivalent to the description `the bearer of "NN"'.

...Fixit...

1.5 Evans on singular thought

Evans develops his theory of singular reference in response to Kripke's causal view. Ostensibly his purpose is to keep the healthier part of descriptivism intact, while endorsing the causal element in reference fixation. However, the theory he proposes is far richer and more detailed view than either descriptivism or the causal theory.

1.5.1 Frege's input

Evans begins by explaining Frege's notion of sense of proper names. Frege's famous Aphla-Ateb example is interpreted as supporting the claim that everyone who understands the sentence `¼ NN ¼' must think about NN in the same particular way. That is, every competent speaker of the language must be able to grasp the same sense of the proper name.

The notion of the sense of x is interpreted as a way of thinking about x. However, thinking about x in a particular way should not rely on one's knowledge of a description uniquely true of x. The notion of sense applies not only to singular terms, but to sentences as well. The sense of a sentence would be a thought. Frege endorsed the following Intuitive Criterion of Difference: the thought associated with a sentence S as its sense is different from the thought associated with a sentence S¢ as its sense, if it is possible for someone to understand both S and S¢ at the same time, but to take different propositional attitudes (e.g. believing) towards them. It is possible for someone to believe in the truth of the sentence:

Byron was a great poet

and not to believe in the truth of the sentence:

Wordsworth was a great poet,

whilst understanding both sentences. Such is the evidence that those sentences have different senses. Simultaneously the semantic notion of sense must be separated from the pragmatic notion of tone. Byron's line:

And there lay the steed with his nostril all wide

has the same sense as:

And there lay the horse with his nostril all wide,

but different tone. Everyone who understands both sentences and believes one of them, must also believe another, and vice versa. The difference between the two sentences lies in the use of different literary styles.

1.5.2 Russell's input

Evans then moves on to take several important insights of Russell's on board. They may be summarised as follows:

The sole role of genuine referring expressions is the identification of an object;
If the referring expression fails to identify the object, the sentence containing it has no semantic content;
Russell's criterion: if the object purportedly denoted by t can be supposed not to exist without making the sentence `¼t¼' meaningless, then t is not a referring expression.

The upshot of this is that the range of genuine referring expressions is restricted to such expressions as `I', `this', `the blue spot in my visual field', and the like.

Evans formulates the following crucial principle supposedly endorsed by Russell:

Russell's Principle.: A speaker can entertain a thought (or judgement) about x only if he knows which individual in the world he is thinking (or judging) about.

The knowing-which requirement relies on discriminating knowledge: the speaker must be able to distinguish x from all other objects in the world. Evans immediately identifies three conditions of discriminating knowledge. The speaker distinguishes x from all other objects when:

He perceives it at present;
He recognises it if presented with it;
He knows distinguishing facts about it.

Russell's Principle is regarded as an antidote to the Photograph Model of mental representation. Suppose Jake used to know two Chinese twins in his childhood. Years later he has vague memories of one of them, the twin A. He may wonder: `How many children does that Chinese girl have now?' The Photograph Model will produce an answer that he automatically thinks about the twin A, even though he cannot distinguish her from the other twin. That is, the Photograph Model sees the causal link as a sufficient condition for determining the object of representation. The Photograph Model thus omits the requirement imposed by discriminating knowledge.

Evans aims at providing a theoretical defence of Russell's Principle and of the relevant conception of discriminating knowledge. He starts with verificationism, and specifically Dummett's version of it. It is not crucial to consider all the intricacies of Evans' critique of Dummett. What is important to note is simply the fact that verificationism relies on demonstrative identification in its explanation of thought. According to the verificationist, to think the proposition `a is F' we must either be presented with a, or else to be able to recognise a if confronted with it. That is, the verificationist model of explanation satisfies the threefold requirements mentioned above.

The polemic with verificationism yields the following constraint on discriminating knowledge:

The Generality Constraint.: If a subject entertains the thought that a is F, then he must be endowed with conceptual resources to entertain the thought that a is G for every property G of which he has a conception.

Thoughts are structured. We can compare, then, the structure of different thoughts such as `John is happy' and `Jim is happy'. Equally we may compare thoughts `John is happy' and `John is friendly'.

On the other hand, the thinker has the conception of a particular property F when he can predicate it of different subjects. In judging that a is F we exercise this ability. But by a symmetrical argument we can say that to have a conception of an individual a we must be able to predicate various properties of him. The thought (judgement) `a is F' must be seen as the product of exercising two abilities: of thinking of a with various properties predicated of it, and of thinking of F as attached to various individuals.

1.5.3 Proper-name-using practices

Evans starts by advertising a new approach which would consist in taking the social nature language seriously and in tracing its influence in the theory of reference of proper names. Ambiguity in the use of proper names prompts us to ask several questions:

What makes a symbol refer to (represent) a particular individual?
Are the speakers' thoughts and intentions dispensable? No. Therefore:
1. How does the fact that `NN' refers to NN (a social fact) depend on the psychology of individuals?
2. How does the psychology of individuals depend on social facts?

The users of `NN' are divided into two major communities. One is the community of producers. The individual NN is introduced to them normally by demonstration, but not necessarily so. They have a capacity to re-identify NN on later occasions. And they do more than just use `NN': they interact with NN, have dealings with him, and so forth. Crucially, the expression does not become a name for x unless it has a currency among those who know x. (Notice the parallel with knowledge by acquaintance.)

The second group of users is the group of consumers. They are introduced into the practice of using the name `NN' either by a description `NN is F', or merely by hearing sentences where `NN' occurs. The initial piece of information a consumer would associate with `NN' is `NN is someone known as "NN"'. The consumer may accumulate several beliefs about NN, but he would still be unable to justify them. There is, therefore, a principled difference between the producer's initiation into the practice and the consumer's initiation into the practice. Both the producers and the consumers are aware of the social (pragmatic) convention governing the name `NN'. It consists in the stipulation `Let this man be called `NN"'. But it is only the producers who from time to time identify the man as NN; it is only the producers who are credited with the ability to justify the convention.

(Generally, a speaker can be introduced into a practice either by a producer, or a consumer. Notice that, although there is some similarity here with the notion of the division of linguistic labour, it is still limited: producers do not depend on consumers in their use of `NN'.)

Evans has some interesting remarks on the criteria of identity of practices. His discussion is too quick. Roughly put, the identity of a practice is shaped by the flow of information passing between the users. If the users regularly and consistently confuse between two individuals they take to be named `NN', there will eventually emerge a single practice of the name `NN'. Alternatively, the same individual may pass under different names, leading a double life. That is the case of Jekyll and Hyde and, we may add, of some criminal adventurers. Here we shall have two practices, even though the individual named is one and the same.

1.5.4 The fixation of reference

The preceding discussion rather naturally yields the reference-fixing rule. The individual x is the referent of the name `NN' if he is regularly and consistently identified as being called `NN' by the producers of the `NN'-using practice. ...

Readings
Evans [1982],14-22, 42-4, 89-105, 373-404

Chapter 2
Descriptions

We continue our enquiry into singular reference. We have looked at one category of singular terms-proper names-and will presently consider definite descriptions.

2.1 Russell's view

One way to approach Russell's view is by noticing a conflict between the grammatical form of the sentences and their logical form. In his analysis of quantification Frege paved the way in breaking the alliance between grammar and logic. Consider the sentences:

Some physicists are rich

(2.1)

and

All philosophers are poor.

(2.2)

Grammatically those sentences have the subject-predicate form. However, if richness is predicated of physicists, it is not clear of which physicist it is predicated. In fact, it does not matter for the truth of the sentence which physicist is rich. Furthermore, suppose the proposition is about some physicists. Is it about rich physicists? This seems plausible; yet, if it were false, it would not have been about anyone.

Those and similar difficulties led Frege to divorce logical form from the grammatical form. The logical form of the sentences above is given by:

$x(Phys(x) ÙRx)

and

"x(Phil(x) É Px).

Russell's idea may be interpreted as divorcing grammar and logic in the case of description sentences. Those sentences have the form `The F is G'. Consider the sentence:

The author of Childe Harold was mad.

(2.3)

It is tempting to think of it as having the subject-predicate form-even more so, than in the case of (2.1) and (2.2). We do not normally suppose that `some physicists' denote physicists, because we do not know whom exactly. But `the author of Childe Harold' may intuitively be taken to denote an individual, namely, Byron. However, according to Russell, the sentence really should be presented as:

$x(CHx Ù"y(CHy É y=x) ÙMx).

This can be abbreviated as:

Mix CHx.

Notice that the sentence above does not have the subject-predicate form: the expression ix Ax is not regarded as a singular term. Rather, we are dealing here with an existential proposition.

In another curious twist Russell distinguishes between primary and secondary occurrences of definite descriptions. The distinction is easiest to understand as a solution of the problem of negative existentials (tailored to the case of descriptions). Consider first the sentence `Byron is not stupid.' It can be paraphrased as `It is not the case that Byron is stupid', but the paraphrase comes down to the same thing. Symbolically, the sentences `a is not F' and `It is not the case that a is F' are synonymous. Both are rendered as Fa in the canonical notation.

Now consider the sentence `The author of Childe Harold was not nice.' Given our analysis of descriptions, we have a choice. The negation can take either the wide scope or the narrow scope. That is, we choose between:

$x(CHx Ù"y(CHy É y=x) ÙØNx).

and:

Ø$x(CHx Ù"y(CHy É y=x) ÙNx).

Since the author of Childe Harold exists, the two sentences are materially equivalent. But now consider the sentence `The author of the Bible was not nice':

$x(Bx Ù"y(By É y=x) ÙØNx).

(2.4)

and:

Ø$x(Bx Ù"y(By É y=x) ÙNx).

(2.5)

Here the sentence (2.4) is false (there is no unique author of the Bible, nice or not nice), but the sentence (2.5) is true.

Similarly we resolve the case of propositional attitudes reports. Consider the sentence `Bush believes that the author of the US Constitution is nice.' We can interpret this sentence as either the one with the description having a primary occurrence:

$x(Cx Ù"y(Cy É y=x) ÙBush believes that Nx).

(2.6)

or the one with the description having a secondary occurrence:

Bush believes that $x(Cx Ù"y(Cy É y=x) ÙNx).

(2.7)

The original English sentence is true. But, provided there is no unique author of the US Constitution, the paraphrase (2.6) is false. Here we have a hint that definite descriptions should normally have a secondary occurrence in the sentences of natural language, and the paraphrase (2.7) is the right one. As yet another example, we may consider the famous `I thought your yacht was larger than it is.'...

Russell's theory offers an effective treatment of four puzzles. First, empty descriptions no longer present any difficulty. Even if there were no author of Childe Harold-if it were created by a clever computer programme-the statement (2.3) remains meaningful, but false. The description does not contribute the individual it purportedly denotes, but a concept expressed by the predicate it contains (the predicate Ax above).

Secondly, granted that existence is not a predicate, the theory explains paradoxes of existential claims. Consider the sentences:

The King of France exists

(2.8)

and:

The King of France does not exist.

(2.9)

If these are subject-predicate sentences, the sentence (2.8) is meaningless (since its singular term denotes nothing), whereas (2.9) still cannot be true (for the same reason). On Russell's account, paradoxes disappear. We represent these sentences respectively as:

$x (Kx Ù"y(Ky É y=x))

and:

Ø$x (Kx Ù"y(Ky É y=x)).

The third puzzle relates to identity sentences. Consider the sentence:

The author of Childe Harold is the same individual asthe author of Don Juan.

(2.10)

If definite descriptions are referring expressions, then their semantic role is filled in by the individual they denote. Thus the semantic content of (2.10) is the triviality that Byron is Byron. On Russell's account, we can eliminate the paradox by interpreting (2.10) as:

$x(CH(x) Ù"y(CH(y) É y=x)ÙDJ(x) Ù"z(DJ(z) É z=x)).

This shows why (2.10) is informative. Interestingly, the sentence like:

The author of Childe Harold is the same individual asthe author of Childe Harold

is also not trivial. It should be interpreted as:

$x(CH(x) Ù"y(CH(y) É y=x)ÙCH(x) Ù"z(CH(z) É z=x)),

which is logically equivalent to:

$x(CH(x) Ù"y(CH(y) É y=x)).

The content of this sentence is also given by:

The author of Childe Harold exists,

which is clearly not a triviality.

Fourthly, there is the puzzle of belief sentences. Consider the sentence:

Bruce believes that the King of France is mad.

We wish to say that such a belief report is true, even though there is no King of France. If the expression `the King of France' were a referring expression, then that would have been difficult to explain. On Russell's account, no special difficulty arises: the sentence `The King of France is mad' is simply an existential sentence, so that the belief report can be true even though the description has no denotation.

Readings
Russell [1919]

2.2 Strawson's critique

Strawson brings attention to the great role context plays in natural languages. The same singular term (or: singular noun-phrase) can be used to refer to different things in different contexts. This seems like a platitude. Yet it supports a distinction that Russell failed to acknowledge. Meanings attach to declarative sentences. Yet truth values attach to their particular utterances. Thus, in one context the sentence:

The thing on the table is green

may express truth on one occasion and possess no truth value on another. A key distinction that Strawson draws-evidently borrowing it from Quine-is between use and mention (the latter also labelled as `referring'). Consider the sentence:

Ankara is pretty.

Here the name `Ankara' is used, whereas the city of Ankara is mentioned-by virtue of the use of the name `Ankara'. Or we could say, the name `Ankara' is used to refer to the city of Ankara.

Just like Quine, Strawson accuses Russell in confusing use and mention, yet with a twist. His claim is that expressions are used to mention (=refer to) different things. Which things they mention depends on context. Sometimes, in an `unsuitable' context expression may fail to mention any thing altogether. Examples may help here. Consider the definite description `the King of France'. According to Strawson, its meaning is fixed by the conventions of the English language. Any sufficiently competent English speaker understands its meaning. Then, in one context, it may be used to mention Louis XIV, in another context-Philip the Fair. In yet another context, whilst the discussion is about France's current political system, it will fail to mention any particular individual. Therefore, the English sentence:

The King of France is fair

is always meaningful, but the utterance of it on different occasions may be true, false, or truth-valueless. There does not seem to be any reason why we should not treat definite descriptions as singular terms, provided we take care to relativise them to suitable contexts. What about Russell's intuition about existential claims involved in definite descriptions? Strawson does not wish to abandon it completely. His complaint is that such a claim is presupposed (or implied) by the assertion. But it is not the meaning of the sentence itself.

Sometimes it is argued that there is no clear winner in the Russell-Strawson exchange. Yes: Russell's account is not faithful to the surface grammar. But: neither is Frege's analysis of existence. There is no reason to be suspicious of the philosophical analysis of the kind Russell advocates just because it violates pre-theoretic intuitions. It seems to me, however, that there are two issues by which we may judge the exchange. One is whether Russell is guilty in confusing use and mention and whether such a confusion has an impact on the theory of descriptions. A second, and related issue, is how well Strawson's own account reflects pre-theoretic intuitions. Suppose an article is published in a classics journal containing the sentence:

The author of The Iliad was blind.

(2.11)

Suppose we discover after the publication that there was no (unique) author of The Iliad. Are we prepared to say, with Strawson, that the utterance changes its truth value upon the discovery? The whole point of Strawson's view seems to be that truth values must be sensitive to the context in which they are made. Then, the original reader of the article can judge the utterance of (2.11) true, while a later reader will judge it truth valueless.

Readings
Strawson [1950]

2.3 Donnellan's account

Keith Donnellan develops a view intended to provide a criticism of Russell's theory. We have first to clarify what his position is.

The key distinction drawn by Donnellan is the distinction between referential and attributive uses of definite descriptions. If a description is used attributively, it determines which individual is talked about-that is, precisely the individual that fits the description. If a description is used referentially, it enables the hearer to pick out an individual even though it might not fit the description. We must explain the mechanism responsible for the twofold role of descriptions. We can identify several ways in which Donnellan attempts to present it.

First, when a description is used attributively, the statement is about whatever satisfies the description. Consider the statement:

Smith¢s murderer is insane.

(2.12)

If made upon the inspection of Smith's corpse, it is about anyone guilty of the act. The speaker may be ignorant or altogether mistaken about the identity of the murderer: he does not have `in mind' anyone in particular.

If, on the other hand, the statement (2.12) is made in a courtroom, in the presence of the alleged murderer, then the description is used referentially. The speaker has `in mind' a particular person in the dock. So, on this first gloss, in using the description referentially the speaker must be able to think about the individual in another way than the one stipulated by the description. In this case, the speaker still refers to that individual in the dock, even if he is not guilty.

Second, the distinction is clarified along the lines of the distinction between de re and de dicto beliefs (375). In the referential use the speaker expresses a de re belief about a particular individual, whereas in the de dicto use he expresses a belief about whoever uniquely fits the description. Here we may be relying on distinguishing between `saying of' and `saying that'. When the speaker says something of an individual, his belief must be de re. But this is not convincing, since one can speak of an individual and still express a de dicto belief, thinking about the individual under a description. However, the referential-attributive distinction cannot be assimilated to de re-de dicto, for reasons mentioned by Kripke.

Third, Donnellan suggests that attributive use introduces an element of generality (379), unlike the case of referential use. In the attributive use, we `weakly' refer, in the sense that the identity of the individual described does not contribute to the identity of the proposition. With referential use, we assert a singular proposition, whose identity depend on the identity of the individual described.

Donnellan's account can be construed as providing a criticism of Russell. Russell's view (and also Strawson's view) is applicable to the attributive use of descriptions. It neglects their referential use. However, such a criticism holds only if the referential-attributive distinction is semantic and if descriptions are to be regarded as ambiguous. Then we could say that the Theory of Descriptions takes care of one meaning of descriptions.

Kripke defends Russell's view as follows. Even if the Theory of Descriptions is incorrect for English, we can imagine a different language-call it `Russell English'-where it applies. And we can imagine that the speakers of Russell English are competent with the referential-attributive distinction. Therefore, the mere existence of such a distinction is not sufficient for refuting the Theory of Descriptions. Three Russell English languages are possible:

Weak: Truth conditions of sentences with definite descriptions are stipulated to coincide with truth conditions of their Russell translations.
Intermediate: Sentences with definite descriptions are paraphrases of their Russell translations. Descriptions are not terms.
Strong: Descriptions are banned.

Kripke argues that even the speakers of the strong Russell English will be able to recognise the referential-attributive distinction.

Readings
Donnellan [1966]
Kripke [1977]

2.4 Incomplete definite descriptions

Another problem for the Theory of Descriptions seems to arise from incomplete definite descriptions. A definite description `the F' is incomplete if there are many Fs. In the sentence:

The door is open

(2.13)

there is an incomplete description `the door'. In making the utterance (2.13) one presumably is not saying that there is only one door in the world and that it is open.

Kripke suggests that incomplete definite descriptions, at least on the occasions like (2.13), function as indexicals. The sentence (2.13) means a sentence such as:

This door is open.

The occurrence `this door' may also be replaced by `the door I have in mind', `the door in the present context', and so forth.

In defending Russell's view, we may first notice that a similar problem arises for proper names, at least so far as we believe in nominal descriptivism outlined above. The sentence:

John Kennedy was shot dead

(2.14)

may also be regarded as problematic, because there is more than one person in the world named `John Kennedy'. The name `John Kennedy' is shared. But in uttering the sentence (2.14) I presumably mean that a unique individual named so was shot dead. So it is natural to suggest that sentences on such occasions are used non-literally. Literally, they respectively mean that there is only one door in the world, and that there is only one individual named `John Kennedy'. Definite descriptions themselves are semantically unambiguous. And they themselves are used literally. It is the whole utterance which is used non-literally.

In criticising Russell we must show that his theory gets the semantic content of the utterance (or: the meaning of the sentence) wrong. But this is far from clear: uniqueness and completeness of descriptions are not semantic facts.

Several other considerations can be mentioned in support of Russell. Unique proper names and descriptions cannot be treated as indexicals. But then, whether or not the description/name is complete/unique affects the semantic analysis of the sentence. And this seems arbitrary.

Secondly, with the sentence where a real indexical occurs, the speaker cannot mean what the sentence means. Indexicals are referentially indeterminate. But this is not at all so with incomplete descriptions.

Finally, Kripke argues that incomplete descriptions are not always used as a shorthand for complete descriptions. This is true; but Russell did not suppose otherwise. Here we note that `the table' and `this table' cannot be synonymous: the expression `this table' is not used attributively.

Readings
Kripke [1977]

Chapter 3
Demonstratives

We have seen that there are promising attempts-either by Evans, Donnellan, or Strawson-to explain the referring role of proper names and definite descriptions by appeal to the facts of linguistic use. When we turn to indexicals, such as `I', `you', `now', `here', it becomes clear that their referring role should be explained in semantic terms. They denote only relative to the context of utterance. But the meaning of a sentence containing an indexical determines truth conditions. So, while the meaning of indexicals remains fixed, they can be used to refer to different individuals.

3.1 Kaplan's account

The suggestion just mentioned is not that the speaker can use the sentences containing indexicals to refer to different individuals. If reference is determined semantically, it cannot be made dependent on speakers' intentions.

The suggestion seems to apply in the case of the indexical `I'. One must use `I' to refer to oneself. This is a key fact about its meaning. And the identity of the speaker is a key fact about the context of its utterance. No description can capture the the meaning of `I'. For instance, `the utterer of these words' will not do, since I can use the description to refer to the person whom I am now quoting.

The suggestion, then, is that the semantics of indexicals consists in a rule specifying how reference is a function of the context of utterance. In the case of `I' such a rule would specify that `I' is to be used to refer to oneself. So a sentence involving indexicals (`indexical sentence') will a have a semantically determined, yet context-dependent truth conditions.

Kaplan's proposal is to assign two semantic properties to an indexical sentence. It may be assigned a semantic property just as it stands, and it can be assigned a semantic property relative to a context. One property is what is said by a sentence, labelled by Kaplan as its `content'. When Clint Eastwood utters the sentence

I am hot,

(3.1)

the content of this utterance is that Clint Eastwood is hot. When Eli Wallach utters the sentence (3.1), the content of his utterance is that Eli Wallach is hot.

The second property assigned to indexical sentences is their `bare' lexical meaning set by the linguistic conventions of their use. Kaplan labels it `character'. The character of an indexical sentence is supposed to determine its content in varying contexts.

Only contents are semantically evaluated. What Clint Eastwood said with his use of the sentence (3.1) may or may not be true. Similarly, the expression `I' may or may not refer to Clint Eastwood. Therefore, characters are regarded as functions from contexts to contents, whereas contents are functions from circumstances of valuation to semantic values. In the case of sentences those semantic values will their truth conditions. Generally, also non-indexical sentences possess character. But their character is not context-sensitive, being represented by a constant function. An eternal sentence would be a good instance of sentences with constant character.

In other words, the character of `I' remains the same for different speakers and different occasions of utterance (contexts). But its content varies.

A further device introduced by Kaplan is the special demonstrative `dthat'. Its purpose is to fix the reference of the expression occurring within its scope. The idea here is to regard descriptions as a demonstration. The expression `dthat[a]' will then refer `directly', without the interference of the sense of `a'. In other words, `dthat' is a demonstrative `that' with the completing singular term (normally, a description) functioning as its demonstration. For example, we shall have analytical equivalence:

`I¢means the same as `dthat[the person who makes this utterance]¢.

Readings
Kaplan [1978]

3.2 Doubts about Kaplan's view

Do other indexicals behave in a way analogous to `I'? Consider the indexical `now'. The sentence:

I am hot now

(3.2)

has no absolute truth conditions. They must vary with context. The relevant contextual parameters will include the speaker and the time of utterance. The context is fully specifiable.

Consider the indexical `he' and the sentence:

He was hot then.

(3.3)

The sentences (3.2) and (3.3) are structurally analogous. A speaker can also use one of them to express what is being said by the other. But there is no clearly and unambiguously specifiable context. The reference of `he' may depend on the speaker's intentions. (We shall return to this point in our discussion of Stalnaker's views.) Similar problems arise with the indexical `you'.

3.3 Dthat

Kaplan has devised a neat example, utilising the demonstrative `dthat', intended to show that reference is fixed independently of the speaker's intentions.

Readings
Kaplan [1978]

3.4 Self-locating knowledge

The distinction between character and content is further utilised in John Perry's illuminating examples. The principal target of his examples is the thesis of the transparency of thought due to Frege. This is the principle that if a speaker X understands a sentence S and a sentence S¢, then, if S and S¢ express the same thought, X must realise this fact. There would be no possibility for the speaker to think that S, but not to think that S¢. Thoughts are not like mountains observed from different location and presenting different appearances. Such cases of different appearance may be possible if thoughts are characterised as objects of propositional attitudes (beliefs or desires). But so far as they are tied to truth and falsity, they must be available to different speakers. This is the background of Frege's realist dictum that thoughts are transmitted through generations.

However, when we attend to certain indexical expressions, things may look otherwise. Perry considers several examples.

Heimson/Hume.: A mad Heimson believes that he is Hume. Hume, not being mad, also believes that he himself is Hume. Heimson is an expert on Hume and knows a lot about him. So Heimson and Hume may agree on every fact about Hume's biography, but still disagree on who is Hume.
Rudolf Lingens.: Lingens is an amnesiac in the Stanford library. He has found and read a biography of himself. So he knows a lot about Lingens. He knows that Lingens is a cousin of a famous spy. He may even know that Lingens is an amnesiac lost in the Stnaford library. But he still does not know that he is Rudolf Lingens. He may try to exit the library and go to a doctor only if he knows that he is the amnesiac lost in the library.

Two difficulties emerge. First, the character/content distinction cuts deeper than we thought. Some contents (thoughts) are not expressible by eternal sentences. They are tied to their indexical character. Second, it is unclear how to individuate those self-locating, or `perspectival', contents. If it is a different truth, for Lingens, that he is Lingens, even though he already knew all the facts about Lingens, should we say that the thought that today is sunny, expressed today, is the same as the thought that yesterday was sunny, expressed tomorrow? Identity criteria for those thoughts become blurred.

Readings
Perry [1977]

Chapter 4
Truth

We shall start rather unconventionally with Colin McGinn's presentation and criticism of correspondence and coherence theories of truth. We shall then move on to consider Tarski's account and modern deflationary views.

4.1 Shades of the past

A certain basic property of truth is accepted by everyone. It can be put thus:

Disquotational Principle.: S is true iff p.

A slightly mysterious formulation is made transparent with particular examples. Then:

`Snow is white¢ iff show is white.

Therefore, `S' in the above principle features as the name of the proposition `p'. By applying the truth-predicate we move from something referring to a proposition (sentence) to something referring to what the proposition is about.

The coherence theory of truth has the following claim:

S is true iff the belief that p coheres with other beliefs.

This may be seen as leading to a version of idealism. For combining the disquotational principle with it will yield:

p iff the belief that p coheres with other beliefs.

In particular:

Snow is white iff the belief that snow is white coheres with other beliefs.

This looks as an idealist claim, because snow is white whether or not there are beliefs in the world, and because snow's whiteness simply does not consist in the coherence of beliefs. Idealism may be true or false, but it cannot be derived by the analysis of the concept of truth.

The correspondence theory sees truth as objective, independent of people's beliefs or judgements. The world is a world of facts, and those facts, independent of our ability to believe or verify them, make our judgements true or false. Truth becomes a relational concept, a relation between judgements (sentences, propositions) and facts. Hence:

S is true iff S corresponds to the fact that p.

Problems arise here with the correspondence relation. Consider the sentence `Snow is white.' According to the correspondence view, we have:

`Snow is white¢ is true iff `Snow is white¢ corresponds to the fact that snow is white.

But of course the sentence `Snow is white' also corresponds to the fact that snow is not white. It all depends on how we conceive of the correspondence relation. It appears as though truth must be built into the relation to get us the right kind of facts. This explains why the correspondence view has the air of intuitive appeal about it.

Readings
McGinn [2000],87-91

4.2 Tarski's theory of truth

It remains controversial whether Tarski's work is intended to refine the correspondence theory, or is more in line with deflationism. Deflationism is a host of several views. Very roughly indeed, they are allied by the idea that the disquotational principle contains all the information one needs for understanding truth. The place of Tarski's theory is not an issue of merely historical significance. Its proper examination may cast light on the substantive issues within the theory itself and within the accounts it inspired.

Tarski's stated goal is to provide a notion of truth suitable for mathematics and logic and such that would be free of paradox. He calls his account `semantic'. The term is clarified in §5 by saying that a semantic relation obtains between language and objects. This may hold for names, but needs re-working for sentences. Certainly Tarski does not intend to talk about a relation between sentences and what they express.

We can do better by looking at semantic predicates. These include the predicates `designate', `is satisfied by', `determines uniquely', and `true'. A predicate F is semantic if and only if the following holds:

F signifies either a property which only a linguistic expression may have, or a relation in which a linguistic expression may stand to something.
The question whether F holds of an expression E of the language L depends on the meaning of E in L.

Truth is a semantic predicate. A theory of truth must be formally correct (not circular) and materially adequate. The latter is meant to rule out a definition such as:

S is true iff 2+2=4.

Essentially a materially adequate definition is such that illuminates our pre-theoretic intuitions and, to an extent, the ordinary use of the term. Tarski's solution of the problem of material adequacy is the following condition:

Convention T.

A formally correct definition of the predicate `true' for a given language L in the English language is materially adequate just in case it implies all sentences that can be obtained from the T-schema:

S is true in L if and only if p

by substituting for `S' an `informative designator' of a declarative sentence of L and for `p' the English translation of that sentence.

A language L for which `true' is being defined is called `object-language'. The language in which `true' is defined-i.e. the English language in the Convention T-is called `metalanguage'. Tarski maintains that the object-language and its metalanguage must be distinct. They need not be separate, though: the metalanguage may be an enrichment of the object-language. A designator of the expression is informative iff someone who understands it should be able to get hold of the expression it designates. Thus "The curfew tolls the knell of parting day" is informative, whereas the definite description `the first line of Gray's Elegy' is not. (Elsewhere Tarski suggests a different method of `structural-descriptive' names.)

Instances of the T-schema are `T-equivalences' such as:

`Ankara is in Turkey¢is true in L iff Ankara is in Turkey.

Here the object-language L is a fragment of English, and the metalanguage is a fragment of English endowed with the truth-predicate applying to the sentences of L.

Now, the truth of a sentence must depend on the semantic behaviour of its parts. The task is to characterise this behaviour. Clearly parts of some sentences have no truth values, as they themselves are not sentences. We need more general concepts.

We consider a language containing sentences of the subject-predicate form.

Compositionality.: Any sentence Fa in this language is true iff the predicate is satisfied by whatever entity the name designates.

Suppose we have a language with names and predicates fixed. Then:

X refers-in-L to Y iff X is a and Y is Ankara, or X is b and Y is Istanbul.
Y satisfies-in-L X iff X is F and Y is pretty, or X is Istanbul and Y is ugly.
S is true-in-L iff S is Fa and Ankara is pretty, or S is Ga and Ankara is ugly, or S is Fb and Istanbul is pretty, or S is Gb and Istanbul is ugly.

4.3 Doubts about Tarski

Two objections must be mentioned.

4.3.1 Antecedent understanding of the object-language

4.3.2 Projectibility

Readings
Tarski [1944]

4.4 Redundancy

The redundancy account of truth holds that `true' and `false' contribute at most to the force of the assertion, but not to its sense. However, Ramsey finds some place for the utility of `true':

[Sometimes] we cannot in ordinary language eliminate the words `true' and `false'. Thus if I say `He is always right', I mean that propositions he asserts are always true, and there does not seem to be any way of expressing this without using the word `true'. ... [But suppose] for a moment that only one form of proposition is in question, say the relational form aRb; then `He is always right' could be expressed by `For all a, R, b, if he asserts aRb, then aRb', to which `is true' would be an obviously superfluous addition. When all forms of proposition are included the analysis is more complicated but not essentially different; and it is clear that the problem is not as to the nature of truth and falsehood, but as to the nature of judgement and assertion, for what is difficult to analyse in the above formulation is `He asserts aRb'. Ramsey, [1931,106-7]

4.5 Minimalism

According to Horwich's view, the minimal theory exhausts the theory of truth. We must specify what the minimal theory is. It is an infinite collection of `axioms' in the bi-conditional form:

The proposition that snow is white is true iff snow is white.
The proposition that grass is green is true iff grass is green.

and other propositions of such a kind. Those propositions are characterised as instances of the following schema:

Equivalence Schema.: The proposition that p is true iff p.

The connective `iff' is no stronger than the material bi-conditional of propositional logic.

The reason why truth is useful lies in its generalisation capacity. Suppose we wish to formulate a law of self-implication:

If snow is white, then snow is white.
If grass is green, then grass is green.
If Ankara is pretty, then Ankara is pretty.

Infinitely more instances of this sort can be added. Armed with the minimalist account, we can say:

The proposition that if snow is white, then snow is white, is true.
The proposition that if grass is green, then grass is green, is true.
The proposition that if Ankara is pretty, then Ankara is pretty, is true.

Finally we can employ propositional quantifiers:

For all x, if x is a proposition and is of the form `if p, then p¢, then x is true.

Truth functions as a useful device of generalisation that facilitates expression. All of the facts whose expression involves the truth predicate may assume no more about truth than what is expressed by instances of the Equivalence Schema. Finally, we need an additional axiom stating that only propositions are true.

We can now explain the fact that `true' allows inferences from `The proposition that S is true' to the relevant instances of `p' with the following example (where S abbreviates the sentence expressing p): p is true
p = the proposition that S
The proposition that S is true
The proposition that S is true S
S.

Readings
Horwich [1998]
Davidson [1996]

4.6 Thick disquotationalism

Taking the Disquotational Principle to be a piece of meaning analysis leads to the redundancy view. We can resist that by taking the left-hand side to express something logically stronger than the right-hand side. So the logical form and ontological commitment of the left is stronger than that of the right.

The left side contains a predicate denoting a property. It also contains a name denoting a proposition. All this is apparently lacking in the right side. Thus adding the truth-predicate increases the expressive power of the language.

Secondly, the right does not always entail the left. The expression `Sherlock Holmes is a detective' does not entail that `Sherlock Holmes is a detective' is true. Nevertheless, the first expression still functions propositionally, although perhaps it is not strictly assertible. Similarly in the general case of truth-value gaps.

The thick disquotationalist view is that truth is a genuine property, but such that it can be specified without reference to any property. Its criterion of application is embodied in the Disquotational Principle; but there is no attempt there to explain truth in terms of some property logically or materially equivalent to it.

The disquotational nature of truth is revealed in two claims:

Truth is a property of propositions that allows to deduce the fact (generally, a state of affairs) stated by the proposition.
Truth is the only such property.

What about falsity? Here we say that the falsity of a proposition allows to deduce facts opposite to that stated by the proposition.

Is truth a unique disquotational property? Take knowledge: if X knows that p then p. The response for this, and similar counter-examples, is that such concepts themselves involve truth.

Perhaps, though, other properties may licence disquotation. Take the predicate `is intelligible' applied to propositions:

The proposition expressed by this sentence is intelligible.

Here we apparently deduce the fact stated by the proposition from the ascribing the property to that very proposition. Similarly for `is written in English'.

But all such counter-examples apply merely in a range of cases. Truth licences disquotation for all propositions.

The unique role of truth is revealed in two further examples. Suppose there is a community not in possession of the concept of truth. Then the members of this community cannot form the beliefs about the world. To do that, they would need to have the concept of truth.

Also, imagine being told that there is a property F such that when it applies to x, another property G applies to y ¹ x. This would be remarkable, and yet this is precisely how truth works in the Disquotational Principle.

Readings
McGinn [2000],92-108

Chapter 5
Meaning

5.1 Indeterminacy of translation

Suppose that you are a speaker of a language L₁ and that you are undertake translating the sentences of another language L₂ into L₁. You must set up a translation manual (or meaning-theory) for correlating the sentences of L₁ with the sentences of L₂. It is plausible to believe that one translation manual is correct, another is incorrect. There must be some `fact of the matter' as to whether the translation succeeds. Quine argues that there is none. We shall now consider two versions of his argument.

5.1.1 The argument from radical translation

A radical translator faces the task of translating a language of a foreign tribe with no additional information available. He has no information from other persons familiar with L₁ and L₂, or from similarities of syntax and vocabulary. All he has is the use made by tribesmen of their language. The issue at stake is understanding of the notion of meaning. To learn what it is for them to mean such-and-such we must see how we come to know that they mean such-and-such.

Of course the meaning of individual words is determined by convention. But those conventions have to be revealed in behaviour. Here is the claim:

Radical translation.: Facts about behaviour are not sufficient in determining whether a certain meaning-theory is correct.

Evans' objection: The theory of meaning remains determined. We endorse the following:

Compositionality constraint.: A theorem fixing the truth-conditions of S should be derivable from the specification of semantic values of its parts in such a way that the speaker could move to understanding S on the basis of exposure to those parts.

The theorist of meaning must assign semantic values to sub-sentential parts in such a way that the observed assent conditions of whole sentences could be accommodated, and in such a way that the assignment is consistent with the occurrences of those parts in all the contexts.

For example, we may hold as axioms of our theory T₁ of meaning the following sentences governing `gavagai':

("x)(x satisfies `gavagai¢iff x is a rabbit).

(5.1)

and `beyaz':

("x)(x satisfies `beyaz¢iff x is white).

(5.2)

A competing theory T₂ would have a different axiom:

("x)(x satisfies `gavagai¢iff x is an undetached part of rabbit).

(5.3)

On the basis of T₁ we can have:

`a is a beyaz gavagai¢is true iff a is white rabbit.

(5.4)

On the basis of T₂ we can have:

`a is a beyaz gavagai¢is true iff a is a white undetached part of rabbit.

(5.5)

But then we cannot account for the situation of black-and-white rabbit. So perhaps we should switch to T₃ with a different axiom for `beyaz':

("x)(x satisfies `beyaz¢iff x is a part of a white animal).

(5.6)

Then we get:

`a is a beyaz gavagai¢is true iff a is a white part of an animal.

(5.7)

But T₃ will fail when the natives talk about white doors or houses.

5.1.2 The argument from theoretical indeterminacy

The argument can be formulated thus:

Equate observation sentences of L₁ with observation sentences of L₂ through inductive equation of stimulus meanings.
To build theoretical sentences we must utilise analytical hypotheses about semantic structure.
The justification of analytical hypotheses comes from matching up relevant observation sentences.
So far as the truth of a physical theory is under-determined by evidence, the translation of the foreigner's physical theory is under-determined by translation of his observation sentences.
Therefore, translation of physical theories is indeterminate-to the extent that physical theories are under-determined by all possible observations.

However, indeterminacy of translation is additional to the under-determination of theories by evidence. This is so, since, although translation of physical theories is under-determined by evidence, there must still be a fact of the matter about physics.

Therefore, once we have made a choice of physical theory, the choice of the translation manual is still open. The premisses 1 and 2 seem sensible. But what about the premiss 3? Read one way, it may simply be saying that the range of analytical hypotheses is not fixed by stimulus meaning. There may be mutually inconsistent hypotheses, each of them being consistent with stimulus meaning. But then we already have the desired indeterminacy, and there is no novel argument here.

Read another way, the inference is valid so far as we assume that the only facts relevant to translation are translations of observation sentences and that those translations are not determined at the level of stimulus meaning.

Readings
Quine [1960],26-35
Evans [1975]
Quine [1970]

5.2 Davidson on meaning

Davidson's ambition is to build a systematic theory of meaning which broadly utilises Tarski's account of truth. We may start with specifying general strictures on the theory of meaning for a natural language L.

Extensional adequacy.: The theory of meaning must generate a theorem which gives the meaning of each sentence of L.
Compositionality.: The theory of meaning must have a finite number of axioms. Each of the generated theorems must exhibit the semantic structure of the sentence in question.
Interpretation.: The theory of meaning for L must allow us to interpret correctly the speakers of L.

A few comments on compositionality would be in place. A theory of meaning which respects this condition can provide an answer to the questions of how we understand an infinite number of sentences by being exposed to only a finite number of them and how in fact it is possible to learn a language.

The idea behind Davidson's view is that all the meanings are given when the truth-conditions of all sentences are given. His argument can be split into several stages:

The theory of meaning must match each sentence S in L with some sentence in the meta-language.
We must avoid the bi-conditional schema:

S means that p,

because we must eliminate the intensional expression in favour of an extensional connective.
We could try this:

S Û p.
But this is not well-formed, since `S' is a name. So the expression on the left-hand side should be converted into a sentence.
An immediate suggestion would be to supplement the name with a predicate:

S is G Û p.
But now it seems that the predicate `is G' will at least be co-extensive with the truth predicate as spelled out by Tarski.

The argument seems plausible, so far as it does not insist on the uniqueness of the truth predicate to play the theoretical role assigned to it.

There are two problems with this account. One is the issue of correct translation. We must hold that the sentence named by `S' is translated by the sentence replaced by `p'. But any such translation is meaning-preserving. So a threat of circularity lurks here.

Another problem is the following rule linking meaning and truth conditions:

[S means-in-L that p] Û [S is true-in-L iff p].

The rule appears to be endorsed by Davidson, when he says that giving truth conditions is a way of giving the meaning of the sentence. However, we can proceed from left to right, but not from right to left. This is so both when we read `iff' as a material and strict bi-conditional. Consider the case of the material bi-conditional. From:

`Snow is white¢ means-in-L that snow is white

we infer:

`Snow is white¢ is true-in-L iff snow is white.

Now, on the other hand:

Snow is white iff grass is green.

Reading our rule above from right to left and using the rule of substitution, we now derive:

`Snow is white¢ means-in-L that grass is green.

It is tempting now to read `iff' as a strict bi-conditional, that is, as `necessarily iff'. This is already bad enough for Davidson's theory, since its ambition was to have an extensional theory of meaning, thereby excluding modal locutions. But even then the rule fails from right to left. We should only consider the bi-conditional:

Snow is white iff snow is white and 2+2=4,

which gives us:

`Snow is white¢ means-in-L that snow is white and 2+2=4.

In order to respond to these objections we must move beyond the programmatic sketch of Davidson, [1967b]. The strategy we explore does not essentially differ from Evans' strategy in rebutting Quine's indeterminacy argument. We need an interpretational theory of truth to guide us in formulating a theory of meaning.

Suppose there is a language English* for which we have built a truth-theory. And suppose that the T-schema for English* yields the clause:

`Ankara is pretty¢is true-in-English* iff snow is white.

Apart from that, English* is like English. Now the question is as to why exactly English* has such a clause. It may, first, be the case that this is an unstructured sentence. Then no interpreter would attribute English* to us (that is, would equate English with English*). For we are sensitive to the individual contributions of words to the content of utterance, to `what is being said'.

Perhaps, then, this clause is derived from some other theorems governing the components of the sentence. Perhaps `Ankara' refers to snow, and `is pretty' is satisfied by white things. But suppose that the predicate `is green' is satisfied by green things in English* and in English. Then we must consider another clause of the T-schema:

`Ankara is green¢is true-in-English* iff snow is green.

And now the interpreter will discover that English speakers believe that `Ankara is green' is true-in-English, but they disbelieve that snow is green. Thus English* and English are not the same language.

Similarly, consider the predicates `has physical shape' and `has physical size'. They are extensionally equivalent:

`x has physical mass¢is true-in-English iff x has physical size.

Nevertheless they do not mean the same. To show this, take the predicate modifier `large'. The compositional approach must explain the link between `is a large f' and `is a f'. So we must avoid saying that:

`x has a large physical mass¢is true-in-English iff x has a large physical size.

This sentence is false, since there are small objects with a large mass, and vice versa.

Therefore, the idea is that the difference in meaning can be accounted on the basis of the difference in truth value, provided there are sufficient constraints on interpretation. The project of applying the T-schema to every sentence of the language eliminates false entries.

The constraint emerging here is a structural one. A revealing theory of meaning must be such that the meaning specification of a sentence S is determined by the meaning specifications for S₁, ¼,S_n just in case it would be possible to come to know what S means on the basis of prior knowledge what S₁, ¼, S_n mean. Now we must rule out cases where the speaker is wired up to the effect that knowing S triggers knowing S₁, ¼, S_n. So must talk about reasoning from knowing what S₁, ¼, S_n mean to knowing what S means. Hence:

Structural constraint.: It would be possible for someone to proceed by rational inductive means from knowing what S₁,¼, S_n mean to knowing what S means just in case the meaning specifications for S₁, ¼, S_n determine the meaning specifications for S.

Readings
Davidson [1967b]
Dummett [1975]

Chapter 6
Modality

The purpose of our discussion of modality is to prepare technical ground for the semantics of belief to be considered later, as well as to deepen our understanding of the earlier discussions of Kripke and Kaplan. We shall none the less say a few words about the puzzling nature of modal concepts.

Forbes [1985]
McGinn [2000],69-86

Chapter 7
Pragmatics

Stalnaker's pragmatic approach to meaning is another incarnation of the idea that the content (meaning) of a statement is to be explained in terms of its truth conditions.

By the term `pragmatics' here we understand a theory of relations between contents of linguistic expressions and contexts of their use. First, context affects content, because `what the expression says' is affected by where and when it is being used. But, less straightforwardly, contents affect contexts. Stalnaker's theory with both kinds of influence.

7.1 Comparison to demonstratives

Earlier we saw an account of context-dependence suggested by David Kaplan. It was designed to give a semantic theory of indexical and demonstrative expressions. The crux of this approach lies in the distinction between the character and the content of expressions. For every expression f, its character is something that can be known by every competent speaker of the language. When f occurs within some utterance on a particular occasion, its character remains constant. The context of an utterance U is its circumstance. Kaplan proposed to identify four parameters of context: the time of utterance, the place of utterance, the utterer, and the world of utterance. Before we specify the context of utterance, the statement containing a demonstrative expression not only lacks a truth-value, but also does not have content. In Kaplan's idiom, it does not say anything. The explanation is that although demonstratives have constant characters, the function which assigns content for every such character is not constant. It is dependent on one or more parameters of the context.

Robert Stalnaker's approach aims to extend Kaplan's strategy to the kinds of statements other than those involving demonstratives. It employs a different conceptual setting. We consider concrete utterance tokens and for every such utterance token different addressees of the utterance along with the utterer himself may form different pragmatic presuppositions. These latter are nothing but propositions assumed by the speaker, who can either seriously believe in them, assume them `for the sake of argument', or perhaps pretend to believe in them. By occupying himself with utterance tokens, Stalnaker envisages a situation where the addressee of my utterance (call him `Jacob') interprets his sensory experience as an act of assertion. Such an interpretation is an integral part of a successful assignment of semantic values to the particular assertion, i.e. of the Jacob's presuppositions. It has several components. In the first place, it must contain some sort of a behavioural theory allowing Jacob to identify the opening of my mouth as an act of assertion, rather than as an act of singing a song or cleaning my throat. Secondly, Jacob should have a lexical theory associating the sequence of noises emanating from my mouth with linguistic items. When I utter a statement-say, `Snow is white'-Jacob's lexical theory should tell him I utter it in English, rather than in some obscure idiolect vocally indistinguishable from English. Plainly lexical theory employs mundane empirical observations, such as the location of the conversation (London as opposed to Beijing), the name of the addresser (`John Smith' as opposed to `Mao Zedong'), and so forth.

Since Kaplan's theory is concerned with utterance types, his goal is not in accounting for the sensory experiences of the speakers which are responsible for treating certain noises and shapes as utterances in a particular language. It is helpful to think of his theory as explaining those actual conversations which contain only genuine assertions in a native language, so that no foreigners or jokers are allowed to take part in the conversation. On several occasions Stalnaker stresses the absence of a conflict between Kaplan's approach and his own:

[These two are] theories that are applied at different stages in the explanation of speech. And they are not competing [theories]: neither can do the job that the other was designed to do. Stalnaker, [1987,123]

7.2 Presuppositions

Stalnaker is concerned with pragmatic presuppositions. To presuppose a proposition in the pragmatic sense is to make certain assumptions about the context of utterance. Pragmatic presuppositions should be distinguished from semantic ones. A proposition P semantically presupposes a proposition Q just in case Q is necessitated by P and is also necessitated by ØP. A proposition P pragmatically presupposes a proposition Q just in case under normal conditions one can infer that a speaker believes that Q from either his assertion or his denial that P. Generally, then, it is propositions that are presupposed. But a more abstract way of representing presuppositions is to think of them as possible worlds. Those worlds must be relevant to what is going on in the conversation. They will constitute the context set.

Example 1 Suppose I come to the class right after the election night and announce, `Bush is the US President'. Among the worlds that are in my context set will be the following:

w₁ = {There is America, America has a president, Kerry won the election, ...}

w₂ = {There is America, America has a president, Bush won the election, ...}

w₃ = {There is America, America has a president, election was invalid, ...}

Presuppositions, in other words, delimit the set of alternative possibilities that that the speakers intend to distinguish with their speech acts. Each participant in the conversation has his own context set, yet a speaker assumes that all participants presuppose everything that he presupposes.

7.3 Two-dimensional semantics

Consider the statement `You are mad' uttered by me in the company of Jon and Ken. The propositional concept associated with my utterance is given by a 3 ×3 matrix displayed in Table 7.1. Different rows correspond to different contexts-that is, to different sets of presuppositions of the parties involved. Different columns correspond to different circumstances of evaluation-that is, to different beliefs of the speakers about the world.

	i	j	k
i	1	0	0
j	1	0	0
k	0	1	0

Table 7.1: `You are mad'

Suppose that in uttering the statement `You are mad' I am talking to Jon. So the content of what I am saying with my utterance is [Jon is mad]. This content and its truth values are represented along the ith line of the propositional concept. I happen to believe that Jon is mad. Thus d_ii=1. Jon does not believe that he is mad. Thus d_ij=0. Ken also does not believe that Jon is mad. Thus d_ik=0. Now Jon understands what I am saying. So, according to Jon, the content of my utterance is also [Jon is mad]. That is why the ith and jth rows are bound to be identical. Now Ken misunderstands me and thinks that I am talking to him. So, according to him, the content of my utterance is [Ken is mad]. Its truth values are represented along the kth row. I do not believe that Ken is mad. Thus d_ki=0. But Jon does believe that Ken is mad. Thus d_kj=1. Finally, Ken does not believe of himself that he is mad. Thus d_kk=0.

We have said that Jon and myself disagree on the fact that Jon is mad. This is reflected in d_ii ¹ d_ij. But this is not the full story. For me, the fact that d_ij=0 is equivalent to saying that if the world were as Jon believes it is, then Jon would not have been mad. This is a disagreement about a counterfactual state of the world. The fact that we disagree about the actual state of the world is rather reflected in the fact that d_ii ¹ d_jj. Since, however, the utterance has the same content for both Jon and myself, the values of d_ij and d_jj have to be the same. Things are different with Ken. For him my utterance has a different content. So the disagreement with Ken cannot be registered at the ith row. It is reflected in the kth row.

We shall now introduce the notion of a diagonal proposition. It is the set of worlds such that the utterance's propositional concept, evaluated at that world, yields a proposition that is true at that world. There are at least two uses for this notion. First, it may be used to capture Kaplan's notion of character. The diagonal proposition is a proposition read-off by the speakers by virtue of their grasp of the lexical component of the utterance. (This claim should be treated with caution, since there are important disanalogies between character and diagonal proposition.)

Another use arises in utterance involving necessary a posteriori and contingent a priori statements. Suppose I utter the statement `Hesperus is Phosphorus'. How to fill in the propositional matrix for this utterance? First, the statement is true, i.e. actually true. As we have seen, names are rigid designators. They denote the same individual in every possible world. So the utterance expresses a necessary truth. On the other hand, if one of the speakers-Ken, for example-is ignorant about astronomical facts, he might believe the statement be false. And since he is convinced by Kripkean arguments about the rigid designation of proper names, he will also believe it to be necessarily false. The resulting propositional concept is displayed in Table 7.2.

	i	j	k
i	1	1	1
j	1	1	1
k	0	0	0

Table 7.2: `Hesperus is Phosphorus'

However, if we leave the analysis there, we shall violate the following:

Principle of effective communication.: A proposition asserted must be the same relative to each world in the context set.

So we may try to eliminate the world k from the matrix. But the resulting matrix will make the utterance appear trivial. For if it is the case that the speaker uttered a necessary truth relative to every participant in the conversation, then he in effect asserted something trivial. We will violate the following:

Principle of non-triviality.: A proposition asserted is always true in some, but not all, of the possible worlds in the context set.

To satisfy both principles we must interpret the content of the utterance as reflected in the diagonal proposition (see Table 7.3).

	i	j	k
i	1	1	0
j	1	1	0
k	1	1	0

Table 7.3: `Hesperus is Phosphorus' diagonalised

Readings
Stalnaker [1970]
Stalnaker [1978]
Stalnaker [1981]
Stalnaker [1987]

Chapter 8
Metaphor

Davidson [1978]
Searle [1979]
Walton [2000]

Chapter 9
Logical form

Wittgenstein [1929]
Davidson [1967a]
Evans [1976]

Chapter 10
Assignments

1 What is the distinction between determination of meaning and fixation of reference? How significant is it for Kripke's argument?

2 `It seems to me that the most athletic Hollywood actor is Austrian.' Paraphrase this sentence when the definite description has a primary occurrence and when it has a secondary occurrence. So far as possible, formalise the paraphrases in the language of predicate logic. Provide detailed explanations.

3 What difficulties does the sentence `I have been wounded' present, according to Frege? Briefly outline Kaplan's solution of the problem.

4 How different is the situation of a radical translator from the situation of an ordinary translator? Discuss one specific example.

References

[Blackburn and Simmons 1999]: S. W. Blackburn and K. Simmons, editors. Truth. Oxford University Press, 1999.
[Davidson 1967a]: D. Davidson. The logical form of action sentences. In Ludlow [1997].
[Davidson 1967b]: D. Davidson. Truth and meaning. In Ludlow [1997].
[Davidson 1978]: D. Davidson. What metaphors mean. In Inquiries into Truth and Interpretation, pages 245-64. Oxford University Press, 1978.
[Davidson 1996]: D. Davidson. The folly of trying to define truth. Journal of Philosophy, 93, 1996. Available via JSTOR.
[Donnellan 1966]: K. S. Donnellan. Reference and definite descriptions. Philosophical Review, 77: 281-304, 1966. Available via JSTOR.
[Dummett 1975]: M. A. E. Dummett. What is a theory of meaning? In Ludlow [1997].
[Evans 1973]: M. G. J. Evans. The causal theory of names. In Ludlow [1997].
[Evans 1975]: M. G. J. Evans. Identity and predication. Journal of Philosophy, 72: 343-63, 1975. Available via JSTOR.
[Evans 1976]: M. G. J. Evans. Semantic structure and logical form. In Ludlow [1997].
[Evans 1982]: M. G. J. Evans. The Varieties of Reference. Clarendon Press, 1982.
[Florensky 1990]: P. A. Florensky. By the Watersheds of Thought. Pravda, 1990. In Russian.
[Forbes 1985]: G. Forbes. The Metaphysics of Modality. Clarendon Press, 1985.
[Frege 1948]: G. Frege. On sense and reference. Philosophical Review, 57: 207-30, 1948. Available via JSTOR.
[Horwich 1998]: P. Horwich. The minimalist conception of truth. In Blackburn and Simmons [1999].
[Humboldt 1988]: W. von Humboldt. On Language. Cambridge University Press, 1988.
[Kaplan 1978]: D. Kaplan. Dthat. In Ludlow [1997].
[Kripke 1977]: S. A. Kripke. Speaker's reference and semantic reference. In Ludlow [1997].
[Kripke 1980]: S. A. Kripke. Naming and Necessity. Basil Blackwell, 1980.
[Ludlow 1997]: P. Ludlow, editor. Readings in the Philosophy of Language. The MIT Press, 1997.
[McGinn 2000]: C. McGinn. Logical Properties. Clarendon Press, 2000.
[Perry 1977]: J. Perry. Frege on demonstratives. Philosophical Review, 86: 474-497, 1977. Available via JSTOR.
[Quine 1960]: W. V. O. Quine. Translation and meaning. In Ludlow [1997].
[Quine 1970]: W. V. O. Quine. On the reasons for the indeterminacy of translation. Journal of Philosophy, 67: 178-83, 1970. Available via JSTOR.
[Ramsey 1931]: F. P. Ramsey. Facts and propositions. In Blackburn and Simmons [1999].
[Russell 1912]: B. A. W. Russell. The Problems of Philosophy. Oxford University Press, 1912. Cited by 1946 edition. Available online.
[Russell 1919]: B. A. W. Russell. Descriptions. In Ludlow [1997].
[Sainsbury 2002]: R. M. Sainsbury. Russell on names and communication. In Departing from Frege. Routledge, 2002.
[Searle 1958]: J. R. Searle. Proper names. Mind, 67: 166-73, 1958. Available via JSTOR.
[Searle 1979]: J. R. Searle. Metaphor. In A. Ortony, editor, Metaphor and Thought. Cambridge University Press, 1979.
[Stalnaker 1970]: R. C. Stalnaker. Pragmatics. In Context and Content Stalnaker [1999].
[Stalnaker 1978]: R. C. Stalnaker. Assertion. In Context and Content Stalnaker [1999].
[Stalnaker 1981]: R. C. Stalnaker. Indexical belief. In Context and Content Stalnaker [1999].
[Stalnaker 1987]: R. C. Stalnaker. Semantics for belief. In Context and Content Stalnaker [1999].
[Stalnaker 1999]: R. C. Stalnaker. Context and Content. Oxford Univeristy Press, 1999.
[Strawson 1950]: P. F. Strawson. On referring. Mind, 59: 320-344, 1950. Available via JSTOR.
[Tarski 1944]: A. Tarski. The semantic conception of truth. Philosophy and Phenomenological Research, 4, 1944. Available via JSTOR.
[Walton 2000]: K. Walton. Existence as metaphor? In A. Everett and T. Hofweber, editors, Empty Names, Fiction and the Puzzle of Existence. CSLI, 2000.
[Wittgenstein 1929]: L. Wittgenstein. Some remarks on logical form. In Ludlow [1997].

File translated from T_EX by T_TH, version 3.72.
On 13 Apr 2008, 17:29.

Philosophy of Language Lecture notes

Sandy Berkovski

Bilkent University Spring 2007

Abstract

Contents

Chapter 1 Proper names

1.1 Kripke on descriptivism

1.1.1 Frege-Russell

1.1.2 The cluster theory

1.1.3 Kripke's modal argument

1.1.4 Kripke's epistemic argument

1.1.5 Kripke's semantic argument

1.2 The causal theory of reference

1.2.1 Baptism and transmission

1.2.2 Direct reference

1.2.3 Two types of causation

1.3 A neo-Russellian response

1.4 Nominal descriptivism

1.5 Evans on singular thought

1.5.1 Frege's input

1.5.2 Russell's input

1.5.3 Proper-name-using practices

1.5.4 The fixation of reference

Chapter 2 Descriptions

2.1 Russell's view

2.2 Strawson's critique

2.3 Donnellan's account

2.4 Incomplete definite descriptions

Chapter 3 Demonstratives

3.1 Kaplan's account

3.2 Doubts about Kaplan's view

3.3 Dthat

3.4 Self-locating knowledge

Chapter 4 Truth

4.1 Shades of the past

4.2 Tarski's theory of truth

4.3 Doubts about Tarski

4.3.1 Antecedent understanding of the object-language

4.3.2 Projectibility

4.4 Redundancy

4.5 Minimalism

4.6 Thick disquotationalism

Chapter 5 Meaning

5.1 Indeterminacy of translation

5.1.1 The argument from radical translation

5.1.2 The argument from theoretical indeterminacy

5.2 Davidson on meaning

Chapter 6 Modality

Chapter 7 Pragmatics

7.1 Comparison to demonstratives

7.2 Presuppositions

7.3 Two-dimensional semantics

Chapter 8 Metaphor

Chapter 9 Logical form

Chapter 10 Assignments

References

Philosophy of Language
Lecture notes

Bilkent University
Spring 2007

Chapter 1
Proper names

Chapter 2
Descriptions

Chapter 3
Demonstratives

Chapter 4
Truth

Chapter 5
Meaning

Chapter 6
Modality

Chapter 7
Pragmatics

Chapter 8
Metaphor

Chapter 9
Logical form

Chapter 10
Assignments