WAB: "Fragments" |
The following contribution was presented first at the Glasgow University Senior Philosophy Seminar on November
21st, 2000. Original publication on WAB's website (2000.12.1).
Two basic approaches to these substitutivity puzzles have emerged in the literature, one which holds that, for certain types of context, our naïve semantics is inadequate, the other which holds that it is pragmatic considerations that have to be factored in if we are to escape the puzzle. Frege, of course, is the classic exponent of the first type of approach. He holds that we must abandon the naïve assumption that the role of a singular term is always just to stand for an object and that typically a singular term has a sense as well as a reference. In the ascription of a propositional attitude, singular terms occurring in the ascribed proposition refer not to their referents, but to their senses, according to Frege. Hence two attitude-ascribing sentences, identical save for their containing different singular terms, are about different things (viz., the different senses of those terms) even though those terms in non-attitude-ascribing sentences refer to the same object.
In the view of many writers, the postulation of senses does no explanatory work: Faced with a problem concerning the difference in cognitive significance between two true sentences 'a = b’ and 'a = a’, to say that the difference lies in the senses of the singular expressions is to offer little more than to say that sense is the I-know-not-what that accounts for the difference. Later writers in the Fregean tradition have attempted to supply some philosophical backbone to the idea that, in certain contexts, a singular terms refers not to an object, but to (depending on the theory) a mode of occurrence, or an aspect, or a time-slice of that object.
The pragmatic approach, by contrast, holds that there is no need to explain the difference in truth-value between a pair of sentences that are alike save for their having different, but co-referring, singular terms because there is no such difference: the two singular terms refer to the same object, the truth-values of the two sentences are the same; it is only our naïve intuitions that inform us, incorrectly, that the sentences differ in truth-value. In one prominent version of this view, our mistake is to confuse the semantic content of the two sentences with their respective pragmatic implicatures.
I shall approach the problem from a direction opposite to that of the theorists mentioned above. My approach will be to discern the principles that speakers use in determining which singular terms to deploy in their conversations. Once these principles are identified and explained, it becomes easy to see why a speaker who is aiming to avoid speaking falsely should choose one singular term over another, even though both terms are generally used to refer to the same individual. Evidently, then, my approach is pragmatic, investigating, as it does, the principles governing language in use. But I agree with the Fregeans in holding that our intuitions about the truth-values of the attitude-ascribing statements are correct.
The "main mistake" made by Wittgenstein's contemporaries, including his Tractatus self (PI § 23), is, he argues, that they investigated language decontextualized, as a non-spatial and non-temporal structure (PI § 108). "It would produce confusion," he says, "if we were to say : the words of the ...sentence ... have a definite sense, and the giving of it, the 'assertion' supplies something additional. As if the sentence, spoken by a gramophone belonged to pure logic; as if here it had the pure logical sense... The sentence, I want to say, has no sense outside the language-game" (RPP I § 488). Taken in one way, this reads as an extreme position -- one favoured by writers who call themselves 'integrationists'. This is an anarchic view -- it maintains that there are no rules or conventions for the use of words; meaning is invented or negotiated 'on the fly'. One problem with this position is that it treats all language-use as freshly minted metaphor or even as Humpty-Dumpty-speak, thereby blurring important distinctions between various types of discourse. It is hardly likely that this is Wittgenstein's position, given his extensive discussion of the rules of language, of language as an institution. He writes: "To obey a rule, to make a report, to give an order, to play a game of chess, are customs (uses, institutions)" (PI § 199) and in a frequently quoted passage he says: "And hence also 'obeying a rule' is a practice. And to think one is obeying a rule is not to obey a rule. Hence it is not possible to obey a rule 'privately': otherwise thinking one was obeying a rule would be the same thing as obeying it" (PI § 202). This is saying no more than that there are public rules for the use of words -- dictionary definitions may be thought of as one species of such rules -- and, of course, the rules are subject to change over time.
One can study language from the "gramophone" perspective -- such is the project of lexicography and pure linguistics -- but (and I take this to be Wittgenstein's point) if ours is a philosophical interest in such human phenomena as reference, communication and meaning, then it is as well to bear in mind that it is we language-users, not words themselves, that refer and communicate, and that generally the meaning of a word is not a property or an object associated with that word, but is our use of it (PI § 43). The point is illustrated at PI § 525. Wittgenstein cites the sentence "After he had said this, he left her as he did the day before". He comments "Do I understand this sentence ? Do I understand it just as I should if I heard it in the course of a narrative ? If it were set down in isolation I should say, I don't know what it's about. But all the same I should know how this sentence might perhaps be used; I could myself invent a context for it". Wittgenstein is not making the trite point that we don't know what the "this" is or who the 'he' and 'her' are. His point is, rather, that in one context, the speaker may be talking about the state of the woman left, in another of the manner of the man's leaving, in another of the mere fact of the leaving's being repeated. None of these senses is present in the "gramophone" (or grammar book) utterance; there is absolutely no warrant for giving any one of them the privileged status of being the sense attaching to the context-less utterance.2 That would be like saying that there is one primary pronunciation in English of the combination of letters 'o-u-g-h' (a composition of the sounds of its component letters?!) with the others derivable from that. On the other hand, there is one primary pronunciation in English of the combination of letters 'a-n-d', and likewise, there are many utterances the meanings of which are pretty much constant across contexts.
It is tempting to suppose that communication is possible because we have at our disposal a fund of words the fixed meanings of which competent speakers understand -- "a symbolism used in an exact calculus" (BB, p.25). But this is altogether to ignore how we actually use words -- not, indeed, in a completely unruly way, but in ways that are highly adaptable to the purposes at hand and to the context. This is not to say that individuals generally use words idiosyncratically, as a private code; rather, meanings are, as applied linguists put it 'negotiated’ between the partners in a conversation. Thus (to take an extreme example) the headline in the sports section of the newspaper which reads 'Supercallyarefantasticcelticareatrocious’ is entirely intelligible to those few people who are afficionados both of Scottish football and Mary Poppins, but is a mystery to all others.
Donald Davidson, in a similar vein, remarked "there is no such thing as a language" (Davidson 1986, p.173) and later elaborated "we cannot define successful communication in terms of shared meanings, practices or conventions since we have no idea what meanings are until we can abstract them from occasions of use". It is the "calculus conception" of language that he is here vilifying. Meanings change over time; there are also novel or idiosyncratic usages such as malapropisms and new metaphors, and a speaker's departures from rules are likely to provoke no greater reaction than a raised eyebrow; sometimes these deviant uses "catch on" and become part of regular speech behaviour. We come to a conversation with a "prior theory" but, once the conversation gets started, we typically make real-time accommodations (a "passing theory"), adjusting to unforeseen peculiarities of our conversational partner and to the surrounding circumstances. This is not to say (nor does Davidson think) that an individual is free to make Humpty Dumptyish innovations and expect to be understood. Likewise, when Wittgenstein says "we don't use language according to strict rules" (BB, p.24), he is not implying that there are no rules, only that there is some slack to accommodate divergency, invention and various sorts of solecism. Words, as Wittgenstein said, are like tools, and competent speakers are like McGyver in being able to efficiently adapt tools to the contingencies of different situations. This competence, as we shall see, is not employed only in emergency, McGyver-like circumstances, but routinely in ordinary conversations.
I propose not to examine these Wittgenstein/Davidson views directly, but to vindicate them by offering a pragmatic solution (one which hinges on an examination of properties of language-in-use) to the familiar problem concerning the failure of substitutivity salva veritate of co- referring expressions in certain contexts. It should be noted, in the light of what has already been said, that there ought to be some unease about the very title of this problem, since, if expressions themselves do not refer, they do not co-refer. But we shall suppress such disquiet, at least temporarily
It can be a tricky business deciding how a report should be framed. If X is reporting to Y something about Z, then X's choice of words must (or certainly should) be based on what X knows or believes that Y knows or believes about the knowledge and beliefs of X himself, and sometimes also on what X knows or believes about the knowledge and beliefs of Z, and about what X takes Y's beliefs about those to be. There may be conflicting pulls on X's selection of words. For example, if I know that you have often seen the dog Spot roaming the streets and that you know its name without knowing that it's my neighbour's dog, then, were I allowed just one short sentence, I might be torn between "My neighbour murdered Spot", thereby telling you of the fate of a dog that you knew and liked, and "My neighbour murdered his dog", forgoing the foregoing information but instead informing you, usefully, that this was a domestic affair, not a matter of Dirk's taking a dislike to a stray. The Overarching Principle is that a speaker should, within reasonable limits, choose those words that will render his report maximally intelligible to the hearer. This principle is in line with the Gricean maxim of manner. It is a defeasible principle since there will be occasions where it might be worth sacrificing a little bit of intelligibility for some other desideratum such as elegance -- prime example: Wittgenstein's Tractatus. Under normal circumstances, operating within 'reasonable limits' involves not bothering to spell things out at huge length just to protect oneself against willful misunderstanding or adolescent logic-chopping.
Different speakers may use the same sentence to mean different things and a great deal of knowledge about a speaker's beliefs, together with a variety of contextual clues, may be needed if we are to elicit what he meant. Wittgenstein (PI § 79) gives the example "Moses did not exist". Someone uttering this sentence could hardly be singling out an individual and adding, in the same breath, that no such individual exists. But what is the speaker saying? The answer is that different speakers, using tokens of the same type sentence, may be saying a variety of different things. Wittgenstein suggests that, in using that sentence, one speaker might mean "The Israelites did not have a single leader when they withdrew from Egypt", another "The leader of the Israelites was not called Moses", another "There cannot have been anyone who accomplished all that the Bible relates of Moses", and so on. In a given context, a speaker might say "Moses did not exist" because he judges, from his knowledge of the beliefs of his hearer and the contextual clues available (including, of course, the immediately preceding conversation), that his hearer will understand which of the above mentioned senses (or some other) he has in mind. It will be noted that, in only the third of the above expansions could any sort of case be made that it features the name 'Moses' acting as a rigid designator representing a historical individual. In the first expansion, the name does not appear at all; in the second, it is mentioned, but not used. Chomsky has claimed that empirical investigation reveals that "[t]here seems to be no general relation [of representation] that holds between expressions of language and parts of the world, so the nature of that relation cannot be the central question of semantics" (Chomsky, 1996, p.38). This appears to be correct. Names have the plainest claim to be representatives of parts of the world yet, as we have just seen they do not always serve that function, and it would require an extensive survey to discover just how frequently names are used as stand-ins for, or representatives of, their bearers.
What should be clear, just from inspection of this one example, is that what a speaker means, on some occasion, when he says "Moses did not exist", is not generally determinable simply by appeal to Gricean implicatures. Other pragmatic elements play an essential rôle.3 The difference between the contribution to interpretation made by the Gricean maxims and that made by other pragmatic features is readily illustrated. If I know that you know Dirk by name and know him to be my neighbour, and I say to you "Dirk, my neighbour, won't lend me a bowl of sugar", then I exploit the maxim of quantity. The apposition indicates to you that I think that neighbours ought to be neighbourly; I convey that message without saying or implying it but by visibly flouting the maxim of quantity. Now, compare that with a situation where I say to you "My neighbour is insane". What I mean on this occasion, is that Dirk is insane -- I am using the definite description referentially. I know that you know that I have neighbours on both sides, but I am confident that you take my statement to be about Dirk. If we have spent the last half hour talking about Dirk and his misdemeanours then I can rely on the Gricean maxim of relevance to convey to you that it is Dirk whom I mean. But if Dirk had not featured in our preceding discussion, I may let you know that it is he whom I have in mind by pointing at him, or at his house, or at Spot (whom you know was murdered by Dirk) lying badly mutilated in the gutter. I need not bother with any such gesturing if I know that you know my other neighbour, Margaret, to be a model of sanity. I make use of such considerations to indicate to you not something I convey in addition to what my words literally mean, but to indicate what I mean, what I am saying. If you hear my words as a statement about Margaret, that was not what I meant, and that statement would be false (whereas the statement that Dirk is insane is almost certainly true).
It is the content of an utterance (not its character) that is true or false, and what I shall be concerned with in what follows is pragmatic determination of content (and, ipso facto of truth-value). Consider this real-life example. My wife once shook me out of sleep and and said, in a worried tone of voice, 'She's up with her ear'. There is, of course, plenty that was implicated or conveyed by my wife's utterance over and above what she meant -- for example, that it was my turn to go through and soothe our daughter. What she meant, however, was that, at the time of utterance (the middle of the night), our daughter had got out of bed, troubled by a lingering ear infection. This is not what my wife implicated, but what she meant, and what she said turned out to be true. It would be extremely unnatural to say that my wife's words literally meant something different from that -- for what does "She's up with her ear" literally mean ? My wife could use so short a sentence because she knew that I would catch on -- and she is not one to waste her breath on long sentences, certainly not at that time of the night.
Instead of asking how names and other singular terms relate to the world (which, if Chomsky is right, would be to commit the fallacy of the complex question), let us ask the less theoretically loaded question "What do speakers use such terms for ?". One important use of a singular term is to draw the attention of a hearer to some object. A sensible speaker will use the most appropriate term to best serve that purpose. What object a speaker succeeds in drawing his hearer's attention to when he uses a singular term may depend on that hearer. It may depend on some complicated facts about that hearer, such as what the hearer believes and believes about others' beliefs, including those of the speaker, or it may depend on something simple, like geographical location or on what was the topic of a recent discussion between speaker and hearer. A speaker normally uses a name or (referentially) a definite description to refer to the object geographically or conversationally in the vicinity of himself and his hearer, not to some object remote from them and their conversation. If I say "John", I am usually referring to the nearest John. It is a convention that when, for example, I referentially use the phrase "the table" when talking to someone, I am, ceteribus paribus, talking about the nearest table not (say) the third nearest. Such conventions (J.L. Austin (1950) called them "demonstrative conventions" for linking a statement to some particular state of affairs) help locate the intended referent.
One of the ways I succeed conversationally is if the object to which I intend to draw an audience's attention by my use of a singular term coincides with the object to which the audience's attention is drawn. Without such coincidence, speaker and audience will be at cross-purposes. In a conversation, there may be no unique referent of a singular term if speaker and hearer take the term differently. It is clear, for example that, for a definite description used referentially, the individual to whom the speaker intends to refer may be different from the one to whom the hearer thinks the speaker is referring. So a circumspect speaker will select a singular term with care if he wishes the hearer to catch his intentions. He may well use a different singular term for drawing the attention of a different hearer to the same individual. As sensitive speakers, we defer to the hearer; our singular terms are used deferentially. For the benefit of my child, I use the phrase "your mum" to refer to the same person I describe to the police as "that woman you wanted to question about the recent shoplifting at Woolworths".
There is a similar distinction between the statement that a speaker intends to make and the statement his audience takes him to be making. The criteria of identity for a statement differ from those for the sentence used to make the statement. The truth value of a statement in which reference to some object is made will depend inter alia on what object is being referred to, and that cannot be "read off" the sentence used; we need to ascertain the referent, by adverting to such factors as context, the intentions of the speaker and the way that the audience takes the speaker's words. Without the necessary meeting of minds, no unique statement can be identified. A simple example, one which does not turn on the use of singular terms, will illustrate the point. If I say to Aziz: "Mother Theresa kicked the bucket", then I would take myself to be saying nothing less than the truth, for in my idiolect that statement means "Mother Theresa died". But Aziz, who does not share my idiolect will say that my statement is false on the grounds that Mother Theresa never kicked anything in her life. Aziz misidentifies what statement I intended to make; my sentence is not enough for him to go on. And a "neutral" observer would surely say that there are two propositions (two contents) here, one intended by me, the other understood by Aziz; that I had intended to state or affirm one proposition while he had interpreted me as stating something different in my use of those very same words. Of course, if I had any sense, I would in the first place have chosen words that Aziz would be unlikely to misinterpret.
Contextual factors frequently intrude in very subtle ways to help fix both reference and meaning. If I say "He finished the bottle", then what "the bottle" is being used to stand for (an object or its contents) may depend on whether I am talking about an omnivore or a baby. I can use the name "Chomsky" to refer to the author who inspired that last example (Chomsky 1995, p.23), but when I say "Chomsky occupies two metres on my bookshelf", I am generally referring to his books, rather than recalling the youth whom I put up in my office before he became a famous linguist. Or consider the statement "The ball is round", uttered during a squash game just as (as it happens) the racket strikes the ball. If the utterance is made by an observer familiar with squash but weak in materials science, the statement is false; if the utterance is made in response to someone unversed in squash who, witnessing the game from a distance, is enquiring about the shape of squash balls, the statement is true (Travis 1996). Or take the sentence "Cuba is similar to Jamaica" (an example of Amos Twersky's). This sentence is true if uttered in a seminar on grain production, but false if uttered in a political science seminar. Conversely, in the latter seminar, "Cuba is similar to N. Korea" is true; in the former false. Another example, due to Chomsky: The sentence "I live near Boston and far from Sydney" is true if I tell that to you, but false if I tell it to a Martian (Chomsky 1996, p.47). In such cases, where we succeed in telling the truth, we are able to do so by using concise sentences rather than convoluted ones because we can rely on our audience, in the particular context of utterance, being able to "fill in" what is missing. Sometimes, such "filling in" requires little more than the mere (mental) adding of a few words; other times, as in Wittgenstein's "Moses" example or the example about my daughter's earache, a proper reconstruction of the speaker's meaning requires a wholesale rephrasing.
If Wittgenstein is correct about the variety of ways in which the sentence "Moses did not exist" is used, then it would be wrong to say that, in that sentence, the name "Moses" stands for the object Moses. What we can say is that, in different uses of that sentence, the name plays a variety of different rôles, and this is evidenced by the variety of things that speakers can mean when using the sentence. For example, if, in a given context, a speaker uses the sentence "Moses did not exist" to mean "The Israelites did not have a single leader when they withdrew from Egypt", all we can say is that the shorter sentence does duty for the longer; it would be foolish to press the question "For what part of the longer sentence is "Moses", in the shorter one, an abbreviation ?". The way that the statement, expressed one way, maps onto the statement expressed the other way is not by straightforward replacement of bits of one sentence by bits of the other. Devices for condensing statements (metonymy is one of the simplest to explain) have evolved over a long period and, though a short sentence may be grammatically simple, its semantics (i.e. a detailed account of what the sentence, used in a given context, means) may be hugely complicated. Language-use develops like an organism -- as Wittgenstein remarked in his early work, "Everyday language is part of the human organism and is no less complicated than it ... from the external form of the clothes one cannot infer the form of the thought they clothe, because the external form of the clothes is constructed with quite another object than to let the form of the body be recognized .... The silent adjustments needed to understand everyday language are enormously complicated" (TLP 4.002). The "object" for which language is constructed is to communicate effectively -- clearly and concisely.
On one occasion, meant one way, the statement "Moses did not exist" may be true; on another occasion, a quite different statement meant by the speaker in using that same sentence may be false. It takes great ingenuity (although it comes easily to us) to draw upon mutual knowledge of the preceding conversation, of our surroundings and our beliefs to ensure that the way an audience understands a speaker's concise statement is just the way that the speaker intends it. So, for example, if we have been conversing about what we both know to be a liberal citizens' committee in Copenhagen anxious to institute an improved civic prostitution service, I may say, truly, "The citizens' committee wants prostitutes to be taught new skills". The sense of that statement is, of course, quite different from that of the statement I would be making if, using that same sentence, I were speaking truly about a conservative committee in Manila advocating the retraining of prostitutes for some more respectable line of work. What the phrase "new skills" is being used to refer to is different in each case. It may be that, in this case, context merely helps the hearer select the intended senses of ambiguous words ("new" as "additional" or "different"; "skill" as "technique" or "occupation"). But context can also be exploited to impose novel sense. The distance between the tip of a man's nose and the tip of his penis is roughly half a metre, but, in Zbigniew's case, it is several thousand kilometres, since he lives in the United States but was circumcised in Poland. There is enough slack in the rules to allow us to extend the use of 'tip' to a detached part, and much verbal humour depends on inventing such acceptable (intelligible) extensions. Of course, in humour we want creative or imaginative extensions of the extension, whereas in our ordinary reporting, the exploitation of the physical or conversational environment for the purpose of securing understanding tends to be much more systematic and pedestrian.
When stating the Overarching Principle, we mentioned the desideratum of not spelling things out at tedious length and, as we have seen, this is often achieved through various condensation devices, some quite standard -- such as the use of pronouns and other proforms -- others which rather more inventively draw on the linguistic and situational context. It should be stressed that condensation is the norm rather than the exception -- the number of "eternal sentences" occurring in ordinary conversation is extremely small. There are very good reasons why our utterances are normally condensed, including
(a) We don't want to bore our conversational partners.
(b) Frustra fit per plura quod potest fieri per pauciora (Ockham's Razor).
(c) Time is money.
(d) We are biologically disposed to speak concisely. For example, as Julius Moravcsik points out, "in view of limitations on attention span and the perceptual mechanism, there must be a premium on brevity and perceptual perspicuity in syntax" (Moravcsik, 1983, p.235)
(1) Lois kissed Superman before she kissed Clark Kent.
One theory to explain why, though (1) is true (since Lois did kiss Superman before she kissed Clark Kent),
(2) Lois kissed Superman before she kissed Superman.
is false, is that the rôle of a referring expression is to stand for some temporal phase of its bearer. But, as Saul shows (1997a, p.104), this proposal runs into trouble with
(3) Superman is Clark Kent.
for here the temporal phases allegedly referred to are different. So this theory would say that (3) is false -- the wrong answer. Saul also demonstrates that a neo-Fregean account, due to Graeme Forbes (1997), which invokes modes of attire, cannot be right. For we should not wish to analyse
(4) Lois slept with Superman before she slept with Clark Kent.
as
(4F) Lois slept with Superman, so-attired, before she slept with Clark Kent, so-attired.
Fixing up the Forbes account would, as Saul shows (1997b), be a complicated and epicyclic project. What it, and the preceding theory (that referring expressions stand for temporal phases) have in common is that they assume that (1) has a fixed meaning (the "gramophone" meaning) which it is for a semantic theory to reveal; they share the defect of neglecting the rôle of context in determining meaning.
Some philosophers, such as Kent Bach (1987), would say that (1) must be false, since Superman and Clark Kent are the same person. Would we want to say that strictly speaking, they are correct ? Clearly not -- there is no "strictly speaking" about it -- language itself is not the arbiter of right and wrong. Linguistic intuitions are not infallible, and they are, to a certain extent, pliable, yet it would be foolish to be receptive to a theory which says that most people don't know how to use their own native language and as a consequence take to be true many statements which are false, and vice versa. People who know the story say that (1) is true, and they are right to do so. If Bach buys his daughter a Clark Kent doll when she had specifically asked for a Superman doll, she would think it a pretty poor joke were Bach to explain that, strictly speaking, he had bought her exactly what she requested. Someone who utters (1) with the intention of uttering a truth is not in error and has not misused language in any way. We use language to say what we want to say, and fail only on rare occasions, for example, when we are mistaken about the common meaning of a word we use. Here is a typical scenario in which, in using (1), the speaker speaks the truth: Martine and I, having just seen the Superman movie, fall into a dispute over whether Lois slipped her arms around the supersuit and kissed its wearer before or after she slipped her arms round the grey suit and kissed its wearer. If, in this case, Martine utters (1) then she is right; she says something true. And neither of us take ourselves to be debating, in effect, the truth-value of (2), even though we both know that "Superman" and "Clark Kent" are names of the same individual. In her use of (1), Martine intends, and I take her to be intending, something like
(1a) Lois kissed the guy when he was doing his Superman thing (wearing the kit etc.) before she ever kissed the same guy when he was dressed as Clark Kent the newspaper reporter.
The (mental) "filling in" required to recover sense (1a) from (1) is obviously quite considerable.
In other contexts, the use of (1) expresses a falsehood. Suppose, for example, we know that Millicent has not seen the movie, has received only a garbled version of the plot and, in particular, is unaware that Superman is Clark Kent. If Millicent now utters (1), we should tell her, correctly, that she has said something false, and would go on the explain that, since Superman and Clark Kent are one and the same person, each Superman-kissing is simultaneous with, is in fact identical to, a Clark Kent-kissing. In this case, what Millicent is saying, and what we take her to be saying is
(1b) (E!x)(E!y)(~(x = y) & x = Superman & y = Clark Kent & Lois kissed x before she kissed y)
or (borrowing an idea of Quine's), making verbs out of names
(1b') (E!x)(E!y)(~(x = y) & x supermans & y clarkkents & Lois kissed x before she kissed y)
where 'to superman' is to do the Superman thing'. While (1a) is a spelling out of what both Martine and I mean, (1b) or (1b') spell out what Millicent means.4 Thus, Sentence (1) is used in different contexts to make different statements with different truth-values, and this is due not to the presence of indexicals or ambiguous expressions in the sentence, but is due to the different circumstances of utterance.
Given the context of the conversation between Martine and myself -- where we each know that the other has just seen the film -- it makes entirely good sense for us to use (1) as a convenient abbreviation for (1a). Although (1a) is more explicit than (1), we would never use it in normal conversation because there are useful and subtle mechanisms, which every competent speaker employs, for engaging audience knowledge of the context in such a way that, though using a short (condensed) sentence, the speaker can be confident that the audience will be able to retrieve the "explicit" meaning. Now, comparing (1a) and (1), two lessons become clear: First, that it is pointless to ask who the names "Superman" and "Clark Kent" stand for in (1), for they do not even feature in (1a) -- this case is therefore very similar to the Wittgenstein example we were considering earlier where, for certain uses of "Moses did not exist", it is pointless to ask what object the name "Moses" stands for; second, that the reason we may have had for inclining us to think that we might be able to substitute "Superman" for "Clark Kent" in (1) -- namely that both names stand for the same object -- has now vanished, for, in this case, they stand for no object at all.
What contribution, then, do the names "Superman" and "Clark Kent" make to the meaning of (1) as used by Martine and me ? The answer, in most general and truistic terms, is that they serve the function of helping to ensure that, taken together with the relevant contextual factors, it is meaning (1a) that is yielded in the context in which Martine and I are conversing. Our use of the names in this context is not to pick out an individual but to concisely paraphrase (1a) by appropriating the way that some other people (including Lois) use those names. But that is by no means the only way in which we can have linguistic fun when there are two names of one individual at our disposal. If I say to you
(5) Diana Spencer was quiet and reserved, but Lady Di was the effervescent Queen of Hearts.
then I have said something true. And the (rather difficult) "spelling out" of (5) which would make my intended sense explicit draws quite differently on context than does the spelling out of (1). That (5) (as just used by me) is true, and not a contradiction, is proof that the names occurring in it are not there being used to denote a single individual. To try to insist that they must be being used to denote something (a sense, a temporal phase etc.), on the grounds that the logical rôle of a name is to denote, is symptomatic of the philosopher's unhealthy craving for generality. Natural languages are full of nooks and crannies; all manner of ingenious ways have developed for doing things with words and the supposition that names behave in a uniform manner is wholly without empirical warrant. The requirement of crystalline purity is only a logician's fantasy (PI § 107).
Although Saul terms her sample sentences which contain no attitude, modal or quotational constructions "simple", they are in fact the most testing ones for any theory designed to cope with the substitutivity phenomenon, for they leave least scope for manoeuvre. My own suggestion has been that there is no problem, because there is no constraint requiring substitution of "co-referentials" salva veritate in sentences, and that supposing that there is such a constraint betrays a fundamental illusion about the project of semantics -- in fact, it commits the "main mistake" identified by Wittgenstein, that of thinking that the semantics of natural languages is the study of language as an abstract system rather than of words used in a context by speakers. It is because of proper reliance on contextual factors which the speaker knows will enable his audience to grasp what he means that the speaker frames a statement in the way that he does. The speaker can expect audience uptake if, through previous experience with that, or a similarly placed audience, the speaker can see that his audience uses words and actions in much the same way as he does himself, and catches on to the same clues. This is, as Wittgenstein puts it, "part of the framework on which the working of our language is based" (PI § 240). But we are wonderfully flexible, and if we have reason to believe that our hearer does not share our form of life (or does not yet share it as is the case with normal young infants just getting to grips with the ways of the words and the world), then we adapt. As Davidson says, "if you are pretty sure that somebody is going to interpret you in a 'non-standard' way, then you are foolish if you don't speak in a non-standard way" (Davidson, 1993, p.119).
Once we are properly sensitive to the ways in which contextual factors are relied upon to ensure audience uptake, we can always see why a speaker might refuse to substitute a "semantically co-referring expression" for the one he used. We can see why such a substitution could alter the sense and the truth-value of the original statement. A speaker of (1), talking to someone else who knows the movie, would refuse to substitute "Superman" for "Clark Kent" simply because he is aware that appeal to context, demonstrative conventions and speaker's knowledge would not allow the audience to retrieve the intended sense (1a) from the resulting sentence. Put positively, a competent speaker intending to convey the message of (1a), knows that uttering (1) is the succinct way of doing just that when in the presence of a competent and knowing hearer.
In a recent paper, Alex Barber (2000) notes that a sensitive analysis of simple sentences should acknowledge that it makes a difference whether utterers of such sentences are in the know or are ignorant with regard to the significance of the singular terms used in their utterances. Barber calls these two types of utterer Cognoscenti and Ignoramuses respectively. He considers the simple sentence
(**) Superman is more successful with women than Clark Kent.
He thinks that this sentence is 'semantically false' but can be used to convey a truth, and that what truth it conveys can be figured out by adverting to the Gricean principles for co-operative conversation. Suppose someone (whom Barber calls CognoscenteA) utters (**). Barber's story about how Cognoscente B figures out the intended implicature of A's utterance goes like this:
A just said something false, knowing that I know this, and knowing that I know she knows, etc. This amounts to a violation of the first maxim of quality. What can A be up to, short of violating the Cooperative Principle? Perhaps it is this. Some are unaware that Superman is Clark Kent. They would, under forseeable epistemic conditions, utter what A just uttered. A must be implicating that just those epistemic conditions obtain that would prompt an ignoramus to utter (**) (though not 'Superman is more successful with women than Superman'). In other words, A is trying to convey to me that Superman/Clark Kent is more successful with women when appearing as Superman -- or, as we might say, is Supermanizing -- than when appearing as Clark Kent.
Apart from the fact that we may doubt whether it is realistic that cognoscenti -- even smart cognoscenti -- go through such a process of reasoning about what their conversational partners think that ignoramuses would say, the account does not work.. For imagine a dialogue between two cognoscenti, one a man who is trying to elicit from a woman her take on what makes for male attractiveness:
Tom: Why is Superman so much more attractive to women than Clark Kent, even though really they're one and the same person?
Mary: Well, it's said that clothes don't make the man, but that's untrue. A drab grey suit is a major turn-off, but put a guy in a supersuit that really shows off his credentials and...bingo.
Surely an ignoramus would not make the utterance that Tom made above. In my view (**) is a true condensed utterance.
We pointed out earlier that, although linguistic intuitions are not sacrosanct, it is foolish to declare that sentences we use successfully in ordinary discourse for ordinary purposes of communication do not have the meaning we think they have just because some theory tells us that they don't. For there is nothing more to the sentences having the meaning we think they have other than their having the use they do have in ordinary successful communication. This seems obvious, and it is curious that a straight thinker like Simon Blackburn has talked himself into believing otherwise. Blackburn concocts a story about a tourist -- let's call him Cedric -- who takes the train up Snowdon and finds the journey disappointing (for it's true that the scenery on that train ride is fairly bland and the slopes are not spectacular). Next day, however, Cedric gets more ambitious and goes climbing up Yr Wyddfa across Crib Goch. He thinks it is great and says so, without realising that Yr Wyddfa is actually Snowdon. According to Blackburn, it is true (but not helpful) to say of Cedric that he thinks that Snowdon is duller than Snowdon (Blackburn, 1984, p.333). But Blackburn is wrong; it plainly is not true. One can, of course, envisage the locals sitting in the pub ridiculing Cedric; one of them might say "Cedric thinks that Snowdon is duller than Snowdon", but none of them seriously believes that Cedric thinks that Snowdon is duller than itself. The data to be explained is that, while they might truly say that Cedric believes that Snowdon is duller than Yr Wyddfa, they would not say (because it's patently false) that Cedric believes that Snowdon is duller than Snowdon. Cedric himself would not assert "Snowdon is duller than Snowdon" and, in reporting his beliefs faithfully, they must say things as he sees them, unless they are out for crude distortion or a cheap joke. The obligation to report someone's beliefs from that person's perspective is what I shall call the "Perspectival Principle". It is a defeasible principle, in that it may be overridden. And another point to make is that, in respecting this principle, we must sometimes refuse, when reporting an attitude, to mimic the sentence used by the person expressing that attitude. For sometimes the person may use the wrong words, and if we want to report what his real attitude is, we should use the right words. Failure to do so is what Joseph Moore (1999) has called "misdisquotation".
Blackburn's mistake is shared by many philosophers who write on this topic. Jennifer Saul, for example, holds that a proper pragmatic account of attitude ascription will vindicate the Salmon-Soames claim that some of our intuitions about the truth-values of sentences are mistaken, and will explain why we are prone to make such errors (Saul 1998). This is completely misguided: our intuitions are not mistaken, so it is not our errors that need to be explained; if anything is to be explained it is how an erroneous philosophical theory establishes so strong a hold that, in order to preserve it, theorists are willing to abandon common sense. The situation here is quite different from cases where claims based on untutored common sense are shown, by mathematics or by science, to be mistaken. Many or even most people may believe that there must be a greatest prime number. They are wrong, and this can be proved. But that is very different from being mistaken about what the words we use mean, for if most people use a word "W" to mean X, then X is what "W" means (or at least it is one of the meanings of "W"); there is no other criterion. This is not to deny that this meaning may fade in time under pressure of various sorts. All this is obvious to anyone who has to update a dictionary. So we cannot say that a person who claims that the statement "If there is life on Mars, then 2+2=4" is false is in error. Most people use "If ... then ... " to signify a ground-consequence relation, so that is what the connective means, or that, at least, is one of its meanings.5
As was the case with the "simple" sentences, it is easy to find non-indexical, non-ambiguous sentences used for making attitude-reports which have different truth values when used to make statements in different contexts. Thus, while
(6) Lois Lane believes Superman is a reporter.
would normally be taken as false, we can, following Jonathan Berg (1988, p.371), imagine a scenario in which (6) is used to express a truth: Two people are marvelling at Superman's ability to conceal his identity. One says to the other, "Look, there's Superman in his Clark Kent outfit; he's incredibly convincing! Everyone thinks he's a reporter -- Jimmy Olson, Mr. White -- why, even that clever Lois Lane believes Superman is a reporter". In this scenario, the "spelled out" version of (6) is
(6a) Lois Lane believes that the guy whom she knows only as "Superman" when he is doing his superheroic thing, but whom you and I know both as "Superman" and "Clark Kent" is a reporter.
Of course, the speaker can avoid this mouthful by relying on his audience's relevant knowledge and their knowledge of Lois' relative ignorance and their knowledge of the devices, either well established or invented "on the fly", which, given this kind of context, permit the shrinking of (6a) down to (6). In this context, it will be noted, we jettison the Perspectival Principle because, with Superman in full view of us, the most salient perspective is our own. Similarly,
(7a) Lois Lane believes that the guy whom she knows only as "Clark Kent" when he is doing his newspaper reporter stuff, but whom you and I know both as "Superman" and "Clark Kent", can fly.
shrinks, all being well, to
(7) Lois believes that Clark Kent can fly
and, as normally used, is false. Nathan Salmon says that, on his account, such a statement is true, but he does, without apparent embarrassment, mark that claim with an exclamation mark (Salmon 1989, p.247). It's as if he realizes that his readers will think he is nuts, but hopes to talk them round into believing that they and ordinary speakers are wrong in the way they make attitude-ascriptions such as (7). As I tried to make clear in my discussion of Blackburn and Saul, this hope is in vain.
Every case of failure of substitutivity of "co-referring expressions" can be explained when we flesh out the problem sentences in the appropriate ways, and see why the fleshing out of what the speaker means is not captured by the result of substituting the "co-referring term" in the original unfleshed sentence. Take the true report
(8) Maggie thinks that Odile is tired.
Mark Richard (1990, p.2) writes, "For Maggie to think that Odile is tired, she must have some representations of Odile and of being tired 'put together' in an appropriate way. In some broad sense of 'sentence', she must employ a mental sentence saying that Odile is tired". Now suppose that, unknown to Maggie, Odile is Maggie's mother, but Maggie thinks that the visibly indefatigueable Catherine is her mother. In this scenario, we are correctly inclined to say that
(9) Maggie thinks that her mother is tired.
is false, even though it results from the true (8) just by substitution of a "co-referring term".
Richard, as we have seen, couches his theory of attitude-ascription and his solution to puzzles such as this within a "language of thought" framework. But this is quite unnecessary. In order to say that Odile is tired, Maggie must, of course, have the appropriate words, but her thinking that Odile is tired need not be split into word-like parts. Maggie's mental representation of Odile (if she has one) may be graphical rather than verbal. And her thinking that Odile is tired may just consists in her taking there to be close enough resemblance in relevant respects between Odile and some other tired individuals. When she thinks that Odile is tired, Maggie stands in a relation not to a proposition or a sentence, but to Odile -- just as she does if she wants Odile not to be tired (where, as Ben-Yami notes (1997, p.85), the verb "wants" has no propositional object). One can surely notice a similarity between one droopy-eyed, slow-moving object and another, without the word "tired" or "tiredness" or any mentalistic counterparts of these being present in one's noticing; lions do. 6
So, in reporting what Maggie thinks, I am under no obligation to quote the words (or translations of the words) from her mental sentence, since there is no such thing. But there is a pragmatic principle to the effect that, other things being equal, I should use something close to those words (or a translation of them) that the reported (the person about whom the report is being made) would herself use.7 We have called this "the Perspectival Principle". The existence of this principle is easy enough to explain: in reporting people's attitudes one generally wants to say things as they see them, and how they see things is how they would say them if they could (and if words are adequate to the task). Departure from this principle is usually the result of pressure from what I called "the Overarching Principle" -- to make oneself maximally intelligible to one's audience, i.e. to the reportee.
Sometimes to use those words that the reported would use would be at the expense of intelligibility to the reportee, so the reporter changes those words. In reporting Maggie's thinking to you, I choose a sentence that will be maximally useful to you. Thus, if neither you nor Maggie know that Odile is her mother -- let us suppose that both of you think that Catherine is (and know that Catherine's name is "Catherine") -- it would be wrong for me to tell you "Maggie thinks that her mother is tired". Wrong, both in the sense of being misleading, and in the sense of being false. For the operative pragmatic principle (in this case the Overarching Principle trumps the Perspectival Principle) dictates how the sentence is to be interpreted -- dictates what statement a speaker, conforming to such principles intends, and what statement an audience should figure him to be intending. Note that, in this case, the usual means for making the report explicitly de re is not much help. For, if I say to you "Maggie thinks of her mother that she is tired" then, so long as you think that Catherine is Maggie's mother, you will take my statement to be false and it is moot (undecidable) whether the statement really is true or false. For decidability and clarity, I should de-condense my statement and say "Maggie thinks, of whom I, but not you, know to be her mother that she is tired". I may actually put it in this way, rather than simply saying "Maggie thinks that Catherine is tired" because I choose to use this occasion to finally reveal a dark family secret.
Here, then, is the prescription: If A would sincerely assert "I ø that Fc", then R, reporting to H, makes a true ascription by saying "A øs that Fc" UNLESS framing the report that way would either be unintelligible to H or would give H a false impression; in which case, the Overarching Principle kicks in, and the true report would be "A øs that Fd". So, for example, if R conveys to H, demonstratively or by some other means that he will be using the name "d" for the item in question, then he must so use it. This is what happens in the Berg example, where R lets H know that he plans to use the term "Superman" for that guy -- the one who, at the time of the demonstration, is wearing a grey suit and spectacles. Again, if R knows that H associates the name "c" with something other than the item about which R is reporting, but that H knows the name "d" of that item, then his true report to H will be "A øs that Fd". That was the case in our report about Maggie, the girl who didn't know the true identity of her mother.
Kripke’s Pierre provides an interesting illustration of the use of these principles. Here, the Reporter (Kripke) and the Hearer (Ourselves) know far more about Pierre’s situation than does the Ascribee (Pierre). So, in reporting Pierre, the Perspectival Principle is trumped. Here, the faithful reporter must use the uncondensed sentence "Pierre believes that London (the parts he has seen in magazines, when he called the city 'Londres’) is pretty". Kripke, in trying to bully us into anwering 'yes’ or 'no’ the question 'Does Pierre or does he not believe that London is pretty?’ commits the Fallacy of the Simple Question. This is just a lawyer’s trick to get a witness to tell a distortion of the truth.
Laurence Goldstein: A Wittgensteinian Approach to Substitutivity Puzzles
Keywords: Attitudes, Pragmatics, Substitutivity, WittgensteinAbstract
There are well known cases where the replacement in a statement of one singular term by another appears to result in a statement with a truth-value different from that of the first even though both singular terms refer to the same object. A pragmatic solution to such puzzles, in the Gricean tradition, argues that this appearance is deceptive: both statements, taken literally, have the same truth-value, but the utterance of one of them is in some way "inappropriate". The most common type of example used to illustrate the "failure of substitutivity" phenomenon involve statements ascribing propositional attitudes. Yet the phenomenon manifests itself in much simpler types of statement, and any decent solution must handle these cases too. The non-Gricean pragmatic solution offered here avoids (as Wittgenstein wished to avoid) the notion of a statement "taken literally", i.e. divorced from its context of use. A speaker's choice of singular term depends on the context of utterance, including who the audience is. What a speaker means generally depends on context thus broadly construed. Very subtle pragmatic principles, possessed by all competent speakers of a language, normally allow an audience to retrieve what a speaker means in any given context. Such principles also dictate that, in general, one cannot substitute so-called "co-referring expressions" salva veritate.0. Introduction
Frege identified a puzzle in the philosophy of language: that one sentence can have a cognitive significance different from another even though the second sentence is just the first with one of its singular terms replaced by a co-referring expression. Thus, the revelation that the Morning Star is the Evening Star agitated the belief-systems of ancient Babylonians in a way that the announcement that the Morning Star is the Morning Star would not have done. To say that a proposition has cognitive significance does not mean that it has the same impact on the cognitive economy of all individuals. Some ancient Babylonians, doubtless, took astronomical discoveries in their stride and promptly understood and embraced them; the more ignorant, suspicious or cynical ones would have held back. If Zog were one of one of the latter, then a contemporary ascription, 'Zog believes that the Morning Star is the Evening Star’ is false, while 'Zog believes that the Morning Star is the Morning Star’ is true. Thus, Frege’s original example easily transforms into one concerning propositional attitude ascription, where substitutivity of co-referential expressions can fail to preserve truth-value. A version of this problem was known in antiquity: You do not know that hooded man; that hooded man is your brother……therefore you do not know your brother. The conclusion is false, but substituting "that hooded man" for the co-referential "your brother" would deliver a truth.Two basic approaches to these substitutivity puzzles have emerged in the literature, one which holds that, for certain types of context, our naïve semantics is inadequate, the other which holds that it is pragmatic considerations that have to be factored in if we are to escape the puzzle. Frege, of course, is the classic exponent of the first type of approach. He holds that we must abandon the naïve assumption that the role of a singular term is always just to stand for an object and that typically a singular term has a sense as well as a reference. In the ascription of a propositional attitude, singular terms occurring in the ascribed proposition refer not to their referents, but to their senses, according to Frege. Hence two attitude-ascribing sentences, identical save for their containing different singular terms, are about different things (viz., the different senses of those terms) even though those terms in non-attitude-ascribing sentences refer to the same object.
In the view of many writers, the postulation of senses does no explanatory work: Faced with a problem concerning the difference in cognitive significance between two true sentences 'a = b’ and 'a = a’, to say that the difference lies in the senses of the singular expressions is to offer little more than to say that sense is the I-know-not-what that accounts for the difference. Later writers in the Fregean tradition have attempted to supply some philosophical backbone to the idea that, in certain contexts, a singular terms refers not to an object, but to (depending on the theory) a mode of occurrence, or an aspect, or a time-slice of that object.
The pragmatic approach, by contrast, holds that there is no need to explain the difference in truth-value between a pair of sentences that are alike save for their having different, but co-referring, singular terms because there is no such difference: the two singular terms refer to the same object, the truth-values of the two sentences are the same; it is only our naïve intuitions that inform us, incorrectly, that the sentences differ in truth-value. In one prominent version of this view, our mistake is to confuse the semantic content of the two sentences with their respective pragmatic implicatures.
I shall approach the problem from a direction opposite to that of the theorists mentioned above. My approach will be to discern the principles that speakers use in determining which singular terms to deploy in their conversations. Once these principles are identified and explained, it becomes easy to see why a speaker who is aiming to avoid speaking falsely should choose one singular term over another, even though both terms are generally used to refer to the same individual. Evidently, then, my approach is pragmatic, investigating, as it does, the principles governing language in use. But I agree with the Fregeans in holding that our intuitions about the truth-values of the attitude-ascribing statements are correct.
I. Language and Language-Use
Wittgenstein passed this verdict on his philosophical contemporaries: "The main mistake made by philosophers of the present generation, including Moore, is that when language is looked at, what is looked at is a form of words and not the use made of the form of words" (LA, p.2).1 Language should not be regarded as a system of signs which latch, in a mechanical way, onto situations (the so-called "calculus" conception (PI § 81) that he had embraced in his early work), but as words used in context, in the performance of various sorts of action, this interweaving of words and action constituting our "language-games" (PI § 7). To imagine a language thus properly conceived, i.e. as the greatly heterogeneous amalgam of our language-games, is to imagine a form of life (PI § 19). A form of life is a manifold of types of activity characteristic of a given species -- for humans, the types of activity that are particularly striking are those that are speech-involving, i.e., our language-games. Our form of life is complicated (PI, p.174) as is evident from the multiplicity and diversity of the language-games in which we engage -- some are listed at PI § 23, with the prefatory remark: "the term 'language-game' is meant to bring into prominence the fact that the speaking of language is part of an activity, or of a form of life". The term "language-game" is a rather unfortunate translation of "Sprachspiel". The latter derives from two verbs, "sprechen" (to speak) and "spielen" (to play). Playing need not be playful, and Wittgenstein certainly does not use the word, in this context, to connote frivolous play; he is alluding to the various rôles that spoken words play in a great variety of human activities.The "main mistake" made by Wittgenstein's contemporaries, including his Tractatus self (PI § 23), is, he argues, that they investigated language decontextualized, as a non-spatial and non-temporal structure (PI § 108). "It would produce confusion," he says, "if we were to say : the words of the ...sentence ... have a definite sense, and the giving of it, the 'assertion' supplies something additional. As if the sentence, spoken by a gramophone belonged to pure logic; as if here it had the pure logical sense... The sentence, I want to say, has no sense outside the language-game" (RPP I § 488). Taken in one way, this reads as an extreme position -- one favoured by writers who call themselves 'integrationists'. This is an anarchic view -- it maintains that there are no rules or conventions for the use of words; meaning is invented or negotiated 'on the fly'. One problem with this position is that it treats all language-use as freshly minted metaphor or even as Humpty-Dumpty-speak, thereby blurring important distinctions between various types of discourse. It is hardly likely that this is Wittgenstein's position, given his extensive discussion of the rules of language, of language as an institution. He writes: "To obey a rule, to make a report, to give an order, to play a game of chess, are customs (uses, institutions)" (PI § 199) and in a frequently quoted passage he says: "And hence also 'obeying a rule' is a practice. And to think one is obeying a rule is not to obey a rule. Hence it is not possible to obey a rule 'privately': otherwise thinking one was obeying a rule would be the same thing as obeying it" (PI § 202). This is saying no more than that there are public rules for the use of words -- dictionary definitions may be thought of as one species of such rules -- and, of course, the rules are subject to change over time.
One can study language from the "gramophone" perspective -- such is the project of lexicography and pure linguistics -- but (and I take this to be Wittgenstein's point) if ours is a philosophical interest in such human phenomena as reference, communication and meaning, then it is as well to bear in mind that it is we language-users, not words themselves, that refer and communicate, and that generally the meaning of a word is not a property or an object associated with that word, but is our use of it (PI § 43). The point is illustrated at PI § 525. Wittgenstein cites the sentence "After he had said this, he left her as he did the day before". He comments "Do I understand this sentence ? Do I understand it just as I should if I heard it in the course of a narrative ? If it were set down in isolation I should say, I don't know what it's about. But all the same I should know how this sentence might perhaps be used; I could myself invent a context for it". Wittgenstein is not making the trite point that we don't know what the "this" is or who the 'he' and 'her' are. His point is, rather, that in one context, the speaker may be talking about the state of the woman left, in another of the manner of the man's leaving, in another of the mere fact of the leaving's being repeated. None of these senses is present in the "gramophone" (or grammar book) utterance; there is absolutely no warrant for giving any one of them the privileged status of being the sense attaching to the context-less utterance.2 That would be like saying that there is one primary pronunciation in English of the combination of letters 'o-u-g-h' (a composition of the sounds of its component letters?!) with the others derivable from that. On the other hand, there is one primary pronunciation in English of the combination of letters 'a-n-d', and likewise, there are many utterances the meanings of which are pretty much constant across contexts.
It is tempting to suppose that communication is possible because we have at our disposal a fund of words the fixed meanings of which competent speakers understand -- "a symbolism used in an exact calculus" (BB, p.25). But this is altogether to ignore how we actually use words -- not, indeed, in a completely unruly way, but in ways that are highly adaptable to the purposes at hand and to the context. This is not to say that individuals generally use words idiosyncratically, as a private code; rather, meanings are, as applied linguists put it 'negotiated’ between the partners in a conversation. Thus (to take an extreme example) the headline in the sports section of the newspaper which reads 'Supercallyarefantasticcelticareatrocious’ is entirely intelligible to those few people who are afficionados both of Scottish football and Mary Poppins, but is a mystery to all others.
Donald Davidson, in a similar vein, remarked "there is no such thing as a language" (Davidson 1986, p.173) and later elaborated "we cannot define successful communication in terms of shared meanings, practices or conventions since we have no idea what meanings are until we can abstract them from occasions of use". It is the "calculus conception" of language that he is here vilifying. Meanings change over time; there are also novel or idiosyncratic usages such as malapropisms and new metaphors, and a speaker's departures from rules are likely to provoke no greater reaction than a raised eyebrow; sometimes these deviant uses "catch on" and become part of regular speech behaviour. We come to a conversation with a "prior theory" but, once the conversation gets started, we typically make real-time accommodations (a "passing theory"), adjusting to unforeseen peculiarities of our conversational partner and to the surrounding circumstances. This is not to say (nor does Davidson think) that an individual is free to make Humpty Dumptyish innovations and expect to be understood. Likewise, when Wittgenstein says "we don't use language according to strict rules" (BB, p.24), he is not implying that there are no rules, only that there is some slack to accommodate divergency, invention and various sorts of solecism. Words, as Wittgenstein said, are like tools, and competent speakers are like McGyver in being able to efficiently adapt tools to the contingencies of different situations. This competence, as we shall see, is not employed only in emergency, McGyver-like circumstances, but routinely in ordinary conversations.
I propose not to examine these Wittgenstein/Davidson views directly, but to vindicate them by offering a pragmatic solution (one which hinges on an examination of properties of language-in-use) to the familiar problem concerning the failure of substitutivity salva veritate of co- referring expressions in certain contexts. It should be noted, in the light of what has already been said, that there ought to be some unease about the very title of this problem, since, if expressions themselves do not refer, they do not co-refer. But we shall suppress such disquiet, at least temporarily
II. Framing Reports
Parties to a conversation typically possess differing knowledge including differing knowledge about what each other knows. If conversation is to be co-operative, speakers should make use of the knowledge that they think each other possesses. So, for example, if I know that you know my neighbour, but don't know that his name is Dirk, I should not report to you that Dirk murdered his dog in those words; I should say to you "My neighbour murdered his dog". The choice of words is pragmatic -- the words are chosen in a way that takes account of the conversational situation. I want to argue that all the problems associated with the substitutivity of co-referring expressions can be solved once we pay attention to the pragmatic principles that guide the manner in which reports are framed. The key to the solution lies in recognizing that one of the complicated aspects of our form of life is that we report events and situations to fellow agents, and tailor our words to the particular capacities and limitations of the particular agents to whom we are reporting. This obvious fact is routinely overlooked in discussions of this subject. Robert Brandom, for example, makes the perfectly proper point that, in investigating attitude-ascription, we need to distinguish "which aspects of what is said express commitments that are being attributed and which express commitments that are undertaken [by the ascriber]" (Brandom 1994, p.173). Likewise, Bart Geurts investigates principles governing what a speaker who makes an attitude-ascription presupposes about the state of the world and about the beliefs of the ascribee (Geurts 1998, pp.551-7). But Brandom and Geeurts completely ignore those aspects of what is said that take account of what the reporter knows, or thinks he knows, about the person to whom the report is being made. The pragmatic principles which guide the framing of such reports are subtle, in the sense that they require us to fine-tune our words to the multi-dimensional particularities of different conversational situations, but the most general of these principles -- what I shall call the "Overarching" and the "Perspectival" principles -- are rather easy to state, and their existence is easy to explain.It can be a tricky business deciding how a report should be framed. If X is reporting to Y something about Z, then X's choice of words must (or certainly should) be based on what X knows or believes that Y knows or believes about the knowledge and beliefs of X himself, and sometimes also on what X knows or believes about the knowledge and beliefs of Z, and about what X takes Y's beliefs about those to be. There may be conflicting pulls on X's selection of words. For example, if I know that you have often seen the dog Spot roaming the streets and that you know its name without knowing that it's my neighbour's dog, then, were I allowed just one short sentence, I might be torn between "My neighbour murdered Spot", thereby telling you of the fate of a dog that you knew and liked, and "My neighbour murdered his dog", forgoing the foregoing information but instead informing you, usefully, that this was a domestic affair, not a matter of Dirk's taking a dislike to a stray. The Overarching Principle is that a speaker should, within reasonable limits, choose those words that will render his report maximally intelligible to the hearer. This principle is in line with the Gricean maxim of manner. It is a defeasible principle since there will be occasions where it might be worth sacrificing a little bit of intelligibility for some other desideratum such as elegance -- prime example: Wittgenstein's Tractatus. Under normal circumstances, operating within 'reasonable limits' involves not bothering to spell things out at huge length just to protect oneself against willful misunderstanding or adolescent logic-chopping.
Different speakers may use the same sentence to mean different things and a great deal of knowledge about a speaker's beliefs, together with a variety of contextual clues, may be needed if we are to elicit what he meant. Wittgenstein (PI § 79) gives the example "Moses did not exist". Someone uttering this sentence could hardly be singling out an individual and adding, in the same breath, that no such individual exists. But what is the speaker saying? The answer is that different speakers, using tokens of the same type sentence, may be saying a variety of different things. Wittgenstein suggests that, in using that sentence, one speaker might mean "The Israelites did not have a single leader when they withdrew from Egypt", another "The leader of the Israelites was not called Moses", another "There cannot have been anyone who accomplished all that the Bible relates of Moses", and so on. In a given context, a speaker might say "Moses did not exist" because he judges, from his knowledge of the beliefs of his hearer and the contextual clues available (including, of course, the immediately preceding conversation), that his hearer will understand which of the above mentioned senses (or some other) he has in mind. It will be noted that, in only the third of the above expansions could any sort of case be made that it features the name 'Moses' acting as a rigid designator representing a historical individual. In the first expansion, the name does not appear at all; in the second, it is mentioned, but not used. Chomsky has claimed that empirical investigation reveals that "[t]here seems to be no general relation [of representation] that holds between expressions of language and parts of the world, so the nature of that relation cannot be the central question of semantics" (Chomsky, 1996, p.38). This appears to be correct. Names have the plainest claim to be representatives of parts of the world yet, as we have just seen they do not always serve that function, and it would require an extensive survey to discover just how frequently names are used as stand-ins for, or representatives of, their bearers.
What should be clear, just from inspection of this one example, is that what a speaker means, on some occasion, when he says "Moses did not exist", is not generally determinable simply by appeal to Gricean implicatures. Other pragmatic elements play an essential rôle.3 The difference between the contribution to interpretation made by the Gricean maxims and that made by other pragmatic features is readily illustrated. If I know that you know Dirk by name and know him to be my neighbour, and I say to you "Dirk, my neighbour, won't lend me a bowl of sugar", then I exploit the maxim of quantity. The apposition indicates to you that I think that neighbours ought to be neighbourly; I convey that message without saying or implying it but by visibly flouting the maxim of quantity. Now, compare that with a situation where I say to you "My neighbour is insane". What I mean on this occasion, is that Dirk is insane -- I am using the definite description referentially. I know that you know that I have neighbours on both sides, but I am confident that you take my statement to be about Dirk. If we have spent the last half hour talking about Dirk and his misdemeanours then I can rely on the Gricean maxim of relevance to convey to you that it is Dirk whom I mean. But if Dirk had not featured in our preceding discussion, I may let you know that it is he whom I have in mind by pointing at him, or at his house, or at Spot (whom you know was murdered by Dirk) lying badly mutilated in the gutter. I need not bother with any such gesturing if I know that you know my other neighbour, Margaret, to be a model of sanity. I make use of such considerations to indicate to you not something I convey in addition to what my words literally mean, but to indicate what I mean, what I am saying. If you hear my words as a statement about Margaret, that was not what I meant, and that statement would be false (whereas the statement that Dirk is insane is almost certainly true).
It is the content of an utterance (not its character) that is true or false, and what I shall be concerned with in what follows is pragmatic determination of content (and, ipso facto of truth-value). Consider this real-life example. My wife once shook me out of sleep and and said, in a worried tone of voice, 'She's up with her ear'. There is, of course, plenty that was implicated or conveyed by my wife's utterance over and above what she meant -- for example, that it was my turn to go through and soothe our daughter. What she meant, however, was that, at the time of utterance (the middle of the night), our daughter had got out of bed, troubled by a lingering ear infection. This is not what my wife implicated, but what she meant, and what she said turned out to be true. It would be extremely unnatural to say that my wife's words literally meant something different from that -- for what does "She's up with her ear" literally mean ? My wife could use so short a sentence because she knew that I would catch on -- and she is not one to waste her breath on long sentences, certainly not at that time of the night.
III. Condensed Utterances
As is well known, the substitution of "co-referring expressions" in a report may result in a statement which possesses a truth-value different from that of the original report. This has long been regarded as a puzzle and various baroque theories have been proposed to solve it; the main focus of attention has been on those statements which report the psychological attitudes of others. Early attempts to solve the puzzle include Frege's suggestion, mentioned earlier, that there is more to the semantics of a singular term than its Bedeutung (variously translated as "reference" and "meaning"), and that, in the propositional clauses of attitude reports, singular terms do not signify their normal Bedeutungen, but their senses. Russell's radical idea was that, surface apearances to the contrary, definite descriptions and ordinary proper names do not refer (and a fortiori do not co-refer). Both Frege and Russell offer "internalist" solutions -- solutions which hold that the identity of the statement made by a speaker is totally determined by the identity of the speaker and the (logically disambiguated) sentence used. This, by Wittgenstein's lights, exemplifies the "main mistake" of looking at a form of words and not the use made of the form of words in a particular context.Instead of asking how names and other singular terms relate to the world (which, if Chomsky is right, would be to commit the fallacy of the complex question), let us ask the less theoretically loaded question "What do speakers use such terms for ?". One important use of a singular term is to draw the attention of a hearer to some object. A sensible speaker will use the most appropriate term to best serve that purpose. What object a speaker succeeds in drawing his hearer's attention to when he uses a singular term may depend on that hearer. It may depend on some complicated facts about that hearer, such as what the hearer believes and believes about others' beliefs, including those of the speaker, or it may depend on something simple, like geographical location or on what was the topic of a recent discussion between speaker and hearer. A speaker normally uses a name or (referentially) a definite description to refer to the object geographically or conversationally in the vicinity of himself and his hearer, not to some object remote from them and their conversation. If I say "John", I am usually referring to the nearest John. It is a convention that when, for example, I referentially use the phrase "the table" when talking to someone, I am, ceteribus paribus, talking about the nearest table not (say) the third nearest. Such conventions (J.L. Austin (1950) called them "demonstrative conventions" for linking a statement to some particular state of affairs) help locate the intended referent.
One of the ways I succeed conversationally is if the object to which I intend to draw an audience's attention by my use of a singular term coincides with the object to which the audience's attention is drawn. Without such coincidence, speaker and audience will be at cross-purposes. In a conversation, there may be no unique referent of a singular term if speaker and hearer take the term differently. It is clear, for example that, for a definite description used referentially, the individual to whom the speaker intends to refer may be different from the one to whom the hearer thinks the speaker is referring. So a circumspect speaker will select a singular term with care if he wishes the hearer to catch his intentions. He may well use a different singular term for drawing the attention of a different hearer to the same individual. As sensitive speakers, we defer to the hearer; our singular terms are used deferentially. For the benefit of my child, I use the phrase "your mum" to refer to the same person I describe to the police as "that woman you wanted to question about the recent shoplifting at Woolworths".
There is a similar distinction between the statement that a speaker intends to make and the statement his audience takes him to be making. The criteria of identity for a statement differ from those for the sentence used to make the statement. The truth value of a statement in which reference to some object is made will depend inter alia on what object is being referred to, and that cannot be "read off" the sentence used; we need to ascertain the referent, by adverting to such factors as context, the intentions of the speaker and the way that the audience takes the speaker's words. Without the necessary meeting of minds, no unique statement can be identified. A simple example, one which does not turn on the use of singular terms, will illustrate the point. If I say to Aziz: "Mother Theresa kicked the bucket", then I would take myself to be saying nothing less than the truth, for in my idiolect that statement means "Mother Theresa died". But Aziz, who does not share my idiolect will say that my statement is false on the grounds that Mother Theresa never kicked anything in her life. Aziz misidentifies what statement I intended to make; my sentence is not enough for him to go on. And a "neutral" observer would surely say that there are two propositions (two contents) here, one intended by me, the other understood by Aziz; that I had intended to state or affirm one proposition while he had interpreted me as stating something different in my use of those very same words. Of course, if I had any sense, I would in the first place have chosen words that Aziz would be unlikely to misinterpret.
Contextual factors frequently intrude in very subtle ways to help fix both reference and meaning. If I say "He finished the bottle", then what "the bottle" is being used to stand for (an object or its contents) may depend on whether I am talking about an omnivore or a baby. I can use the name "Chomsky" to refer to the author who inspired that last example (Chomsky 1995, p.23), but when I say "Chomsky occupies two metres on my bookshelf", I am generally referring to his books, rather than recalling the youth whom I put up in my office before he became a famous linguist. Or consider the statement "The ball is round", uttered during a squash game just as (as it happens) the racket strikes the ball. If the utterance is made by an observer familiar with squash but weak in materials science, the statement is false; if the utterance is made in response to someone unversed in squash who, witnessing the game from a distance, is enquiring about the shape of squash balls, the statement is true (Travis 1996). Or take the sentence "Cuba is similar to Jamaica" (an example of Amos Twersky's). This sentence is true if uttered in a seminar on grain production, but false if uttered in a political science seminar. Conversely, in the latter seminar, "Cuba is similar to N. Korea" is true; in the former false. Another example, due to Chomsky: The sentence "I live near Boston and far from Sydney" is true if I tell that to you, but false if I tell it to a Martian (Chomsky 1996, p.47). In such cases, where we succeed in telling the truth, we are able to do so by using concise sentences rather than convoluted ones because we can rely on our audience, in the particular context of utterance, being able to "fill in" what is missing. Sometimes, such "filling in" requires little more than the mere (mental) adding of a few words; other times, as in Wittgenstein's "Moses" example or the example about my daughter's earache, a proper reconstruction of the speaker's meaning requires a wholesale rephrasing.
If Wittgenstein is correct about the variety of ways in which the sentence "Moses did not exist" is used, then it would be wrong to say that, in that sentence, the name "Moses" stands for the object Moses. What we can say is that, in different uses of that sentence, the name plays a variety of different rôles, and this is evidenced by the variety of things that speakers can mean when using the sentence. For example, if, in a given context, a speaker uses the sentence "Moses did not exist" to mean "The Israelites did not have a single leader when they withdrew from Egypt", all we can say is that the shorter sentence does duty for the longer; it would be foolish to press the question "For what part of the longer sentence is "Moses", in the shorter one, an abbreviation ?". The way that the statement, expressed one way, maps onto the statement expressed the other way is not by straightforward replacement of bits of one sentence by bits of the other. Devices for condensing statements (metonymy is one of the simplest to explain) have evolved over a long period and, though a short sentence may be grammatically simple, its semantics (i.e. a detailed account of what the sentence, used in a given context, means) may be hugely complicated. Language-use develops like an organism -- as Wittgenstein remarked in his early work, "Everyday language is part of the human organism and is no less complicated than it ... from the external form of the clothes one cannot infer the form of the thought they clothe, because the external form of the clothes is constructed with quite another object than to let the form of the body be recognized .... The silent adjustments needed to understand everyday language are enormously complicated" (TLP 4.002). The "object" for which language is constructed is to communicate effectively -- clearly and concisely.
On one occasion, meant one way, the statement "Moses did not exist" may be true; on another occasion, a quite different statement meant by the speaker in using that same sentence may be false. It takes great ingenuity (although it comes easily to us) to draw upon mutual knowledge of the preceding conversation, of our surroundings and our beliefs to ensure that the way an audience understands a speaker's concise statement is just the way that the speaker intends it. So, for example, if we have been conversing about what we both know to be a liberal citizens' committee in Copenhagen anxious to institute an improved civic prostitution service, I may say, truly, "The citizens' committee wants prostitutes to be taught new skills". The sense of that statement is, of course, quite different from that of the statement I would be making if, using that same sentence, I were speaking truly about a conservative committee in Manila advocating the retraining of prostitutes for some more respectable line of work. What the phrase "new skills" is being used to refer to is different in each case. It may be that, in this case, context merely helps the hearer select the intended senses of ambiguous words ("new" as "additional" or "different"; "skill" as "technique" or "occupation"). But context can also be exploited to impose novel sense. The distance between the tip of a man's nose and the tip of his penis is roughly half a metre, but, in Zbigniew's case, it is several thousand kilometres, since he lives in the United States but was circumcised in Poland. There is enough slack in the rules to allow us to extend the use of 'tip' to a detached part, and much verbal humour depends on inventing such acceptable (intelligible) extensions. Of course, in humour we want creative or imaginative extensions of the extension, whereas in our ordinary reporting, the exploitation of the physical or conversational environment for the purpose of securing understanding tends to be much more systematic and pedestrian.
When stating the Overarching Principle, we mentioned the desideratum of not spelling things out at tedious length and, as we have seen, this is often achieved through various condensation devices, some quite standard -- such as the use of pronouns and other proforms -- others which rather more inventively draw on the linguistic and situational context. It should be stressed that condensation is the norm rather than the exception -- the number of "eternal sentences" occurring in ordinary conversation is extremely small. There are very good reasons why our utterances are normally condensed, including
(a) We don't want to bore our conversational partners.
(b) Frustra fit per plura quod potest fieri per pauciora (Ockham's Razor).
(c) Time is money.
(d) We are biologically disposed to speak concisely. For example, as Julius Moravcsik points out, "in view of limitations on attention span and the perceptual mechanism, there must be a premium on brevity and perceptual perspicuity in syntax" (Moravcsik, 1983, p.235)
IV. Failure of Substitutivity in Simple Sentences
The acid test for any proposed solution to the substitutivity problem is not propositional attitude reports (which offer great opportunities for arcane and epicyclic theoretical manoeuvres), but the kind of "simple" sentences, recently discussed by Jennifer Saul, which do not sustain substitutivity salva veritate. A virtue of Saul's papers (1997a,b) is that they remind us that, while propositional attitude reports may present particularly tricky cases, the phenomenon of "failure of substitutivity salva veritate of co-referring terms" is present not just in reports of attitudes, but in simpler cases, such as my report of Dirk and his dog. Since the root phenomenon is the same, any theory which purports to explain failure of co-referentiality but which applies only to reports of attitudes, is wrong. Consider one of Saul's examples:(1) Lois kissed Superman before she kissed Clark Kent.
One theory to explain why, though (1) is true (since Lois did kiss Superman before she kissed Clark Kent),
(2) Lois kissed Superman before she kissed Superman.
is false, is that the rôle of a referring expression is to stand for some temporal phase of its bearer. But, as Saul shows (1997a, p.104), this proposal runs into trouble with
(3) Superman is Clark Kent.
for here the temporal phases allegedly referred to are different. So this theory would say that (3) is false -- the wrong answer. Saul also demonstrates that a neo-Fregean account, due to Graeme Forbes (1997), which invokes modes of attire, cannot be right. For we should not wish to analyse
(4) Lois slept with Superman before she slept with Clark Kent.
as
(4F) Lois slept with Superman, so-attired, before she slept with Clark Kent, so-attired.
Fixing up the Forbes account would, as Saul shows (1997b), be a complicated and epicyclic project. What it, and the preceding theory (that referring expressions stand for temporal phases) have in common is that they assume that (1) has a fixed meaning (the "gramophone" meaning) which it is for a semantic theory to reveal; they share the defect of neglecting the rôle of context in determining meaning.
Some philosophers, such as Kent Bach (1987), would say that (1) must be false, since Superman and Clark Kent are the same person. Would we want to say that strictly speaking, they are correct ? Clearly not -- there is no "strictly speaking" about it -- language itself is not the arbiter of right and wrong. Linguistic intuitions are not infallible, and they are, to a certain extent, pliable, yet it would be foolish to be receptive to a theory which says that most people don't know how to use their own native language and as a consequence take to be true many statements which are false, and vice versa. People who know the story say that (1) is true, and they are right to do so. If Bach buys his daughter a Clark Kent doll when she had specifically asked for a Superman doll, she would think it a pretty poor joke were Bach to explain that, strictly speaking, he had bought her exactly what she requested. Someone who utters (1) with the intention of uttering a truth is not in error and has not misused language in any way. We use language to say what we want to say, and fail only on rare occasions, for example, when we are mistaken about the common meaning of a word we use. Here is a typical scenario in which, in using (1), the speaker speaks the truth: Martine and I, having just seen the Superman movie, fall into a dispute over whether Lois slipped her arms around the supersuit and kissed its wearer before or after she slipped her arms round the grey suit and kissed its wearer. If, in this case, Martine utters (1) then she is right; she says something true. And neither of us take ourselves to be debating, in effect, the truth-value of (2), even though we both know that "Superman" and "Clark Kent" are names of the same individual. In her use of (1), Martine intends, and I take her to be intending, something like
(1a) Lois kissed the guy when he was doing his Superman thing (wearing the kit etc.) before she ever kissed the same guy when he was dressed as Clark Kent the newspaper reporter.
The (mental) "filling in" required to recover sense (1a) from (1) is obviously quite considerable.
In other contexts, the use of (1) expresses a falsehood. Suppose, for example, we know that Millicent has not seen the movie, has received only a garbled version of the plot and, in particular, is unaware that Superman is Clark Kent. If Millicent now utters (1), we should tell her, correctly, that she has said something false, and would go on the explain that, since Superman and Clark Kent are one and the same person, each Superman-kissing is simultaneous with, is in fact identical to, a Clark Kent-kissing. In this case, what Millicent is saying, and what we take her to be saying is
(1b) (E!x)(E!y)(~(x = y) & x = Superman & y = Clark Kent & Lois kissed x before she kissed y)
or (borrowing an idea of Quine's), making verbs out of names
(1b') (E!x)(E!y)(~(x = y) & x supermans & y clarkkents & Lois kissed x before she kissed y)
where 'to superman' is to do the Superman thing'. While (1a) is a spelling out of what both Martine and I mean, (1b) or (1b') spell out what Millicent means.4 Thus, Sentence (1) is used in different contexts to make different statements with different truth-values, and this is due not to the presence of indexicals or ambiguous expressions in the sentence, but is due to the different circumstances of utterance.
Given the context of the conversation between Martine and myself -- where we each know that the other has just seen the film -- it makes entirely good sense for us to use (1) as a convenient abbreviation for (1a). Although (1a) is more explicit than (1), we would never use it in normal conversation because there are useful and subtle mechanisms, which every competent speaker employs, for engaging audience knowledge of the context in such a way that, though using a short (condensed) sentence, the speaker can be confident that the audience will be able to retrieve the "explicit" meaning. Now, comparing (1a) and (1), two lessons become clear: First, that it is pointless to ask who the names "Superman" and "Clark Kent" stand for in (1), for they do not even feature in (1a) -- this case is therefore very similar to the Wittgenstein example we were considering earlier where, for certain uses of "Moses did not exist", it is pointless to ask what object the name "Moses" stands for; second, that the reason we may have had for inclining us to think that we might be able to substitute "Superman" for "Clark Kent" in (1) -- namely that both names stand for the same object -- has now vanished, for, in this case, they stand for no object at all.
What contribution, then, do the names "Superman" and "Clark Kent" make to the meaning of (1) as used by Martine and me ? The answer, in most general and truistic terms, is that they serve the function of helping to ensure that, taken together with the relevant contextual factors, it is meaning (1a) that is yielded in the context in which Martine and I are conversing. Our use of the names in this context is not to pick out an individual but to concisely paraphrase (1a) by appropriating the way that some other people (including Lois) use those names. But that is by no means the only way in which we can have linguistic fun when there are two names of one individual at our disposal. If I say to you
(5) Diana Spencer was quiet and reserved, but Lady Di was the effervescent Queen of Hearts.
then I have said something true. And the (rather difficult) "spelling out" of (5) which would make my intended sense explicit draws quite differently on context than does the spelling out of (1). That (5) (as just used by me) is true, and not a contradiction, is proof that the names occurring in it are not there being used to denote a single individual. To try to insist that they must be being used to denote something (a sense, a temporal phase etc.), on the grounds that the logical rôle of a name is to denote, is symptomatic of the philosopher's unhealthy craving for generality. Natural languages are full of nooks and crannies; all manner of ingenious ways have developed for doing things with words and the supposition that names behave in a uniform manner is wholly without empirical warrant. The requirement of crystalline purity is only a logician's fantasy (PI § 107).
Although Saul terms her sample sentences which contain no attitude, modal or quotational constructions "simple", they are in fact the most testing ones for any theory designed to cope with the substitutivity phenomenon, for they leave least scope for manoeuvre. My own suggestion has been that there is no problem, because there is no constraint requiring substitution of "co-referentials" salva veritate in sentences, and that supposing that there is such a constraint betrays a fundamental illusion about the project of semantics -- in fact, it commits the "main mistake" identified by Wittgenstein, that of thinking that the semantics of natural languages is the study of language as an abstract system rather than of words used in a context by speakers. It is because of proper reliance on contextual factors which the speaker knows will enable his audience to grasp what he means that the speaker frames a statement in the way that he does. The speaker can expect audience uptake if, through previous experience with that, or a similarly placed audience, the speaker can see that his audience uses words and actions in much the same way as he does himself, and catches on to the same clues. This is, as Wittgenstein puts it, "part of the framework on which the working of our language is based" (PI § 240). But we are wonderfully flexible, and if we have reason to believe that our hearer does not share our form of life (or does not yet share it as is the case with normal young infants just getting to grips with the ways of the words and the world), then we adapt. As Davidson says, "if you are pretty sure that somebody is going to interpret you in a 'non-standard' way, then you are foolish if you don't speak in a non-standard way" (Davidson, 1993, p.119).
Once we are properly sensitive to the ways in which contextual factors are relied upon to ensure audience uptake, we can always see why a speaker might refuse to substitute a "semantically co-referring expression" for the one he used. We can see why such a substitution could alter the sense and the truth-value of the original statement. A speaker of (1), talking to someone else who knows the movie, would refuse to substitute "Superman" for "Clark Kent" simply because he is aware that appeal to context, demonstrative conventions and speaker's knowledge would not allow the audience to retrieve the intended sense (1a) from the resulting sentence. Put positively, a competent speaker intending to convey the message of (1a), knows that uttering (1) is the succinct way of doing just that when in the presence of a competent and knowing hearer.
In a recent paper, Alex Barber (2000) notes that a sensitive analysis of simple sentences should acknowledge that it makes a difference whether utterers of such sentences are in the know or are ignorant with regard to the significance of the singular terms used in their utterances. Barber calls these two types of utterer Cognoscenti and Ignoramuses respectively. He considers the simple sentence
(**) Superman is more successful with women than Clark Kent.
He thinks that this sentence is 'semantically false' but can be used to convey a truth, and that what truth it conveys can be figured out by adverting to the Gricean principles for co-operative conversation. Suppose someone (whom Barber calls CognoscenteA) utters (**). Barber's story about how Cognoscente B figures out the intended implicature of A's utterance goes like this:
A just said something false, knowing that I know this, and knowing that I know she knows, etc. This amounts to a violation of the first maxim of quality. What can A be up to, short of violating the Cooperative Principle? Perhaps it is this. Some are unaware that Superman is Clark Kent. They would, under forseeable epistemic conditions, utter what A just uttered. A must be implicating that just those epistemic conditions obtain that would prompt an ignoramus to utter (**) (though not 'Superman is more successful with women than Superman'). In other words, A is trying to convey to me that Superman/Clark Kent is more successful with women when appearing as Superman -- or, as we might say, is Supermanizing -- than when appearing as Clark Kent.
Apart from the fact that we may doubt whether it is realistic that cognoscenti -- even smart cognoscenti -- go through such a process of reasoning about what their conversational partners think that ignoramuses would say, the account does not work.. For imagine a dialogue between two cognoscenti, one a man who is trying to elicit from a woman her take on what makes for male attractiveness:
Tom: Why is Superman so much more attractive to women than Clark Kent, even though really they're one and the same person?
Mary: Well, it's said that clothes don't make the man, but that's untrue. A drab grey suit is a major turn-off, but put a guy in a supersuit that really shows off his credentials and...bingo.
Surely an ignoramus would not make the utterance that Tom made above. In my view (**) is a true condensed utterance.
V. Attitude Reports
Propositional attitude reports are no different from simple statements with respect to the substitutivity problem. The pragmatic considerations guiding the speaker's choice of words may be complex and sometimes conflicting, as we indicated at the beginning of this paper, and spelling out the speaker's intended meaning may be tiresome. Given the subtle pragmatic principles governing a speaker's choice of singular term, there should be no expectation that the substitution of one such term for another, even when in some contexts both terms can be used to pick out the same individual, will preserve truth.We pointed out earlier that, although linguistic intuitions are not sacrosanct, it is foolish to declare that sentences we use successfully in ordinary discourse for ordinary purposes of communication do not have the meaning we think they have just because some theory tells us that they don't. For there is nothing more to the sentences having the meaning we think they have other than their having the use they do have in ordinary successful communication. This seems obvious, and it is curious that a straight thinker like Simon Blackburn has talked himself into believing otherwise. Blackburn concocts a story about a tourist -- let's call him Cedric -- who takes the train up Snowdon and finds the journey disappointing (for it's true that the scenery on that train ride is fairly bland and the slopes are not spectacular). Next day, however, Cedric gets more ambitious and goes climbing up Yr Wyddfa across Crib Goch. He thinks it is great and says so, without realising that Yr Wyddfa is actually Snowdon. According to Blackburn, it is true (but not helpful) to say of Cedric that he thinks that Snowdon is duller than Snowdon (Blackburn, 1984, p.333). But Blackburn is wrong; it plainly is not true. One can, of course, envisage the locals sitting in the pub ridiculing Cedric; one of them might say "Cedric thinks that Snowdon is duller than Snowdon", but none of them seriously believes that Cedric thinks that Snowdon is duller than itself. The data to be explained is that, while they might truly say that Cedric believes that Snowdon is duller than Yr Wyddfa, they would not say (because it's patently false) that Cedric believes that Snowdon is duller than Snowdon. Cedric himself would not assert "Snowdon is duller than Snowdon" and, in reporting his beliefs faithfully, they must say things as he sees them, unless they are out for crude distortion or a cheap joke. The obligation to report someone's beliefs from that person's perspective is what I shall call the "Perspectival Principle". It is a defeasible principle, in that it may be overridden. And another point to make is that, in respecting this principle, we must sometimes refuse, when reporting an attitude, to mimic the sentence used by the person expressing that attitude. For sometimes the person may use the wrong words, and if we want to report what his real attitude is, we should use the right words. Failure to do so is what Joseph Moore (1999) has called "misdisquotation".
Blackburn's mistake is shared by many philosophers who write on this topic. Jennifer Saul, for example, holds that a proper pragmatic account of attitude ascription will vindicate the Salmon-Soames claim that some of our intuitions about the truth-values of sentences are mistaken, and will explain why we are prone to make such errors (Saul 1998). This is completely misguided: our intuitions are not mistaken, so it is not our errors that need to be explained; if anything is to be explained it is how an erroneous philosophical theory establishes so strong a hold that, in order to preserve it, theorists are willing to abandon common sense. The situation here is quite different from cases where claims based on untutored common sense are shown, by mathematics or by science, to be mistaken. Many or even most people may believe that there must be a greatest prime number. They are wrong, and this can be proved. But that is very different from being mistaken about what the words we use mean, for if most people use a word "W" to mean X, then X is what "W" means (or at least it is one of the meanings of "W"); there is no other criterion. This is not to deny that this meaning may fade in time under pressure of various sorts. All this is obvious to anyone who has to update a dictionary. So we cannot say that a person who claims that the statement "If there is life on Mars, then 2+2=4" is false is in error. Most people use "If ... then ... " to signify a ground-consequence relation, so that is what the connective means, or that, at least, is one of its meanings.5
As was the case with the "simple" sentences, it is easy to find non-indexical, non-ambiguous sentences used for making attitude-reports which have different truth values when used to make statements in different contexts. Thus, while
(6) Lois Lane believes Superman is a reporter.
would normally be taken as false, we can, following Jonathan Berg (1988, p.371), imagine a scenario in which (6) is used to express a truth: Two people are marvelling at Superman's ability to conceal his identity. One says to the other, "Look, there's Superman in his Clark Kent outfit; he's incredibly convincing! Everyone thinks he's a reporter -- Jimmy Olson, Mr. White -- why, even that clever Lois Lane believes Superman is a reporter". In this scenario, the "spelled out" version of (6) is
(6a) Lois Lane believes that the guy whom she knows only as "Superman" when he is doing his superheroic thing, but whom you and I know both as "Superman" and "Clark Kent" is a reporter.
Of course, the speaker can avoid this mouthful by relying on his audience's relevant knowledge and their knowledge of Lois' relative ignorance and their knowledge of the devices, either well established or invented "on the fly", which, given this kind of context, permit the shrinking of (6a) down to (6). In this context, it will be noted, we jettison the Perspectival Principle because, with Superman in full view of us, the most salient perspective is our own. Similarly,
(7a) Lois Lane believes that the guy whom she knows only as "Clark Kent" when he is doing his newspaper reporter stuff, but whom you and I know both as "Superman" and "Clark Kent", can fly.
shrinks, all being well, to
(7) Lois believes that Clark Kent can fly
and, as normally used, is false. Nathan Salmon says that, on his account, such a statement is true, but he does, without apparent embarrassment, mark that claim with an exclamation mark (Salmon 1989, p.247). It's as if he realizes that his readers will think he is nuts, but hopes to talk them round into believing that they and ordinary speakers are wrong in the way they make attitude-ascriptions such as (7). As I tried to make clear in my discussion of Blackburn and Saul, this hope is in vain.
Every case of failure of substitutivity of "co-referring expressions" can be explained when we flesh out the problem sentences in the appropriate ways, and see why the fleshing out of what the speaker means is not captured by the result of substituting the "co-referring term" in the original unfleshed sentence. Take the true report
(8) Maggie thinks that Odile is tired.
Mark Richard (1990, p.2) writes, "For Maggie to think that Odile is tired, she must have some representations of Odile and of being tired 'put together' in an appropriate way. In some broad sense of 'sentence', she must employ a mental sentence saying that Odile is tired". Now suppose that, unknown to Maggie, Odile is Maggie's mother, but Maggie thinks that the visibly indefatigueable Catherine is her mother. In this scenario, we are correctly inclined to say that
(9) Maggie thinks that her mother is tired.
is false, even though it results from the true (8) just by substitution of a "co-referring term".
Richard, as we have seen, couches his theory of attitude-ascription and his solution to puzzles such as this within a "language of thought" framework. But this is quite unnecessary. In order to say that Odile is tired, Maggie must, of course, have the appropriate words, but her thinking that Odile is tired need not be split into word-like parts. Maggie's mental representation of Odile (if she has one) may be graphical rather than verbal. And her thinking that Odile is tired may just consists in her taking there to be close enough resemblance in relevant respects between Odile and some other tired individuals. When she thinks that Odile is tired, Maggie stands in a relation not to a proposition or a sentence, but to Odile -- just as she does if she wants Odile not to be tired (where, as Ben-Yami notes (1997, p.85), the verb "wants" has no propositional object). One can surely notice a similarity between one droopy-eyed, slow-moving object and another, without the word "tired" or "tiredness" or any mentalistic counterparts of these being present in one's noticing; lions do. 6
So, in reporting what Maggie thinks, I am under no obligation to quote the words (or translations of the words) from her mental sentence, since there is no such thing. But there is a pragmatic principle to the effect that, other things being equal, I should use something close to those words (or a translation of them) that the reported (the person about whom the report is being made) would herself use.7 We have called this "the Perspectival Principle". The existence of this principle is easy enough to explain: in reporting people's attitudes one generally wants to say things as they see them, and how they see things is how they would say them if they could (and if words are adequate to the task). Departure from this principle is usually the result of pressure from what I called "the Overarching Principle" -- to make oneself maximally intelligible to one's audience, i.e. to the reportee.
Sometimes to use those words that the reported would use would be at the expense of intelligibility to the reportee, so the reporter changes those words. In reporting Maggie's thinking to you, I choose a sentence that will be maximally useful to you. Thus, if neither you nor Maggie know that Odile is her mother -- let us suppose that both of you think that Catherine is (and know that Catherine's name is "Catherine") -- it would be wrong for me to tell you "Maggie thinks that her mother is tired". Wrong, both in the sense of being misleading, and in the sense of being false. For the operative pragmatic principle (in this case the Overarching Principle trumps the Perspectival Principle) dictates how the sentence is to be interpreted -- dictates what statement a speaker, conforming to such principles intends, and what statement an audience should figure him to be intending. Note that, in this case, the usual means for making the report explicitly de re is not much help. For, if I say to you "Maggie thinks of her mother that she is tired" then, so long as you think that Catherine is Maggie's mother, you will take my statement to be false and it is moot (undecidable) whether the statement really is true or false. For decidability and clarity, I should de-condense my statement and say "Maggie thinks, of whom I, but not you, know to be her mother that she is tired". I may actually put it in this way, rather than simply saying "Maggie thinks that Catherine is tired" because I choose to use this occasion to finally reveal a dark family secret.
VI. How to Make True Attitude Reports
I can now summarize the prescription for making true attitude reports. This is a prescription which, I claim, we all follow with almost 100% success. In order to simplify things a bit, let us limit attention to non-nested ascriptions of attitude (so we won't bother with locutions like "David believes that Posh fears that Brooklyn hopes that Manchester United will win the European Cup". We restrict ourselves, then, to cases in which a reporter R is telling his hearer H that some person A (the ascribee) øs that some item has a particular property F. Let us suppose, further, that (as in all the examples I can think of that are used in the literature) there are just two singular terms "c" and "d" for that item, where it may not be known to all three parties R, A and H that there are these two singular terms, or that these are two singular terms for a single item. We shall choose "F" so that, if A øs that Fc and sincerely asserts "I ø that Fc" she would dissent from "I ø that Fd", and vice-versa.Here, then, is the prescription: If A would sincerely assert "I ø that Fc", then R, reporting to H, makes a true ascription by saying "A øs that Fc" UNLESS framing the report that way would either be unintelligible to H or would give H a false impression; in which case, the Overarching Principle kicks in, and the true report would be "A øs that Fd". So, for example, if R conveys to H, demonstratively or by some other means that he will be using the name "d" for the item in question, then he must so use it. This is what happens in the Berg example, where R lets H know that he plans to use the term "Superman" for that guy -- the one who, at the time of the demonstration, is wearing a grey suit and spectacles. Again, if R knows that H associates the name "c" with something other than the item about which R is reporting, but that H knows the name "d" of that item, then his true report to H will be "A øs that Fd". That was the case in our report about Maggie, the girl who didn't know the true identity of her mother.
Kripke’s Pierre provides an interesting illustration of the use of these principles. Here, the Reporter (Kripke) and the Hearer (Ourselves) know far more about Pierre’s situation than does the Ascribee (Pierre). So, in reporting Pierre, the Perspectival Principle is trumped. Here, the faithful reporter must use the uncondensed sentence "Pierre believes that London (the parts he has seen in magazines, when he called the city 'Londres’) is pretty". Kripke, in trying to bully us into anwering 'yes’ or 'no’ the question 'Does Pierre or does he not believe that London is pretty?’ commits the Fallacy of the Simple Question. This is just a lawyer’s trick to get a witness to tell a distortion of the truth.
- Austin, J.L. 1950. Truth. Proceedings of the Aristotelian Society Supp. Vol. 24, pp.111-28.
- Bach, K. 1987. Thought and Reference. Oxford: Clarendon Press.
- Barber, A. 2000. A Pragmatic Treatment of Simple Sentences. Analysis 60, pp.300-308.
- Ben-Yami, H. 1997. Against Characterizing Mental States as Propositional Attitudes. The Philosophical Quarterly 47, pp.84-89.
- Berg, J. 1988. The Pragmatics of Substitutivity. Linguistics and Philosophy 11, pp.355-370.
- Blackburn, S. 1984. Spreading the Word. Oxford: Oxford University Press.
- Brandom, R. 1994. Reasoning and Representing. In M. Michael and J. O'Leary-Hawthorne (eds), Philosophy in Mind. Dordrecht: Kluwer Academic Publishers, pp.159-178.
- Chomsky, N. 1995. Language and Nature. Mind 104, pp.1-61.
- Chomsky, N. 1996. Powers and Prospects. Boston MA: South End Press.
- Davidson, D. 1986. A Nice Derangement of Epitaphs. In R. Grandy and R. Warner (eds), Philosophical Grounds of Rationality. Oxford: Oxford University Press.
- Davidson, D. 1993. Reply to Andreas Kemmerling. In R. Stoecker (ed.), Reflecting Davidson. Berlin: Walter de Gruyter, pp.117-119.
- Forbes, G. 1997. Reply to Saul. Analysis 57, pp.109-14.
- Geurts, B. 1998. Presuppositions and Anaphors in Attitude Contexts. Linguistics and Philosophy 21, pp.545-601.
- McDowell, J. 1994. Mind and World. Cambridge MA: Harvard University Press.
- Moore, J. 1999. Misdisquotation and Substitutivity: When not to Infer Belief from Assent. Mind 108, pp.335-365.
- Moravcsik, J. 1983. Natural Languages and Formal Languages: A Tenable Dualism. In R. Cohen and M. Wartofsky (eds), Language, Logic and Method. Dordrecht: D. Reidel Publishing Company, pp.225-239.
- Recanati, F. 1993. Direct Reference. Oxford: Blackwell.
- Richard, M. 1990. Propositional Attitudes. Cambridge: Cambridge University Press.
- Salmon, N. 1989. lllogical Belief. In J. Tomberlin (ed.), Philosophical Perspectives 3: Philosophy of Mind and Action Theory. Atascadero CA: Ridgeview Publishing Company.
- Saul, J. 1997a. Substitution and Simple Sentences. Analysis 57, pp.102-8.
- Saul, J. 1997b. Reply to Forbes. Analysis 57, pp.114-18.
- Saul, J. 1998. The Pragmatics of Attitude Ascription. Philosophical Studies 92, pp. 363-389.
- Stich, S. 1976. Do Animals have Beliefs? Australasian Journal of Philosophy 57, pp.15-28.
- Strawson, P.F. 1986. "If" and "É". In R. Grandy and R. Warner (eds), Philosophical Grounds of Rationality . Oxford: Oxford University Press.
References
| 1 | I use the following abbreviations for Wittgenstein's works: TLP 1961: Tractatus Logico-Philosophicus, trans. D.F. Pears and B.F. McGuinness. London: Routledge and Kegan Paul. BB 1958: The Blue and Brown Books. Oxford: Blackwell. LA 1970: Lectures and Conversations on Aesthetics, Psychology and Religious Beliefs, ed. C. Barrett. Oxford: Blackwell. PI 1958: Philosophical Investigations (2nd edition), edd. G.E.M. Anscombe and R. Rhees, trans. G.E.M. Anscombe. Oxford: Blackwell. RPP I 1980: Remarks on the Philosophy of Psychology, Volume I, ed. G.E.M. Anscombe and G.H. von Wright, trans. G.E.M. Anscombe. Oxford: Blackwell. | |
| 2 | It might be claimed, that Wittgenstein's example sentence has three (or more) meanings. In what follows, we shall be alert to the possibility of multiple ambiguity, but aware also of the danger of multiplying ambiguities beyond necessity when a simpler, more plausible account can be given of how what a speaker in one context means by a certain sentence may be different from what a speaker in another context means by that same sentence. | |
| 3 | For an excellent, extended defence of a distinction between sentence meaning, what a speaker communicates (where appeal needs to be made to the Gricean maxims) and what a speaker says or means (where one needs to appeal to other pragmatic, contextual elements), see Recanati (1993), esp. p.236 ff.. | |
| 4 | (1a) and (1b) are relatively context-insensitive, when compared to (1). | |
| 5 | Grice, to the contrary, holds that "if … then …" means just the truth-functional conditional of the propositional calculus. For an elegant refutation of Grice, see Strawson (1986). | |
| 6 | John McDowell (1994, p.3) would say that the the lion merely has perceptual sensitivity to features of its environment, whereas we have that but in a "special form", ours being "taken up into the ambit of spontaneity" (i.e., concept-enriched), this being symptomatic of our operating within what McDowell, following Wilfrid Sellars, terms the "space of reasons". I cannot here engage with the subtleties of McDowell's position, but would remark that it is hard to separate the reason that the lion picks out a weak member of a herd from the "special form" of reason that we would have for doing the same were we similarly hunting prey. | |
| 7 | This sentence is surrounded by hedges because (i) we normally have to guess what words people would use to express their attitudes; (ii) the person reported may speak only one language, the words of which don't translate exactly into words of the reporting language; (iii) we may be reporting attitudes of a creature that has no language, and here the scope for guesswork and error are considerable. If my dog grabs every bone that he can lay his teeth on, but runs a mile at the sight or the smell of a sardine, then, although I agree (with Stich (1976), for example) that the dog doesn't have our concept of a bone, I should be much happier to report his excited tail-wagging behaviour, when he sniffs the bone I'm hiding behind my back, with the sentence "He thinks that I'm going to give him a bone" than with the sentence "He thinks that I'm going to give him a sardine". |