The paratactic theory revisited and revised

Content and context: the paratactic theory revisited and revised

Ian Rumfitt

We all use, and apparently understand; claims like (1) Galileo said that the earth moves.

What is more, this claim appears to entail (2) Among the things said by Galileo was that the earth moves.

Reading this last sentence straight, so to speak, one is led to conclude that among the things-that-there-are are things-that-are-said, things traditionally called propositions. If that much is right, then it seems that an important part of any systematic investigation into natural language will be to determine which propositions are expressed by which declarative utterances.

The thought just adumbrated is likely to be familiar, for after a period in which they were anathematized as “creatures of darkness”,(1) propositions have lately been enjoying a revival among the philosophically orthodox. It is no longer fashionable to dwell upon the difficulties that might attend the “straight” reading of sentence (2). Rather, attention has turned to the nature of the propositions that are there said to exist. In particular, we are asked to consider whether there are fruitful theoretical identifications to be made between propositions and either: sets of possible worlds in the style of Stalnaker; set-theoretical combinations of objects and properties in the style of Russell; or combinations of elements in some “third realm”, in the style of Frege. A perusal of the writings of the most influential parties to this controversy, however, soon reveals a fundamental assumption which they share:

It is astonishing what language can do. With a few syllables it can express an incalculable number of thoughts … This would not be possible, if we could not distinguish parts in the thought corresponding to the parts of a sentence, so that the structure of the sentence can serve as a picture of the structure of the thought.(2)

If I may wax metaphysical in order to fix an image, let us think of the vehicles of evaluation – the what-is-said in a given context – [the] propositions … as structured entities looking something like the sentences that express them. For each occurrence of a singular term in a sentence there will be a corresponding constituent in the proposition expressed.(3) Neither of these passages is coherent unless it makes sense to speak of the thought, or the proposition, that such-and-such a declarative utterance expresses. They are, then, manifestations of the widespread belief that, if there are thingsthat-are-said at all, then exactly one thing-that-is-said will be expressed by each significant declarative utterance. Whether one holds this belief will affect greatly the way in which one formulates projects and problems in formal semantics and the philosophy of language more generally. And determining its truth or falsity is an important pre-requisite for any assessment of the accounts of the nature of propositions” just mentioned. It may, then, be worth returning to the question of how to elucidate the notion of proposition, in the hope of settling this matter. We could well begin by examining the semantical structure of the simple report (1) and, in particular, the constituent phrase “that the earth moves”.

I. The structure of indirect reports


Even in speaking of this string of words as a constituent, I am making an assumption that some would deny. A. N. Prior,(4) for example, held that the particle that” in (1) – the complementizer” in the linguists’ argot – belongs semantically with the verb, so that (1) is to be divided along the lines:

Galileo / said that / the earth moves.

On this view of the matter, “to say” is precisely not a transitive verb; rather, it is a part of speech that takes a noun-phrase at one end and a complete clause at the other. This view is interesting, but it would take me too far from my theme to examine it properly. So I will simply assume that the words “that the earth moves” form a phrase as they appear in (1); also, that the phrase they form is a noun-phrase.

One question that plainly needs to be asked about this noun-phrase is what sort of item it designates. But there is another question to be asked, whose answer is largely independent of the answer to the first. On any view of the matter, the phrase “that the earth moves” is not simple; its significance somehow relates to the significance of its parts. We need to ask what that relation might be.

In addressing this question, one naturally turns for guidance to other complex noun-phrases. The semantics of functional noun-phrases of various kinds, such as “Mary’s father’s mother” and “{Hesperus, the Moon}”, is reasonably well understood. As models for the putative noun-phrase “that die earth moves”, however, these are in a crucial way unhelpful. As the term “functional noun-phrase” suggests, the designatum of (for example) the set-theoretical name “{Hesperus, the Moon}” is a function of the designata of its components, “Hesperus” and “the Moon”. Specifically, it is the set whose elements are precisely those designata. The fact that the designatum of a name of the form “{a, b}” is a function of the designata of the names “a” and “b” brings certain consequences. In particular, since Hesperus is the same planet as Phosphorus, it follows that the designatum of the functional name “{Hesperus, the Moon}” is the same as the designatum of “{Phosphorus, the Moon}”. So if the name “that Hesperus is bright” were understood on the model of “{Hesperus, the Moon}” we should expect its designatum to be the same as that of “that Phosphorus is bright”.

It would, I think, be foolish to claim that this consequence is obviously wrong. Treatments of “that”-clauses which have this consequence – specifically, neo-Russellian treatments in which the designatum of “that Hesperus is bright” is indeed a set-theoretical construct (typically, the pair ) – have recently been elaborated in some detail,(5) and it may turn out in the end that the air of initial implausibility that inevitably attaches to any such treatment can be dispelled.(6) Nevertheless, any unprejudiced investigation into this matter is unavoidably impeded if one’s understanding of “that”-noun-phrases simply presumes that the implausibilities can be dispelled. So at this stage we want a parsing of a “that p” noun-phrase which is neutral on that question, and that is not provided by treating such a noun-phrase on the model of a functional name.


It is one of Davidson’s contributions to this topic to have provided a treatment that is appropriately neutral.(7) Confronted by a typical simple indirect report such as

(1) Galileo said that the earth moves, he proposes to discern an unwritten full-stop after the word “that”, so that the business part of the report (i.e. the claim assessed for truth or falsity) consists of a simple relational sentence

Galileo said that.

In Davidson’s own version of this “paratactic” proposal, the relation obtains between the man Galileo and the reporter’s subsequent utterance of “The earth moves”, where “utterance” means “event of utterance” – the unrepeatable producing of certain words on a certain occasion.(8) The report, then, will be true just in case the man Galileo stands in the saying relation to the reporter’s subsequent utterance. However, variants are possible. In particular, one might suppose that the purportedly demonstrative “that” finds its reference by “deferred ostension”, the utterance actually produced by the reporter being but a sample of the item to which he properly refers.(9) Indeed, Davidson’s own treatment of direct reports(10) provides a model for this. Parity with his theory of indirect speech leads one to expect him to construe

Galileo said “Eppur si muove”


Galileo produced an utterance equiform with this utterance:

Eppur si muove, but in fact he does not. The particular utterance of “Eppur si muove” is understood as being merely an exemplar, the reference so to speak passing through it to reach the string of words that the sample exemplifies.(11) In the case of indirect speech, the item indirectly ostended could very well be the proposition that the utterance expresses.

I will eventually argue that we need to emend Davidson’s theory along some such lines as these. But what we immediately need to note is that both the echt Davidsonian theory and the suggested Ersatz provide the neutral treatment of “that”-noun-phrases that was sought. Suppose a reporter truly says “Tom says that Hesperus is bright”. Then, according to Davidson, Tom will stand in the saying relation to the reporter’s utterance of “Hesperus is bright”; according to the emended version of the theory, he will stand in a different saying relation to the proposition that that utterance expresses. Further consideration may show that standing in Davidson’s saying relation to an utterance of “Hesperus is bright” involves being so related to an utterance of “Phosphorus is bright”; and it may show that the proposition expressed by uttering “Hesperus is bright” is the same as that expressed by uttering “Phosphorus is bright”. Equally, though, it may not; and if it turns out that one can stand in the saying relation to one of these utterances without being so related to the other, a paratactic treatment of the noun-phrase “that p” will not be embarrassed. This is why the paratactic parsing is such an attractive initial proposal when discussing these matters, for it allows us to decide them as they surely ought to be decided: by analysis of the concept of saying, and/ or of the conditions of identity of things that are said. In any event, these matters are not settled in advance, one way or the other, by the very parsing of the “that”-noun-phrase.


All the same, we first need to recognize (or so I think) that the paratactic proposal itself stands in need of quite radical reformulation.

The first thing to notice is that Davidson’s own presentation of the proposal is unnecessarily extreme. “Sentences in indirect discourse”, he tells us, “wear their logical form on their sleeves (except for one small point). They consist of an expression referring to a speaker, the two-place predicate ‘said’, and a demonstrative referring to an utterance” (1984, p. 106). In his hands, then, the paratactic proposal involves the claim that every instance of the “that” of indirect speech (every instance of the complementizer) is also an instance of the equiform demonstrative pronoun. Just this claim, however, is open to serious objection. Specifically, the claim is tenable only if the constraints on the project of discerning “logical forms” are implausibly lax. Davidson appears to suppose that any assignment of logical form to sentences which fits into a truth theory of the right sort, and which delivers interpretative T-theorems, is as good as any other, and good enough to be correct. One might, though, expect something more. In particular, one might expect the structure postulated in accounting for the interpretation of an utterance to bear some reasonably close relation to the structure postulated in accounting for its well-formedness.

If one comes to semantics with any such expectation, then the paratactic theory, as Davidson formulates it, is inevitably utterly unprepossessing. It is simply not plausible that instances of complementizers are instances of the equiform demonstrative pronoun. Whatever etymology might teach us,(12) there is next to nothing in common grammatically between (on the one hand) a pronoun that combines with a common noun to form a noun-phrase which can then stand in various sentential positions, and (on the other hand) a particle that takes a finite clause to produce an object for a verb like “say” or “believe”. It may, then, be worth reformulating the paratactic proposal in a manner that shows why the claim that instances of the complementizer are instances of the demonstrative is really an adiaphoron.

Suppose that our lexicon did contain a separate entry for “that” as complementizer – “[tha]comp” as it might be – entry which had a grammatical distribution very different from the distribution of a demonstrative pronoun. Then at the levels of structural description relevant to well-formedness (i.e. at “D-structure” and “S-structure” in the linguistic paradigm that now predominates), the structure of “Galileo said that the earth moved” could be something like:(13)

Within this paradigm, however, there are many reasons(14) for supposing that the interpretation of an utterance will take place under a structural description of a different kind: descriptions which collectively constitute a distinct level, “LF”, and which are derived from S-structures by the application of certain specifiable movement rules. And then – without in any way back-tracking from the idea that the complementizing occurrences of that” enjoy a distinctive syntactical distribution – we might formulate Davidson’s proposal as the suggestion that one of the processes through which LF-structures are got from S-structures is a process (we might call it arboreal fission) which converts the structure shown above into this one:

Here the complementizing “hat” gives way to a demonstrative pronoun which just happens, in English,(15) to be equiform with it.(16)

Indeed, the formulation of the paratactic proposal as a conjecture about the relation between specific levels of grammatical description brings with it, or so I think, a further advantage. Suppose that we hear the report “John said that someone heard the shooting of the hunters” in circumstances which render it incorrigibly indeterminate whether the clause that follows the complementizer meant “Someone heard the hunters shooting” or “Someone heard the hunters being shot”. Then, in the first place, the string of words uttered in making the report (i.e. the string of words “John said that someone heard the shooting of the hunters”) is ambiguous. And, in the second place, the report itself is ambiguous: since there is no telling what the reporter attributed to John, there is no telling what the reporter himself is saying.

It is hard to see how Davidson’s version of the paratactic proposal can account for this second fact. He will construe the problematical report thus:

John said that.

Someone heard the shooting of the hunters and he can point out that in the circumstance imagined, the second utterance perforce remains unconstrued. What is unclear is why this should generate an ambiguity in the report. The business part of the report, after all, is supposed to be simply the claim John said that”, where “that” is taken to pick out an utterance. And in the circumstance described there is indeterminacy neither in what “John said that” means nor in which utterance is the denotation of “that”. Granted: since it is indeterminate how that second utterance is to be construed, it will be indeterminate whether John stands in Davidson’s saying relation to it, so Davidson is well placed to account for an indeterminacy in the truth value of the report. But nobody should be trying to account for that. The problem was to account for an ambiguity – that is, an indeterminacy in what was said. An ambiguity, so understood, must not be confused with an indeterminacy in the truth value of an unambiguous claim.

If, however, we will take but one simple (if somewhat radical) step, the present version of the paratactic proposal can escape from this difficulty. The level LF, we may suppose, is a level of description at which any structural ambiguities in the utterance being described are resolved; different possible disambiguations are represented by different LF-trees. Let us suppose, too, that such structural disimbiguation is accomplished prior to fission. Then, letting the trees be the LF-trees which correspond to the readings “Someone heard the hunters being shot” and “Someone heard the hunters shooting”, there will be two descriptions of our problematical report, as shown in Figure 4.

The ambiguity, we can say, consists in its being indeterminate which of these different LF-descriptions represents the reporter’s intentions in making the problematical report. One will overcome the ambiguity if and only if this indeterminacy is eliminated.

It may seem, though, as if this explanation is bought at too high a price. Specifically, it may be wondered what can possibly remain here of the idea that the function of the demonstrative “that” is to pick out an utterance. This version of the proposal, it will be said, succeeds only in converting the paratactic theory into a version of the proposal whereby “A said that such-and-such” relates the speakerA to a linguistic type: in this instance, a tree structure at a level which is supposed to provide representations of all the possible disambiguations of the sentence types uttered. Then – it will be concluded – the proposal becomes liable

to the familiar litany of objections to which any such theory is subject, and which “On Saying That” itself magisterially rehearses. I mean: failure to account for the context dependence of words after the “that” of indirect speech; (17) failure to account for the possibility that the same tree structure might mean different things in different languages or idiolects; etc. etc.

Figure 4

I agree that these advantages of Davidson’s treatment are not lightly to be cast aside. However, we can keep them, while retaining the solution to the problem about ambiguity, if only we will take the advertised radical step. We have long been accustomed to the idea that tree-structures such as those labelled by “S[sub.1]” and “[S.sub.2]” might represent disambiguations of an ambiguous string. What we need to introduce, I think, is the idea that we could also use such tree-structures to speak unambiguously. With that idea in play, we can imagine that the utterances in question are actually effected by inscribing (or otherwise “pronouncing”) the trees [S.sub.1] and [S.sub.2]. So, if the report “John said that someone heard the shooting of the hunters” receives the interpretation

Figure 5

then the report (so interpreted) will be true just if A stands in the two-place saying relation to the ensuing utterance of [S.sub.1]. Since any such utterance has a determinate interpretation (viz. “Someone heard the hunters being shot”) there is no obstacle to speaking of a determinate saying relation here.

II. The indirect saying relation


In actually deciding whether a particular utterance is properly reported as a saying that such-and-such, we obviously need (in addition to a parsing of reports) some elucidation of the conditions under which a speaker stands in the saying relation to an utterance, and the second famous element of Davidson’s theory is the prototype of just such an elucidation. So, referring to his example

Galileo said that the earth moved, which of course he parses as

The earth moves.

Galileo said that,

Davidson explains the transitive verb or two-place predicate “say” here as the existential closure of

a three-place predicate which holds of a speaker (Galileo), an utterance of the speaker (“Eppur si muove”), and an utterance of the attributer (“The earth moves”). This predicate is from a semantic point of view a primitive … [but] … it is also worth observing that radical interpretation, if it succeeds, yields an adequate concept of synonymy as between utterances.(18)

Now a theory of “interpretation” for a speaker (in Davidson’s rather technical sense of that word) is a theory, knowledge of which would confer on one who had it the capacity to understand that speaker’s utterances, and it is a cardinal Davidsonian thesis that such a theory can assume the form of a truth theory (cf. especially Davidson 1984, p. 171). Accordingly, we may gloss the proposal just cited by stipulating that:


A says that.



([exists] u) Say(A, u, that))(19)


([exists] u) There is a theory, knowledge of which would enable one to understand those of A’s utterances which are made using the linguistic abilities operative in making u, and which canonically entails this:(20)

True(u) = p) or, more briefly,

([exists] u) (There is a theory, interpretative for A as he produces u, which canonically entails this:

True(u) = p).

This dovetailing of the theory of indirect speech and the theory of understanding is, I think, highly attractive. It reflects what I take to be the insight that what is said in the course of an utterance is nothing other than what somebody who understands the utterance understands to be said. However, the dovetailing cannot be effected just as stated. The letter “p” in the formulation just given presumably goes proxy for the string of (English) words that the reporter utters. But the paratactic proposal is no longer being defended as a thesis about that string; rather, it is being defended only as a conjecture about a structural description (more or less remote from the surface form), under which the utterance receives its interpretation.

Nevertheless, once we take the suggested radical step, we are in a position to reinstate the dovetailing. All we need do is conceive the truth theory as stated in a language whose sentences are LF-representations, and then stipulate that:


[A said [that].sub.s]


([exists] u) (There is a theory, interpretative for A as he produces u, which canonically entails this:

[[u is [True].sub.s] [[iff].sub.c [[T].sub.s][.sub.s]) where [[T].sub.s] now goes proxy for an LF-tree, and where [[…][sub.s] [iff][sub.c] [—][sub.s]] sub.s] is is an LF-tree in which […][sub.s], the sentential connective [iff][sub.c], and [—][sub.s] are the principal constituents.

In fact, it will assist matters later if a minor reformulation is made at this stage. Let us say that a declarative utterance u’ R-relates(21) to a declarative utterance u – u’Ru for short – just in case (i) u’ is an instance of the LF-tree T and (ii) there is a theory, interpretative for A as he produces u, which, when applied to u itself in the context of u’, canonically entails:

[[u is True][sub.s] [iff][sub.C] [T][sub.s][sub.s].

With this definition, we can replace Davidson’s stipulation by the formula:


[A says that][sub.s]



There’s an utterance of A’s to which that one R-relates.


A couple of features of the explanation of R-relatedness require comment.

In the first place, it will be noted that the explanation proposes a theory of a truth-predicate, “True”, which applies to individual sayings – that is, to those individual, unrepeatable utterances or speech episodes in the course of which something gets said. When context-dependent expressions are in question (as they will be in this paper) such a truth-predicate is, I think, the predicate it is best to axiomatize when casting in truth-theoretical form a theory knowledge of which suffices for understanding, for it is the individual utterances that somebody will either understand or fail to understand. It must be well understood, however, that the adoption of such an approach requires a generous construe of the word “theory”. For sure, such a “theory of interpretation” will include axioms which specify (in the familiar way) the semantic contribution of words (sc. word types) and thereby the truth conditions of sentences (sc. sentence types). But if it is to deliver theorems which tell when particular speech episodes are True or not, the theory must also include axioms which record information about the contexts of those episodes. So, one can truly report the saying u, which is Dr Lauben’s uttering of the words “I have been wounded”, as a saying that Lauben has been wounded, and u is indeed True just in case Lauben has been wounded. But any theory that delivers that last biconditional must record not merely the generic sense of the English word “I” – that is to say, the fact that any utterance of “I” designates its utterrer – but also the specific contextual fact that the utterer of u was Lauben. There is, however, nothing surprising about the claim that knowledge of some such context-specific fact is needed if one is to understand the utterance in question.

This aspect of the formulation is designed to take account of the fact that the utterance being reported may contain expressions (like “I”) whose contribution to the Truth-grounds of the utterance depends upon features of its context. The italicised clause in the formulation of R-relatedness (i.e. the notion that the theory is applied in a particular context) is designed to take account of the fact that the same holds good of the utterance produced in order to make the report.

How it does so is best seen by example. Suppose Tony says “Rumfitt is a fool”. Then, given that the context determines that that particular utterance of the name is indeed used to designate me, I can truthfully report:

Tony said that I am a fool.

Can the present version of the paratactic theory account for this? We start, as usual, by parsing “Tony said that I am a fool” as

[I am a fool][sub.s]

Tony said that,

which is now glossed as

[I am a fool][sub.s]

There’s an utterance of Tony’s to which that one R-relates. Let us call the heralding utterance of this report u’. By the explanation of R-relatedness, the report will be true just in case there’s an utterance u of Tony’s and a theory which (i) is interpretative for Tony’s speech at the time of u, and (ii), when applied to u in the context of u’, entails:

[[u is True][sub.s] [iffl[sub.c] [I am a fool][sub.s]][sub.s].

The force of the italicised clause should now be clear. Among the features that constitute the context of u’ is the fact that I, Rumfitt, was its producer. So long, then, as we understand the theory to be propounded in a particular context, there is no indeterminacy in using an indexical expression such as “I” to give the Truth-conditions of an utterance. In particular, the context of u’ determines (among other things) who is speaking, and thereby determines what has to be the case for a saying in which “I” is used to be true. This elucidation of the Truth conditions of such reports does deliver the results that we expect. Tony, we are supposing, does say “Rumfitt is a fool”, using the name “Rumfitt” to designate me. Familiar compositional principles will then deliver the theorem

[[u is True][sub.s] [iff][sub.C] [I am a fool][sub.s]][sub.s],

on the basis of such axioms as

In the context of u, the name “Rumfitt” denotes me.

What is more, a set of axioms of which this is one can give an entirely adequate understanding of Tony’s speech. Thus may the truth of the report be explained.

The present theory of indirect speech is already perfectly placed to handle reports made using indexicals. All that is needed is the admission of indexical expressions into the language in which we formulate theories, knowledge of which would confer understanding, and the coeval admission that at least some T-theorems must themselves be understood as propounded in particular contexts. But since the relevant context is always easily recoverable (invariably being the context of the report itself) we have yet to find a reason why such an admission should be problematical.


It is, however, an admission that Davidson himself has been reluctant to make. In his paper “Reply to Foster”, we read him stipulating that

In general, an adequate theory of truth uses no indexical devices, and so can contain no translations of a very large number and variety of sentences,(22)

and while (to my knowledge) he does not explicitly repeat this stipulation elsewhere, some such restriction appears to guide his conception of what an adequate truth theory might be.

There is a weak reading of this restriction which does not conflict with the proposed treatment of reports made using indexicals. The axioms of the truth theory capable of dealing with the English indexical “I” will deliver such theorems as:

(u) (z) (Of (u, “I have been wounded”) & By (u, z) [contains]

True(u) = z has been wounded)

in which, indeed, no index cal devices are used. But the second conjunct of the quoted sentence makes it clear that Davidson intends his restriction to apply, not just to the proper axioms of the truth theory, but more generally to the metalanguage from which any instantiating singular term will be drawn. We have to ask what could possibly motivate such a restriction. Reports like “He said that I was a fool” and “She said that that man over there had threatened her” may, in context, possess Truth conditions which are as determinate as are those of any report. If, like Davidson himself, one sets about developing theories, knowledge of which would confer understanding, it is natural to hope to use the notion of such a theory to elucidate these Truth conditions. But, as I hope to have shown, the natural way of doing just that on a paratactic theory requires the presence of indexical expressions in the metalanguage. What reasons could there be, then, for working within a metalanguage which is restricted precisely in lacking such expressions?

On so a fundamental a question of method in semantics, it is disappointing that Davidson himself does not advance any argument in favour of the restriction he favours. One might conjecture that he was moved to impose it in order to save the

idea that the “meanings” assigned by the theory are generic features of words which are instantiated on a variety of different occasions of utterance. But of course the present treatment involves no departure from that idea. All it involves is the recognition that that generic knowledge might need supplementing with knowledge of the particular context of utterance if it is to yield knowledge sufficient for understanding, and that, surely, is something that must be recognized on any view. The point here can be made, I think, with any indexical expression, but perhaps it emerges most strikingly in respect of the second-person pronoun. I might know as much as can be known about the generic use of the English word “you”, but if somebody says to me “You are in danger”, I will not fully have understood his utterance unless I know that I was indeed the addressee. That knowledge is certainly not generic, but is specific to the particular context of utterance.(23)

Tyler Burge, however, has a more intricate argument for Davidson’s restriction:

Whereas the object-language user freely relies on context to complete the semantical interpretations of his sentences, the meta-theorist should not follow suit in explicating truth conditions. For no reasonable explication of the contribution of demonstrative constructions to the truth conditions of containing sentences will result from simply carrying the construction over into the metalanguage. What is needed there is sufficient generality to account for the referential variability of these constructions. (1974, p. 212)

The last two sentences of this passage plainly allude to a treatment of the semantics of demonstratives that was briefly sketched by Davidson himself in “Truth and Meaning”, but which was rejected there in part for the very same reason:

No logical errors result if we simply treat demonstratives as constants [Davidson here cites Quine]; neither do any problems arise for giving a semantic truth definition. “|I am wise’ is true if and only I am wise”, with its bland ignoring of the demonstrative element in “I”, comes off the assembly line with “|Socrates is wise’ is true iff Socrates is wise” with its bland indifference to the demonstrative element in “is wise” (the tense) … [But in objection to this] it could also be fairly pointed out that part of understanding demonstratives is knowing the rules by which they adjust their reference to circumstance; assimilating demonstratives to constant terms obliterates this feature. (1984, pp. 33-34)

This attack is indeed conclusive against a theory in which (for example) the axiom governing “I” is simply

“I ” denotes me.

However, such a theory is certainly not what I am defending. The axiom governing “I” which is needed to generate the theorems we are postulating, theorems such as

(u) (z) (Of (u, “I have been wounded”) & By (u, z) [implication]

True(u) [equivalent] z has been wounded), will be along the lines of

(v) (x) (z) (Of (v, “I”) & By (v, z) [implication] des (v, x) [equivalent] x = Z)(24)

and such an axiom (unlike the principle: “I” denotes me) precisely does record “the rule by which |I’ adjusts its reference to circumstance”. What is being suggested is that we can maintain a beneficial correspondence between the truth theory, the theory of understanding and a theory of indirect speech by recognizing that such theories are themselves propounded in particular contexts, and that indexical expressions need not, then, be alien to the languages in which those theories are expressed. But that suggestion consists with a proper appreciation of the inadequacy of the principle

“I” denotes me

as a description of the semantics of “I”.

There is a sense in which Burge’s argument reflects a penchant for imposing rather crude formal tests of adequacy – a penchant which afflicted the early development of Davidsonian semantical theory. Justly wanting to rule out from serious consideration the so-called “trivial” truth theory, whose axioms were all instances of the form

Tr (“p”) [equivalent] p

it was common for early theorists to impose as a formal requirement on an adequate truth theory that it be finitely axiomatized, somewhat as Burge suggests that we rule out the axiom

“I” denotes me

by formally eliminating indexical expressions from the metalanguage. But problems of this nature surely do not warrant such draconian restrictions. The trivial truth theory may be ruled out from serious consideration simply because it is trivial; i.e. because it contributes nothing to the resolution of the problem that presumably drove us to start constructing semantical theories in the first place – the problem, I mean, of coming to see how our understanding of words can combine to compose our understanding of whole sayings. Similarly, the axiom

“I ” denotes me

fails for exactly the reason Davidson and Burge give: it fails to record the way “I” varies in reference with circumstance. But there is no reason to erect a monument to this failure by completely eliminating indexical expressions from the metalanguage.


One might, indeed, risk drawing a broader moral. The reader may have wanted to protest that at least the statement

In the context of u, the name “Rumfitt” denotes Rumfitt generates content-ascriptions which are of wider currency than those generated by the statement:

In the context of u, the name “Rumfitt” denotes me The second statement, unlike the first, can figure only in one man’s truth theory.(25) The truth of this protest is incontestable; but one can and should question its purported significance. For the fact that names enjoy wider currency than a word like “I” does not, I think, gainsay the fact that contextual factors are equally important in determining what a name designates,(26) so that whatever difference might obtain between the statements just displayed, it is not that the first, unlike the second, is eternal. If that much is right, there is no real prospect of finding a specification of the content of sayings made in a metalanguage which is completely free from indexicality. For if the sayings whose content is to be specified involve referring expressions, then it is hard to see how a metalanguage in which they could be reported could itself manage without such expressions.(27) Yet if names are tainted, then surely all the familiar categories of referring expression are touched with the brush of indexicality.

To abandon the quest for a context-independent specification of what so-and-so says is, I think, to depart in a crucial respect from the Fregean conception of the content of an utterance. Of course, with his concrete ontology of utterances standing in various content-conferring relations, Davidson himself is already far away from the baroque metaphysics of Fregean Gedanken, but his insistence upon indexical-free specifications of truth-conditions may perhaps be understood as a descendant (within his more austere framework) of Frege’s notion that the content of a saying is an item in a timeless, objective “third realm”. Yet if part of what characterizes the timelessness and objectivity of such a realm is the possibility of specifying the contents in it using only eternal sentences, then the quest for it would appear to be chimerical. Rather, in thinking about content, we need to take into account the context of the content-specifying report as carefully as we do the context of the saying which is being reported. In particular, we need to recognize that the truth or falsity of C’s report to D

A told B that such-and-such

might depend upon a relationship between the way C and D exploit their common linguistic and contextual knowledge and the way A and B exploited their common knowledge in transmitting and receiving the original message. I will return to this briefly at the end of the paper.

III. Deferred ostension and the notion of a proposition


Whether one utterance reports (i.e. R-relates to) another can depend, then, on contextual features of both the reporting utterance and the utterance being reported. As explained, we can re-cast Davidson’s idea that the report relation might be explicated with reference to an interpretative truth theory in a fashion that takes account of this. But in the various r-evisions of the theory so far, there is one original Davidsonian notion to which we have clung: viz. that the item picked out by the demonstrative “that” (at whatever grammatical level that demonstrative might appear) is an individual, unrepeatable speech episode. That notion, however, generates difficulties of its own, and it is time to see how to overcome those.

The basic difficulty here shows up in a number of ways, but perhaps it emerges most simply in the “counting problem” for Davidson, constructed by McFetridge 1975).(28) Suppose that the only words that Galileo uttered under his breath in the course of an interview with Bellarmine were “Si muove la terra”. Then, it seems, we can truly say: “On that occasion, Galileo said just one thing under his breath, viz. that the earth moved”. The aptness of the locution “viz.” in this context makes it plain that the “say” here is the “say” of indirect speech. But if the “say” of indirect speech is a transitive verb, relating a speaker to a demonstrated utterance, we will surely expect the things said to be utterances too. And, as McFetridge observes, just this is incompatible with the thought that Galileo said just one thing under his breath on that occasion. For consider the dialogue:

A: The earth moves.

B: Galileo said that.

A: The earth moves.

B: That’s another thing Galileo said. McFetridge 1975, p. 3 1)

If Davidson’s treatment were correct, B’s final comment would be perfectly in order. And yet, surely, we twentieth-century reporters cannot so easily falsify the claim: “On that occasion, Galileo said just one thing”.

As McFetridge recognizes, the obvious move to make (a move, it was noted at the outset, entirely in the spirit of the paratactic enterprise) is to suppose that the things said are picked out by deferred ostension”. Even in quotidian discourse, examples of this abound. Pointing to my own copy of St. Augustine’s Confessions,

I might say “Of all books, that one has had the most regrettable influence on the culture of the West”; I am understood as making my predication of the book-type that the immediately demonstrated book-token instantiates, and not of the book-token itself. Our question must then be: is it possible to identify or to construct a kind of object (presumably abstract) one instance of which will “stand behind” distinct utterances of the words The earth moves” (among other things) in the way that one book type, the Confessions, stands behind various distinct volumes?


McFetridge himself offers an interesting answer:

[Suppose that] a class of utterances [are] sufficiently alike for there to be at least one utterance which reports all and only the members of the class. Such an utterance binds the utterances it reports into a unity, so I shall call it a binder of the [class, and call such a class a proposition]. Now it seems plausible to maintain that if an utterance reports any utterance then it reports itself i.e. the relation reports is reflexive. Thus the binders of propositions will be members of the propositions they bind. My claim … is that the demonstrative “that” in attributions of sayings is characteristically uttered in the presence of an ensuing utterance, but that its referent is not that utterance but the set of utterances bound by that utterance. (McFetridge 1975, p. 141-2)

McFetridge himself does not attempt to say anything very substantial about the report relation, but his suggestion may be adapted to the present account of it. In an utterance of the LF-trees of Figure 6, and where u’ is the component utterance of the second tree, [S.sub.1], the proposal is to construe “that” as denoting the class of utterances reported by u’, i.e. {u: u’Ru}. Evidently, just one such class may have many members.

This proposed emendation, however, while elegant and ingenious, is on the wrong track. McFetridge himself dismisses the obvious difficulty over (for example) “There are many things that nobody has ever said”, suggesting that insofar as this is true, the quantifier may be construed substitutionally; but even if that were so, which I doubt, a less tractable problem remains. For suppose – a supposition for which there are powerful arguments(29) – that a set has to have the members that it has in fact, and consider the counterfactual:

([alpha]) Had Charles spoken at all, he would have said that the earth moves. On the present elaboration of the paratactic theory, this ought to mean:

([beta]) Had Charles spoken at all, he would have produced an utterance which belongs to that proposition: the earth moves,

where the demonstrative “that” is understood as passing through the unrepeatable utterance to reach the set that utterance binds. But which set is that, exactly? If we suppose that demonstrative expressions are “rigid” in retaining their actual designations even in modal and counterfactual contexts, we must conclude that the designation of the demonstrative here is the set of all utterances to which the reporter’s utterance of the earth moves” actually R-relates. And surely, no matter how Charles had spoken, he could not have produced an utterance which belonged to that set. For, if the assumption about sets is correct, then the only utterances which could belong to the set are those that actually do belong to it, and an utterance that Charles might have made but in fact did not will hardly be among those. A claim in the form of ([beta]), then, is demonstrably always false; but ([alpha]) could easily be true.

What generates this difficulty, at bottom, is the fact that sets are constructions from their members: the identity of a set is determined by the identities of its members, or by its lack of members, if it is the empty set. And the way to escape from these problems is to choose as the object of the deferred ostension not a set but a type – an abstract entity whose identity does not depend upon the identity of its members, but instead depends upon the condition which a token meets or would have to meet in order to instantiate it. In specifying the kind of type which is helpful to us here, the R-relation will play a crucial elucidating role; but it will do so as part of a specification of the condition for an object to instantiate types of this kind, not as part of any set-theoretical construction.

The specification that is needed here, I think, proceeds in two stages. 30 The first stage is the identification of a type whose tokens are individual utterances, i.e. the identification of an act. Specifically, we introduce the notion of a propositional act by stipulating that the condition for an utterance u to be a token of the propositional act bound by u’ is that u’ report u, i.e. u’Ru. (It is understood that this same condition applies in counterfactual circumstances, even when the utterances which there meet the condition are not those which actually meet it.) Supposing, now, that the utterance u’ is the utterance which heralds a given paratactic report, one might be tempted to suppose that the denotation of the demonstrative in that report could be the propositional act bound by u’; but this, I think, is not quite right. The denotation of the demonstrative ought, ex officio, to be the thing said (in the sense of the verb “say” relevant to indirect speech); but the thing said by producing a declarative utterance surely cannot be identical with the thing done by that production. The thing done is: saying that such-and-such is the case. But the thing said is simply: that such-and-such is the case.

Nevertheless, the identification of propositional acts brings us very close to the identification of propositions themselves, for we can introduce propositions in one-to-one correspondence with propositional acts. That is, we can stipulate:

The proposition associated with the propositional act A1

the proposition associated with the propositional act A2


to do [A.sub.1] is to do [A.sub.2]

i.e. iff

act [A.sub.1] = act [A.sub.2].

Then, in a paratactic report heralded by utterance u’ we may, I think, construe the “that” as denoting the proposition associated with the propositional act bound by u’; for short, the proposition bound by u’. So, a simple report in the form

A said that p is construed along the lines:


A said that proposition

where the nature of the deferred ostension is as explained. Of course, on this account of the matter, the truth condition of a simple report will only be as clear as it is clear which token utterances (i.e. which sayings) are instances of which propositional acts. And it will be clear which sayings instantiate which acts only to the degree that it is clear how sayings are R-related. But that relation in turn was elucidated by reference to a theory, knowledge of which would confer understanding of the utterances of the relevant speaker at the relevant time. So, we can only be clear what a man said (i.e. which indirect reports of his speech are true) to the degree that we are clear how to understand his utterances at the time he made the speech being reported. This dependency is very far from being unwelcome.

The present conception of indirect reports validates our initial inference from

(1) Galileo said that the earth moves to

(2) Among the things said by Galileo was that the earth moves. The inference is validated in a way that permits the thing said by Galileo to be the very same as the thing said by Copernicus. For the identity of the thing said goes with the identity of a type, an act: and utterances by both Galileo and Copernicus could very well have been tokens of one and the same propositional act. 3.

The notion that the “that” of parataxis denotes the proposition bound by the heralding utterance u’ obviously requires that there be one and only one such proposition, i.e. that there be one and only one such act. Since the act in question is defined by a condition on u’, this requirement will be met if the condition is determinate, i.e. if it is determinate which utterances u’ reports. Perhaps this will be determinate only against the backdrop of some particular scheme of interpretation;” making that assumption, though, there is no reason to baulk at uniqueness.

The uniqueness in question, however just consists in this: that (one and) only one proposition is bound by any given utterance. It is crucial not to confuse the relation “u binds p” with another relation “u expresses p”. A definition of this latter relation, which I hope is fair to those philosophers who have used it, may be given in the style: utterance u expresses proposition p iff

(i) p is associated with a propositional act A and

(ii) u is an instance of A.

We noted at the outset the prevalence of the assumption that each declarative utterance expresses one and only one proposition, but the present definition suggests that this assumption (in particular, its component assumption of unique expression) is much less secure than one might at first suppose. Let us call a philosopher non-Russellian if he thinks that there can be utterances of “a is F” and “b is F” which do not R-relate, even though the terms “a” and “b” co-denote. Then it is, I think, demonstrable that any consistent non-Russellian needs to be prepared to deny uniqueness of expression.

Consider the situation described in Robert Louis Stevenson’s story The Strange Case of Dr Jekyll and Mr Hyde, in which Dr Jekyll, a rich and respectable London doctor, produces a drug by taking which he can convert himself into a form entirely under the control of his baser impulses, in which guise he frequently appears in London, assuming the name “Mr Hyde”. Suppose that one of the characters, A, says (let this be utterance [u.sub.1]):

Mr Hyde murdered Sir Danvers Carew.

Then, I think, if a non-Russellian is ever going to be right, he is going to be right in this: that [u.sub.1] is not properly reported as a saying that Dr Jekyll murdered Carew. That is, an utterance U2 of the sentence

Dr Jekyll murdered Sir Danvers Carew does not R-relate to [u.sub.1].

However, we must set against this another fact. Mr Hyde himself – i.e. Dr Jekyll himself – can correctly report the remark u, by saying:

A said that I murdered Carew.

What is more, had A said [u.sub.2], he (Jekyll) could also so report that remark. Accordingly, if [u.sub.3] is an utterance by Hyde, i.e. by Jekyll, of

I murdered Carew then our three utterances am R-related as follows:

[u.sub.3] R [u.sub.1], [u.sub.3] R [u.sub.2], but not [u.sub.1] R [u.sub.2] nor [u.sub.2] R [u.sub.1].

It will not trouble the non-Russellian that some of the u1 are R-unrelated even though each predicates the same attribute of the same object. The failure to be so troubled is what makes him a non-Russellian. But he needs to recognize that this system of R-relations is inconsistent with the hypothesis that each utterance expresses a unique proposition. For let [A.sub.i] be the propositional act bound by [u.sub.1]. Then [u.sub.1] is an instance of [A.sub.3], as well as of [A.sub.1] But [A.sub.3] is not the same act as [A.sub.1]. They cannot be identical (as types) for in the case described they have different tokens. In particular, since [u.sub.3] R [u.sub.2], [u.sub.2] is an instance of [A.sub.3]. But since not [u.sub.1] R [u.sub.2], [u.sub.2] is not an instance of [A.sub.1]. The utterance [u.sub.1], then, is an instance of distinct propositional act, and therefore expresses distinct propositions.

The thought that indexical expressions create special problems for non-Russellians is a commonplace of current discussions.(32) But these discussions are shallow insofar as they ignore the possibility that (i) the non-Russellian might be dead right whilst (ii) the unique expression principle is false. And if (ii) holds good then, even though there may be things that are said, it will not do to foist on the semanticist the project of systematically determining the proposition that each utterance expresses.

A thorough investigation of this possibility must be left for another occasion, but it may be worth mentioning one way in which the conclusions of [subsections] 11 bear upon the debate about the viability of non-Russellian treatments of the attitudes generally. An issue which justly looms large in that debate is the question, how to describe the state of belief of Kripke’s character Peter, who assents to an utterance [U.sub.p] of “Paderewski was musical” when he takes the speaker to be referring to the pianist, but dissents from an utterance [u.sub.s] of “Paderewski was musical” when he takes the speaker to be referring to the statesman, not realizing that one man was both pianist and statesman (Kripke 1979, pp. 265-6). An apparently compelling principle connecting belief with assent(33) appears to show that Peter both believes and disbelieves that Paderewski was musical, a conclusion which, according to Kripke, misdescribes Peter’s state of mind by representing him as having contradictory beliefs.(34) Indeed, the difficulty here has led Kripke to wonder whether “the apparatus of ‘propositions’ does not break down in this area”(Kripke 1980, p. 21).

Now everybody who uses that apparatus recognizes that the identity of the proposition picked out by a phrase of the form “that such-and-such” will in general depend upon features of the context in which the phrase is uttered. This is obvious when the feature in question is a gross feature of the context, as it is in the case of the phrase “that I am a fool”. But under the present way of individuating propositions, which ties the individuation to a contextually embedded interpretative truth theory, the identity of the proposition picked out by such a phrase might depend upon aspects of the context which are far finer than (for example) the identity of the theory’s proponent. In particular, it might very well turn out that a truth theory which delivered the theorems

True([u.sub.r]) [equivalent] Paderewski was musical and

True([u.sub.s]) [equivalent] Paderewski was musical

was interpretative for Peter only when propounded in a context where the communicative intention manifested in the theorist’s first use of the name was to be understood as referring to the pianist, while that manifested in the second was to be understood as referring to the statesman.(35) In particular, we should not have an interpretative theory if this association between communicative intentions and occurences of the names were reversed. Now suppose that [u’.sub.p] is an utterance by the reporter of “Paderewski was musical” in which the reporter intends to be understood as referring to the pianist; and analogously for [u’.sub.s]. Then the various utterances would be R-related as follows:

[u’.sub.p] R [up;.sub.u’], R [u.sub.s;] but not [u’.sub.s] R [u.sub.p;] and not [u’.sub.p] R [u.sub.s].

Under these assumptions about interpretativeness, the proposition bounded by [u’.sub.p] will not be identical with the proposition bounded by [u’.sub.s]. That is to say: the proposition that Paderewski was musical uttered with a communicative intention of the first kind) will not be identical with the proposition that Paderewski was musical (uttered with an intention of the second kind).

This, I think, removes any threat of paradox here. If the communicative intention made manifest in the use of a name is relevant to the content of a saying involving that use, then it will not do in cases like this to formulate the principle linking comprehending assent with belief in terms of sentence types. What we need is something along the lines of:

(B) If A sincerely and comprehendingly assents to (dissents from) u, and if

u expresses the proposition that p, then A believes (disbelieves) that p.

By what was just shown, it makes no sense to ask which, of Peter’s utterances express the proposition that Paderewski was musical unless that question is itself understood as asked in a particular fine context, i.e. with particular communicative intentions. (Asked in a context of the first type, the answer will be [u.sub.p;] asked in the second, it will be [u.sub.s].) Furthermore, (B) will validate the attribution “Peter believes that Paderewski was musical and disbelieves between the two occurrences of the name in the report. As explained, this shift will mean that the proportions picked out in the two halves of the attribution are not the same.

I conclude that the present apparatus of propositions does not break down over this case. For although the apparently problematic attribution is validated by (B), that attribution does not amount to a claim that Peter both believes disbelieves a single proposition.(36) (1) Notably by Quine, of course, whose phrase this is. (2) Frege 1919, p. 390. (3) Kaplan 1989, P. 494. (4) Cf. his 1971. For interesting discussion of this view, cf. Rundle 1979, pp. 287-93. (5) Cf. especially Salmon 1986. Name notwithstanding, the origins of such treatments are more appropriately sought among the sands of Greece than among the snowfields of Mont Blanc. For it is an Aristotelian doctrine (albeit one propounded only in passing) that two men “say one thing” whenever they predicate the same attribute of the same object. Cf. De Interpretatione 18a1 3; Posterior Analytics 93b35; Poetics 1457a28. (6) Inevitably, I mean, if one assumes that the verbs “say”, “believe”, “hope” etc. express relations between, on the one side, a thinker or speaker and, on the other side, whatever is the designatum of a “that-clause”. For then if the designatum of “that Hesperus is bright” is identical with that of “that Phosphorus is bright”, one will be unable to say/believe/hope that Hesperus is bright without saying/believing/hoping that Phosphorus is. And that, at least at first blush, is implausible. (7) Davidson 1969. (8) Cf. Davidson 1971, n. 12. (9) The idea of using Quine’s notion of deferred ostension in emending Davidson’s theory of indirect speech is due to Ian McFetridge. See his 1975, which I discuss in [sections] Ill below. (10) Davidson 1979. (11) Cf. Davidson 1984, p. 9 1: I refer to an expression, but I do it by way of indicating an embodiment of those words in an utterance. For an interesting anticipation, cf. Wittgenstein, Philosophical Investigations, [sections] 16: when I say to someone: “Pronounce the word ‘the”‘, you will count the second “the” as part of the sentence. Yet it has a role just like that of a colour-sample in [a previous] language-game; that is, it is a sample of what the other is meant to say. Of course, Davidson differs from Wittgenstein precisely in not counting the second “the” as, semantically, part of the sentence. (12) Davidson cites with approval the O.E.D. conjecture that the complementizing use of “that” descended from the demonstrative. Cf. also Jespersen 1949, Part III, [sections] 2.3. But the suggestion remains just that: an etymological conjecture. (13) Neglecting the tenses. (14) For general discussion of them see May 1985. (15) That this is accidental puts us out of the reach of the objection that Davidson’s theory must be wrong because it doesn’t, for example, generalize to cover French. Cf. for this objection Schiffer 1987, at p. 125. (16) Although I cannot deal with it fully, it may be worth mentioning another influential recent objection to the paratactic proposal, which derives from the fact that elements inside the “that” clause often interact with elements in the main reporting sentence in ways that are relevant to interpretation. Thus James Higginbotham 1986, p. 39: Suppose I write on the blackboard (36) and, pointing to it say (37): (36) He is a nice fellow. (37) Every boy that. In my judgement, it is quite impossible to get this activity to convey that every boy a good opinion of himself; but just this is one of the things I can surely convey by saying (38): (38) Every boy that he is a nice fellow. (Just to ensure that Higginbotham’s point bears directly on present concerns, I have replaced “believes” by “says” as indicated.) I think, however, that this objection has much less force than first appears. (36) can be thought of as an open sentence (cf. “x is a nice fellow”), as can the second tree we will get when we divide the S-structure of “Every boy that he is a nice fellow”. Then, so far as I can see, we have freedom of theoretical manoeuvre in deciding when speaker A stands in the saying relation to an utterance of “x is a nice fellow”, and the example only encourages us to decide in favour of the answer: when A predicates being a nice fellow”, or something tantamount to that, of himself. For a similar treatment, showing how this thought might develop to cover more complicated cases, see Hornsby 1977, which elaborates the final paragraph of Quine 1969.) It may be thought that Higginbotham’s point is that the “he” in (36) is the so-called “indirect reflexive” pronoun – Castaneda’s “he*”. This is unlikely, in view of the fact that Higginbotham 1989 contains powerful arguments against the existence of such a pronoun. But to anybody inclined to raise the objection in that form, the same sort of reply may be given: A will stand in the saying relation to an utterance of “x* is a nice fellow” just when A says something tantamount to “I am a nice fellow”.

(17) On this point, cf. especially McDowell 1980, [section]X. (18) Davidson 1969, in a footnote added for the reprint in Davidson 1984, at p. 104. (19) “Say” with a capital “S” being the three-place predicate postulated by Davidson in the last quotation. (20) For the notion of canonical entailment, cf Peacocke 1976, p. 188, note (b). Is it going to matter if there is no uniquely best such theory? This question raises two sets of issues. The discussion below is going to exploit the fact that the theories of different interpreters might reflect their different perspectives on the utterances being interpreted. But these differences will reflect no disagreement about interpretation; the various theories, taken in their various contexts, will be compatible, even if none has priority over any other. A far deeper question is whether there will be different schemes of interpretation, that is, sets of contextualised theories that interpret A’s speech in incompatible ways. I will write as if there are not; but if there are, the notions introduced below, including crucially those of R-relatedness and of the proposition bound by an utterance, must be taken to be relative to a particular choice of scheme. (Cf. also n.31 below.) (21) “R” for “reports”. (22) At Davidson 1984, p. 175. “Adequacy” here amounts to passing the “formal and empirical” tests whose semantical importance the paper is devoted to championing. Chief among the “empirical” tests” is conformity to the so-called Principle of Charity.

(23.) The quantified conditional form of the generic theorem for “I have been wounded”, displayed above, shows how the axioms and some of the theorems of the semantical theory will be interpersonal in the sense of being attributable to any competent English speaker. As explained, however, such generic theorems may need to be instantiated by indexical terms before they yield a T-theorem, knowledge of which would suffice for understanding a particular utterance. Unfamiliar as this may be, I see no awkwardness: the interpersonal axioms and theorems will delimit whatever is common to the speech of diverse speakers, whereas the differences between the instantiations will reflect their different perspectives on a single utterance. (24.) Writing “des(v,x)” to mean that the event v of uttering a particular noun-phrase picks out or designates the object x. (25.) Note, though, that because both these statements can figure as axioms in my truth theory for Tony’s speech, the theorems “u is true iff Rumfitt is a fool” and “u is true iff I am a fool” can both be canonical consequences of that theory. (Cf. note 20 above.) (26.) I cannot argue for this here. There are excellent arguments in Cohen 1980. (27.) For how else could one report what a saying was a saying about? By description, perhaps. But if one looked down that road for a way out of the present difficulty, one would first have to confront Strawson’s argument (Strawson 1959, p. 20) that descriptions need to be anchored to objects by indexicals on pain of indeterminacy in circumstances of “massive duplication”. In any case, precisely because the description “The F” imports conceptual material lacking in a singular term “a”, the sentence “The F is G” is ill-placed to specify the content of an uttering of the form “a is G”. (28.) For further paradoxes for Davidson which seem to call for a similar solution, v. Burge 1986, especially [subsections]IV, pp. 200-206. (29.) For a variety of arguments, see Wiggins 1980 Ch. 4, Fine 1981, and Parsons 198 3, pp. 306-8. (30.) The general idea of the specification that follows was suggested to me in conversation by Professor David Wiggins, for which I am grateful to him. (31) In the sense of “scheme” introduced in n. 20 above. Convinced as he is that there will be a multiplicity of such schemes, Quine supposes that this constitutes an argument against the whole idea of positing propositions. For, insofar as we take such a posit seriously, we thereby concede meaning, however inscrutable, to a synonymy relation that can be defined in general for eternal sentences of distinct languages as follows: sentences are synonymous that mean the same propositions. (Quine, 1960, pp. 205-6) I am not sure what Quine means by “taking the posit seriously”, but it is surprising that the future author of “Ontological Relativity” does not consider the possibility that the ontology of propositions might be relative to an ideology of “analytic hypotheses” – an ideology which (or so he is prepared to contemplate in Quine 1969) might well be formulated as a truth theory. (32) Cf. especially Perry 1977; and Kaplan 1989, p. 487. (33) The so-called “disquotational principle”, Kripke 1979, pp. 248-9: If a normal English speaker, on reflection, sincerely assents to p”, then he believes that p.” Since dissent from “p” is tantamount to assent to its negation, and since to disbelieve (the proposition) that p is to believe that not p, this principle entails: If a normal English speaker, on reflection, sincerely dissents from “p”, then he disbelieves (the proposition) that p. (34) “[Peter] cannot be convicted of inconsistency [sc. of “letting contradictory beliefs pass”]; to do so is incorrect”. (Kripke 1979, p. 257) of “letting contradictory be pass’]; to do so is incorrect”. (Kripke 1979, p.257)” (35) Note that the theorist could form these communicative intentions even though he knew that pianist and statesman were one, and knew that his audience knew this.


Burge, T. 1974: “Demonstrative Constructions, Reference and Truth”. Journal of Philosophy, 71, pp. 205-23. 1986: “On Davidson’s |Saying That”‘, in LePore ed., 1986, pp. 190-208. Cohen, L. J. 1980: “The Individuation of Proper Names”, in Philosophical Subjects: Essays Presented to P. F. Strawson, Z. van Straaten, ed., Oxford: Oxford University Press, pp. 140-63. Davidson, D. 1969: “On Saying That”, in Davidson and Hintikka, eds., 1969, pp. 158-74. Reprinted in Davidson 1984, pp. 93-108. 1971: “Eternal vs. Ephemeral Events”. Nous, 5, pp. 179-89. Reprinted in Davidson 1980, pp. 189-203. 1979: Quotation”. Theory and Decision, 11, pp. 27-40. Reprinted in Davidson 1984, pp. 79-92. 1980: Essays on Actions and Events. Oxford: Oxford University Press. 1984: Inquiries into Truth and Interpretation. Oxford: Oxford University Press. Davidson, D. and Hintikka, J. eds., Words and Objections: Essays on the Work of W.V.Quine. Dordrecht: Reidel. Fine, K. 198 1: First-Order Modal Theories 1: Sets”. Nous, 15, pp. 177-205. Frege, G. 1919: “Gedankengefuge”, translated as “Compound Thoughts”, in Frege, Collected Papers, B. McGuinness, ed., Oxford: Blackwell, 1984, pp. 390-406. Higginbotham, J. 1986: “Linguistic Theory and Davidson’s Program in Semantics”, in LePore ed., 1986, pp. 29-48. 1989: “Reference and Control”. Rivista di Linguistica, 1, pp. 301-26. Hornsby, J. 1977: “Saying of”. Analysis, 37, pp. 177-85. Jespersen, 0. 1949: A Modern Grammar of English on Historical Principles. London: George Allen and Unwin. Kaplan, D. 1989: “Demonstratives”, in Themes from Kaplan, J. Almog et al. eds. New York: Oxford University Press, pp. 481-563. Kripke, S. 1979: “A Puzzle About Belief’, in Meaning and Use, A.Margalit, ed. Dordrecht: Reidel, pp. 239-83. 1980: Naming and Necessity. Oxford: Blackwell. LePore, E. ed. 1986: Truth and interpretation: Perspectives on the Philosophy of Donald Davidson. Oxford: Blackwell. May, R. 1985: Logical Form: its Structure and Derivation. Cambridge, Massachusetts: M.I.T. Press. McDowell, J. H. 1980: Quotation and saying that”, in Reference, Truth and Reality, M.Platts, ed. London: Routledge, pp. 206-37. McFetridge, 1. G. 1975: “Propositions and Davidson’s Account of Indirect Discourse”. Proceedings of the Aristotelian Society, 76, pp. 131-45. Parsons, C. 1983: “Sets and Modality”, Ch. 11 of Mathematics in Philosophy. Ithaca: Cornell University Press, pp. 298-341. Peacocke, C. A. B. 1976: “Truth Definitions and Actual Languages”, in Truth and Meaning, G. Evans and J. H. McDowell, eds. Oxford: Oxford University Press, 1976, pp. 162-88. Perry, J. 1977: Frege on Demonstratives”. Philosophical Review, 86, pp. 474-97. Prior, A.N. 1971: Objects of Thought. Oxford: Oxford University Press. Quine, W.V.O., 1960: Word and Object. Cambridge, Massachusetts: M.I.T. Press. 1969: “Reply to Davidson”, in Davidson and Hintikka eds., 1969, pp. 333-35. Rundle, B. 1979: Grammar in Philosophy. Oxford: Oxford University Press. Salmon, N. 1986: Frege’s Puzzle. Cambridge, Massachusetts: M.I.T. Press. Schiffer, S . R. 1987: Remnants of Meaning. Cambridge, Massachusetts: M.I.T. Press. Stevenson, Robert Louis 1886: The Strange Case of Dr Jekyll and Mr Hyde. London: Longman. Strawson, P. F. 1959: Individuals. London: Methuen. Wiggins, D. R. P. 1980: Sameness and Substance. Oxford: Blackwell.

COPYRIGHT 1993 Oxford University Press

COPYRIGHT 2004 Gale Group